TEXT, TIME, AND CONTEXT
Studies in Linguistics and Philosophy Volume 87 Managing Editors KAI VON FINTEL, M.I.T., Cambridge LISA MATTHEWSON, University of British Columbia Editorial Board JOHAN VAN BENTHEM, University of Amsterdam GREGORY N. CARLSON, University of Rochester DAVID DOWTY, Ohio State University, Columbus GERALD GAZDAR, University of Sussex, Brighton IRENE HEIM, M.I.T., Cambridge EWAN KLEIN, University of Edinburgh BILL LADUSAW, University of California at Santa Cruz TERRENCE PARSONS, University of California, Irvine
For further volumes: www.springer.com/series/6556
TEXT, TIME, AND CONTEXT Selected Papers of Carlota S. Smith Edited by
RICHARD P. MEIER University of Texas at Austin
HELEN ARISTAR-DRY Eastern Michigan University and
EMILIE DESTRUEL University of Texas at Austin
123
Editors Richard P. Meier University of Texas at Austin Department of Linguistics 1 University Station B5100 Austin TX 78712 USA
[email protected]
Helen Aristar-Dry Eastern Michigan University Institute for Language Information & Technology 2000 Huron River Dr., Suite 104 Ypsilanti MI 48197 USA
[email protected]
Emilie Destruel University of Texas at Austin Department of Linguistics 1 University Station B5100 Austin TX 78712 USA
[email protected]
ISBN 978-90-481-2616-3 e-ISBN 978-90-481-2617-0 DOI 10.1007/978-90-481-2617-0 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2009926837 © Springer Science+Business Media B.V. 2009 No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)
Preface
This project was begun in February 2007. We, Richard Meier and Helen AristarDry, had the privilege of knowing Carlota Smith as a mentor and friend for many years. In 2006 Richard had broached the possibility of hosting a small conference in honor of Carlota’s long career, but even though Carlota was an avid organizer of almost-yearly conferences on some aspect of cognitive science or semantics or pragmatics she would have none of it. Instead she wished to publish this collection of her papers. So, during a visit of Helen’s to Texas, we—that is, Carlota, Helen, and Richard—sat down to plan this book. We agreed on the parameters of the volume— specifically, to focus on Carlota’s work on tense, aspect, and discourse. Thus, her early work on syntax would not be represented, nor would much of her work on language acquisition. And we agreed on the basic organization of the book, and the particular papers to be included in each major section. This was important progress, but it was the only progress that we would make during the spring semester of 2007. The spring of any academic year is hectic: Richard was in his first year of being department chair, Helen had the ever-pressing duties of the Linguist List, and— most crucially—Carlota was quite ill. Carlota had been diagnosed with cancer in June 2005. She had undergone surgery and chemotherapy and was enduring considerable pain. But through it all she had remained an active member of the Department of Linguistics. She continued to teach very successful graduate seminars on her area of research, actively advised doctoral students, and participated in all aspects of departmental governance. After she had recovered from her first round of surgery, Keith Walters—then a faculty member at UT—joined Richard to interview Carlota at her home. It was John Robertson— Carlota’s husband—who suggested that we conduct this interview; he knew that interviews of longtime law school faculty had proven to be invaluable in understanding the history of UT’s School of Law. Our conversation with Carlota proved to be a fascinating oral history. Hers was an important life in modern linguistics inasmuch as she was one of the very first women to work in generative linguistics. This interview places Carlota’s research in a historical and biographical context that is rarely made clear in academic volumes such as this one. For that reason we publish a slightly redacted, and extensively annotated, version of that interview here. Carlota finished out that spring semester of 2007. She spoke eagerly of her extensive travel plans for the summer—a trip to New York City to see family and friends v
vi
Preface
(and art and theater), an annual family vacation on Martha’s Vineyard, a trip to Arizona for the summer institute of the Navajo Language Academy, even a trip to Japan where she had never been before. She was scheduled to teach the following semester, although Richard was not confident that she would be able to do so. The last time he spoke to Carlota was exactly a week before she died;1 she was having lunch at a campus restaurant with Katrin Erk, an assistant professor in Linguistics. Richard had previously appointed Carlota to serve as Katrin’s “mentor.” The conversation between Katrin and Carlota was animated, as was the conversation with Richard and his wife Madeline when they happened to sit down at a nearby table. But Richard noted that Carlota barely ate. The following Monday, Carlota was—as usual—in her office, meeting with a doctoral student. Richard saw her there, but— because he had no particular sense of urgency—he saw no reason to interrupt their conversation to greet Carlota or to ask how she was doing. That was the last time Richard saw Carlota. She went into a steep decline later that evening, entered the hospice on Tuesday, and died on Thursday May 24. She was 73 years old and was survived by her husband John Robertson, her daughter Alison Smith, her son Joel Smith, and her grandchildren Sylvia and Ari. A Biographical Sketch. Carlota was born in New York City on May 21, 1934. Her family was Jewish but decidedly secular. She grew up in Greenwich Village, when it was still a Bohemian neighborhood. Her father’s career was an unusual one: born in New York City as Charles Phillips, he joined the Communist Party. Under the name of Manuel Gómez, he was one of the founders of the Mexican Communist Party. But Stalin’s rise caused him to become disenchanted with the Party. By 1930 he was writing an investment column for the Wall Street Journal under the byline of Charles Shipman. Shipman would later be Carlota’s maiden name. Charles Shipman’s autobiography, It Had to Be Revolution, was published in 1993 by Cornell University Press. Carlota’s mother Sylvia was an actress who was a member of the Group Theater. We are told by Carlota’s longtime friend Jane Stern that Sylvia was very literate and, as an actress, very precise in her language. Carlota received her bachelor’s degree from Radcliffe College in 1955. In the late 1950’s, she was married to a faculty member, David Smith, at Swarthmore College. Another faculty wife was Lila Gleitman, whose husband Henry was then a psychology professor at Swarthmore. Neither Lila nor Carlota was content to be a suburban housewife, notwithstanding the fact that both were apparently lucky enough to have a Swarthmore undergraduate named Barbara Hall (later Barbara Hall Partee) as one of their babysitters. Lila was already a graduate student in linguistics at the University of Pennsylvania and was studying with Zellig Harris, a pioneer of modern linguistic thought. 1 Helen’s last visit with Carlota had occurred three weeks before. She had come to Austin to discuss the book. And Carlota not only invited Helen and her husband to stay at her house but also insisted on having mutual friends over for champagne and caviar. Only later did Helen realize that Carlota was already under hospice care. Those who knew Carlota will recognize in this incident not only her zest for living, but also her knack for creating an elegant occasion out of even the most daunting circumstances.
Preface
vii
On Carlota’s recounting, it was Lila who decided that Carlota would make a good assistant for Harris. So Carlota became a research assistant and then a doctoral student in the Department of Linguistics at Penn. Zellig Harris had directed Noam Chomsky’s doctoral dissertation (Chomsky 1955) and would later direct Carlota’s as well (Smith 1967). In 1961, Carlota spent a year away from Penn studying at the Massachusetts Institute of Technology in Cambridge, MA, where she was one of the very first woman students to work with Chomsky. After receiving her M.A. (1964) and Ph.D. (1967) at the University of Pennsylvania, Carlota joined the faculty of The University of Texas at Austin in 1969, where she was a member of the Department of Linguistics until her death. Interestingly, her first contact with UT’s faculty seems to have been with David Hakes, who was a developmental psychologist working on first language acquisition. She was initially hired as a “faculty associate” under an enormously successful NSF training grant held by Win Lehmann and Gardner Lindzey. Other faculty members in Linguistics and Psychology who came to UT under the aegis of this grant were Donald Foss, Phil Gough, Harvey Sussman, and Robert Wall.2 Carlota served as the chair of the department from 1981 to 1985. During this time she was active in efforts to establish a Center for Women’s Studies. She was director of UT’s Cognitive Science Center from 1987 to 1994. In 1991, she was named the Dallas TACA Centennial Professor in the Humanities. TACA stands for “The Auction Center for the Arts.” Given Carlota’s deep interests in the visual arts, this seems a very appropriate professorship for her to have held. Carlota’s Research Career. A close look at Carlota’s CV reveals four important strands to her work, three of which are represented in this volume: (1) English syntax, (2) Child language development, (3) The syntax and semantics of tense and aspect, and (4) Discourse interpretation. Carlota’s earliest work was on the syntax of English. Her first publication (“A Class of Complex Modifiers in English”) dates from the year she spent at MIT and appeared in the journal Language (1961). Carlota began that paper with the observation that adjectives can only occur postnominally with English indefinite pronouns: e.g., Bob would like something spicy for dinner, but not Bob would like spicy something for dinner. She also observed that postnominal adjective phrases can readily occur in noun phrases (NPs) with an indefinite determiner, but not in NPs with a definite determiner. So, the sentence I bought a book yellow with age is fine, but not I bought your book yellow with age. Her solution was to derive pre- and
2 Years later Richard would first meet Carlota at another institution of American academic life to which Gardner Lindzey made fundamental contributions: the Center for Advanced Studies in the Behavioral Sciences in Palo Alto. In 1986, Carlota was there on sabbatical from UT, as was Peter MacNeilage, who was also then a member of UT’s Linguistics Department. Richard was a postdoctoral fellow at Stanford. In the Spring Semester of 1986, Richard had had his on-campus interview for the position he now holds at UT. But because both Carlota and Peter were away from Austin, his interview with them was instead on the veranda of the Center, overlooking the Stanford campus and, in the distance, the southern reaches of San Francisco Bay. Academic life has its moments.
viii
Preface
post-nominal adjectivals from relative clauses (which are obviously postnominal). Relative clauses were themselves derived transformationally by merging two sentences that each contained the same noun. She then proposed transformational rules of relative clause reduction and adjective preposing. Relative clause reduction was blocked if two conditions were met: the head of the relative clause was definite and the adjective occurred with a complement. The rule of adjective preposing was obligatory, but was blocked by indefinite pronouns. On this early transformational account, the sentence She has a green hat was derived from a sequence of two sentences: She has a hat. The hat is green. Transformational rules of relative clause formation, relative clause reduction (“whiz deletion”), and adjective preposing yielded the desired result. In 1964 she published a second paper in Language, one that is still extensively cited. This paper further developed the model proposed in her 1961 paper. In this second paper, she noted a variety of interesting dependencies between the determiners of nouns and the kinds of relative clauses that can modify those nouns. Again she proposed a transformational account of the facts. Inasmuch as linguistics has long moved on from the early analyses of transformational grammar, the particular solutions that Carlota proposed in these papers would not be adopted now. However, her papers remain important because of their clear exposition of distributional regularities in the structure of English NPs, and their discussion of the syntax of modifiers within those NPs. During the 1960s, Carlota developed a second strand of work in first language acquisition. The issue of how a child (or indeed any idealized learner) could acquire a first language given the linguistic input available to children was front and center in work in generative grammar (Chomsky 1959, 1965). In 1969, Carlota published, along with Elizabeth Shipley and Lila Gleitman, an influential experimental study of how children acquire English as a first language; in ensuing years she would publish several more papers on child language development. The issue in Shipley et al. (1969) was the relationship between competence and performance in the child. More specifically, they wondered about the extent to which the rather “primitive”looking utterances of the young child reflect his/her actual linguistic competence. To examine this question, they looked at children’s responses to simple commands. As it turned out, the relationship between competence and performance seemed to change across development, such that “telegraphic” speakers, but not one-word speakers, were more likely to respond appropriately to well-formed commands (i.e., the kind they would hear from their parents) than to commands whose syntax was consistent with the child’s own linguistic production. In her subsequent work on first language acquisition, Carlota was concerned with issues of linguistic complexity. In Smith (1970), she argued that the distribution of semantic content across the sentence determined children’s accuracy in an elicited imitation task; specifically, sentences in which one phrase— whether an NP or a VP—dominated a disproportionate fraction of the sentence’s “information-carrying elements” were likely to be difficult for children. This was a property that she referred to as “compression” in 1970, but that she subsequently termed “density” in other work examining sources of complexity in linguistic
Preface
ix
performance (Smith 1988; Smith and van Kleeck 1986). In her acquisition work, she would also examine the acquisition of tense and aspect; it is this acquisition work that we sample here (Smith 1980 and 1993, “The Acquisition of TimeTalk: Relations Between Child and Adult Grammars” and “The Acquisition of Tense: Bootstrapping into Syntax”). Starting in the mid-1970s, Carlota embarked on what was certainly her most important line of research. In many papers—the first of which was published in 1975—and in a very important 1991 book (The Parameter of Aspect), she analyzed how languages encode time and how they encode the ways events and situations occur over time. How did she get into work on tense and aspect? As she explained in the interview that we publish here, this research interest was an outcome of her year in France in the early seventies. English and French are historically related in many different ways, yet have quite different temporal systems. As virtually the first lesson in schoolbook French makes clear, French present tense verbs can be translated into English either using the simple present tense (with its characteristic habitual interpretation) or using the present progressive. Thus, depending on context, the sentence Nous parlons à Marie can be translated either as “We speak to Mary” or “We are speaking to Mary.” Unlike English, French has two past tenses, the imparfait and the passé composé (ignoring the largely antiquated passé simple), which differ in whether the endpoints of an event or situation are “visible.” Thirdly, English— unlike French—does not have a true future tense, but instead uses the modal will. All these differences are ones that she would subsequently explore; for example, in Smith (1986a), reprinted here as “A Speaker-Based Approach to Aspect”, Carlota compared the English and French aspectual systems. In Smith (2007), reprinted here as “Tense and Context in French”, she examined expressions of futurity in French and English. Carlota’s work on the expression of time in language is notable because of its empirical foundation in careful analyses of a number of quite different languages, including not only English and French, but also Russian, Mandarin, and Navajo. Her work on Russian, which was the result of a collaboration with Gilbert Rappaport, is published as a chapter in her 1991 book. In Smith and Erbaugh (2005), reprinted here as “Temporal Interpretation in Mandarin Chinese”, Carlota and her co-author Mary Erbaugh discussed the ways in which time is conveyed in Mandarin, a language that has no grammatical tense. Carlota considered the aspectual system of Navajo in a 1996 paper, “Aspectual Categories in Navajo,” that we reproduce here. Through her many years of research on Navajo, she became a member of the Navajo Language Academy, a group that seeks to further the study of Navajo, to keep Navajo from becoming endangered, and to provide training in linguistic research to members of the Navajo Nation. With the development of her two-component theory of aspect, Carlota found her niche as a graduate teacher. Her model of temporal aspect provided an excellent framework for graduate students seeking to analyze the temporal systems of an array of languages, including under-described languages that are so much the focus of research in UT’s Linguistics Department. The students working with her produced insightful descriptions of the temporal systems of Korean (Ahn 1995),
x
Preface
ASL (Rathmann 2005), Mandarin (Yeh 1993; Ren 2008), Q’anjob’al (Mateo Toledo 2008), and Iquito (Lai, 2009).3 The fourth major strand in Carlota’s work consisted of her analyses of discourse. At different periods, she investigated topic and focus (Smith 1971, 1986b, 1991, 1998), backgrounding and foregrounding (Smith 1995; 1999a, b), and subjectivity (Whitaker and Smith 1984; Smith 2002a, b), as well as orienting her important work in tense and aspect toward the interpretation of sentences in discourse. Indeed, throughout her career, sentence interpretation within coherent texts sparked many of her most original syntactic and semantic observations. And in her last book, it functioned to bind her other interests together: Modes of Discourse, published in 2003 by Cambridge University Press, drew on her previous analyses of viewpoint and situation type, discourse topic, foregrounding, and subjectivity to produce a multi-faceted characterization of five modes, or genres, of discourse. Two of the principles labeled “key insights” in the book are that “Linguistic meaning is often due to a group of forms rather than to a single form” (p. 10) and “Grammatical terms often have two different functions: conveying information, and giving cues to text structure” (p. 11). These principles reflect her insistence on the importance of discourse context in syntactic and semantic interpretation— something that she stressed as early as 1977, when there was almost no generative work on the interaction between discourse and syntax.4 One of her earliest articles on tense (“The Vagueness of Sentences in Isolation”) emphasized the contribution of surrounding sentences to apprehension of the temporal reference of a given sentence. Many sentences, she noted, are underspecified as to temporal reference; and what appears to be a semantic property of the sentence may in fact be an artifact of discourse interpretation. Similarly, her important 1980 article “Temporal Structures in Discourse” (also in this volume) made the point that the interpretation of Reference Time and Event Time often depends on the ‘capture’ of one sentence by another sentence that precedes it in the discourse.5 See Barbara Partee’s introduction to Part IV for further discussion of the links between discourse and tense interpretation in Carlota’s work. Just as Carlota viewed the temporal interpretation of sentences as heavily dependent on the surrounding discourse, so she viewed the discrimination of discourse types as heavily dependent on temporality. In Modes of Discourse (Smith 2003),
3 Co-directors or—after Carlota’s passing—the directors of some of these dissertations included Nora England (Lai, Mateo Toledo), Richard Meier (Rathmann), and Stephen Wechsler (Ren). 4 The field of discourse analysis at this time was still grounded in ethnography and sociolinguistics. And, despite the work of a number of functionalists (see, for example, Givon 1975; Li and Thompson 1976; Hopper 1979; and Hopper and Thompson 1980), generative syntax had not yet integrated many insights drawn from the interpretation of connected discourse. 5 In the 70’s she most often appealed to Reichenbach’s notions of Reference Time, Speech Time, and Event Time for explanation of the temporal properties of sentences in discourse. Later she formalized these properties in the framework of Discourse Representation Theory, a theory particularly congenial to her approach because of its precise explication of context.
Preface
xi
the modes are distinguished first by the temporality of their characteristic situation types (i.e., dynamic, static, or atemporal) and secondly by the type of progression, or text advancement, that they exhibit. The modes characterized by dynamic or static situation types (Narrative, Report, and Description) form the group of “temporal modes,” and their progression is either temporally or spatially based.6 Carlota’s focus on these types of progression serves as her springboard for a new treatment of background; usually considered a temporal phenomenon, background is presented in Modes of Discourse as any deviation from the expected mode of progression. In that book and in earlier papers (for example, her 1984 paper co-authored with Jeanne Whitaker on the French author Gustave Flaubert, “Some Significant Omissions: Ellipses in Flaubert’s Un Coeur Simple”, which we reprint here.), Carlota sought to bring the analytic tools of linguistics to the humanistic study of literature. Carlota was always an appreciative reader of fiction, alert to the textual properties of literary narrative. It was Carlota’s interest in literature—rare among linguists at that time—that was one of the things that drew Helen to her when she (Helen) was a graduate student in the English Language and Linguistics program at UT Austin in the 70’s. During her scholarly career, Carlota often explicated the syntactic and pragmatic principles which underlie literary effects, a topic that she returned to late in her career through the organization of several conferences.7 Thus Modes of Discourse bears witness to the many topics and fields that attracted the interest of her wide-ranging intellect, at the same time as it demonstrates the coherence among her ideas. Concluding Thoughts. As we said at the outset of this essay, Carlota was a dear friend and a valued mentor to us both. In this book we remember the fact that Carlota was an enormously broad intellect. She crossed disciplinary boundaries to link the Department of Linguistics to a range of other departments at the University of Texas, including Psychology, Philosophy, Asian Languages, French & Italian, and others. She pursued her research in the face of turmoil in her personal life in the early 70s and in the face of the cancer that marked her last years. Through the last weeks of her life, one of her greatest concerns was to build the future of the department in which she had spent her entire career. She was an engaging friend, who enjoyed food and wine and prepared memorable meals for friends and family. Her abiding love, loyalty, and respect for her husband, family, friends, and colleagues was matched
6 The temporal modes, Narrative, Report, and Description, are subdivided on the basis of progression type: Narrative, for example, progresses with bounded events and explicit temporal adverbials; whereas Report has temporal progression centered on Speech Time, and Description, though a temporal mode, typically has spatial progression. The atemporal modes, Information and Argument, both have metaphorical motion through the text domain. 7 A member of the Society for Text and Discourse, Carlota organized a symposium on Subjectivity in Texts for the 2002 meeting, having previously organized workshops on text and discourse Structure at the University of Texas in October 2000 and again in March 2006. Proceedings from the last workshop on discourse can be found on-line at: http://www.utexas.edu/cola/depts/ linguistics/workshops/4DW/
xii
Preface
only by the love, loyalty, and respect that she received in return. Until her very last day, hers was a life that was truly well-lived.
References Ahn, Y. O. (1995). The Aspectual and Temporal Systems of Korean from the Perspective of the Two-Component Theory of Aspect. Doctoral dissertation, University of Texas at Austin. Chomsky, N. (1955). Transformational Analysis. Doctoral dissertation, University of Pennsylvania. Chomsky, N. (1959). Review of Verbal Behavior by B. F. Skinner. Language, 35, 26–58. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Givon, T. (1975). Negation in language: Pragmatics, function, ontology. Stanford University Working Papers on Language Universals, 18, 59–116. Hopper, P. J. (1979). Some observations on the typology of focus and aspect in narrative language. Studies in Language, 3, 37–64. Hopper, P. J., and Thompson, S. A. (1980). Transitivity in grammar and discourse. Language, 56, 251–299. Lai, I. (2009). Time in the Iquito Language. Doctoral dissertation, University of Texas at Austin. Li, C. N., and Thompson, S. A. (1976). Subject and topic: A new typology of Language. In C. N. Li (Ed.), Subject and Topic (pp. 458–489). New York: Academic Press. Mateo Toledo, E. (2008). The Family of Complex Predicates in Q’anjob’al (Maya): Their Syntax and Meaning. Doctoral dissertation, University of Texas at Austin. Rathmann, C. (2005). Temporal Aspect in American Sign Language. Doctoral dissertation, University of Texas at Austin. Ren, F. (2008). Futurity in Mandarin Chinese. Doctoral dissertation, University of Texas at Austin. Shipley, E., Smith, C. S., and Gleitman, L. R. (1969). A study in the acquisition of language: Free responses to commands. Language, 45, 322–343. Shipman, C. (1993). It Had to be Revolution: Memoirs of an American Radical. Ithaca, NY: Cornell University Press. Smith, C. S. (1961). A class of complex modifiers in English. Language, 37, 342–365. Smith, C. S. (1964). Determiners and relative clauses in a generative grammar of English. Language, 40, 37–52. Smith, C. S. (1967). Restrictions on English Transformations: The Combinatorial Possibilities. Doctoral dissertation, University of Pennsylvania. Smith, C. S. (1970). An experimental approach to children’s linguistic competence. In J. Hayes (Ed.), Cognition and the Development of Language, 109–133. New York: John Wiley Smith, C. S. (1971). Sentences in discourse: An analysis of an essay by Bertrand Russell. Journal of Linguistics, 7, 213–235. Smith, C. S. (1980). The acquisition of TimeTalk: Relations between child and adult grammars. Journal of Child Language, 7, 263–278. Smith, C. S. (1986a). A speaker-based approach to aspect. Linguistics and Philosophy, 9, 97–115. Smith. C. S. (1986b). Sentence topic in texts. Studies in the Linguistic Sciences, 15, 187–203. Smith, C. S. (1988). Factors of linguistic complexity and performance. In A. Davison and G. M. Green (Eds.), Linguistic complexity and text comprehension: Readability issues reconsidered (pp. 247–279). Hillsdale, NJ: Erlbaum. Smith, C. S. (1991). A valediction for sentence topic. In C. Georgopoulos and R. Ishihara (eds.), Interdisciplinary Approaches to Language (pp. 545–565). Dordrecht: Kluwer. Smith, C. S. (1993). The acquisition of tense: Bootstrapping into syntax. Proceedings of the Conference, Early Cognition and the Transition to Language. Center for Cognitive Science, University of Texas at Austin.
Preface
xiii
Smith, C. S. (1995). Activity sentences in narrative: States or events? In P. Amsili, M. Borillo, and L. Vieu (eds.), Time, Space and Movement: 5th International Workshop, Part A: 193–206. Toulouse: Institut de Recherche en Informatique de Toulouse. Smith, C. S. (1998). The uses of focus. Proceedings of the Workshop on Text Structure. University of Texas at Austin. Smith, C. S. (1999a). Cues to the small structure of texts. In M. Plénat, M. Aurnague, A. Condamines, J.-P. Maurel, C. Molinier, and C. Muller (Eds.), L’Emprise du sens: Structures linguistiques et interprétations (pp. 289–303). Amsterdam: Rodopi. Smith, C. S. (1999b). Activities: States or events? Linguistics and Philosophy, 22, 479–508. Smith, C. S. (2002a). Perspective and point of view. In H. Hasselgard, S. Johansson, B. Behrens, and C. Fabricius-Hansen (Eds.), Information Structure in a Cross-Linguistic Perspective (pp. 63–79). Amsterdam: Rodopi. Smith, C. S. (2002b). Accounting for subjectivity (point of view). In B. Nevin (ed.), The legacy of Zellig Harris. Volume 1: Philosophy of Science, Syntax, and Semantics (pp. 137–163). Amsterdam: John Benjamins. Smith, C. S. (2003). Modes of Discourse. Cambridge: Cambridge University Press. Smith, C. S. (2007). Tense and context in French. Cahiers Chronos, 16, 1–21. Smith, C. S., and Erbaugh, M. (2005). Temporal interpretation in Mandarin Chinese. Linguistics, 43, 713–756. Smith, C. S., and van Kleeck, A. (1986). Linguistic complexity and performance. Journal of Child Language, 13, 389–408 Yeh, M. (1993). Aspectual Viewpoint, Situation Type, and Temporal Location: A Unified Analysis of –Guo in Mandarin. Doctoral dissertation, University of Texas at Austin. Whitaker, J., and Smith, C. S. (1984). Some significant omission: Ellipsis in Flaubert’s ‘Un Coeur Simple.’ Language and Style, 14, 251–292.
Richard P. Meier, Austin TX, USA Helen Aristar-Dry, Ypsilanti MI, USA
Acknowledgments
The editors wish to acknowledge the financial assistance to this project provided by the College of Liberal Arts of The University of Texas at Austin. Those funds were drawn from the Dallas TACA Centennial Professorship in the Humanities that Carlota had held prior to her death. Robert D. King also contributed funds from his professorship. We particularly wish to thank the assistance of Carlota’s friends and family. John Robertson, Jane Stern, and Joel Smith gave us very helpful comments on the preface. It was John Robertson’s idea that we interview Carlota in 2005. Brian C. Price, formerly the senior administrative associate in the Department of Linguistics, was the cameraman for that interview. Liberal Arts Instructional Technology Services provided technical assistance in the preparation of the DVD of the interview. Gina Pollard of the staff of UT’s Department of Linguistics carefully retyped the manuscripts for two chapters: “The Vagueness of Sentences in Isolation” and “Temporal Structures in Discourse”; she also helped with the scanning and re-scanning of Carlota’s publications. Our thanks also to Leslie J. Crooks, the departmental manager in Linguistics, for her frequent administrative assistance on this project. We thank Mary Erbaugh and Jeanne T. Whitaker for permission to reprint the papers on which they are co-authors. Mary also proofread, and corrected, proofs for Chapter 13. Sarah Wagner and Elyssa Ann Winzeler helped us enormously in proofreading the manuscript. Lastly, we thank the following publishers for giving us reprint rights: Cambridge University Press: “The Acquisition of TimeTalk: Relations Between Child and Adult Grammars” published in Journal of Child Language, 7 (1980), 263–278. “Sentences in Discourse: an Analysis of an Essay by Bertrand Russell” published in Journal of Linguistics, 7 (1971), 213–235. Center for Cognitive Science, The University of Texas at Austin: “The Acquisition of Tense: Bootstrapping into Syntax”, distributed as part of the 1993 Proceedings of the Conference on Early Cognition and the Transition to Language.
xv
xvi
Acknowledgments
Chicago Linguistic Society: “The vagueness of sentences in isolation” published in the Proceedings of the Chicago Linguistic Society, (1977), 568–577. John Benjamins Publishing Company “Accounting for subjectivity” published in B. Nevin (ed.), The Legacy of Zellig Harris. Amsterdam: John Benjamins (2002), 137–163. Max Niemeyer Verlag: “Temporal structures in discourse” published in C. Rohrer (ed.), Time, Tense, and Quantifiers. Tübingen, Germany: Niemeyer (1980), 355–375. MIT Press: “The domain of tense” published in J. Guéron and J. Lacarme (eds.), The Syntax of Time. Cambridge, MA: MIT Press (2004), 597–620. Mouton de Gruyter: “Temporal interpretation in Mandarin Chinese” published in Linguistics, 43 (2005), 107–146. Plenum Publishing Co.: “The temporal reference of the English futurate construction” published in Cognition and Communication, 16 (1983), 81–96. Queens College Press: “Some significant omission: Ellipsis in Simple.’” Language and Style, 14 (1984), 251–292.
Flaubert’s
‘Un
Coeur
Springer: “A speaker-based approach to aspect” published in Linguistics and Philosophy, 9 (1986), 97–115. “Activities: States or events?” published in Linguistics and Philosophy, 22 (1999), 479–508. “The syntax and interpretation of temporal expressions in English” published in Linguistics and Philosophy, 2 (1978), 43–100. University of Chicago Press: “Aspectual categories in Navajo” published in International Journal of American Linguistics, 62 (1996), 227–263. All of Carlota Smith’s papers that we reprint here have been lightly copyedited. Across the volume, we use a common style of headings. To the extent possible, bibliographic references have been corrected and/or updated. We have also attempted to correct minor errors (e.g. obvious reference to the wrong example number) that we are confident Carlota would have wanted corrected.
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xv
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xix
Interview with Professor Carlota S. Smith . . . . . . . . . . . . . . . . .
xxi
Carlota S. Smith: Publications . . . . . . . . . . . . . . . . . . . . . . .
lxi
Part I
Aspect
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Manfred Krifka
3
A Speaker-Based Approach to Aspect . . . . . . . . . . . . . . . . . . . Carlota S. Smith
7
Aspectual Categories in Navajo . . . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith
25
Activities: States or Events? . . . . . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith
61
Part II
Tense
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jacqueline Guéron
89
The Syntax and Interpretation of Temporal Expressions in English . . . Carlota S. Smith
95
The Temporal Reference of the English Futurate . . . . . . . . . . . . . Carlota S. Smith
147
The Domain of Tense . . . . . . . . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith
161
Tense and Context in French . . . . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith
183
xvii
xviii
Contents
Part III The Acquisition of Tense Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Richard P. Meier The Acquisition of Time Talk: Relations Between Child and Adult Grammars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith The Acquisition of Tense: Bootstrapping into Syntax . . . . . . . . . . . Carlota S. Smith Part IV
205
209 225
Discourse Structure and Discourse Modes
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Barbara H. Partee
241
Sentences in Discourse: an Analysis of a Discourse by Bertrand Russell Carlota S. Smith
249
The Vagueness of Sentences in Isolation . . . . . . . . . . . . . . . . . . Carlota S. Smith
273
Temporal Structures in Discourse . . . . . . . . . . . . . . . . . . . . . Carlota S. Smith
285
Temporal Interpretation in Mandarin Chinese . . . . . . . . . . . . . . Carlota S. Smith and Mary S. Erbaugh
303
Part V
Context and Interpretation
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Helen Aristar-Dry
345
Some Significant Omissions: Ellipses in Flaubert’s Un Coeur Simple . . Jeanne T. Whitaker and Carlota S. Smith
349
Accounting for Subjectivity (Point of View) . . . . . . . . . . . . . . . . Carlota S. Smith
371
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
395
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
399
Contributors
Helen Aristar-Dry Institute of Language Information and Technology, Eastern Michigan University, Ypsilanti, MI, USA Mary S. Erbaugh University of Oregon, Eugene, OR, USA Jacqueline Guéron Institut du Monde Anglophone, Université de Paris 3 – Sorbonne Nouvelle, Paris, France Manfred Krifka Institut für deutsche Sprache und Linguistik, Humboldt-Universität zu Berlin, Berlin, Germany Richard P. Meier Department of Linguistics, University of Texas, Austin, TX, USA Barbara H. Partee Department of Linguistics, University of Massachusetts Amherst, Amherst, MA, USA Carlota S. Smith Formerly of Department of Linguistics, University of Texas, Austin, TX, USA Jeanne T. Whitaker Wheaton College, Norton, MA, USA
xix
Interview with Professor Carlota S. Smith
This interview was conducted at Carlota Smith’s house in Austin on July 7, 2005 by Richard P. Meier and Keith Walters (formerly at UT Austin, now on the faculty of the Department of Applied Linguistics, Portland State University). Brian Price was the cameraman. The transcript was revised, edited, and annotated by Emilie Destruel and Richard P. Meier.
A Year in France Keith: Were you teaching English in France or linguistics? Carlota: I was teaching linguistics. Keith: Linguistics, I see. xxi
xxii
Interview with Professor Carlota S. Smith
Carlota: Yeah, that was a break. Keith: Mm-hmm. Carlota: I taught there for almost two years in the seventies in Paris.1 At that time, there were very few linguists. Richie Kayne2 was there. Carlota: Nicolas Ruwet3 was around. And so they actually got me a job. Maurice Gross4 got me a job on a – on a claim that there was no French person who could do it. And I insisted. I just did everything in French. I wouldn’t speak to any French person in English. Richard: Oh, you did all your teaching in French. Carlota: Yes, which was probably quite mean to the students the first semester or after that. [Talking to cat. . .] He can go out. See, he’ll do anything. Max, come here. No. Keith: It’s okay. Cats will never behave. Carlota: Well, not predictably anyway. Keith: I mean, they never do what you want them to do. Carlota: That’s why we love them. Keith: They will never follow orders. Carlota: Maxie, you are so bad. Keith: They won’t follow orders. That’s what I should have said. Carlota: So, yeah. And so I got to know Richie [Kayne] well that year and Nicolas and various. . . . and Maurice, of course. And I taught at Vincennes. So this was, you know, ’73 or some year like that. So it was still the aftermath of the ’68 bit.5 Keith: Right. Carlota: So Vincennes is the middle of the woods. You took a bus to the end of the line. Then. . . No, you took the Metro to the end of the line. You took a bus into the woods.
1 The
University of Paris VIII at Vincennes. his PhD from Paris VIII in 1976, now a professor at New York University. 3 (1933–2001). Rouwet taught at Paris VIII. Much of his early influence came through his book ‘Introduction a la grammaire générative’ in 1967. 4 (1934–2001). Founder of the Laboratoire d’Automatique Documentaire en Linguistique at the University of Paris VII and the Centre National de la Recherche Scientifique. 5 In May 1968, a series of protests began with student strikes at universities in Paris, following confrontations with university administrators and the police. The De Gaulle government’s attempts to quash these strikes resulted in inflamed battles and protests, finally causing the collapse of the government. 2 Received
Interview with Professor Carlota S. Smith
xxiii
Richard: So it was kind of wrecked, wasn’t it, at the time, the campus? Carlota: Oh, it was amazing. I never understood the value of money before that period, not in education. I mean, there was no money – like, we had no offices. Keith: I don’t think French professors have offices even now, do they? Carlota: Yeah. I mean, I was lucky, because Maurice Gross had one of the few – he had a CNRS lab.6 He had the only linguistics library, actually, in Paris. I had an office there. So once a week, I would just take the metro and then the bus out to Vincennes and you would actually. . . Well, there were blackboards. You would bring chalk from the lab. And then just before class, I would go into the ladies room and get a wet paper towel and bring it in for an eraser, because of course there was no eraser. There actually wasn’t a lot of paper. Books were stolen. But when were you there? Keith: No, no. This is just very much like the situation I found in Tunisia. Carlota: Yes, I found it similar in Tunisia. Keith: And, you know, certainly, because traditionally, and of course then all the colonies got this, there’s the responsable des classes7 in high schools, and what his or her job is, is to bring an eraser and chalk. Carlota: Oh, they didn’t have that in Paris. [All laugh.] Keith: With the university, I was thinking, maybe France had given this up or maybe by that time it just didn’t happen at the university. Carlota: However they figured it out. Keith: And then they had to bring the book called The cahier des classes,8 in which you noted down what you’d had taught for the lessons, so that when the inspecteur or inspectrice came, he or she could verify that you were indeed following the program. Carlota: You wrote down what you were supposed to do. Keith: Exactly. Carlota: The other thing about Vincennes was it was an old army barracks. And in the hallways, they were selling stolen books, food, you know, stolen items, and you name it. So it was a zoo. So only the really serious people would learn anything. And the ones who were really serious figured out you had to get on at Maurice’s lab [laughs] and read the linguistic books there. [laughing] It was a
6 See
footnote 4. person in charge of bringing chalk, etc. to the class. 8 The class book. 7 The
xxiv
Interview with Professor Carlota S. Smith
very strange experience in some ways. But I did end up speaking pretty good French. And the first year, my kids went to boarding school, because I knew they’d never – I knew Paris well enough to know that if they were with me, they’d never learn French, because you don’t get invited to kids’ houses and stuff. Keith: Oh, right. Carlota: So they went to this boarding school. Then the second year, they were with me in Paris. . .and went to a lycée. Richard: And the boarding school was in French the first year? Or was it. . .? Carlota: Oh, yes. Richard: Yeah, it was entirely in French. Carlota: It happened that a friend of mine, a French friend, his father was one of the founders of the school. . .he fixed me up, he said that. . . Keith: Mm-hmm. Carlota: So when we left, French people thought they were French. Keith: Right, of course. Carlota: They were just the right age. Keith: Kids, exactly, yeah. Carlota: They were ten to twelve, some of that time. Keith: That’s great. Richard: So that kind of makes sense of how you got involved in French. Carlota: That’s right.
Chinese, Navajo and Russian Richard: But how did you decide to work on Chinese and Navajo along the way too? Carlota: Oh, well, Chinese. I’d always sort of liked Chinese, because of Chinese restaurants when I was a little girl in New York. My father was very interested in things Chinese and we went down to Chinatown about once a month and had dim sum and wandered around Chinatown. Everyone in the restaurant knew us. The men used to make little animals for me out of drop dip dough that you make dim sum with. We used to go in the kitchen. So, I just had a very long association and a sort of pleasant association. And, you know, it’s different from English, right? And the Navajo was more fortuitous. I was looking for a language that was not Indo-European and that was different from Chinese. I guess Kluwer, the publisher, loved, you know, [unintelligible], adored American
Interview with Professor Carlota S. Smith
xxv
Indian languages. And I had read a lot about Hopi for some reason, you know, Whorf’s language. Didn’t want to work on Hopi. I don’t even remember. It was just. . . I mean, I was looking for an American Indian language. Richard: You’re saying that Kluwer actually suggested it? Carlota: Well, the editor. Richard: The editor in part suggested it? Carlota: Yeah. I mean, they. . . Well, I suggested it and they leapt on it. Richard: Oh, I see. So while you’re – this is while you were doing your book on aspect.9 Carlota: Yeah, that’s right. Well, I had to do Russian, because you can’t be serious about aspect unless you do Russian. Richard: Right, Right. Carlota: Fortunately, I got Gil [Rappoport] to work with me. So that’s the only chapter that people worked with me on. Keith: Right. Because you actually studied Chinese here. Carlota: Yes, I did. I’d already – I’d audited a lot of Chinese. Keith: I remember Jocelyn Liu10 talking about it. Carlota: Yes, that’s right. Exactly. Then I got Jeannette Faurot who was then in the department too.11 Keith: Yes, right. Carlota: She was wonderful. I bombarded her with questions. I’m quite bad about that. I almost lost a friend when I was bombarding somebody else about Navajo. I just constantly asked questions and gave them things and said, “What about this?” She was very nice. [laughs] And somebody else. And so I was just. . . I think I heard Navajo was mysterious and maybe I knew that there was pretty good documentation of Navajo, because there are wonderful dictionaries. Richard: Uh-huh. Right. Carlota: And I really couldn’t have done it without the dictionaries.
9 The
Parameter of Aspect. (1991). Dordrecht: Kluwer. doctoral student, Department of Linguistics, UT Austin. Ph.D. 1987. 11 PhD from UCBerkeley in 1972. Professor Emeritus Center for Asian Studies and Department of Asian Studies at UT Austin. 10 Former
xxvi
Interview with Professor Carlota S. Smith
Getting into Linguistics Carlota: So I’m going to answer your question about getting into linguistics and stuff. Which is sort of a good story, I think, showing the difference between now and then also. So as an undergraduate for some mysterious reason I took a course with [Roman] Jakobson. We read The Raven in 15 languages or something. I don’t think I learned anything, [laughs], but I did have exposure to Jakobson. And then I was a faculty wife at Swarthmore12 and I had a young daughter. I really couldn’t stand doing nothing, so I was looking for a job. So I looked in publishing, which I thought might be something I would end up doing. I had always thought that. But, you know, Philadelphia is not full of publishing jobs and. . . oh, you asked about feminism, and so in a couple of places they would say, “Why yes, we do have a woman, but she’s – our woman is on leave right now, you know, maternity leave, and then she’ll come back, so we don’t need another one, right?” [all laugh] That was the inference. So I sort of couldn’t find a job, so I was just trumpeting it. Keith: This is in the publishing industry? Carlota: Right. So I was telling. . . And Swarthmore is a small college. So anyway, in short, so Lila Gleitman,13 who was a faculty wife at the time, Henry Gleitman was at Swarthmore, one day she said. . . And I didn’t know her very well, but I’d seen her, you know, and talked to her a certain amount. We liked each other. She said, “You know, I think you might like doing what I do.” Maybe this was May of some year or other. And I said, “Oh, you know, tell me more.” She told me a little bit. And, you know, a week later or something, she marched me down to Penn,14 introduced me to Zellig Harris,15 and I was hired for the summer to be a research assistant. And my job, which I was perfectly capable of doing was to categorize English verbs for Harris’s research project, which was in principle on mechanical translation. This is whatever we are – we’re in ’59 or ’60 now. So these people had been doing this for quite some time – [laughs] – this project. So I knew nothing about linguistics, and I loved it! See now, Vendler16 was on that project. So we used to sit around. . . Richard: Vendler was on the project too? Carlota: Yeah. There were all sorts of people.
12 College
in Swarthmore, PA.
13 PhD from University of Pennsylvania in 1967. Now emeritus professor of psychology, University
of Pennsylvania. of Pennsylvania, Philadelphia, PA. 15 (1909–1992). PhD from UPenn in 1934. Began teaching at UPenn in 1931. Founded the Linguistics department there in 1946. 16 (1921–2004). PhD from Harvard in 1959. Founding member of the philosophy department at the University of Calgary. 14 University
Interview with Professor Carlota S. Smith
xxvii
Richard: What was he at that time? He was a student? Carlota: He was a research associate or something. I don’t know exactly what his job was. He must have been sort of like me. I don’t think he – he wasn’t teaching. Maybe he was sort of. . . He must have been on leave. I don’t think he was teaching philosophy. He was a philosopher, a former Jesuit priest. He was married to Helen at that time, Helen Vendler, because I got to know her then. So we would sit around at lunch and say things like, “Oh, I saw the funniest verb. I found the funniest verb this morning.” And everybody would say, “Really?” [laughs] And you would tell them you’d found this verb, like behave, you know, where you have to have the reflexive afterwards and it has to agree. So I loved it! So at the end of the summer, I decided to go to graduate school. It couldn’t have been simpler and my whole graduate school. . .. See the point I wanted to make is that there was money. Harris had money. So he could just hire these people as assistants on a long-shot and if it didn’t work out. . . Richard: On the recommendation of another graduate student. Carlota: No, she was a graduate student. Richard: She was. Carlota: Lila was a graduate student. Richard: Yeah, but just on the recommendation. Carlota: Exactly. Richard: Lila kind of walks in and says. . . Carlota: And Lila didn’t know me that well. I mean, she knew me a little. Richard: Yeah. Carlota: We’d had some good conversations, you know. And exactly, just out of the blue. Keith: And this was federal money? Carlota: This was federal money. Keith: To support this translation project, that’s what I assumed. . . Carlota: This is NSF money.17 Keith: Yes. Carlota: That’s the point I wanted to make. Okay. Both of my interesting career things have to do with that. That NSF money comes around. My whole graduate school was paid for by – I was a research assistant the whole time.
17 U.S.
National Science Foundation.
xxviii
Interview with Professor Carlota S. Smith
Richard: By Harris’s grant the whole time? Carlota: On Harris’s grant the whole time. Richard: So you never taught while you were in graduate school? Carlota: No. Keith: Can you talk a little bit about what Penn was like at that time and what it was like to work with Harris and what was going on in Linguistics at that time? Because as Richard’s question pointed out, some pretty amazing people were there then. Richard: Mm-hmm. Keith: And some pretty amazing things happened. Carlota: Exactly. Keith: In terms of consequences for the history of the discipline. Carlota: Exactly. And P.S. or parenthesis, that’s why we really have to talk to Win [Lehmann],18 because he was prescient. I mean, he was able to see. He hired people for the Linguistics Department [at UT] . . .. It was just a few years later. Okay. So, while Haj [Ross]19 was there [at Penn] some of the time, we got to be very good friends. I guess he is probably the most well-known person. Well, okay. Harris was an extraordinary presence. He was a wonderful teacher. He, in fact, wasn’t interested in my work and I never worked closely with him. So, everything I learned from him I learned as a student or as a research assistant, because he would give me papers to write and I would write research. Little research jobs to do, and I would do them. He was totally intense and he rarely talked about the competing, the beginning of competing theory. So he was teaching us his version of transformational grammar, which was of course what Chomsky learned with him and then changed. So Harris, the main thing – well, not the main thing. Harris was a very insightful person, who had been brooding about this stuff for years. So every time we came up with something, he knew already what it was. I remember he once had me do a little paper on – I don’t know what to call them – modifiers, I guess, something like this. I eventually noticed rather. And he was reading this and said, “Well, where is it?” Then he found it. . . “Oh, I thought of it!” But he knew that it was supposed to be there. [laughs] He was just. . . He had tremendous knowledge. And he was really interested in. . . His way of training was sort of modeling. You know, he didn’t ask questions. So none of us learned to teach. [all laugh] None of the teachers were teachers in the current sense of the word, nothing remotely Socratic ever happened, you know. [laughter] Or even where teachers sort of talk about how 18 Winfred P. Lehmann (1916–2007). PhD in Germanic Linguistics from the University of Wiscon-
sin. Joined UT Austin in 1949 as an Associate Professor in Germanic Linguistics. in 1967. Was a student of Zellig Harris at UPenn and of Chomsky at MIT. Now at the University of North Texas.
19 PhD
Interview with Professor Carlota S. Smith
xxix
they got to things. No, it wasn’t like that. It was really modeling and demonstrating, but it was wonderful. I mean, you just got a tremendous sense of excitement about the enterprise. And also, although he didn’t teach us about other theories, he didn’t – he wasn’t ideological about his theory. He didn’t sort of warn us against them or trash them in any way. He just wasn’t interested, you know. [laughs] So all we did was his stuff and I guess you could say that was a big flaw in the program. So there was a tremendous sense of intensity, and this – and Lila of course was there, and this gallant little band working. I guess that was another sense. Well, Henry Hoenigswald20 was a wonderful teacher. He was a historical linguist who was very traditional. So it was sort of an ecumenical group, come to think of it. Somebody named Southworth,21 who did morphology. Henry Heesch did logic and stuff. He was a holdover traditional European and did things that way. Zeno [Vendler] didn’t teach. I don’t think he taught in the Philosophy Department. He might have. I think I would have known, because we were kind of good friends. Richard: Who were the other students besides Lila? Carlota: Well, I’m just trying to think. There was somebody named Jim whose last. . . Oh, well, Bill Watt, William Watt, who is still a friend. [. . .] He was at Irvine22 for a long time. He’s now retired. But he’s kind of stopped doing linguistics and started doing semiotics and stuff long ago. Liz Shipley,23 who Lila and I worked with, was around, she was in Psychology. And Henry, of course, was in Psychology. Oh, Myrna Gopnik24 – Richard: Oh, of course. Carlota: – was around. She was a friend of mine. So, I mean, it was a small group, a very small group. Richard: But choice. Carlota: I can’t think of anyone else. Yeah, self-chosen or randomly chosen or something.
Going to MIT Carlota: So then one day, I can’t remember the order in which these things occurred. Chomsky came to give a talk one day. And the other day, there was this little manuscript on Harris’s desk, which I for some reason picked up and 20 (1915–2003). 21 Professor
Emeritus of South Asian Linguistics, University of Pennsylvania. of California at Irvine. 23 PhD in 1961 from UPenn in Psychology. 24 Now Professor of Linguistics at McGill. 22 University
xxx
Interview with Professor Carlota S. Smith
asked if I could borrow it. It was Syntactic Structures.25 I was just absolutely thrilled by Syntactic Structures. I mean, it was just – I loved it, and it was a whole different way of thinking, and it completely blew everything else out of my mind, in terms of how to approach linguistics; although, I really didn’t have that formal background, but I was able to understand enough and I had some of it. And as I said, I don’t remember which order Chomsky talked about his work and I read this. No idea. But then, another random event, my then husband, David Smith, wanted to spend the year in Cambridge the following year. So Harris wrote Noam and said, you know, “This person is coming.” So I spent the second year of graduate school at MIT. And that changed and determined everything I did in linguistics. MIT was teeny. [Edward] Klima – Ed Klima was there.26 Lees would come from time to time, you know, from the Midwest. He was at Illinois and sort of commune [laughs] and give a talk or something.27 Maybe he was even there for part of the time. And Haj. I guess maybe – yeah, I think Haj wasn’t there when I was there. Ted Lightner was there.28 Toward the end, Paul Postal29 started hanging around a lot. He wasn’t officially there yet. So it was really just beginning. But we all understood that this was it. Richard: Mm-hmm. Carlota: I think maybe [George] Lakoff30 would pop in from time to time, too, because he was in Indiana doing his dissertation work. So we felt, I mean. . . Chomsky always assumed this stance of beleaguered, but it was just that. I mean, you know, it was already. . . I mean, I guess he’d already published not only Syntactic Structures, but the Review of Skinner.31 Keith: Right. Right. Carlota: And, you know, everybody understood that this was it. This was just the most exciting thing that had happened in a long time. Keith: It’s long seemed to me for people outside the discipline of linguistics it was the Review of Skinner that really was the important intellectual milestone. Richard: Oh, I think that’s true for people in psychology. Keith: And even in many other disciplines, I think, it’s just sort of. . . Carlota: Yeah, it reverberated. Keith: Exactly. Just entire edifices just crumbled. 25 First
published in 1957 (Mouton). Faculty member, MIT, 1959–1967. 27 Lees, Robert B. (1922–1996). Professor of Linguistics at University of Illinois and later at Tel Aviv. 28 Ph.D. 1965, MIT. 29 PhD in 1963 from Yale. Now professor at NYU. 30 PhD in 1966 from Indiana University. Now professor at UC Berkeley. 31 Review of Skinner’s Verbal Behavior. Language, 35 (1959), 26–58. 26 (1931–2008).
Interview with Professor Carlota S. Smith
xxxi
Carlota: They did. And, you know, it’s so interesting, when I taught Plan II, Cognitive Science Approach to Linguistics, I’d given that article to the students. Some like it and some don’t. It’s dated. Keith: Right, yes. Carlota: But it’s still terrific. Keith: Mm-hmm. Carlota: I mean, to me, anyway. And some of them see it. For some of them, it’s sort of old news. It’s like I’ve just been rereading Augie March. Well, I have very mixed feelings. [laughs] Which I adored when I was 17. [laughs] Some of it’s old news and some of it’s just totally compelling still. Keith: Yes. Carlota: So that was a very exciting year for me. And I didn’t understand these things at all, but I wrote a paper, which Chomsky sent to Language. That was my first publication.32 Richard: Chomsky sent it to Language? Carlota: So, of course, they published it. Richard: What does that mean that Chomsky sent it to Language? Carlota: Well, you guess. It meant they published it. That’s what it meant. [laughs] Richard: No, no, no, no. I mean, you mean he just sent it and then didn’t talk to you about it? Carlota: Well, no, no. No, no, no. We had worked on it, everybody. . . I mean, it was. . .. Okay, I see. So Chomsky was a lecturer. We all went to everything. Umm, we all went to everything. So if there was a talk. . .. And that still bothers me about this department. Everybody doesn’t come to the colloquium. Somebody gave a colloquium, we all went – everybody. Richard: Mm-hmm. Carlota: Oh, Morris Halle,33 of course, was there. How could I forget Morris? Well, because I wasn’t. . .[laughs, unintelligible] Richard: And was Jakobson present a lot?
32 ‘A
class of complex modifiers in English.’ Language, 37 (1961), 342–365. at MIT since 1951, now emeritus.
33 Professor
xxxii
Interview with Professor Carlota S. Smith
Carlota: No. He was at Harvard and he was not present. Roger Brown34 was there, and I was sent by Chomsky to be the deputy from this group to Roger Brown’s seminar. Richard: Oh, that’s funny. [laughs] Carlota: And I acted as such, and then when I got back to Penn and started working with Lila, and Liz Shipley on acquisition,35 so that sort of worked out. Richard: So Chomsky was kind of sending emissaries to William James Hall or whatever to contact. Carlota: Yes, that’s right. No, no. Well, Brown, I think he had. . . I don’t remember. it seems to me he had a seminar at MIT. I think I’d remember. I’m not sure. Because I was an undergraduate in Cambridge, and it’s all very familiar to me, so I really don’t remember if I went over, because I was auditing Quine’s36 class that year at Harvard, so I went over a lot, you know, to Harvard anyway, so I really don’t remember. I think it was at MIT. Richard: So Chomsky sent your paper to. . . Carlota: Anyway, so the point is the apprenticeship that I really had in linguistics, besides me listening to Harris was this, I would write, draft this paper, and they would all read it: Morris, Ed Klima, Chomsky, Lees, whoever else, you know, they wandered in. And then they would tear it to pieces and I would write it again. Then they would all read it again. This happened several times. At the end, it wasn’t so bad. [laughs] And he probably told me he was sending it, but I didn’t understand what that meant. So I won’t say entre-nous in this setting, but really, the second paper37 I published in Language I did to show myself I could do it without him. Richard: Mm-hmm. Carlota: When I got back to Penn. [laughs] Because obviously, you can’t live on it. And by then, I understood. Keith: The system. Carlota: What it meant. Richard: What you are also saying is that the first – he wasn’t beleaguered even in 1961 or whenever you sent that first paper.
34 (1925–1997).
PhD in Psychology from the University of Michigan 1952. For most of his career he was on the faculty of the Department of Psychology at Harvard, except 1957–1962 when he was on the faculty of MIT. 35 ‘A study in the acquisition of language: Free responses to commands.’ Language, 45 (1969), 322–343. 36 (1908–2000). From 1930 to 2000, Quine was professor at Harvard in Philosophy. 37 ‘Determiners and relative clauses in a generative grammar of English.’ Language 40 (1964), 37–52.
Interview with Professor Carlota S. Smith
xxxiii
Carlota: Well, when I am saying that I assume, I know nothing about it. I mean, I – I. . . It was probably reviewed. I haven’t a clue. First of all, it was a long time ago. Second of all, well, for whatever reason, I actually – I do not remember ever receiving reviews or being asked to revise or anything. It certainly couldn’t have been perfect; although, it did get a lot of scrutiny. I mean, there were probably four or five iterations of this process that I told you guys about. So it really got a lot of attention. Keith: Before–before it was sent. Carlota: Before it was sent to Language. We had all these people. I mean, think about it. It’s amazing. Richard: Right. Carlota: And that’s because there were so few of us. Keith: Right. Exactly. Yes. Carlota: And I guess Chomsky saw this would be kind of a good article for his point of view. It was called “A Class of Complex Modifiers of English.” And actually, the second paper has been still – people still read it. Nobody reads that [first] one anymore. But it was sort of – it led up to this point. It wasn’t bad. So that year was obviously a very exciting year. There was something else I was gonna. . . Maybe I. . . I already said Roger Brown. Well, you want to ask me anything else about it? Keith: I would like to ask you something about – especially because you were at Penn and then you were at MIT. To what extent were you or other women thinking about gender issues and aware of how few of you there were [laughs] in a context? I mean, you know, because it’s. . . Carlota: Well, I’m glad you add that. Keith: Exactly. [all laugh] Carlota: Because I was not unaware. The truth is at the time I liked it fine. I was aware of it; I didn’t mind. It was only later that I came to feel differently. At the time, sure, I was so used to being the only woman. Keith: Right.
Returning to Penn Carlota: And I suppose I was treated differently. I didn’t feel I was treated badly. But, you know, if there were five students and there were all these people reading your work, obviously, I wasn’t treated badly. At least, it was not obvious
xxxiv
Interview with Professor Carlota S. Smith
to me that I was, you know. [laughs] I think I was just treated like some weird creature, you know. But as I say, I liked it fine. People were perfectly nice to me. And then when I got back to Penn, which, of course, I had to go back to real life after this year, intellectually a somewhat interesting thing happened, which was Henry Heesch asked me maybe to rewrite that first paper or maybe the second – after I wrote the second paper, which served as my MA thesis and was published in Language and sent by me to Language. Henry Heesch asked me to redo it completely in the Harris framework, and I refused. [laughs] So that was sort of independent, right? [laughs] Keith: Slaying one’s father and getting on with one’s life. [laughs] Carlota: Well, no. I don’t know that Harris was close enough to be my father. Keith: To be a father, ok. Carlota: But still, it was sort of getting on. Yeah, slaying. You know, saying, well, this is all very nice, but you got me for me. Richard: It wasn’t on your agenda anymore. Keith: Exactly. Carlota: No, it wasn’t. And I think before then I had learned and loved it, but I hadn’t embraced it, so it wasn’t – it didn’t – it wasn’t as if I had shifted an allegiance. I hadn’t really felt allegiance. I was still seeing what the field was. Keith: Mm-hmm. Mm-hmm. Carlota: And also, as I said, I read Syntactic Structures in the spring of the year before I went to MIT, so I already understood that this was what excited me. So when I went to MIT, I knew what I was doing. It wasn’t just, oh, well, I find myself here. I mean, I did find myself there, but by then I knew. So then, I worked with Lila and Liz and we did a couple of papers together.38 And I divorced David Smith. So that was the second time I was a working woman. Because Henry Heesch, with whom I obviously had a mixed history, was extremely nice to me and hired me. I really think it was only because he understood my personal situation. He hired me to be a – what was I? – some kind of TA, which meant I took notes on his lectures and wrote them up for the students. I never [laughs] taught a class in my entire graduate school. Nor did anyone else by the way. I mean that just wasn’t seen as something. . . Keith: That was how graduate school worked at that time. Carlota: That’s right. You didn’t train people to teach. So then I had to finish my dissertation obviously. And by this time, Harris was virtually not there at all. He 38 Shipley et al. (1969) and ‘Old and new ways not to study comprehension: Comments on Petretic
and Tweney’s experimental review of Shipley, Smith and Gleitman.’ Journal of Child Language 5 (1978), 501–519.
Interview with Professor Carlota S. Smith
xxxv
was a University Professor. He had a project in New York he was working on, and he was in Israel a lot and was known as Zealot the Carpenter as you may have heard. He loved to be known as Zealot the Carpenter. One doesn’t know if people who called him that understood his other life or not. Can’t help thinking in Israel they must have, you know. But anyway, there was this conceit of a zealot carpenter. So the point is, he was never around. And I still remember the rage I would feel when I would have an appointment with him to talk about my dissertation and then he would call that morning or that afternoon, and by this time, I had two young children and had to get babysitters and stuff like that, and he would cancel. He would call from New York. He would sound like a little boy who’s been naughty and knows that he’s going to get away with it, right? [laughs] What could I do, you know? So I’m convinced he never read a word I wrote. But Chomsky said that he thinks the same thing about his dissertation. [laughs] So I’m in good company, right? Keith: Right. Exactly, yes. Carlota: So then, well, I had to get a job, because I didn’t have any money. And I got a job. And this is the next female thing. I got a job in the English Department at Penn, which I later realized was partly probably due to sexual harassment in the end. That is. . . I don’t think I need to tell this story for the camera. [. . .] But it wasn’t the same for women, shall we say. So that was a very strange experience. So, I had never taught before, right? Keith: Mm-hmm. Carlota: And they said in the interview, in the hiring discussions, “Well, we really want our graduate students to learn something about linguistics.” Well, since I’d been a major – a literature major undergraduate, it seemed perfect. And then they changed their minds over the summer and decided they didn’t [laughs] want their graduate students to learn linguistics. So I was teaching – I don’t know – Introduction to English Literature or something. Keith: Mm-hmm. Richard: Right. Carlota: And then I stayed sane because the Linguistics Department had me teach a course that I made up as I went along on stylistics. I also was teaching History of English, which I had lied and said that I knew something about it, [laughs], which I knew nothing about at all. But there’s nothing makes you learn faster than having to teach a class. [laughs] We’ve all been there. Richard: Right. Carlota: But now, of course, one suggests – one does that deliberately. Keith: Exactly. Carlota: They would like to learn about ‘x,’ so let’s put it on the syllabus. But at the time, it was terrifying, because I did have two young children and I was a
xxxvi
Interview with Professor Carlota S. Smith
divorced mother and blah, blah. So after. . . But I was still. . . Well, you know, the children were seeing their father and everything, so I wasn’t quite ready to leave that area. Richard: What year was this? Carlota: Let’s see. I don’t know. Maybe ’67, ’66, some year like that. Oh, that’s right. I got my degree in ’67, so I guess we’re ’65. Richard: Oh this is before you had your degree, right? Carlota: No, no. I started teaching after, so it was ’67. Richard: ’67.
Going to Texas Carlota: Yeah, or ’68. That’s right, ’68, ’69, ’70, around that time. So then. . . Oh, I was invited to a conference due to my friend Watt, who by then had a job in Pittsburgh – I’m sure due to him – and I gave a paper on a sort of linguistics and acquisition that was quite well received and was in a volume of papers. This is relevant to the next event. And so then I decided it was really – after two years of this, I mean, I didn’t belong in the English Department. They weren’t going to hire me. And I didn’t even want to be hired in that – in the Linguistics Department, because that was not the sort of work I was interested in. They weren’t a generative linguistics department. They sort of were. They were very eclectic, but they were – you know, Harris had basically stopped being there and there was no dominating – dominant presence who had replaced him, and so it wasn’t that interesting by then to me, so I started looking for a job. And I did get a job at Berkeley, but then for personal reasons I decided to come to Texas. And there was this man that I – Ted Lightner was coming here. So he said, “I’ll get you a job.” I said, “No! I will get myself a job or not.” So by this time I was beginning to, you know. . . [laughs] Well, [unintelligible], by then it was very clear to me that there was a lot to do here and whatever little thing I could do I was going to do. So completely by chance. . . This is the next good story about money and everything and chance. Completely by chance I received a letter the previous week or something from somebody in the Psychology Department here, David Hakes,39 asking for a reprint of the paper that I gave at the conference at Pitt, which is still a good paper, I think. So I wrote him this. . . I didn’t know this person, Dave Hakes. So I wrote him this wonderful letter. I said, “Dear Professor Hakes. . .”, you know, enclosed is the reprint. . . First paragraph. Second paragraph, “For personal reasons, I want to come to Texas. I
39 (1934–1982). Ph.D. 1961, Psychology, University of Minnesota. Developmental psychologist in
Department of Psychology, UT Austin, from 1961 to 1982.
Interview with Professor Carlota S. Smith
xxxvii
want a job.” Third paragraph, “If so, I need to know right away because I have this other job at Berkeley.” Well, they called me the next week and they said, “We have this NSF money, training grant, faculty associate. Do you want to be in Psychology or Linguistics?” So I had the sense to say Linguistics. [laughs] Because I couldn’t have stayed in Psychology, because I didn’t know any Psychology. [laughs] But I mean, that could never happen now, right? Keith: Yeah. Carlota: I mean, they just had money. They had money for faculty. Harvey [Sussman] came as a faculty associate. I think Peter MacNeilage40 came as a faculty associate. I’m not sure about that. Richard: I know Harvey did. Carlota: Right. I think maybe Harvey, because we came the same year. Cameraman: Lee? Richard: Lee did too? Carlota: Lee [Baker] is earlier. He was already here when I got here. But I think that’s right. I think it was him. So it was just extraordinary. Richard: I’m just curious about one detail. When you wrote to Hakes, was this in the usual time of year when one would apply for jobs or was this. . .? Carlota: It was the spring. So I’d already. . . Richard: You’d already gotten this job at Berkeley. Carlota: I must have accepted it, because apparently they were furious, livid, when I. . . Richard: So it was March or April. I’m just kind of. . . Carlota: Yeah, I think it was March. That’s right. That’s important for the story. It was March or April. I don’t remember which. And the thing was. . . Well, in retrospect, I think it was a wise professional decision, but the thing that. . . I had been to Berkeley, by the way. I had talked to people there and spent time, because of the acquisition stuff I was doing. Dan Slobin41 was there. Richard: Mm-hmm. Carlota: What’s her name? Sue. . . Richard: Ervin-Tripp.
40 Professor
of Psychology at UT Austin. from Harvard in 1964, in Social Psychology. Now Professor Emeritus at UC Berkeley in Psychology.
41 PhD
xxxviii
Interview with Professor Carlota S. Smith
Keith: Sue Ervin-Tripp,42 right Carlota: Yeah, those people. So I talked to people a lot there, but the job they offered me was not in Linguistics. It was in some umbrella department. I’ve forgotten the name. Do you remember? Richard: Oh, that rhetoric [department where Brian] MacWhinney43 got his degree? [. . . .] Carlota: Yeah, it was probably that. And I was dubious about that. So that, you know, however, I didn’t have any other jobs, so I accepted. But so I wasn’t. . . And I understood that they weren’t doing what I was interested in doing. But I also really liked acquisition and I probably would have stayed with acquisition had I gone over there. Anyway, so, yeah, it was late. [laughs] It was March or April. And probably Harvey was already signed up, I don’t know. But, I mean, I just don’t think that could happen today. Richard: No. Carlota: I mean, maybe some people have training grants. I mean, there are training grants. Richard: There are training grants. Carlota: But still, the flexibility to. . . I don’t know. Richard: Not in linguistics that I know of. Carlota: Right. Richard: Maybe there are some in psychology. Carlota: Maybe they usually have one faculty associate, but Win was able to wrangle it for two that year. I don’t know what he did. Richard: Mm-hmm. And the training grants wouldn’t lead to a faculty position. They would. . . You would leave after a couple of years. Carlota: Yeah. Well, in those days, it’s sort of appalling to realize how lackadaisical I was at least. I mean, it just sort of happened after two years of this that I was on. And I was in France when I got tenured. I mean, I didn’t do anything. [laughs] You know, I had published a few things and I don’t know. One just didn’t think about it.
42 PhD form University of Michigan in 1955, in Social Psychology. Now Professor Emeritus at UC
Berkeley in Psychology. at Carnegie Mellon in Psychology.
43 Professor
Interview with Professor Carlota S. Smith
xxxix
Richard: Mm-hmm. And one other detail. When you wrote to David Hakes, was he the one that you negotiated everything with? Was it David Hakes or did Win ever enter the picture or anybody? Carlota: I think that maybe he. . . He was a complete stranger to me. I think perhaps he talked with someone on the phone. So he was not. . . Richard: Lehmann? Carlota: No, no. I’m sorry. David. Richard: Oh, it was David Hakes. Carlota: I think it was David who called me. I’m not sure about that, but I think so. I don’t think I met Win till I came here. And Win, I’ll just. . . One funny thing. I remember one year he got me money to go somewhere. You know, some – it was a professional trip. Some overseas trip somewhere. And I was expressing surprise and gratitude. He said, “Well, better you than some physicist.” [all laugh] Such a typical Win threat – remark, I mean. So that’s all that story. And then when I got here, of course, everybody was doing stuff I was interested in. So it was a very good match. So I don’t know. Want to ask me any other questions?
University of Texas at Austin Keith: Do you want to talk about UT now? Richard: Sure. Why not. Keith: Could you talk a little bit about who was here and what was going on and sort of how Linguistics was becoming a department or had become a department? Carlota: It had become a department. Keith: Right, ok. Carlota: And there was a stellar array of people. So Lee Baker,44 Bob King,45 Bob Harms,46 Win Lehmann. I was the only woman come to think of it, and then, huh, somebody else. Well, of course, Peter. Peter and Harvey and I came, and, am I forgetting anyone? Richard: Did you say Emmon?
44 (1939–1997).
PhD in Linguistics from University of Illinois in 1968. Joined the Department of Linguistics at UT Austin in 1968 as an NSF Faculty Associate. 45 PhD 1965 from University of Wisconsin. 46 PhD 1960 from University of Chicago. Professor Emeritus at UT Austin.
xl
Interview with Professor Carlota S. Smith
Carlota: Emmon. Emmon. That’s right. Emmon [Bach] was sort of the main – probably the most distinguished at the time of the linguists. Lee was more, much more junior. I mean Lee was younger than I am whereas Emmon was older than I am. So just in terms of generations. I don’t think I’m forgetting. . . Richard: Do you remember-? Keith: David DeCamp47 came? Carlota: He was there. Keith: And had he been hired in Linguistics or was he in English first or something. . . I don’t know anything about his situation. Carlota: I think he may have been in them both. Keith: Ok. Carlota: He was certainly very much a presence in Linguistics. He was sort of the connection to your world or what you are to the world of language use. Because. . . Keith: Exactly, yes. Carlota: Because really, at that time, Bob King wasn’t interested in such things. So everybody was very formal in their orientation, except of course for Harvey, who was always on the edge, you know, by choice. And Peter, of course, did his thing, which at the time was not at all what it is now. So it was kind of side by side with formal linguistics. It wasn’t taking issue with it, you know, or questions. So David DeCamp was really important. . .[inaudible]. Richard: You didn’t mention Arch Hill, did you, earlier? Carlota: He’d just retired. So I knew him only very slightly. He would occasionally come in and one would see him. But since I had no occasion really to get to know him. Richard: I didn’t realize he’d retired that early. Carlota: Well, if he hadn’t, he almost had. Richard: Yeah. Carlota: Because he wasn’t around very much. I mean, Hill Library existed. And I think he would pay attention to it. Maybe he was around, but was not so well. Richard: It was already called the Hill Library at that point? Carlota: Yes, it was. He [Hill] certainly wasn’t part of things. [. . . .] Well, he represented the earlier paradigm, as did Win. And of course, Win was quite 47 (1927–1979).
PhD from UC Berkeley in 1953. Joined UT Austin in 1959.
Interview with Professor Carlota S. Smith
xli
extraordinary, I think, in a sense, I don’t know, but I assume basically it was his decision to hire these people. So he understood, Win did, somehow that this was what it was. Richard: You said. . . And I was kind of interested before when you were talking about the training grant, that’s why I asked about David Hakes. Carlota: Yes. Richard: Was it David Hakes who decided to hire you and hire. . .? Carlota: I don’t know. Richard: Which is really a fairly extraordinary decision for someone who’s in developmental psychology. Carlota: You’re absolutely right. I have no idea. Richard: And, you know, David Hakes is kind of – is a somewhat revered person in developmental psychology here. The library here in development psychology is the Hakes Library. And he died [in 1982] before I arrived. But I never heard this association with Hakes and any association in your arrival here. And I just find it very interesting that he might have been at work in. . . At least in passing a letter onto. . . [Lehmann]. Maybe it was only passing on the letter. Carlota: That’s right. Well, after all, I was. . . I mean, he must. . . I think he knew that I wasn’t a psychologist, because my paper, that paper, it wasn’t a psychology paper. [laughs] It was a linguistics acquisition paper. Richard: Although, Shipley, Smith and Gleitman was very well cited by psychologists. Carlota: Right. Richard: So he would have been well aware of that paper. Carlota: That’s right. I think that had already been published. But this other one was just me. I went in a few other directions as well, so it probably was a pretty representative sample of what I was doing or some of what I was doing. Keith: And I would also think at that time, at least as I understand the history of the discipline, I guess you’d say the intellectual history of American social sciences more broadly, there was this belief that linguistics had something to offer, and that what was going on in linguistics was going to transform work in at least certain social sciences at a very fundamental level. Carlota: Yes, that’s a really good point. Keith: And I certainly think that psychology very much. . . Carlota: Was a perfect example. Keith: Exactly. Psychology may be the prime example.
xlii
Interview with Professor Carlota S. Smith
Carlota: Yes, that’s a great point. Because that was before the sort of what I like to call the kind of lover’s disappointment, where it’s like when the downfall of the derivational theory of complexity. That was the end – Richard: Yes. Carlota: – of the sort of romance of psychology with generative grammar, you know, where everything was gonna be sort of a little gene in the head or had a little synapse. Synapse. In fact, I remember somebody [. . .] telling me once in all seriousness that they expected to find the synapse for each transformation. [. . . .] Carlota: And I think after that there was, you know, sort of an extreme reaction the other direction, such that people weren’t interested in linguistic theories. They didn’t want to know. You know – Carlota: – didn’t want to be disappointed or seduced again. [laughs] Keith: Right, Right. Carlota: So that’s really an important point, Keith, yes. And actually when people, you know, since then – later, I got into some cognitive science and I still am – people definitely talk about Chomsky and linguistics as probably the second important thing that happened to start cognitive science as a field. First being, you know, communication theory, information theory by George Miller and Shannon and stuff. Keith: Shannon and Weaver’s stuff. Carlota: Right. And George Miller,48 too. Keith: Exactly. Yes. Carlota: The magic number seven. Keith: Exactly, yes. Carlota: But then this second step. . . sorry. Richard: Then there’s also, I guess, Jerry Bruner’s work at that time, too.49 Carlota: Yes. And actually some of Roger Brown’s work. It all fed into the same. Richard: Bruner is one of the [two] with [George Miller]. . . who got the Cognitive Science Center going at Harvard.50 Carlota: That’s right. You’re absolutely right. So it all did connect.
48 Professor
Emeritus in Psychology at Princeton University. Professor of Psychology at NYU. 50 Harvard Center for Cognitive Studies. 49 Research
Interview with Professor Carlota S. Smith
xliii
Keith: Right. Right. Carlota: Very much. And people felt it here, and I think there was a tremendous sense of pride and excitement, because at that time, people really could discover things just by looking at something nobody had looked at. And that’s essentially what happened to me. [laughs] Nobody looked at things from this lens, this perspective. Keith: Mm-hmm. Carlota: And so you could discover something every day, you know. And in fact, I remember – going back to MIT – I wrote this little paper about conjunctions. It was just a little paper with some observations. But he said, “Well, you know, you may as well write it up or publish it somewhere.” And it is published somewhere. “You know, because these are interesting observations.” Keith: That hadn’t been made before or hadn’t been documented before. Carlota: Or in this perspective. I mean, everybody from the area knew it, you know. . . [laughs] Keith: Right. Carlota: So it’s important to remember it’s a matter of perspective. Keith: Mm-hmm. Carlota: And a way of, what do you do with these facts? A way of using them. But so that was a tre-. . . And then early on, there was this big conference, the second Texas conference. I think it’s the second Texas, where Chomsky talked and Lakoff. It was just the war – the beginning of the wars of the world. You know, the generative versus interpretive semantics wars. And Chomsky was here, and a lot of people were here: Bach and Harms published that book together.51 Keith: Mm-hmm. Carlota: With Fillmore. You know, a very important paper still.52 Keith: Right. Carlota: So there was a tremendous sense that we were part of things. There. . . At that time, I don’t think there was so much the sense of the last twenty years that MIT is sort of Mecca, you know. And, I mean, I remember one year I taught at Brown [University] as a visiting professor. I used to characterize it as scuttling off to hear the latest word from Mecca, because they’d all – certain acolytes would sort of rush down for Chomsky’s class every Tuesday.
51 Bach, E., and Harms R. T., Eds. (1968). Universals in Linguistic Theory. New York: Hott, Rine-
hart & Winston. Case for Case’. In Bach & Harms (1968).
52 ‘The
xliv
Interview with Professor Carlota S. Smith
Richard: Mm-hmm. Carlota: And then others would never touch it, you know. It was just very ideological. Richard: That’s interesting you think there was less of that in the early seventies with conceiving of MIT. . .? Carlota: Yes, because things were happening in more places. Richard: In more places. Carlota: Well, Emmon was a really important person at the beginning. In generative linguistics. He was very active yet. And he was very. . . He worked with students. He did all kinds of things. And Lee, you know, was beginning to be known, too, so things were happening here. And I’m trying to think where else. Maybe not so many other places. Richard: Of course, Chicago, you know. Carlota: Well, Chicago was never generative. Richard: I was just thinking for the generative semantics stuff. [. . . .] Carlota: So there was a tremendous, you know, palpable sense of excitement that we [at UT] were doing interesting things. There were very high expectations. Keith: Mm-hmm. Carlota: Yeah, I don’t know. Any more questions?
Women’s Studies at UT Keith: Let me, let me, again, one of the things that. . . We’ve been talking about the linguistics part when you came to UT. Can you talk a little bit about the women’s studies issue? Carlota: Oh, yes, of course. Keith: Because you played a very important role in sort of getting that – I mean, legitimating that interest here and then helping ensure that there came to be a center for what was then women’s studies. Now it’s Woman and Gender Studies. Carlota: Oh, I played a little role [inaudible]. Keith: Well, if you could try. Carlota: Okay. Sure. Yeah, I would like to talk about. Thank you. Well, I gradually, you know, as a divorced, single mother, etc., I did begin to think about these
Interview with Professor Carlota S. Smith
xlv
things. And I, you know, went to college in the fifties and I went to an Ivy League school, and we were all taught to be – well, we used to say later we were all taught to have sort of salons for our intellectual husbands. You know? [chuckles] So we were supposed, we’d be all these things and not use them in any sense. And incidentally, my Radcliffe class just had its 50th reunion, which I was unable to go to, but my closest friend – still my closest friend, my roommate there, you know, was there and has reported it to me in detail. And I have the biography. There’s a book they published with everybody’s biographies, and a lot of women went on to do really interesting things. So they didn’t stay there. Keith: Right. Carlota: You know. Well, so I didn’t actually ever go to any of these consciousness raising groups or anything collective. I thought that was entirely internal. But one sort of autobiographical fact, when I was growing up, I was in a very political, intellectual environment in New York. And there was what I came to feel, was the illusion that women were taken seriously and had serious jobs. Keith: Mm-hmm. Carlota: And there used to be a tremendous amount of political and literary discussion all the time in my house. When I was a teenager, I remember sort of looking fiercely at them all, noticing that the women would sort of say their piece and everybody would listen and then the men would discuss. I mean, I’m oversimplifying, but that’s what I felt about it. And I don’t think I was altogether wrong. Keith: Right. Carlota: So they all meant it. They didn’t live it though. [all chuckle] Or they believed it, but they didn’t mean it. Some combination like that. So it was – had been in my mind for most of my life. So then when I sort of went through my own private consciousness raising, I decided – and I had a daughter also, you know, so obviously these things were very important to me. And I worried about the students at the time. When I first came and I think for the first few years, most of the linguistic students were men. The proportion is completely reversed now. Keith: Exactly. Carlota: And I worried about it and I tried to think what – you know, where along the line could women be trained or could women’s training equip them better for what they would have to face. And I began to – in this same spirit in which I looked at these groups when I was a teenager, I would notice how women behaved at the University of Texas. So for example, so remember, Richard, there was a very smart man in [a department at UT] who didn’t get tenured. He was married to a really beautiful woman with long hair who was very smart, too [who was also a professor at UT].
xlvi
Interview with Professor Carlota S. Smith
Richard: You’re talking about [the woman]. [. . . .] Carlota: Yes. I’m really talking about her. So for example, one day, I remember there was some meeting or some committee in my office. I was chair, I guess. So people came, and we were maybe waiting for somebody. And the woman, who’s very smart, very productive, sort of an aggressive woman, was sitting there smiling at some jerk, some male, and, you know, asking what he did. You know, practically blinking her eyes; although, I don’t think she was, but I was just appalled. And I’m sure I did it, too; although, I did it less and less. You know, this sort of reflexive feminine cocktail party behavior. At a dinner party, you know, “Tell me what you do. What are you working on?” Keith: Well, what Pamela Fishman calls the “interactional shit work” that women do. I mean, that’s her phrase – Carlota: Exactly. Oh, it’s perfect. Keith: – for exactly that kind of role that woman are often assigned in this society. Carlota: You’ve come to the end of some tape or other. Keith: I would like to ask another thing. Do you remember reading Lakoff’s Language and Women’s Place?53 Carlota: Of course. Keith: As you may know, it has been recently been republished. Carlota: Oh, I didn’t know. Keith: With notes by her. And then there are comments by, oh, 15 or 20 linguists writing now about how her work has influenced – Carlota: Oh. Keith: – their current thinking about language in general. Carlota: Oh, I must have a look. Do you have it? Keith: Yes, I have it. Carlota: Can I borrow it sometime? Keith: Yes, yes, yes, yes, yes. And so I was just thinking I would be very interested in hearing what you recall about reading Language and Women’s Place.
53 Lakoff,
Robin T. (1975). Language and Woman’s Place. New York: Harper & Row. Revised and expanded edition: Lakoff, Robin T. (2004). Language and Woman’s Place: Text and Commentaries, edited by Mary Bucholz. New York: Oxford.
Interview with Professor Carlota S. Smith
xlvii
Carlota: Yeah. Right, right. Okay. Let me think a minute. Well, I had mixed feelings about it. I thought it was programmatic. Keith: Mm-hmm. Carlota: It certainly didn’t sort of demonstrate. And I thought some that was oversimplified, which of course turned out to be the case, like the stuff about tag-questions. Lots of interactional uses. Keith: Right. Carlota: And but . . .uh. . . I thought it was wonderful. I mean, my main feeling was yes. [Keith chuckles] You know, yes, there is a pattern of behavior and much of it is linguistic and intonational, and it does sort of cohere into a woman’s style, a powerless style. I thought even though things were wrong – may have been wrong, she was basically right. And I still think that. And I think all the research that it, you know, inspired is testament to that, you know, because it was really very new. Nobody had – I mean, DeCamp, people like that were more traditional. Keith: Oh, yes. Exactly. Right. Carlota: You know, and he – well, she really just identified things. I think it was a research agenda – Keith: Right. Exactly. Right. Carlota: – that she identified. And I thought it was just terrific. And I think I had people read it. I never did that kind of work myself. But I was very interested in it and I followed a lot of the research about it. And I haven’t done this for a long time, but at one point, I was teaching a course called “Language and Power” before you – Richard: Oh, I remember that. Carlota: – came and starting teaching your stuff. Keith: Right, yes. Carlota: And we definitely read not only Lakoff, but then a lot of the studies that followed up: some of the studies about pronouns, you know, the generic he. Keith: Right. Exactly. Sure. Carlota: Yeah. All that stuff. So yes, I think I understood it right away that it was important and that it was basically right, you know. And so then, you know, at UT, I just did whatever I could, so. . . Oh, I know what it was. I happened to be chair of the department at a time – at the time when women studies was being discussed. And I very deliberately said to myself, “I am going to do whatever I can to promote this.” And I was on the committee. Betty Sue Flowers54 and 54 Formerly,
Professor of English at the University of Austin. Former director, Lyndon Baines Johnson Library and Museum.
xlviii
Interview with Professor Carlota S. Smith
various people. And I definitely, you know, whatever small wave I could bring, I did. So that was not chance at all. Keith: Right. Carlota: And then that one year I was head of the Woman. . . Faculty Women’s Caucus, that was much later. So yes, I’ve certainly been aware of it. And there is this tradition, maybe less so now that most senior women figured, they made it on their own, so why worry? Keith: Right. Carlota: So, I think the sense that a lot of women cared about this mattered. And now, I have to say I’m not doing anything on the Women’s Faculty Caucus. I partly feel that it’s really. . . When I first got on it, there were a lot of junior women who were very active or recently tenured women. And in the past few years, it’s seemed to me that it’s mostly people who’ve been around forever. [. . .] Women who’ve been doing this. And so it’s sort of, there we all are talking to each other. It’s sort of dead. Richard: Yeah. My impression has been that junior women in Linguistics have not typically been very involved in that organization. Carlota: No, no. And senior women. . . I remember taking Nora [England]. And she just said, “Oh, that was really boring.” [laughs] And it was. She was exactly right. I took her to the opening tea or something like that. Keith: Can you talk a little bit about how you think the situation of women as faculty or as students – how that situation has changed at UT during the time that you’ve been here? I mean, as you mentioned earlier, the demographics of the student population has certainly changed. Carlota: Reversed. Keith: It’s reversed. And do you have any observations you would share about that shift or what it’s meant or how aware people had been of what might have been going on while that transition took place? Carlota: I’m not sure people have the faintest idea, if by people you mean students. Keith: Okay, students. Faculty. . . Carlota: Faculty. . . Keith: Administrators. [laughs] Carlota: Administrators, I don’t know. I think administrators were not unaware, but I’m not sure they did anything much. In fact, I don’t know that they did. [talking to the cat] Just ignore it. It’s the only thing to do. Otherwise, you really do spend all your time rushing round. Oh, I do a lot [clapping her hands] [laughs] especially now since I’m here. Oh, well, I might as well. But the men are out. Well, you know, when I first became active in this, I noticed something that we
Interview with Professor Carlota S. Smith
xlix
all noticed – that the problem wasn’t for women getting a job. Women are perfectly good at writing dissertations. Keith: Mm-hmm. Carlota: And women are very good at doing what they’re supposed to do. Richard: Right. Carlota: You know, so they did what you’re supposed to do to get a job, and they would get a job. It’s then that the problem sets in. Partly the biological clock, and, you know, husbands or significant others who have jobs, too, and, you know, all that stuff, which is still there. Well, no, I take it all back about the administration, because of course the whole maternity leave efforts have been recognition of that very point. Keith: Mm-hmm. Though even that, as I understand it, is quite problematic. Carlota: It’s very. . . From what I hear, it’s very problematic and it’s all. . . [talking to the cat] NO! He actually understands ‘no’ said in a very firm tone. [laughs] Richard: I’m sure he understands everything you say. [laughs] Carlota: Yes, that’s right. He’s translating it into, what. . .Mesopotamian or something [laughs] Yeah, it seems to be much too dependent on the individual department. Keith: Exactly. Yes. Right. Carlota: And so, you know, in some departments women are afraid to ask, because what will people think? And others, they ask and they’re sort of discouraged. You know, then in some, they’re encouraged maybe. So completely random or completely accidental. Keith: “Can you get pregnant next semester rather than this semester?” or something. [laughs] Carlota: Yes. Well, I’m asked sometimes, and I always say the sooner the better getting pregnant, because if you wait, the longer you wait, the closer you are to tenure and blah, blah, blah, and it’s harder and you’re older. Richard: And so was it helpful for you that you arrived here with your children already? Carlota: Of course. Indeed. And people say, “Oh, how could you have done all that in graduate school?” Well, I had no choice, right? So I did it, and it was done. Richard: Right. And how old were Joel and Allison55 when you got here? Do you remember? If you can’t, we’ll figure it out.
55 Carlota’s
children.
l
Interview with Professor Carlota S. Smith
Carlota: Should be able to do the arithmetic. Richard: Yeah, if it’s sixty. . . Carlota: Maybe six and eight or something. Richard: So they’re – they’re kind of – they’re already into school. Carlota: They’re already in school. They went to school here. Richard: So as soon as you arrived here, they were off to whichever school it was they went to. Carlota: That’s right. That’s right. And one more thing, I did not do much linguistics in the first few years I was here. I remember one day thinking, hmm, why is everybody making more money than I am? Keith: Right. Carlota: And then thinking, oh, well, I haven’t published anything lately, have I? Because I would get home in the afternoon. Try to get home in the afternoon. And then I’d sort of pull myself together and started doing some work. It was a lot better. So it – it wasn’t all done. You know, it was very. . . I mean, when you’re a single mother, you have two children, there is – there is a lot to do. Keith: Right. Richard: Tremendous. Carlota: I’m certainly not sorry. I was just lucky. I was just incredibly lucky that I got tenured, actually, because I didn’t try to get tenure, and I did my best work much later, except maybe those first articles. So it was just a lax system, you know. Richard: A very different system. Different time. Carlota: Different time. Keith: A very different system. Carlota: Well, I think Win really did believe in me. Richard: Mm-hmm. Carlota: I know that. . . I mean, I did do a little, but I didn’t do much. And I think, you know, he must have written good letters, but I haven’t a clue. I wasn’t here. I didn’t make any – I didn’t put together a file. Richard: But nobody – but nobody told you, oh, silly thing to go off to France that year. Carlota: Oh, no. Richard: So they weren’t saying, “Oh, you must stay here.” Carlota: Oh, no, no, no. I wasn’t. . . I wasn’t doing anything unusual.
Interview with Professor Carlota S. Smith
li
Keith: Right. Carlota: Everyone got tenure or something. [laughs] Keith: Right. Carlota: At least that was my perception. I think in Linguistics everyone did. So I wasn’t alone. But anyway, back to . . . So certainly the business about getting pregnant, you know, there’s been some effort, perhaps not entirely – well, definitely not entirely successful, but some institutional effort. And the Women’s Faculty Caucus and I think other people too in the administration are still trying to figure out what to do. Keith: Right. Carlota: And the thing that’s most shocking is that the numbers – the patterns are not any different basically. Keith: Mm-hmm. Right. Carlota: And how can that be? It is a mystery, at least it is to me. Now I think my sense is that women I know are much more professional after they got their first job than they used to be. It seems to me 15–20 years ago, you would read about first-rate dissertations and then you’d never hear of the woman again and you’d hear about the men again. I remember noticing that and talking to people about it. I think at that Cornell conference, that was one of the points I was particularly concerned with. I don’t think you’d find that anymore. In fact, I have two. I just did a tenure review for somebody, a woman, and I was very impressed with the consistent work. And, you know, think of our own women. They know what you’re supposed to do. Keith: Yes. Right. Exactly. Carlota: Yeah. So I think there is a difference. I think women are more professional in their first years in jobs. [. . . .] Richard: There’s one other thing I wanted to ask about. . .and I guess that’s all. Carlota: Yeah. Richard: And that. . . And that just is on terms of the trajectory of your research, I mean, now. Carlota: Oh, yes, you asked me about aspect. Richard: Yeah. Now, I mean, now much of your research is on tense and aspect. Carlota: Yeah, I’m really trying to get away from it. I’ve been doing it forever. Richard: Very firm. Carlota: 20 years. Richard: But how did you. . .?
lii
Interview with Professor Carlota S. Smith
Carlota: Yeah, I’ve wondered that. I think it was when I was in France. And the difference between French and English, because it’s sort of intriguing. They are so close! But they happen to be really different in that area. Richard: Mm-hmm. Carlota: And the truth is. . . When I was there, I regarding teaching as an interruption in my main activity, which was learning French. So I just loved learning French, and I read everything! I read comic books, I read leftwing papers, rightwing papers, novels, magazines. I watched television. I talked to everybody all the time. You know, the smallest exchange in the grocery store was fun, because I would pick up a little something. Keith: Mm-hmm. Right. Carlota: And I just – so I did get very involved in the language and I did learn a lot about it. Richard: And so the difference in the aspectual systems. . .[inaudible]. Carlota: I’ve wondered myself, and I think that’s the answer. I think I got interested then. Richard: Mm-hmm. Carlota: And then somehow. . . Oh, I guess the first paper I did was actually about the auxiliary have,56 and then I kind of went on from there. But I haven’t ever been so disciplined as to say, “Oh, now I’m going to work on ‘x.’” Do you do that? Do you do that? I don’t think people do. You know, you find that you’re working on ‘x.’ Richard: I mean a new topic. Yeah, I mean, this kind of. . . You hear English professors sometimes. They finish one book project for tenure and then they pick another project that seems, to my mind, light years different than the previous one. So it’s very, you know, different. . . Carlota: Well, I remember having sort of post-partum depression after the book, my first book, the aspect book. And I remember talking to a friend, Fred Wiseman, the filmmaker, and he said, oh, that used to happen to him all the time, and now he just makes sure to get a new project in mind before the end of the film to avoid that. Richard: Mm-hmm. Keith: Well, I have a question that relates to a phrase that you used just when we were talking about things at UT and about the tenure situation. You said that you observed that you had done your best work long after you had gotten tenure. What would you characterize as your best work? And what do you like about it?
56 ‘A
Theory of Auxiliary Have in English.’ Indiana University Linguistic Club. 1976.
Interview with Professor Carlota S. Smith
liii
Carlota: Oh God. Well, I do think that early work was kind of good actually. But since then, well, I guess the aspect book. I guess I like both my books, because I think their modes of discourse is very original, I think.57 Keith: Right. Exactly. Carlota: It’s not like anything else. So I kind of like that. Keith: Mm-hmm. Carlota: And well, actually, Chomsky once said to me thousands of years ago that, you know, if you really go into anything in detail, you get into the whole language. And the thing about aspect is that I’ve done that, I think or not entirely but a lot of it. So it has taken me into all sorts of sub-sys – like a determiner system, which might seem very far from aspect, but it really isn’t if you think about generics and situations. Keith: Right. Exactly. Yes. Carlota: So it’s been my way into cross-linguistic work, which I never used to do. And it’s been my way into semantics. So it’s just taken me very, very far, so it’s been great fun for that reason. So I guess what I like about the work is the sort of [unintelligible] or how that’s been realized in the work. Keith: Mm-hmm. Carlota: And I think it is sort of. . . It’s been fun to try to do general theories and see if they work or not. So the old things. Keith: Right. Well, I mean, one of the things, as I told you, that I’ve always admired about your work is the way that you bring things together. I mean, ok, syntax and acquisition. You’ve long had an interest in discourse that’s taken a lot of forms. Things from stylistics to your book to other sorts of things that somehow bring together issues of linguistic theory or questions that formalists ask and applied them to context or situations that your average formalist doesn’t spend much time thinking about. Carlota: Yeah, I’m not formal enough for most formalists. Keith: Mm-hmm. But it’s that sort of bringing these things together that I think is such an achievement. Carlota: Oh, I see. Well, maybe that’s true. Yeah, because I sort of – I know how to do basic formal, but it isn’t really the way I think. Keith: Mm-hmm. Carlota: Except whenever I formalize something, I learn something.
57 The
Parameter of Aspect (1991/1997) and Modes of Discourse (2003).
liv
Interview with Professor Carlota S. Smith
Keith: Right. Carlota: So I know not only do you have to do it to be taken seriously, but you have to do it to make sure you’ve thought of everything or at least you can be as complete as possible. Actually, I did do something about discourse early, early. I wrote that article about Bertrand Russell’s style.58 Have you ever seen that? Keith: Mm-hmm. Carlota: So that was sort of way anticipating a lot of surfacy interpretative stuff. Because that was when everybody was talking generative and I kind of realized, oh, that is not of any use here, you know. [laughs] Keith: Right. Carlota: And I published it in some British journal, which I realized would be the right place. Keith: Right. Carlota: But yeah, that is – that is true. I have been able to combine things, you know, not as much as I like. I mean, I don’t know enough languages, for example, and that’s a real defect. Richard: But you’ve worked on a fair number of languages as it turns out. Carlota: Well, what I try to do is stay with those same ones. . . The thing is, for the stuff I work on, you really have to go into detail. Keith: Exactly. Carlota: It’s partly research style. I’m just not interested in 50 languages, you know, “x” in 50 languages, but I’m sure, I know it’s useful to do that. It just doesn’t amuse me. I could do it. So I stayed with. . . I mean, I didn’t stay with Russian. Keith: Right. Carlota: I might have gone back to Russian if I hadn’t gotten sick actually, because of this conference I was supposed to go to. I had thought of doing something last week. I was thinking about going back to Russian. But the reason is I didn’t know it enough. Keith: Right. Carlota: And with Chinese, you know, I can’t chat, but I really know a lot. And the same with Navajo. So as. . . And of course, I do know French pretty well, so I just keep recycling. I mean, that’s what I was working on this spring, the discourse mode stuff in Navajo and Chinese.
58 ‘Sentences in discourse: an analysis of an essay by Bertrand Russell.’ Journal of Linguistics, 7 (1971), 213–235.
Interview with Professor Carlota S. Smith
lv
Keith: Right. Carlota: So it’s just turned into my research strategy. And I think it’s all I could do, you know, because I’m not going to learn. And I have these fantasies of learning, you know, Hebrew, which is very different or going back to Latin which I used to really enjoy thousands of years ago, and maybe I will, but I probably won’t do any work on those languages. So it’s just been sort of faute de mieux in a way. You know, once I got interested in the general questions, I understood quickly that I couldn’t do it for lots of languages. Keith: Right. Right. Yes. Carlota: You know, it just wasn’t possible. Well, you do the same thing, Keith. I mean, you work on languages you know about. Keith: Exactly. Right. Yeah. Carlota: And you, too, Richard, right? Richard: Right. Carlota: You think so? Richard: Well, I think it’s. . . I asked this question about how you get into – got into aspect, because I think it’s interesting to trace these trajectories. Carlota: How did you get into ASL? Richard: Well, getting into ASL was just being offered a research assistantship by Ed Klima. There is no – no story, no romance there at all. Carlota: That’s like me and Lila. Richard: Yeah, like you and Lila. And – and everyone thinks when one works with sign languages or with deaf people that there must be some family reason. Keith: Right. Carlota: Oh, really? Richard: No, there has to be. There has to be. Carlota: Oh. Richard: Or, “Your parents are deaf. You must have deaf relatives.” Keith: Right. Carlota: I see. I guess that’s plausible. Richard: And there’s kind of a reverse sexism here, because for a male to work with deaf people – Carlota: Right. Richard: – it’s awfully like a helping profession type thing.
lvi
Interview with Professor Carlota S. Smith
Keith: Exactly. Richard: I remember talking to somebody on a plane, you know, the usual kind of discourse about how you got into – to how I got into working in sign. And he – he – clearly, this man thought it was odd that, you know, a grown man – Carlota: [laughs] What are you doing here? Richard: – would be working with deaf people, so, and sign languages. So, but enough. Carlota: Aren’t you furious? Richard: It is. It is. Carlota: I mean, it’s just so annoying. Richard: Yeah. And then also just on how particular – how we come to explore particular ideas or topics. There’s a paper that I’m thankfully just about to finish up the revisions on and send off that I’ve been working on for so long. And anyway, the idea for this paper is something that I know dates from 1980 or 1981. When I was a graduate student, I scribbled down some notes – Carlota: Really? That’s exciting. Richard: – about – about particular kinds of errors that children made in learning to sign. And I never forgot about it, but it wasn’t anything I explored until actually it happens that Gene Mirus brought up a kind of related notion. I don’t know how many years ago by now. But it was just something I’d had these cards. And I think I was able to even find the 3 × 5 cards. Carlota: That is so wonderful. Richard: And I knew one of the references still. And it just, you know, it just kind of lays there for a very long time. Keith: Right. Carlota: Germinating. Richard: Yeah, germinating or just waiting for the right time to germinate. Carlota: Mm-hmm. Yes. Richard: And it’s just – it’s kind of fascinating. I was interested when we were talking about Penn that Zeno Vendler was there and you obviously often cite Zeno Vendler. Carlota: I do, yes. Richard: And you mentioned with regard to Lila working, what you and she were working on verbs, and of course in her syntactic bootstrapping model – Carlota: That’s right. It was all to do. . .[inaudible]
Interview with Professor Carlota S. Smith
lvii
Richard: – verbal – the semantics with verbs. Carlota: That was a summer project that year. Richard: Yeah. And so I don’t know what she would say about that, but it’d be interesting to ask Lila – Carlota: It would be interesting. Richard: – how much of this interest now with the lexical semantics, complement structures of verbs, how much of that dates to this project with Zellig Harris. Carlota: Well, you know, it’s interesting, because, as I said, the stated goal – like, Harris had no interest in it – was mechanical translation. He thought – he just was getting money. I mean, he wa-. . . The reason he had no interest is he thought it was wildly premature. He just thought you have to have a good formal grammar before you could talk about the other. So it wasn’t that he had no. . . But anyway, so I think for any language to document seriously you gotta have that. So in a way it’s a very traditional question. It’s not just a computational question or just an acquisition question. It’s just a sort of structuralist question Keith: Right. Right. Carlota: As well as a traditionalist question. It just is always there. A propos of being annoyed, I remember at the doctor, there was, I had a conversation with a doctor that annoyed me in perhaps a similar way, where we were talking about getting the results of some test quickly. He said, “Well, you’re always nice, aren’t you?” And I just thought, “What?! You know, no, I’m not always nice!” [all laugh] And, you know, this assumption, this nice lady. I’m not going to shriek at you! You’re the doctor, you know.” [laughs] And I actually don’t shriek at people very often professionally. In fact, I hope – I think probably not much. But, you know, the idea that’s sort of automatic. And I think women – back to your question, Keith – I don’t think there is so much of that kind of social shit work behavior, if I, you know, see women around in the halls, you know, profe – you know. Keith: Right. Carlota: I. . . Do you? I mean, I don’t. . . Keith: I think the professional context it’s changed – there’s a – you may know an article by Maltz and Borker59 from the eighties. They take a lot of literature about male and female differences in Anglophone countries. And they make the argument that boys and girls are raised in very different cultures. Boys are raised in hierarchical cultures.
59 Maltz,
Daniel L., and Borker, R. A. (1982). A cultural approach to male-female miscommunication. In J. J. Gumperz (Ed.), Language and Social Identity, 417–434. New York: Cambridge University Press.
lviii
Interview with Professor Carlota S. Smith
Carlota: Oh, yes. Yes. Keith: Looking at how they interact on the playground. Carlota: Right. I remember that. Keith: Versus for girls, it’s the culture of the bedroom. You and your best friend go off and spend time in someone’s bedroom. Carlota: Yes. Keith: Maybe a different best friend every week, but you gotta have a best friend. Carlota: Right. And you have sort of little dishes. Keith: Exactly. And you know. . . Carlota: I never had any of that. Keith: [Deborah] Tannen very much took this ball and ran with it. Carlota: She did. Keith: Very successfully. But I have taught that article to undergraduates. And one of the things that I make a point of doing is, you know, we get in a circle and we’re talking. And for about 15 minutes, I try to drop out of the discussion. And I just note down who does what in terms of who interrupts, who takes credit for what’s said, who builds on what is said previously, and then I share the results with the class. Generally, the students are saying, “Oh, this has changed. This is back in the eighties. Feminism happened. We don’t have these problems anymore.” [laughs] Carlota: But in fact we do. Keith: But in fact you find a lot of those things still going on. Carlota: Same damn thing. No, that’s what I find teaching undergraduates. Usually, there are a couple of exceptions. Usually, there are a couple of women who are smart, who like to talk. Keith: Right. Carlota: But except for them, it goes the other way. Keith: Right. Carlota: There are plenty of smart women, who if you – that is, women who know the answer if you ask them. But in terms of initiating things, no. Keith: And so I think that many of the young women do come to appreciate, okay, they have to develop a wider range of skills. In other words, if they are going to be able to move into these professional contexts, they are going to need some skills. But there again, what we see is this asymmetry. Men are rarely forced to develop a range of skills –
Interview with Professor Carlota S. Smith
lix
Carlota: That’s right. Keith: – other than the one that – the ones that they’re really comfortable to using. Carlota: The obvious ones. Keith: Yes. Carlota: Right. Keith: The ones that they’ve been socializing into. Carlota: They’re programmed for. Richard: And I think there have been times in our department where women – younger women obviously had lots of skills and those – Carlota: They’re exploited, right? Richard: – those skills. . . What? Carlota: Exploited. Richard: Those skills – anyone understood why they were exploited. Because the alternative would be to give the task to someone – Carlota: Didn’t want to do it. Richard: – who wouldn’t have – wouldn’t have done it well. Keith: Right. Wouldn’t have taken it seriously. Carlota: Well, when I was chair, I remember noting to my rage and somewhat amusement, [Prof. X] used that technique when you would ask him to do something. He would just do a terrible job. Keith: Right. Carlota: Secure in the knowledge that you wouldn’t ask him to do more things. Well, I did anyway, but I was careful not to ask him much, because he really would do a terrible job. Keith: Mm-hmm. Right. Carlota: And it’s a very effective technique. Of course, there’s also saying no, – Richard: Yeah. Carlota: – which women are less likely to do – Keith: Right. Exactly. Yes. Carlota: – than men. Keith: Right. Carlota: Although, it seems to me with the Women’s Faculty Caucus, the young women that are there are – obviously it’s self-selection, you know.
lx
Interview with Professor Carlota S. Smith
Keith: Mm-hmm. Sure. Carlota: So they are more professional. And also, I think, a lot of women in that certain context – Keith: Exactly. Carlota: – say, of all women, it’s a classic girl’s school phenomenon, where they know what they’re supposed to do, you know, in a sort of certain hospitable context. Keith: Right. Carlota: But they might not do that at all in a different context. Keith: Yes. Right. Carlota: And that’s very hard, I think. You have to decide you’re not going to do it. [laughs] And it takes a while. Because deciding is only the beginning. [laughs] Keith: It’s the un-learning part.
Carlota S. Smith: Publications
Books Smith, C. S. (2003). Modes of discourse. Cambridge: Cambridge University Press. Smith, C. S. (1991). The parameter of aspect. Dordrecht: Kluwer. Paperback edition, 1993; 2nd paperback edition, 1995. 2nd edition, revised (hard-cover and paperback), 1997.
Papers Smith, C. S. (in press). Tense and aspect: Time in language. In C. Maienborn, K. von Heusinger, & P. Portner (Eds.), Semantics: An international handbook of natural language meaning. Berlin: Mouton de Gruyter. Smith, C. S. (2007). Tense and context in French. In E. Labeau, C. Vetters, & P. Caudal (Eds.), Sémantique et diachronie du système verbal français. Cahier Chronos 16. Amsterdam: Rodopi. Smith, C. S. (2007). Tense and temporal interpretation. Lingua, 117, 419–436 Smith, C. S., & Perkins, E. (2007). Temporal inference and zero-marked clauses in Navajo. In M. Becker & A. McKenzie (Eds.), UMOP 33: Proceedings of the 3rd Conference on the semantics of underrepresented languages in the Americas (pp. 121–133). Graduate Linguistics Student Association, University of Massachusetts Amherst. Smith, C. S., Perkins, E., & Fernald, T. (2007). Time in Navajo: Direct and indirect interpretation. International Journal of American Linguistics, 73, 40–71. Smith, C. S. (2005). Aspectual entities and tense in discourse. In P. Kempchinsky & R. Slabakova (eds.), Aspectual inquiries (pp. 223–238). Dordrecht: Springer. Smith, C. S. (2005). Event structure and morphosyntax in Navajo. In N. Erteschik-Shir & T. Rapoport (Eds.), The syntax of aspect: Deriving thematic and aspectual interpretation (pp. 190–212). Oxford: Oxford University Press. Smith, C. S., & Erbaugh, M. (2005). Temporal interpretation in Mandarin Chinese. Linguistics, 43, 713–756. Smith, C. S. (2004). The domain of tense. In J. Guéron & J. Lacarme (Eds.), The syntax of time (pp. 597–620). Cambridge, MA: MIT. Smith, C. S., Perkins, E., & Fernald, T. (2003). Temporal interpretation in Navajo. In J. Anderssen, P. Menéndez-Benito, & A. Werle (Eds.), SULA 2: Proceedings of the Second Conference on the Semantics of Underrepresented Languages in the Americas. Graduate Linguistics Student Association, University of Massachusetts Amherst. Smith, C. S. (2003). Tense and aspect. In L. Nadel (Ed.), Encyclopedia of cognitive science (Vol. 4., pp. 361–367). London: Nature Publishing Group. lxi
lxii
Carlota S. Smith: Publications
Smith, C. S. (2002). Perspective and point of view. In H. Hasselgard, S. Johansson, B. Behrens, & C. Fabricius-Hansen (Eds.), Information structure in a cross-linguistic perspective (pp. 63–80). Amsterdam: Rodopi. Smith, C. S. (2002). Accounting for subjectivity. In B. Nevin (Ed.), The legacy of Zellig Harris (pp. 137–163). Amsterdam: John Benjamins. Smith, C. S. (2001). Discourse modes: Aspectual and tense intepretation. In J. Busquets, F. Lambert, & A. Le Draoulec (Eds.), Cahiers de grammaire n 26, sémantique et discours. Toulouse: ERSS et Université de Toulouse-Le Mirail. Smith, C. S., & Erbaugh, M. (2001). Temporal information in sentences of Mandarin. In Xu Liejiong & Shao Jingmin (Eds.). New views in Chinese syntactic research – International symposium on Chinese grammar for the new millenium (pp. 514–542). Hangzhou: Zhejiang Jiaoyu Chuban she. Smith, C. S. (2001). The Navajo prolongative and lexical structure. In A. Carnie, E. Jelinek, & M. A. Willie (Eds.), Papers in Honor of Ken Hale. Working Papers on Endangered and Less Familiar Languages. Vol. 1: Studies in endangered languages. Cambridge, MA: MIT Working Papers in Linguistics. Smith, C. S. (2001). Discourse mode: an interesting level of local structure. Proceedings of the third workshop on discourse structure. Department of Linguistics, University of Texas at Austin. Smith, C. S. (2000). The semantics of the Navajo verb base. In T. Fernald & P. Platero (eds.), The Athabaskan languages: Perspectives on a Native American language family (pp. 200–227). Oxford: Oxford University Press. Smith, C. S. (1999). Activities: States or events? Linguistics & Philosophy, 22, 479–508. Smith, C. S. (1999). Closed systems in texts. Proceedings of the Workshop on Text Structure. University of Texas at Austin. Smith, C. S. (1999). Cues to the small structure of texts. In M. Plénat, M. Aurnague, A. Condamines, J.-P. Maurel, C. Molinier, & C. Muller (Eds.), L’Emprise du sens: Structures linguistiques et interprétations, Mélanges de syntaxe et de sémantique offerts à Andrée Borillo par un groupe d’amis, de collègues et de disciples. Plénat, M., (Eds.) Amsterdam: Rodopi Publishers. Collection “Faux titre”, n 174. Smith, C. S. (1998). The uses of focus. Proceedings of the Workshop on Text Structure. University of Texas at Austin. Smith, C. S. (1997). Review article, Travaux du cercle linguistique de Prague, Nouvelle Serie. Hajicova, E., Leska, O., Sgall, P. Studies in Language, 21, 675–686. Smith, C. S. (1996). Aspectual categories in Navajo. International Journal of American Linguistics, 62, 227–263. Smith, C. S. (1995). Activity sentences in narrative: States or events? In P. Amsili, M. Borillo, & L. Vieu (Eds.), Time, space and movement: 5th international workshop. Toulouse: Institut de Recherche en Informatique de Toulouse. Smith, C. S. (1995). The range of aspectual situation types: Derived categories and a bounding paradox. In P. M. Bertinetto, V. Bianchi, J. Higginbotham, & M. Squartini (Eds.), Temporal reference, aspect and actionality. Vol. 2: Typological perspectives (pp. 105–124). Turin: Rosenberg & Sellier. Smith, C. S. (1995). The relation between aspectual viewpoint and situation type. Published electronically, ERIC Database. Smith, C. S. (1995). Review of M. Axelrod, The semantics of time: Aspectual categorization in Koyukon Athabaskan (Lincoln: University of Nebraska Press). American Anthropologist, 97, 393–394. Smith, C. S. (1994). Aspectual viewpoint and situation type in Mandarin Chinese. Journal of East Asian Linguistics, 3, 107–146. Smith, C. S. (1994). Pragmatic principles in coreference. In B. Lust, M. Suner, & J. Whitman (Eds.), Syntactic theory and first language acquisition: crosslinguistic perspectives. Vol. 2: Binding, dependencies, and learnability (pp. 335–354). Hillsdale, NJ: Erlbaum.
Carlota S. Smith: Publications
lxiii
Smith, C. S. (1993). The acquisition of tense: Bootstrapping into syntax. Proceedings of the conference, early cognition and the transition to language. Center for Cognitive Science, University of Texas at Austin. Smith, C. S. (1992). Another look at the temporal system of English. In M. Aurnague, A. Borillo, M. Borillo, & M. Bras (Eds.), Proceedings, 4th International workshop on semantics of space, time, and movement (pp. 153–165). Toulouse: Groupe “Langue, raisonnement, calcul,” Centre National de Recherche Scientifique. Smith, C. S. (1991). A valediction for sentence topic. In C. Georgopoulos & R. Ishihara (Eds.), Interdisciplinary approaches to language: Essays in honor of S.-Y. Kuroda (pp. 545–565). Dordrecht: Kluwer. Smith, C. S. (1990). Psycholinguistics: A research review. In E. C. Polomé (ed.), Research guide on language change (pp. 175–216). Berlin: Mouton de Gruyter. Smith, C. S. (1990). Event types in Mandarin. Linguistics, 28, 309–336. Smith, C. S. (1990). From graduate school to tenure. In A. Davison & P. Eckert (eds.), The Cornell lectures: Women in the linguistics profession (pp. 239–244). The Committee on the Status of Women in Linguistics, Linguistics Society of America. Smith, C. S. (1989). Event types in Mandarin. In M. Chan & T. Ernst (Eds.), Proceedings of the Third Ohio State University Conference on Chinese Linguistics (pp. 215–243). Indiana University Linguistics Club. Smith, C. S. (1988). Factors of linguistic complexity and performance. In G. Green & A. Davison (Eds.), Linguistic complexity and text comprehension (pp. 247–279). Hillsdale, NJ: Lawrence Erlbaum Associates. Smith, C. S. (1987). The viewpoint aspects of Mandarin Chinese. Technical Report, Center for Cognitive Science, University of Texas. Smith, C. S. (1987). Notes on aspect in Mandarin Chinese. Texas Linguistic Forum, 28, 91–187. Smith, C. S. (1987). The information needed for inference. Behavioral and Brain Sciences, 10, 733–734. Weist, R., & Smith, C. S. (1987). On the temporal contour of child language: A reply to Rispoli and Bloom. Journal of Child Language, 14, 387–392. Smith, C. S. (1986). A speaker-based approach to aspect. Linguistics and Philosophy, 9, 97–115. Smith, C. S. (1986). Sentence topic in texts. Studies in the Linguistic Sciences, 15, 187–203. Smith, C. S., & van Kleeck, A. (1986). Linguistic complexity and performance. Journal of Child Language, 13, 389–408. Smith, C. S. (1986). The parameter of aspect. In S. Choi, D. Devitt, W. Janis, T. McCoy & Z. Zhang (Eds.), Proceedings of ESCOL ’85 (Eastern States Conference on Linguistics) (pp. 210–231). Whitaker, J., & Smith, C. S. (1984). Some significant omission: Ellipses in Flaubert’s ‘Un coeur simple.’ Language and Style, 14, 251–292. Smith, C. S. (1983). The temporal reference of the English futurate construction. Cognition and communication, 16, 81–96. [Reprinted in: E. Tasmowski & D. Willems, eds. (1983). Problems in syntax (pp. 273–288). New York: Plenum Press]. Smith, C. S. (1983). A theory of aspectual choice. Language, 59, 479–501. Smith, C. S. (1982). Aspect and aspectual choice. Texas Linguistic Forum, 19,167–100. Smith, C. S. (1981). Learnability and the problem of productive lexical rules. In C. L. Baker & J. McCarthy (Eds.), The logical problem of language acquisition (pp. 151–164). Cambridge, MA: MIT Press. Smith, C. S. (1981). The futurate progressive: not simply future + progressive. Proceedings of the Chicago Linguistic Society, 17, 369–382. Smith, C. S. (1981). Semantic and syntactic constraints on temporal expressions. In P. Tedeschi & A. Zaenen (Eds.) Syntax and Semantics. Vol. 14: Tense and aspect (pp. 213–239). New York: Academic Press. Smith, C. S. (1980). The acquisition of time talk: Relations between child and adult grammars. Journal of Child Language, 7, 263–278.
lxiv
Carlota S. Smith: Publications
Smith, C. S. (1980). Temporal structures in discourse. In C. Rohrer (ed.), Time, tense, and quantifiers (pp. 355–375). Tübingen: Niemeyer. Smith, C. S., & Whitaker, J. (1979). Interpreting ellipses in a text. Pragmatics Microfiche. Smith, C. S. (1978). Capturing temporal structure. CUNY Forum, 5–6, 75–87. Smith, C. S. (1978). Constraints on temporal anaphora. Texas Linguistic Forum, 10, 76–94. Smith, C. S. (1978). The syntax and interpretation of temporal expressions in English. Linguistics and Philosophy, 2, 43–100. DiPaolo, M., & Smith, C. S. (1978). Cognitive and linguistic factors in language acquisition: The use of temporal and aspectual expressions. In P. French (ed.), The development of meaning (pp. 338–351). Hiroshima: Bunka Hyoron. Gleitman, L. R., Shipley, E. F., & Smith, C. S. (1978). Old and new ways not to study comprehension: Comments on Petretic and Tweney’s (1977) experimental review of Shipley, Smith & Gleitman (1969). Journal of Child Language, 5, 501–519. Smith, C. S. (1977). The vagueness of sentences in isolation. Proceedings of the Chicago Linguistic Society, 13, 568–577. Department of Linguistics, University of Chicago. Smith, C. S. (1978). The temporal interpretation of complements in English. Texas Linguistic Forum, 6, 150–168. Smith, C. S. (1976). A theory of auxiliary have in English. Indiana Linguistics Club. Smith, C. S. (1975). Review article, ‘Cognitive development and the acquisition of language,’ T. Moore (Ed.), Language, 2, 303–317. Smith, C. S. (1975). The analysis of tense in English. Texas Linguistic Forum, 1, 71–89. Smith, C. S. (1975). A new approach to auxiliary have in English. Proceedings of the Northeast Linguistic Society. Smith, C. S. (1974). Paraphrase. Le langage et l’homme, 26, 22–29. Smith, C. S. (1973). Paraphrase and performance. Recherches Linguistiques, 1, 35–52. Smith, C. S. (1972). On causative verbs and derived nominals in English. Linguistic Inquiry, 3, 136–168. Smith, C. S. (1971). Sentences in discourse: an analysis of an essay by Bertrand Russell. Journal of linguistics, 7, 213–235. Smith, C. S. (1970). Report on for-to complements in English. Linguistics Research Center, University of Texas at Austin. Smith, C. S. (1970). An experimental approach to children’s linguistic competence. In J. Hayes (ed.), Cognition and the development of language (pp. 109–133). New York: John Wiley. [Reprinted in: In C. Ferguson & D. Slobin (Eds.) (1973). Studies of child language development (pp. 497–521). New York: Holt, Rinehart, & Winston.] Smith, C. S. (1969). Ambiguous sentences with And. In D. Reibel & S. Schane (eds.), Modern studies in English: Readings in transformational grammar (pp. 75–79). Englewood cliffs, NJ: Prentice Hall. Shipley, E., Smith, C. S., & Gleitman, L. (1969). A study in the acquisition of language: Free responses to commands. Language, 45, 322–343. [Reprinted in: L. Bloom (Ed.) (1978). Readings in the development of language (pp. 347–370). New York: John Wiley. Smith, C. S. (1968). Some ramifications of the notion of subject in English transformations. Discourse analysis project report, University of Pennyslvania. Smith, C. S. (1964). Determiners and relative clauses in a generative grammar of English. Language, 40, 37–52. Smith, C. S. (1961). A class of complex modifiers in English. Language, 37, 342–365.
A Speaker-Based Approach to Aspect Carlota S. Smith
1 Introduction The aspect of a sentence presents a situation in a certain light, contributing to the point of view conveyed by a sentence. The contrast between perfective and imperfective viewpoint is one component of sentential aspect, and another component is the type of situation talked about. It is well-known that the components interact in some way. One familiar example involves entailment: entailments differ for imperfective sentences, depending on the type of situation talked about.1 In spite of examples like this the relation between situation type and perspective is not very well understood. I will present an analysis which deals with the contribution of both components to sentential aspect. An essential factor of the analysis is the role of the speaker, who is responsible for the choices that indicate the point of view of a sentence. One straightforward sketch of a speaker talking about a situation in the world2 might look roughly like this: the speaker determines the situation type of an actual situation, according to basic categories such as those distinguished by Aristotle and others as Activity, Achievement, Accomplishment and State. The speaker constructs a sentence with the linguistic forms appropriate to the type of situation and chooses the aspectual perspective – simple or progressive in English – in which to present the situation. There are two points to notice about this rough sketch. The situation type of the sentence is determined by the actual situation, and the speaker’s choice is limited to the simple or progressive perspective. The sketch has the virtue of simplicity, and something like it is implied by many discussions of aspectual perspectives.
1 So for example, from John was swimming, one can conclude John swam, whereas from John was building a wall one cannot conclude John built a wall. Swimming is an Activity, building a wall an Accomplishment. These matters are discussed in Vendler (1967), Kenny (1962), Dowty (1977), and Vlach (1981) (among others). I assume familiarity with recent work on situation type, where “situation” is as in note 2. It shows (among other things) that sentences that talk about different types of situation have consistent syntactic and semantic properties. 2 I use “situation” as neutral between event, state, etc.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_1,
7
8
C.S. Smith
However this simple sketch fails to explain some interesting facts about aspect, notably what I will call non-standard aspectual choices. Certain ways of talking about a situation are quite standard for a given language, so much so that they may not be seen as choices. But there are other ways of talking that deviate from the standard. The existence of somewhat deviant sentences brings out the element of choice involved in the construction of all sentences, standard and non-standard.3 The sentences of (1) illustrate one type of non-standard aspectual choice in English. (1)(a) John was really liking the play. (b) Amy is resembling her great-uncle. (c) That cake is looking done. These sentences talk about states, or stative situations. Now standardly in English statives are incompatible with progressive aspect:4 ∗ I am loving you, ∗ He is knowing the answer. (1a–1c) run counter to the standard, since all of them have progressive aspect. Sentences like this are relatively frequent, especially in the spoken language. They suggest the activity and temporary quality that is characteristic of an event rather than a state. Intuitively we can say that they present states as if they were events. Sentences like this present a serious problem for the account sketched above: it is not clear how to provide for a speaker’s treating a state as if it were an event. Does the speaker categorize the actual situation in contradictory ways, as a state and an event? Are the properties characteristic of different situation types given a different status, attributed to the situation differently? A second set of examples shows that the phenomenon of nonstandard choice involving aspect has quite a wide range. In the sentences of (2), situations usually categorized as activities (a type of event) are presented as if they were states. (2)(a) The wheel is in motion. (b) The birds were in flight. (c) A meeting is in progress. These sentences have the main verb and complement characteristic of certain types of statives.5 They too involve a non-standard aspectual choice, in which a situation is presented in a somewhat unusual manner. It seems clear that an account of aspectual choice should explain how it is that speakers can talk about a situation from more than one viewpoint. The picture sketched above must be augmented to allow for the possibility of different perspectives on the same situation. Also needed is a way of allowing for situation
3 Dowty
(1979) has a useful discussion of the main types of states. Comrie (1976), Lakoff (1966), Leech (1970), for discussions of stativity. 5 By “standard” I mean a normal and unmarked way of talking about situations in the world; no reference to a dialect type, such as standard or non-standard, is intended. 4 See
A Speaker-Based Approach to Aspect
9
types themselves to contribute to the aspectual viewpoint of a sentence. I suggest that it is the speaker who relates actual situations to situation types. Speakers do this by using linguistic forms – that are associated with idealized situation types – to talk about actual situations. This new picture differs from the first in that the speaker, rather than the actual situation, determines the situation type used in a given sentence. I suggest that when speakers talk about actual situations they invoke abstract representations, or idealized situation types. The idealized situation types are abstractions that represent the properties characteristic of different situations. They do not depend on particular languages, but rather correspond to the basic categorization of situations that humans make on the basis of their perceptual and cognitive faculties. This is surely the sort of categorization with which Aristotle was concerned. Languages differ as to the situation types that they distinguish grammatically; the properties that play a role in the aspectual system of languages differ too. When speakers talk about an actual situation, they present it as an exemplar of an idealized situation type by using the linguistic forms associated with that type. There are standard and nonstandard presentations, and speakers make a choice whenever they talk about an actual situation. In English it is standard to present situations such as knowing, loving, resembling as states (cf. the sentences of (1)); and to present situations involving flying and moving as events (cf. the sentences of (2)). These matters are discussed more fully in my recent papers on aspect and choice.6 The diagram in (3) shows the components involved when a speaker talks about a situation.
(3)
Actual Situation
x
Linguistic Forms
Idealized Situation Types
a, etc. b, etc.
A B
Speaker
Grammar
It is the relation between these components that is in question. In the speaker-based proposal of this paper, the grammar of a language relates linguistic forms and idealized situation types; the speaker of a language relates an actual situation to, for example, an idealized situation type by using the linguistic forms associated with that situation type. The list of idealized situation types that speakers and languages recognize must include habituals and generics, and perhaps others not explicitly part of current classifications.7 Only habituals will be discussed here. Habituals often appear in discussions of aspect (see Comrie (1976) for examples); this is not surprising, since 6 Smith 7 See
(1982, 1983). Carlson (1983) for discussion of generics and habituals.
10
C.S. Smith
treating a situation as habitual involves the choice of a certain perspective. Many languages have special morphemes that are associated with the habitual; in our terms, with the habitual idealized situation type.8 In languages that do not distinguish habituals and inceptives with special morphemes, such as English, there are syntactic and semantic characteristics that do distinguish them (see Smith (1982), Freed (1979) for discussion). The notion of idealized situation type needs to be worked out more fully, of course. Still, even at this preliminary stage a case can be made for abstract representations of this kind. I have shown here that idealized situations types allow for a fairly satisfactory speaker-based analysis of aspect. The approach successfully distinguishes actual situations from linguistic representations, generally and in specific sentences. Furthermore it leads to a simple account of both standard and nonstandard aspectual choices.
1.1 Situation Type and Viewpoint Aspect Sentential aspect presents a situation from a certain point of view, through situation type and perspective. Both contribute to the aspect of the sentence. To clarify let us distinguish situation aspect and viewpoint aspect. Situation aspect involves the linguistic forms and meanings associated, for a given language, with idealized situation types. In English the linguistic forms are verb classes, particles, and types of complements; the term “Aktionsarten” is sometimes used to refer to these classes. Viewpoint aspect involves the forms and meanings associated with the perfective and imperfective perspectives. The focus of these perspectives differs somewhat from language to language, as will become clear in the following discussion. The viewpoint aspects of English are simple and progressive, indicated by the unmarked verb form and progressive auxiliary. Simple aspect presents a situation from a perfective viewpoint, discussion in Section 1.2 below. Progressive aspect presents an imperfective view. Traditionally this viewpoint is said to present a situation from the inside, without regard to its beginning or ending. Somewhat more precisely, the progressive indicates a time (moment or interval) that includes neither the initial nor final endpoint of a situation.9 Relations between situation and viewpoint aspect must be investigated language by language. In English, viewpoint aspect is dependent on situation aspect: the progressive viewpoint is limited to certain types of situation aspect, and the simple viewpoint varies in interpretation according to type of situation (see the following section). Consider now the well-known restriction of the progressive to non-stative sentences. I will show that the restriction follows from semantic analysis of the stative situation type and of the progressive. I summarize very briefly an analysis
8 Vendler
(1967) points out that habituals have much in common with states.
9 This formulation accords with traditional accounts such as those summarized in Lyons (1968) and
recent model-theoretic analyses such as those of Dowty (1979).
A Speaker-Based Approach to Aspect
11
developed in the references of note 4. The analysis turns on the notion of the endpoints of a situation, as included in idealized situation types and viewpoints. An event in its entirety includes the changes of state that initiate and conclude it, that is, its endpoints.10 The progressive perspective focuses on a time that excludes the endpoints of a situation. Now, this focus indirectly involves endpoints: it only makes sense to talk about excluding endpoints for a situation that is expected to have them. Thus progressive viewpoint implies a situation for which endpoints are expected, at least in principle. The idea that a situation has endpoints in principle is realized by representing them in the idealized situation type for that situation. The idealized situation types of events have both initial and final endpoints among their properties. This reflects the intuition that the notion of an event includes its beginning and ending. The point is uncontroversial for Achievements and Accomplishments, which involve completion; I argue in the aforementioned analysis (Smith, 1983) that Activities, which involve termination, also have endpoints in principle. For events, then, the properties of the situation types match the requirements of the progressive viewpoint. States are unlike events in that the notion of a state does not include its endpoints. The endpoint of a state involves a change into or out of that state, and is not part of the state itself. The idealized situation types of states represent the first and last moments of a state; endpoints, which are changes of state, are not included. Since the endpoints of a state are not part of its idealized situation type, and the progressive require their presence, the progressive is not available to sentences presenting stative situations.11 There is thus no direct relation (pace Vlach, 1981) between statives and progressives. The stative is a situation type, the progressive is a viewpoint aspect that is not available to sentences of the stative type.
1.2 Simple Aspect in English I will now show that the distinction between situation aspect and viewpoint aspect allows an elegant solution to a problem concerning simple aspect in English. As noted above, simple aspect is generally said to present a perfective viewpoint, where the perfective indicates a situation as a whole (that is, complete with initial and final endpoints). The problem is that several classes of sentences with simple aspect have a different interpretation. The data, which shows different interpretations of simple aspect, are quite familiar; but I know of no attempt to explain them. Consider the aspectual interpretation of the following sentences:
10 An event starts at its initial endpoint and stops at its final endpoint, in this view. See Bennett and
Partee (1972) for a definition of events that includes endpoints. (1977) comes to the same conclusion, by a rather different route.
11 Taylor
12
(4) (5) (6) (7) (8)
C.S. Smith
James fixed the clock. Mr Ramsey reached the lighthouse. Lily strolled along the beach. The visitors lived in London. Prue always swam in the pond.
(4–6) have the expected interpretation: events are presented as having both initial and final endpoints, termination or completion depending on the types of event. But notice that (7) and (8), which do not talk about events, have a different reading. The state talked about in (7), and the habitual of (8), can be taken as continuing into the present. This statement can be tested by forming conjunctions such as those of (9). In (9) the sentences (7) and (8) are conjoined with sentences asserting that the situation may continue into the present, and the conjunctions are felicitous. (9) John lived in London and he may still live there. Prue always swam in the pond and I’m told that she still does. It is perhaps natural to understand (7) and (8) in isolation as talking about situations that no longer obtain; although I am not sure that this is a useful criterion, since a sentence in isolation is not entirely natural. But in any case the conjunctions above show that (7) and (8) can be interpreted in more than one way, according to context. This means that the interpretation of completion is not linguistically indicated. In Gricean terms, the interpretation of completion arises from the conversational context: the possibility of its cancellation shows that it is due to conversational implicature. Consider now the conjunctions of (10), which involve (4) and (6) above and are distinctly odd: (10)(a) James fixed the clock and he may still be fixing it. (b) Lily strolled along the beach and she may still be strolling. The conjuncts of (10) are incompatible: the first conjuncts present events as definitely completed, and the second conjuncts assert that they are not completed. In these sentences the implicature of completion is conventional, that is, indicated linguistically, and it cannot be cancelled. To return to simple aspect, the problem exposed by these examples is that certain sentences do not have the perfective interpretation although they have the simple verb form usually associated with that interpretation. The problem sentences are stative and habitual; sentences that talk about accomplishments (4), achievements (5), and activities (6) have the expected perfective interpretation.12
12 There are some examples of sentences involving activities that seem to contradict this claim, e.g.
Mary looked out the window; half an hour later she was still looking. However, I suggest that these are inceptive in interpretation and therefore do not contradict this analysis.
A Speaker-Based Approach to Aspect
13
The different interpretations can be accounted for by recognizing a central meaning of simple aspect that is realized differently for each situation type. As the central meaning I propose the following: simple aspect presents a situation in its entirety. This meaning is interpreted according to the endpoint properties of the situation in question. Since different situation types have different endpoint properties, the central meaning is interpreted differently for each situation type. In other words, interpretation of simple viewpoint aspect varies with situation aspect. If the situation type includes endpoints, they are included in the “simple” viewpoint of that situation type; e.g. achievements are presented as completed or terminated. If the situation type does not include endpoints, the situation in its entirety does not include them either; e.g. a state or habitual state may not still obtain. The proposed central meaning is close to the traditional notion of perfectivity, but not identical to it. The difference lies in the way states and other related situation types are treated. The idealized situation types of states, as noted above, do not include endpoints; and the viewpoint of simple aspect presents states without endpoints. Intuitively, whereas an event in its entirety is seen as having endpoints, a state as a whole is not seen as having them. This seems exactly right, and conforms to the way statives with simple aspect are interpreted. To see that this analysis correctly predicts the actual interpretations discussed above, consider (4–8) above. The first three talk about events, which have endpoints in their situation types. Achievements (sentence (5)) and Accomplishments (sentence (4)) involve completion, and therefore endpoints; Activities (sentence (6)) are represented as terminated rather than completed, I have argued, and so their idealized situation types also have endpoints. In the proposed analysis simple aspect presents a situation with the endpoint properties of its idealized situation type. Therefore simple aspect should, for these sentences, include endpoints; and it does, as the conjunctions of (10) show. Sentence (7) is a stative, and States do not have endpoints in their idealized situation type. Therefore simple aspect should not include endpoints; and it does not, as the conjunctions of (9) show. Although the idealized situation type of Habituals has not been developed here, it is quite well known that Habituals are very much like States (see Vendler (1967) on this point, for instance). The examples of (9) suggest that the notion of an Habitual does not include its endpoints, and I shall assume that this is the case. Then the interpretation of (8) should be like that of (7) in the relevant respect, and it is. Distinguishing situation aspect from viewpoint allows an analysis which varies one of these while holding the other constant. As the discussion shows, this appears to capture precisely the nature of simple aspect in English. Formally the analysis could be implemented by allowing access, for interpretation of simple aspect, to representations of situation types. An alpha rule would be appropriate here. If situation type x has y endpoint properties, simple aspect linguistically presents the situation with y endpoint properties.13
13 “Linguistically
indicate” or “present” is used in this paper for Grice’s conventional implicature.
14
C.S. Smith
1.3 Summary of Part 1 I have sketched very quickly an analysis of sentence aspect that distinguishes situation aspect and viewpoint aspect. In English viewpoint aspect is dependent on situation aspect. Progressive aspect, which corresponds roughly to the idea of the imperfective, is limited to certain situation aspects. The viewpoint of simple aspect presents a situation in its entirety, a central meaning that is close to the traditional notion of perfectivity; it varies in realization according to the endpoint properties of the situation. A speaker-based account of aspect is proposed, in which the features of situations are represented by idealized situation types. There are at least six situation types, the four discussed by Aristotle-Ryle-Vendler, habituals, and generics. Idealized situation types are associated with particular linguistic forms in the grammar of individual languages. Certain features may play an important role in a language, as endpoint properties do in English. In talking about an actual situation a speaker presents it as an exemplar of a given situation type. This occurs when the speaker uses the linguistic forms associated with that situation type. Such an account allows for both standard and non-standard aspectual choice, and brings out the important point that both involve the speaker’s choice of the point of view from which a situation is presented linguistically.
2 Extending the Analysis The approach developed above will first be applied to French, a language whose aspectual system is rather different from English; I then consider briefly some other languages. The discussion of French focuses on the imparfait and the passé composé. Both are past tenses and they are quite similar to the English progressive and simple past, respectively.14 The differences as well as the similarities emerge clearly when we analyze French along the same lines as those used for English. I assume the kind of analysis sketched above in which idealized situation types are related to linguistic forms. The situation types are the same as those considered for English; recall that idealized situation types do not depend on particular languages. I also assume without discussion that linguistic forms are associated with particular situation types in French.15 This enquiry will be directed toward making precise the central meanings of the viewpoint aspect imparfait and passé composé. The relation of viewpoint aspect to situation aspect will be considered, and the interpretation of non-standard aspectual choices.
14 See
Rohrer (1977). passé simple, another past tense of French, is not discussed here. The passé simple is quite limited in present-day French, so much so that few native speakers have clear intuitions about its use (Maurice Gross, personal communication).
15 The
A Speaker-Based Approach to Aspect
15
2.1 Imparfait One widely-respected French grammar, Grevisse (1953), gives the following general description of the imparfait: En géneral, l’imparfait indique, sous I’aspect de la continuité, un fait qui était encore inachevé au moment du passé auquel se report le sujet parlant; en particulier il indique un fait passé en train de se dérouler au moment ou se produit un autre fait. . ..
This is entirely compatible with the traditional notion of the imperfective, and with the English progressive as well. We now ask how this general notion is realized in French; the aim is to find a precise central meaning that accords with standard and non-standard uses of the imparfait. Some examples of standard sentences with the imparfait perspective are given in (11).16 (11)(a) L’enfant pleurait. (b) Quand l’oncle Jean a frappé à sa porte à minuit, elle lisait toujours; sa soeur dormait. (c) Il entrait dans un magasin. (d) Ils bâttissaient une cabine. (e) La mer était calme. (f) Le ciel était couvert; il allait pleuvoir. (g) Sa mère lui racontait des contes de fées tous les soirs. From these sentences it is evident that the imparfait is available for the full range of situation types, not excluding stative. The situation types represented in (11) are as follows: Activity – (a), (b); Achievement – (c); Accomplishment – (d); State – (e), (f); Habitual – (g). Particularly striking in the context of the English analysis are the last three examples, which are different from their English counterparts. The progressive of course is not available for statives, while the imparfait is available; and English habituals standardly have simple aspect.17 The perspective of the imparfait apparently presents all types of situations as continuing. The interpretation with conjunctions like those used above for English 16 The
examples are taken from several standard texts. does find English habituals with the progressive but they have a slightly marked, nonstandard flavor much like English progressive statives:
17 One
John is reading the newspaper regularly. Mary is playing tennis every day. They are much improved by the addition of adverbials: Mary was feeding the cat that week. The direct translation of (11g) is quite bad: His mother was telling him stories of contes de fees every evening. The most appropriate English version of sentences like these has would rather than the progressive: His mother would tell him in the evenings about the . . .
16
C.S. Smith
is consistent with this analysis. Consider for example the conjunctions of (12–14); in each a sentence in the imparfait is conjoined with an assertion that the situation continues into the present.18 (12) L’éte passé ils bâtissaient une cabine; peut-être qu’ils 1a bâtissent encore. (13) Ce matin elle chantait; peut-être qu’elle chante encore. (14) Il croyait aux fantômes quand il était petit, et il y croit maintenant. Since these conjunctions are good, we are justified in concluding that the imparfait presents situations as continuing. We note further that the perspective is available for all situation types, and its interpretation does not vary with situation type. The imparfait lacks the association with endpoints characteristic of the progressive. The difference is not surprising since the imparfait is not restricted to situation types that have final endpoints. In view of this we predict that the imparfait lacks the connotations of activity and of the temporary that also characterize the progressive. These connotations arise through the interaction of the progressive with the idealized situation type of events, I have shown.19 This perspective suggests a progression to the eventual final endpoint, and thus has an active, temporary note; for example, (16) suggests a more temporary situation than does (15). (15) Martha lived in Paris. (16) Martha was living in Paris. Now consider a corresponding pair of sentences in French: (17) Marthe a vécu à Paris. (18) Marthe vivait à Paris. Neither (17) nor (18) has the connotation of activity and/or temporariness typical of the progressive. The difference between them is clear-cut and of an entirely different nature. (17) talks about a situation that is over, whereas (18) talks about a situation that may or may not still obtain. In order to get a clearer idea of the imparfait, we will consider non-standard uses of this perspective. Non-standard choices can be useful for pointing up central aspectual meaning, as demonstrated above for the progressive. A number of nonstandard examples will be presented; we look for a consistent understanding that differentiates them from standard presentations of the situations in question. The examples that follow are drawn from Ducrot (1979).20 Although my analysis is some what different from his, the two are not, I think, incompatible. In his paper 18 I
would like to thank David Birdsong, Jean Lowenstamm, Dina Sherzer, Jeanne Whitaker for judgments and discussion of the French examples. 19 See the references of note 6. 20 cf. Ducrot (1979). This is an admirable discussion of examples.
A Speaker-Based Approach to Aspect
17
Ducrot gives a number of sentences in the imparfait that, as he puts it, require a special context. The following examples are taken from his paper; they are in most cases given contexts which make them plausible: (19) C’est incroyable, le semaine dernière il pleuvait plusiers fois et, cette semaine, il n’y a pas eu une goutte d’eau. (20) Les iranienns s’en prennent maintenant aux interêts americains. Hier deux banques americaines étaient saccagées. (21) L’année dernière je démenageais. (22) Idiot que je suis! L’année dernière j’achetais un appareil de photo dont je n’avais nul besoin et, cette année, je n’ai même pas de quoi de me payer le cinéma. In all of the examples, situations that would standardly be taken as having final endpoints are presented as continuous throughout the period talked about – that is, without endpoints. In (19) for example the period of “last week” is taken as a whole, and the many rainstorms in some sense cover the period; in (20) moving is presented as taking up an entire year, which is plausible if we include planning, recovering, etc. In Ducrot’s formulation, the events characterize the period talked about regardless of where their actual endpoints fall. We now have three kinds of information about the imparfait: its general meaning, the fact that it is available for all situation types, and the examples of non-standard sentences. All focus on the continuing nature of the situation in question, suggesting that it should be considered the central meaning of the imparfait. A formal statement of this perspective would make reference to change of state rather than endpoints, so that it holds for all situation types. One possible formulation might be the following: The imparfait presents a continuing situation x at a time T, with the condition that there is no sequence in T of times, t1 , t2 , such that at t1 x is true and at t2 x is not true. We have now specified a central meaning for the imparfait that is similar but not identical to that of the progressive. This is exactly what is desired since the two perspectives are very close. The considerable overlap comes out clearly when we consider English progressive versions of the non-standard choices of the imparfait. Some are barely acceptable, other are much better. Direct translations of (19–22) follow: (23)(a) (cf. 19) It’s amazing, last week it was raining many times and this week there hasn’t been a drop of water. (b) (cf. 20) The Iranians are now going after American interests. Yesterday two banks were being despoiled. (c) (cf. 21) Last year I was moving. (d) (cf. 22) What an idiot I am! Last year I was buying a camera that I didn’t need and, this year, I don’t even have enough for the movies.
18
C.S. Smith
2.2 Passé Composé The passé composé is generally taken as indicating perfective aspect21 and its general meaning accords with the standard understanding of this viewpoint. Grevisse says that “Le passé composé indique un fait achevé a une epoque determinée ou indeterminée du passée.” As with the imparfait, we consider now the precise meaning of this viewpoint in French. Examples from standard sources show that the passé composé is appropriate for the full range of situation types. (24)(a) Elle a travaillé dix heures ce jour-là. (Activity) (b) Il est entré et s’est assis à son bureau. (Accomplishment) (c) La guerre a éclaté. (Achievement) (25)(a) Marie a été riche. (State) (b) Nous sommes sorties tous les jours. (Habitual) The aspectual interpretation of these sentences is consistent (unlike that of English): the situation of whatever type is presented in its entirety, that is, as completed or terminated. Conjunction confirms this interpretation. Sentences in the passé composé cannot be conjoined with assertions that the situations talked about continuing: the result is contradictory, as the following examples show. (26)(a) L’été passée ils ont construit une cabine; peut-être qu’ils la construisent encore. (b) Ce matin Marie a chanté; peut-être qu’elle chante encore. (c) Jean a été malade hier soir et il est malade maintenant. There is a class of sentences that may seem to controvert the claim that the passé composé presents situations as complete or terminated. These sentences talk about states that may be taken as continuing into the present. However, they are Achievements rather than Statives: they present a change of state, into the state in question. The situation that may continue is the resulting state, not the change into that state. (27)(a) (b) (c) (d)
21 The
Elle a été fachée quand Jean a cassé l’assiette.22 Marie a été heureuse à la vue de son fils. Tout d’un coup, j’ai compris! A ce moment il a su la vérité.
passé composé is discussed in Grevisse (1953), Kamp (1979), and many other places. Its perfective interpretation is not controversial. 22 A reflexive – Elle s’est fachée – would be a more natural way to say this in French; however this sentence does have the interpretation given, in this context.
A Speaker-Based Approach to Aspect
19
The adverbials of (c) and (d), which are characteristic of Achievements, indicate that the suggested analysis of these sentences is correct. Sentences like this may often be translated into English with get or become. Consideration of a wide range of sentences with the passé composé makes it clear that this viewpoint presents a situation in its entirety, that is, as having ended. The sentences of (28) represent other types of passé composé, some of them distinctly non-standard and some borderline cases. Generally it can be said that all these examples present situations as if they had ended, even though in some cases the situations may be a continuing one. (28)(a) (b) (c) (d)
Il l’a aimé pendant trois ans. Paul a véçu sans argent; peut-être qu’il vit maintenant sans argent. Le président a parlé à la television hier soir; et ce matin, il parle toujours. Il l’a aimée; peut-être qu’il l’aime encore.
I will discuss the interpretation of these examples in some detail. Both (a) and (b) present a situation as if it had ended, by talking about a distinct time period that has ended. Thus in (a), he may still love the person in question; the sentence focuses on a three-year period that has definitely ended. Similarly, (b) indicates that at a period now over, he loved her; he may love her now, but the present involves a distinct period of time and thus a distinct situation. The passé composé of (c) is taken as presenting a change of state into the Activity of talking; it is not ungrammatical to use the passé composé inceptively for activities, but it is somewhat unusual.23 Finally consider (d) which on the account given above should be contradictory. Some informants interpreted this sentence as focusing on a period of time that has ended, as in (a) and (b); one informant suggested that (d) would be appropriate for talking about a person who is unpredictable, even seriously unstable, so that in effect the period during which he loved someone would be discontinuous for that person. This explanation accords perfectly with the notion that the passé composé presents situations in their entirety. The passé composé takes a situation as having final endpoints, and when used in a non-standard way it presents a situation in this way although the final endpoint may not actually have occurred. This viewpoint corresponds closely to the traditional notion of the perfective. It may be stated formally for all situation types by specifying change of state rather than endpoints, perhaps roughly as follows: The 23 I
am indebted to David Birdsong for helpful discussion of this point. A literary example of an inceptive Activity sentence in the passé composé, cited by Birdsong, appears in the prose poem “L’Aube” of Rimbaud’s. J’ai embrassé l’aube d’été. Rien ne bougeait encore au front des palais. L’eau était morte. Les camps d’ombres ne quittaient pas la route du bois. J’ai marché, reveillant les hâleines vives et tiedes, et les pierreries regarderent, et les ailes se leverent sans bruit.
20
C.S. Smith
passé composé presents a situation x at time T such that for t1 , t2 included in T, not-x at t1 and x at t2 ; and for t3 , t4 , included in T and following t1 and t2 , x at t3 and not-x at t4 . It should perhaps be noted again that the past tenses are the only tenses of French that offer a grammaticized aspectual choice. The other tenses are consistently neutral in regard to aspect.
3 Comparisons of Aspectual Systems We are now in a position to compare the French and English aspectual systems. Both have viewpoint aspects that correspond roughly to the traditional notions of perfective and imperfective. The viewpoints of the two languages differ slightly in central meaning and in realization, as demonstrated above. To summarize: The imperfective viewpoint in English, the progressive, takes a situation at a time that includes neither its initial nor final endpoint; it is limited to event situations, whose idealized situation types include endpoints. The imparfait, the imperfective viewpoint of French, takes a situation at a time during which no change of state occurs; it is available for all situation types. The perfective viewpoint in English, simple aspect, presents a situation in its entirety; the viewpoint varies in interpretation depending on the endpoint properties of situations. The French perfective viewpoint considered here, the passé composé, presents situations as complete, including the change of state attendant on beginning and ending. The interpretation is consistent for all situation types. Considering familiar closely-related languages, English has an unusually rich system of viewpoint aspect. The French system is typical of Romance languages: they mark aspect obligatorily only in the past, and situation and viewpoint aspect do not interact.24 The Germanic languages do not typically code aspect. One might think, therefore, that the English system is aberrant, the aberration due perhaps to the disappearance of the distinction between verbal and nominal forms with -ing.25 However there is evidence to the contrary. Anglo-Saxon, the language from which English developed, had a grammatically marked distinction between stative and non-stative situation aspect; this point is made in Aristar and Dry (1982). If they are correct, then the general role of situation aspect in English cannot be attributed to the loss of the nominal (although the present system might have evolved in the stages outlined by Visser (1973), of course). The dependencies noted above are certainly not unique to English. In Russian, for example, situation aspect is related to viewpoint aspect: the perfective viewpoint is available only for certain situation types.
24 Spanish
is somewhat of an exception: in addition to an imperfective past there is a progressive in Spanish that is restricted as to situation type, much as the English progressive is restricted. 25 Visser (1973) discusses and demonstrates the development of the progressive.
A Speaker-Based Approach to Aspect
21
Ranging somewhat farther afield, we find that situation aspect plays a grammatical role in languages of various families. The following somewhat tentative remarks result from work with grammars of the languages in question and/or discussion with people who are intensely familiar with these languages. They do not constitute analysis, but rather should be taken as a demonstration that the distinctions noted in the first part of this paper recur in languages of different families. The languages mentioned below were chosen essentially at random and do not constitute a study of the areas or language families involved. It should be noted that some difficulties arise in dealing with information in grammars. Grammars frequently refer to stative, etc., verbs, while the examples and discussion may indicate that the situation aspect of a sentence is involved. However one cannot assume that the appearance of certain aspectual morphemes indicates something about sentential aspect; the problem is especially important for languages in which multiple suffixes contribute in a complex way to word meaning. The African languages Afar (Cushitic) and ChiBemba (Bantu) restrict the viewpoint aspects available to certain situation aspects.26 Viewpoint aspects of ChiBemba include adverbial focus aspect, and continuous aspect; neither of these is available for statives. Habituals on the other hand can have adverbial focus. Arabic has a more restricted paradigm for statives than for non-statives, as does Coptic. Navajo, an Athabaskan language, also distinguishes situation types grammatically, and the stative type appears to be severely restricted.27 Navajo has a class of verb bases, called neuter, that correspond quite closely to statives. Neuter bases are conjugated in only one paradigm (and therefore few viewpoints), whereas the possibilities are generally quite varied. Among the viewpoints generally available are perfective, progressive, imperfective, iterative, and habitual. This brings another dimension along which languages differ: certain aspectual properties of situations are indicated differently in different languages, that is as either situation or viewpoint aspect.28 In Hopi, a Uto-Aztecan language, statives have a more limited set of aspectual possibilities than do non-statives (Kalectaca, 1978). Seneca, an Iroquian language, limits the aspect of statives to one type, descriptive aspect (Chafe, 1967). According to Boas (1909) this is typical of Iroquian languages; Haas (1942) suggests something similar for Tunica. Jacaltec and Quiché, Mayan languages, distinguish syntactically between statives and non-statives. In Jacaltec stative clauses, unlike others, have no aspect marking; they also have a distinctive form for negation (Craig, 1977). Quiché has a separate and highly restricted set of inflectional endings for statives. A language of a different family, Lalana
26 I
am indebted to Anthony Aristar for information about Cushitic; for ChiBemba, see Givon (1972). 27 The information about Navajo comes from Young and Morgan (1980), Reichard (1951), Sapir and Hoijer (1967). 28 For a discussion of different linguistic indications of evidentiality, an aspectual or aspect-related notion, see Woodbury (1981).
22
C.S. Smith
Chinancotecan, has at least two grammatical constructions that distinguish statives from other situation types.29 It cannot be assumed that stative and other aspectual terms have the same interpretation in all of these languages. In fact they may focus in different languages on different properties of situation types, and interact with some others, as Merlin (1980) argues convincingly.30 Merlin shows in particular that the notions of stative and non-stative interact with the property of agentivity in many languages. Such variation is not surprising, given the general approach outlined in this paper. I have suggested that languages can be expected to vary in the properties of idealized situation types that are focused by viewpoint aspect. In fact the detailed study of French and English explicated just such differences in the central meanings of the perfective and imperfective viewpoint aspects of those languages. I would hope that future research can further illuminate this kind of similarity and difference among aspectual systems, which can lead to a better understanding of aspect, point of view, and the notion of idealized situation type. There are certain advantages and disadvantages in beginning an inquiry of this nature with English and related languages. Our knowledge of English allows the study of some fairly subtle phenomena. On the other hand, the relation between aspect and point of view is clearer in languages such as Hopi and Navajo, and some African languages. Typically in these languages morphemes for imperfective and perfective pattern distributionally with the evidential, the dubitative, and other indications of point of view. Acknowledgements I would like to thank David Birdsong, Dina Sherzer, Jeanne Whitaker, for helpful discussions of the material in this paper.
References Aristar, Anthony and Helen Dry: 1982, ‘The Origins of Backgrounding Tenses in English’, Proceedings of the Chicago Linguistics Society, Chicago, Illinois. Bennett, Michael and Barbara Partee: 1972, Toward the Logic of Tense and Aspect in English, Systems Development Corporation, Santa Monica, California. Boas, Franz: 1909, Notes on the Iroquois Language, Putnam Anniversary Volume. Brunot, Ferdinand: 1922, La pensée et la langue, Masson, Paris. Carlson, Gregory: 1983, ‘Logical Form: Types of Evidence’, Linguistics and Philosophy 6, 295–318. Chafe, Wallace: 1967, ‘Seneca Morphology and Dictionary’, Smithsonian Contributions to Anthropology Vol. IV, Smithsonian Press, Washington, D.C. Comrie, Bernard: 1976, Aspect, Cambridge University Press, Cambridge, England. Craig, Colette: 1977, The Structure of Jacaltec, University of Texas Press, Austin, Texas. DeLancey, Scott: 1981, ‘An Interpretation of Split Ergativity and Related Patterns’, Language 57, 626–657. 29 I
am indebted to Frank Trechsel and Bob Mugele for information about Quiché and Lalana Chinancotecan, respectively. 30 DeLancy (1981), in a discussion of patterns related to split ergativity, considers aspectual interpretation in various languages and some connections between aspect and point of view.
A Speaker-Based Approach to Aspect
23
Dowty, David: 1977, ‘Towards a Semantic Analysis of Verb Aspect and the English Imperfective Progressive’, Linguistics and Philosophy 1, 45–77. Dowty, David: 1979, Word Meaning and Montague Grammar, Reidel, Dordrecht. Ducrot, Oswald: 1979, ‘L’lmparfait en français’, Linguistische Berichte 60, 1–23. Freed, Alice: 1979, The Semantics of English Aspectual Complementation, Reidel, Dordrecht. Givon, Talmy: 1972, ‘Studies in ChiBemba and Bantu Grammar’, in Studies in African Linguistics, Supplement 3, University of California Publications, Berkeley, California. Grevisse, Maurice: 1953, Le bon usage, Duculot, Paris, p. 563. Haas, Mary: 1942, ‘Tunica’, Handbook of American Indian Languages, Vol. IV, J. J. Augustin, New York. Kalectaca, Milo: 1978, Lessons in Hopi, University of Arizona Press, Tucson, Arizona. Kamp, Hans: 1979, ‘Events, Instants and Temporal Reference’, in R. Bäuerle, U. Egli, and A. von Stechow (eds.), Semantics from Different Points of View, Springer, Berlin. Kenny, Anthony: 1962, Actions, Emotions and Will, Humanities Press, New York. Lakoff, George: 1966, Stative Verbs and Adjectives in English, Harvard Computation Laboratory Report NSF-17, Cambridge, Mass. Leech, Geoffrey: 1970, Towards a Semantic Description of English, Indiana University Press, Bloomington, Indiana. Lyons, John: 1968, Theoretical Linguistics, Cambridge University Press, Cambridge, England. Merlin, Francesca: 1980, ‘Split Intransitivity: Functional Oppositions in Intransitive Inflection’, unpublished paper, University of California at Berkeley. Malotki, Ekkehart: 1979, Hopi-Raum, Gunter Narr, Tübingen. Reichard, Gladys: 1951, Navajo Grammar, J.J. Augustin, New York. Rohrer, Christian (ed.): 1977, On the Logical Analysis of Tense and Aspect, Gunter Narr, Tübingen. Ryle, Gilbert: 1949, The Concept of Mind, Hutchinson’s University Library, London. Sapir, Edward and Harry Hoijer: 1967, The Phonology and Morphology of the Navaho Language University of California Publications in Linguistics 50. Smith, Carlota: 1978, ‘The Syntax and Interpretations of Temporal Expressions in English’, Linguistics and Philosophy 2, 43–99. Smith, Carlota: 1982, ‘The Temporal Reference of the English Futurate’, Cognition & Communication 16, 81–96. Smith, Carlota: 1983, ‘A Theory of Aspectual Choice’, Language 59, 497–501. Taylor, Barry: 1977, ‘Tense and Continuity’, Linguistics and Philosophy 1, 199–220. Vendler, Zeno: 1967, Linguistics and Philosophy, Cornell University Press, Ithaca, New York. Visser, Frederick: 1973, An Historical Syntax of the English Language, E. J. Brill, Leiden. Vlach, Frank: 1981, ‘The Semantics of the Progressive’, in P. Tedeschi and A. Zaenen (eds.), Syntax and Semantics: Tense and Aspect, Academic Press, New York. Woodbury, Anthony: 1981, ‘Evidentiality in Sherpa Verbal Categories’, unpublished paper, University of Texas. Young, Robert and William Morgan: 1980, The Navajo Language: A Grammar and Colloquial Dictionary, University of New Mexico Press, Albuquerque.
Aspectual Categories in Navajo Carlota S. Smith
1 Introduction Aspect is the semantic domain of temporal point of view in language. The linguistic expressions that a speaker chooses present situations in a particular temporal light. For example, in talking about a bird flying, I may say in English The bird flew, The bird was flying, or The bird was in flight. The first two sentences present a dynamic event from different viewpoints; the third presents a static situation. These are aspectual distinctions. The domain of aspect includes such categories as perfective and imperfective viewpoint, and situation types such as event and state.1 Recent work has shown that covert linguistic categories are associated with events and states in the grammars of many languages; I call them “situation types,” using “situation” as a general term for events and states. In this article I ask whether such categories hold for the Navajo language and, if so, how they are expressed. The question is particularly interesting because Navajo has a set of morphological categories known as aspectual in the literature. I ask how these morphological categories relate to situation types and seek to determine their other functions in the aspectual system of the Navajo language. In Section 2 I introduce the notion of situation type and some relevant facts about the structure of Navajo; Section 3 discusses correlates of temporal properties in Navajo; Section 4 considers the situation types of Navajo; Section 5 discusses how the conventional aspectual categories of Navajo function in the aspectual system.
1 The discussion assumes the framework of my two-component theory of aspect. This view recognizes two independent components in the aspectual system of a language: viewpoint (e.g., perfective, imperfective, etc.) and situation type (e.g., state, durative event, etc.). Situation types categorize a situation according to its temporal features; viewpoint presents all or part of the situation. Sentences give information of both components. The theory is developed and applied to five languages, including Navajo, in Smith (1991). This article is based on the analysis presented there, but it is substantially different in many respects.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_2,
25
26
C.S. Smith
2 Background 2.1 Situation Types 2.1.1 Temporal Classification of Situations Situations have long been classified according to their internal temporal features. Aristotle distinguished between static and dynamic situations; later scholars have added the features of duration and telicity.2 Telicity distinguishes events which have a natural goal or endpoint, with a concomitant change of state, from events with arbitrary endpoints. These classifications are based in human perceptual and cognitive abilities. The interesting point for linguistic analysis is that sentences which present situations of different types have reflexes in the grammar of a language. In his classic article (1957), Vendler showed that a cluster of syntactic properties characterizes sentences that present states and events of each type. Situation types are very general semantic categories representing classes of idealized situations, organized according to their semantic temporal features. These features have grammatical correlates: the sentences that realize each situation type have a set of distributional properties, e.g., co-occurrence with certain adverbials, verbs, aspectual viewpoints, and other forms. Thus the situation types are covert linguistic categories, in the sense of Whorf (1956). Situation types must be established separately for each language. The temporal features ±Dynamic (Dynamic/Static), ±Durative (Durative/Instantaneous), and ±Telic (Telic/Atelic) play a role in many languages. The features can be expressed compactly with plus and minus values, but for perspicuity I usually refer to the two positive values, as above. (1) presents a list of situation types according to these features, with some examples of each. These situation types hold for English, and for a number of other languages.3 (1) Situation Types States: static (own the farm, know the answer, love Mary) Events: dynamic Activities: durative, atelic (laugh, push a cart, walk by the river) Accomplishments: durative, telic (build a house, walk to work, learn Greek) Semelfactives: instantaneous, atelic (tap, cough, flap a wing) Achievements: instantaneous, telic (burst a balloon, reach the top)
2 The internal temporal structure of situations is discussed in Ryle (1949), Vendler (1957), and Kenney (1963). Recent classifications include Dowty (1979) and Smith (1991). 3 The classification recognizes two crosscutting distinctions, ±Telic and ±Durative. This differs slightly from Vendler and Dowty, both of whom make the distinction explicitly for durative events, but they posit two subtypes of Achievements which correspond to Achievement and Semelfactive here.
Aspectual Categories in Navajo
27
The examples present clear cases of each situation type. The temporal features and their grammatical correlatives are discussed more fully in Section 3 below. In talking about a situation in the world, a speaker invokes a given situation type by using the linguistic forms that are associated with it. The verb and its arguments, or verb constellation, convey the concept of a situation, just as a noun, apple for instance, conveys the concept {apple}. Verb constellations are associated with the situation types of a language. Since certain temporal semantic features are intrinsic to a given situation concept, they are expressed by the verb constellation. In the discussion, I use brackets for verb constellation and braces for concepts: so the verb constellation [Mary walk by the river] expresses the concept {Mary walk by the river}, which underlies the sentence Mary walked by the river. For simplicity, I sometimes refer to the situation types associated with it, e.g., Mary walked by the river is an Activity sentence of English, with the properties of dynamism, duration, atelicity. For a particular language, a situation type is grammaticized if the verb constellations that express it have a consistent and unique set of distributional properties. To investigate the situation types of Navajo, I look at sentences that express intuitively clear cases. For instance, {burst a balloon} is a clear case of an instantaneous event, one that has no duration in principle. Therefore, I consider the verb base that expresses {burst a balloon} in asking whether the feature ±Durative has grammatical correlates in Navajo. It is necessary to be very cautious with the Navajo examples, since translations are often less than satisfactory. However, the requirement that sentences have the consistent distributional properties of their situation types should keep us from straying too far afield. 2.1.2 The Range of Situation Types and Some Lexical Distinctions People can talk about situations from more than one point of view, and this flexibility is essential to the aspectual component of language. Speakers may present a situation as a whole, with a broad view. Or they may take a narrow view, talking about the endpoints or the middle of a situation. One situation may be seen as a subpart of another, as belonging to a pattern of situations (habituals), as a member of a class of situations (generics); there are also nonstandard aspectual choices; the general topic is discussed in Smith (1993/1995). I have suggested that the broad view of a situation represents a basic-level categorization for English and some other languages (Smith 1991). This may not be the case for Navajo; see the discussion in Section 4 below. Languages convey broad and narrow views of a situation in various ways. In English, the broad view is usually given in a simple sentence, e.g., Mary built a sandcastle, and the narrower views with verbs or phrases that have the simple sentence as a complement, for instance, Mary began building a sandcastle; Mary is in the process of building a sandcastle; Mary finished building a sandcastle. Russian has verbal prefixes which narrow the view of a situation, as in govorit’ “speak” and zagovorit’ “begin to speak” there are also lexical verbs, e.g., nachinat’ “begin”, konchat’ ‘end’, among others.
28
C.S. Smith
Sentences with narrow views may differ in situation type from those with a broad view. The initial and endpoints of a durative situation, for instance, are instantaneous events in themselves, changes into and out of a durative situation (Freed 1979 and Smith 1991). Thus, sentences presenting endpoints are associated with a situation type with the property +Instantaneous and have the distributional properties characteristic of such sentences. There is no change in situation type, however, for sentences that present an internal stage of a durative situation. Morphemes such as begin and stop give a narrow view of a situation: this is their function. In contrast, other lexical morphemes contribute to determining the type of situation presented. It is useful to have terms for this distinction. I refer to morphemes that give a narrow view of a situation as “superlexical” morphemes, and other (relevant) morphemes as “lexical” morphemes. Thus the English inceptive verbs begin, start, the continuatives in the process of, continue, and the terminatives stop, finish are all superlexical morphemes. Narrow-view morphemes, then, make a different kind of semantic contribution than do other lexical morphemes. The distinction has been noted by many scholars, in many different terms. Discussing the Russian system, for instance, Forsyth calls certain morphemes “procedural” because they “leave unaltered the basic meaning of the original verb.” In contrast, other morphemes “modify the meaning of a verb to produce a lexical derivative. . .a new verb denoting a type of action different from that denoted by the original verb” (1970:19). Thus, in the Russian examples above, the prefix za- of zagovorit “begin to speak” is a procedural morpheme, while the prefix of ugovorit’ “convince” is lexical because it contributes to a verb denoting a different sort of situation. This is precisely the distinction between “lexical” and “superlexical” morphemes. The term “aspectual” has also been used for narrowview morphemes (Freed 1979). The distinction between lexical and superlexical will be useful later, as we consider the information about situations that is conveyed by verb constellations. A third type of morpheme should also be recognized: there are morphemes which are required formally in a given context, but which make no identifiable semantic contribution. I call such morphemes “formal” to distinguish them from those which contribute to the semantics of a verb form or verb constellation.
2.2 Navajo Preliminaries 2.2.1 The Navajo Verb The Navajo verb composite contains most of the grammatical information in a sentence. The verb has an intricate structure, built up in layers with an abstract verb root as the basis of the whole. The verb composite consists of a verbal unit with a series of prefixes and other forms; the prefixes have fixed positions, both hierarchical and sequential. Following Young and Morgan (1987), three levels of the verb composite can be distinguished. The first is the verb theme, which contains the verb root, classifier, and certain tightly bound thematic elements. At the next level is the verb base,
Aspectual Categories in Navajo
29
where the verb root is realized concretely as a set of stems. The base consists of a verb stem, often with prefixes conveying lexical, adverbial, and thematic concepts, including plurality.4 The next level includes the pronominal and conjugational prefixes, which are hierarchically outside the verb base. The hierarchical structure is this: (2) Verb Theme: Verb Base: Verb Composite:
[classifier [root]] [prefixes . . . Theme [classifier+root/stem]] Base [pronom. & conjug. prefixes [Base]] VComp Theme
The linear order of the prefixes does not correspond to their hierarchical order. I assume that the verb composite is derived by adding affixes in hierarchical rather than linear order, following Speas (1986:228). The form in (3) exemplifies a verb composite and its analysis according to the scheme above. In the actual string, the outer prefixes are adverbial and derivational, indicated by “pref.” The inner prefixes are conjugational (cjg), and pronominal, with a subject prefix (subj) and a classifier (cl). The apostrophe indicates glottal stop (’); the acute accent indicates high tone (ó); the hook indicates nasalization (o˛). (3) na’ashkó˛˛ó’ ‘Iswim around’ na’a0 sh - ł- kó˛ó ˛’ 1 2 3 4 5 6 pref + pref + cjg + subj + cl + stem around (1+2), impf (3), I (4), swim (5+6) verb base: [na . . . łkó˛ó ˛’] ‘swim around’ The verb base is cited with the stem shape of the first person and the imperfective viewpoint, following Young and Morgan (1987). Mode is expressed by a conjugational prefix (including 0) and a distinctive stem. In (2) the combination of morphemes 3 and 6 conveys the Imperfective. The aspectual viewpoint of a sentence is signaled in the mode category. There are three modes which convey aspectual viewpoint: the Perfective, Imperfective, and Progressive. The other modes, which contrast with them, are the Usitative or Customary, the Iterative, the Future, and the Optative; they will not be discussed here. The perfective and imperfective are each indicated by one of four conjugational mor phemes: ni, yi, si, 0.5 4 Thematic prefixes are associated with particular verb roots; they appear in the same set of positions as adverbial and other derivational prefixes. There are other treatments of the Athapaskan verb: Kari recognizes three levels—the Lexical, Derivational, and Inflectional and Postinflectional—which he presents in a detailed model (1990:39). The verb base is formed at the first two of these levels. McDonough (1996) suggests that there are two main constituents to the Navajo verb, a tense-subject constituent and a verb stem. 5 Certain verb bases, and classes of verb bases, require particular conjugational morphemes (Young, Morgan, and Midgette 1993:863–74). Recall that there are four morphemes for both the
30
C.S. Smith
Aspectual viewpoints focus a situation from a particular temporal perspective, like different camera lenses. Just as the camera lens makes a scene available for a picture, viewpoint focuses and makes semantically visible all or part of the situation talked about in a sentence. The spans of perfective and imperfective viewpoints differ in how much they include of the situation focused. Perfectives generally focus a situation as a whole, with a closed perspective. In contrast, Imperfective and Progressive viewpoints focus part of a situation, with an open perspective. Somewhat more precisely, the Perfective focuses a situation with initial and final endpoints, while Imperfective viewpoints focus an interval of a situation without endpoints (Comrie 1976 and Dahl 1985). The notions of situation type and viewpoint are formalized in Smith (1991). The perfective viewpoint in Navajo focuses closed, nonstative situations and tends to convey a strong punctual feeling (Young and Morgan 1987). The imperfective viewpoints—Imperfective and the Progressive—present open, nonstative situations and are often associated with durativity. The Progressive appears with a relatively limited set of verb bases of a particular category, discussed below; it conveys a sense of dynamism and immediacy (Midgette 1987). The examples below have superscripts which give the viewpoint of the verb composite: Perfective (Perf), Imperfective (Impf), Progressive (Prog). The situation types of this article are realized at the level of the verb base.
2.3 Verb Lexeme Categories Many verb bases in Navajo realize a set of discontinuous morphological categories which combine with the theme to form the verb base. The categories are recognizable by distinctive patterns of root/stem variation; some also have distinctive prefixes, and some occur with particular viewpoint morphemes. These distinctive morphological categories are known as “aspectual categories” in Athapaskan linguistics. To avoid confusion between different notions of aspect, I use the term Verb Lexeme Category (VLC) for these categories. The prefix and verb stem of nonstative verb bases are partially determined by VLC; stative bases are discussed in Section 4.1. Each VLC has a characteristic
Imperfective and Perfective viewpoints. The facts are quite complex; I summarize some of them briefly. Bases of the Continuative, Conclusive, some Distributive, and some Reversative VLCs require the 0-Imperfective/si-Perfective conjugation pattern. Moreover, if the plural morpheme da appears, it overrides other dependencies and always appears with the si-Perfective. These facts suggest that the si-Perfective is related to plurality. But another VLC involving plurality, the Repetitive, takes either si-or yi-Perfective, depending on the particular verb base. Bases of the Durative VLC require a 0 -Imperfective/yi-Perfective conjugation pattern; bases of the Transitional require a yi-Imperfective/yi-Perfective pattern. The other VLCs have varied conjugation patterns. Whether these dependencies reflect semantic temporal features is unclear (cf. n. 14).
Aspectual Categories in Navajo
31
stem pattern, and some also require a particular prefix. (4) presents the skeletons of five different VLC verb bases formed with the root [’AH1 ] ‘movement of a flat object’ (Young, Morgan, and Midgette 1992:6). The bases differ in stem shape and prefix, or both. (4) [’áád], [na . . . ’ah], [’ad], [yi . . . ’ad], [’al] Actual verb bases that realize these VLC skeletons may have other prefixes as well. The VLCs have semantic names in the literature: the bases above belong to the categories known as Momentaneous, Continuative, Repetitive, Semelfactive, and Cursive, in that order. One question that arises is whether the semantic contribution of a VLC to the meaning of a verb base accords with its name; another is whether all VLCs have consistent semantic meaning. These issues are addressed below. The VLCs fall into two groups, A and B, on distributional grounds (Young and Morgan 1987 and Midgette 1987). The groups differ in occurrence with direct durative verbs and adverbials, and with the imperfective viewpoint. Group A VLCs do not appear with direct durative verbs and adverbials, and their occurrence with the imperfective viewpoint is limited. In contrast, the VLCs of group B appear with direct duratives and have no constraints on appearance with the imperfective viewpoint. This distinction suggests a correlation of type B VLCs with durativity; the possibility will be explored below. (5) lists the VLC categories discussed here, organized into the groups A and B. They are listed with their semantic names and abbreviations used in the superscripts of examples. (5) VLCs according to Groups A Momentaneous (Mom) Transitional (Tran) Reversative (Rev) Semelfactive (Sem)
B Conclusive (Con) Continuative (Ct) Cursive (Cur) Durative (Dur)
Distributive (Dis) Diversative (Div) Repetitive (Rep)
The list includes the main VLCs posited by Young and Morgan (1987) and Kari (1979). Unlike the other VLCs, the Cursive is associated with a single Mode: it requires the Progressive viewpoint, and the Progressive appears only with this VLC. The recognition of these morphological categories is a considerable analytic achievement. In fact, only recently were the VLCs shown to involve both prefixes and stem shapes, in ground-breaking work by Kari (1976) and Hardy (1979). Similar categories exist in other Athapaskan languages; see Kari (1979) for discussion of Ahtna and Axelrod (1993) for Kuyokon. The morphology and phonology of the VLCs are discussed in the references above; I will not consider them here. The VLCs are considered at the level of general temporal features and situation types in Section 3; their lexical functions and specific contributions to meaning are
32
C.S. Smith
discussed in more detail in Sections 4 and 5. I argue that three types of VLCs can be identified on semantic grounds: lexical VLCs, superlexical VLCs, and a third group, which I call formal. The formal VLCs are heterogeneous, without a consistent temporal feature or other semantic contribution to the verb base.
3 Grammatical Correlates of Temporal Features in Navajo In this section I consider the temporal features that distinguish the situation types presented above, ±Dynamic, ±Durative, and ±Telic, and ask whether Navajo has grammatical correlates of these features, including VLCs.
3.1 Temporal Features I begin with some brief comments about how the temporal features are expressed grammatically in English. Each feature has particular co-occurrence possibilities, or syntactic correlates. Appendix A gives a more complete discussion, with examples, of the Durative and Telic features. The Static/Dynamic feature distinguishes stative from nonstative (event) sentences, e.g., Mary owned the farm and Mary bought the farm. Event sentences are compatible with expressions of agency and volition, but statives are not, for semantic reasons, across languages. Thus, stative imperatives, and statives with certain adverbs, are usually semantically ill formed (indicated by ∗ ):∗ Be tall,∗ She carefully owned the farm. There are also language-specific grammatical correlates of the state/event distinction. For instance, in English, event sentences appear neutrally with the imperfective (progressive) viewpoint and with pseudo-cleft do. But stative sentences are ill formed in these syntactic contexts:∗ I am knowing the answer,∗ What she did was know the answer. In French, stative sentences do appear with the imperfective, although not with the pro-verb faire. In Russian, all situation types appear with the imperfective viewpoint, but the perfective is available only for nonstatives. Together these contrasts result in different distributional possibilities for stative and event sentences, for a given language. The feature Durative/Instantaneous distinguishes classes of event sentences. There are four recognized linguistic correlates of this feature, which are given below. They are forms or interpretations which occur with sentences of durative events, but not with sentences of instantaneous events (Mittwoch 1980 and Smith 1991). The correlates can be expected to hold across languages because they are semantically based. Direct duration: Sentences that present durative situations are compatible with direct durative adverbials, such as for an hour, in an hour. The adverbials give the duration of the event in question, e.g., John played in the sandbox for an hour, Mary walked to school in an hour. Sentences of instantaneous events behave
Aspectual Categories in Navajo
33
differently with these adverbials. The in adverbials have an ingressive interpretation: they pertain to an interval before the actual event, as in Mary won the race in an hour. The for adverbials, on the other hand, are odd in some sentences of instantaneous events, e.g.,∗ The balloon burst for five minutes. For others, they trigger the interpretation of a durative event with internal stages, e.g., John coughed for an hour. Inceptives and terminatives: Sentences of durative situations are compatible with inceptive and terminative morphemes, e.g., Mary began/finished building a sandcastle. These sentences present the beginnings and endings of the events. But with the same morphemes, sentences of semantically instantaneous situations either are ill formed or have a different interpretation. Thus The bomb started to explode presents stages preliminary to the actual event; ∗ The bomb finished exploding is semantically ill formed. (I ignore special cases such as reports of slow-motion films; cf. Mittwoch 1980.) Indirect duratives: Sentences of durative situations are compatible with indirect durative adverbials, which imply duration rather than explicitly stating it such as slowly, quickly: John slowly opened the door. The adverbial pertains to the actual event as it unfolds in time. With sentences of instantaneous situations, such adverbials are incompatible or have a different interpretation. Thus ∗ The balloon burst slowly is semantically odd; the adverb in John slowly reached the top pertains to preliminary stages as he approaches the top, not to the actual event of reaching it. Imperfective viewpoints: Imperfectives focus internal intervals of durative situations, but preliminary stages of instantaneous situations. For instance, Mary was walking focuses an interval of a walking situation, but Mary was winning the race focuses an interval preliminary to the instantaneous event of winning. The distributional correlates presented above are based on the logic of duration. Situations that take time have beginnings and endings; they can occur slowly or quickly, and they have internal intervals. It follows that sentences of semantically durative situations are compatible with the linguistic forms that express these notions. It also follows that sentences presenting instantaneous situations are not compatible with those linguistic expressions, or have different interpretations. The distributional constraints presented above are therefore semantic in nature. The third feature that often appears in temporal classifications is that of telicity. The Telic/Atelic feature classifies events according to whether or not they have a goal or natural endpoint. Telic events are completed when the natural endpoint is reached and the change of state occurs, while atelic events can stop at any time. There are distributional correlates for the distinction in English and some other languages. Telic verb constellations are compatible with verbs and adverbials of completion such as finish and in an hour, and are odd with forms of simple duration such as stop and for an hour. Compare, for instance, They built a summerhouse in a year and ?They build a summerhouse/drew a circle for a year.
34
C.S. Smith
In contrast, atelic verb constellations are compatible with verbs and adverbials of simple duration such as stop and for an hour, and are odd with forms of completion. Another feature of telicity is the interpretation of the adverbial almost in telic and atelic sentences. The scope of the adverbial almost in a telic durative sentence is ambiguous: Jane almost built a sandcastle may mean that she almost started to build one, or that she almost completed the building. In contrast, Jane almost walked in the park has only one meaning, that the event almost started. These contrasts are semantically based: the notion of completion is intrinsic to a telic event, irrelevant to an atelic event. There are similar grammatical correlates of the feature of telicity in many languages; see Appendix A for a more complete discussion.
3.2 The Grammatical Realization of Temporal Features in Navajo I now consider whether and how these temporal features are expressed in Navajo. I show that the features ±Dynamic and ±Durative have grammatical correlates. The feature ±Telic does not, because the language does not offer contrasts between verbs and adverbials of completion and simple duration.
3.2.1 The Feature Dynamic/Static This semantic feature is grammatically salient in Navajo. It distinguishes between verb bases that present stative and dynamic situations: in other words, between states and events. There are two main syntactic correlates of the distinction, viewpoint and VLC. Stative and dynamic verb bases have different co-occurrence possibilities with the aspectual viewpoints. Verb bases of events allow a choice of aspectual viewpoint: they may appear with either the Perfective or Imperfective, or the Progressive, although the latter is more limited.6 In contrast, stative verb bases appear with a single fixed viewpoint morpheme which does not have the usual contrastive meaning (cf. Section 4.1 below). Stative and dynamic verb bases also differ in their occurrence with VLCs. Dynamic bases must have one of the VLCs, whereas stative bases do not allow them. These distributional facts about stative verb bases are well known (cf. Hoijer 1949).
6 The Progressive viewpoint in Navajo appears only with verb bases of the Cursive VLC. The Cursive is discussed in Sections 4 and 5 below (see also note 16).
Aspectual Categories in Navajo
35
3.2.2 The Feature Durative/Instantaneous This semantic feature separates durative events from events that are instantaneous in principle. I now investigate the feature of duration in Navajo, using the diagnostics presented in Section 3.1 above. I demonstrate that there is a systematic distinction between verb bases associated with durative event concepts and bases associated with instantaneous event concepts. The bases used for each test present reasonably clear cases of event concepts that are durative or instantaneous. The durative concepts include {work on the hogan}, {eat an apple}; the instantaneous concepts include {catch a glimpse of}, {kick something}, {burst something}, {leave}; additional examples are also given. States are ignored here, since the feature of duration does not distinguish states from other situation types.7 Direct duration: Verbs and adverbials of direct duration are compatible with durative verb constellations and incompatible with instantaneous verb constellations. The Navajo forms of direct duration include verbs which refer to the passing of time, e.g., hodíina’ ‘time passed’, í’íi’Რ‘the day passed’, ahéé’ílkid ‘the clock hand slid in a circle’, and the adverbials shiidá˛a˛’dii ‘for a long time’, shá bíighaah ‘commensurate with the sun: all day long’. These forms require verb bases of group B VLCs and an imperfective viewpoint in the verb composite. (6) and (7) illustrate this; the verbs appear as the main verbs of complex sentences, with the event presented in a subordinate clause; -go is the subordinating suffix: (6a) shá bíighah hooghan binaashnish ‘All day, I work on the hogan’ (Impf B:Ct) (6b) shiid’dii na’nízhozhígóó shił’oolwoł ‘For a long time, I was driving toward Gallup’ (Prog B:Cur) (7) náháshgodgo shee ’i’ íí’a ‘I hoe + the day passed: I spent the day hoeing’ (Impf B:Rep. . .Perf A:Mom) The verb bases all express semantically durative events {I hoe}, {I drive toward Gallup}, {I work on the hogan}. I know of no cases where direct durative expressions appear with a verb base that expresses an instantaneous event.
7 In fact, stative sentences have the semantic temporal property of duration (Smith 1991). In Navajo, Statives are felicitous with direct duratives. They do not appear with the three other correlates of duration because of other factors: inceptives and terminatives by definition involve a change of state and are thus incompatible with statives. Adverbials of indirect duration are incompatible with statives because they imply successive stages; states consist of an undifferentiated period. Finally, as noted above, the imperfective viewpoint does not appear as such with stative bases.
36
C.S. Smith
Group A bases cannot appear in sentences like these (cf. Midgette 1987:100).8 These facts suggest that verb bases of group B present durative events, and that verb bases of group A present instantaneous events. Indirect duratives: Adverbials that imply temporal duration are known as indirect duratives. They are compatible with verb bases that have the semantic feature of duration and pertain to an internal interval of the event (cf. Section 3.1). The indirect duratives of Navajo include t’áá kóódígo “slowly”, t’áá tłehee “slowly and cautiously”, t’áá tdee “slowly”. Some sentences with these adverbials appear in (8). ˛á ˛’ (8a) t’áá kóódígo bilasáana yiyá ‘I slowly ate an apple’ (Perf B:Dur) ˛z (8b) shichidí t’áá tłehee niníłbá˛á ‘I parked my car slowly and carefully’ (Perf A:Mom) (8c) t’áá tłehee shidá’ák’eh binaagóó béésh’adishahí násht’ih ‘I slowly, carefully put a barbed wire fence around my field’ (Impf A:Rev) ˛a ˛dee bíláta’iigaii (8d) t’áá ta ‘It bloomed slowly = gradually became white at the tip’ (Perf A:tr) The adverb pertains to an internal interval of the situation in all of these sentences. The situations {I eat an apple}, {I park my car}, etc., have the semantic feature of duration. The durative verb bases are of both type A and type B VLCs. Adverbials of indirect duration are compatible with instantaneous verb bases given the interpretation that they indicate an interval preliminary to the event; the sentences of (9) have this interpretation. ˛dee tłehee dah diisháágo,. . . (9a) t’a˛a ‘I was slowly leaving,. . .’ (Impf A:Mom) ˛dee ’adzíítáál (9b) t’a˛a ‘I slowly kicked it away (e.g., a ball)’ (Perf A:Mom) (9c) t’áá tłehee séłdoh ‘I slowly, carefully burst it (e.g., a balloon)’ (Perf A:Sem) (9d) t’áá tłehee séłkah ‘I slowly, carefully shot at it’ (Perf A:Sem) 8 There are certain exceptional group A verb bases which can appear with direct durative verbs. These bases indicate events with the semantic property of duration and, crucially, do not have a related group B form. The following example, which has the base [’ahidi. . .kaał], is cited by Midgette (1987:78):
˛’ shizhé’é chizh ahidiłkaałgo i’íí’á My father firewood he chop + sun go down (Impf A:Mom. . .Perf A:Mom) ‘My father chopped firewood all day long’ Sentences like this do not represent a systematic possibility in the language. There are many other verb bases which also indicate durative events and have no group B related forms. But few of them seem to be acceptable to native speakers in this construction (Sally Midgette and Robert Young, personal communication).
Aspectual Categories in Navajo
37
The interpretation of (9d), for instance, is that I was slow in getting out my gun or in raising the gun to shoot, not that the shooting itself was slow. The events {I shoot at it}, {I kick it}, etc., are semantically instantaneous; the VLCs are of group A. The interpretation of adverbials of indirect duration distinguishes between verb bases according to the semantic feature of duration. With this criterion we find verb bases of durative events with VLCs of both group A and group B, and instantaneous verb bases of group A. The interpretation of the imperfective viewpoint is also a diagnostic for the semantic feature of duration. There are consistently different interpretations, depending on whether durative or instantaneous events are involved (cf. Section 3.1). Imperfective viewpoints typically focus the internal stages of durative events. The examples of (10) present sentences with the imperfective viewpoint, with the interpretation of internal focus. ˛ (10a) bilasáana yishá ‘I’m eating the apple’ (Impf B:Dur) (10b) hashb˛i’ ‘I’m building a hogan’ (Impf B:Con) (10c) niishgááh ‘I’m heating it’ (Impf A:Tr) (10d) shoo, nagháí tsé bitł’ááhdé˛˛é ’ tł’iish ch’é’nééh ‘Look, there’s a snake crawling out from under that rock’ (Impf A:Mom) Semantically, the events of these verb bases are durative: {I cry}, {I build a hogan}, etc. They have bases of group A and group B. With instantaneous events, the imperfective viewpoint focuses stages that are preliminary to the event itself. Such sentences are presented in (11).9
9 These sentences exemplify some of the contexts in which imperfectives may appear with bases of group A. (11a) and (11b) are bases for which no group B form exists. (11c) and (11d) are adversatives, appearing in contexts which suggest, in the absence of evidence to the contrary, that an event did not occur or continue. In this case we take it that the person probably did not kick the ball, and that I probably did not leave. The notion of adversative is a pragmatic one, because the relevant contexts cannot be identified on a syntactic basis. Imperfectives may also appear in backgrounded contexts:
taah yish’aah n´ t’éé’ bił néshj˛í˛íd ‘When I was putting it into the water, I slipped and fell’ (Impf A. . . Perf A) The first event, presented imperfectively, is backgrounded; there is no relevant group B form. These examples are noteworthy because they show that group A bases may occur with the imperfective viewpoint. The constraint against the imperfective viewpoint with group A bases depends partly on the availability of a group B form. Group A forms can be used in many contexts if no group B form exists; related verb bases are discussed in Section 4.3 below.
38
C.S. Smith
(11a) yiistééh ‘I catch sight of it’ (Impf A:Tr) (11b) yiishdoh ‘I burst O’ (Impf A:Sem) (11c) yiyiital n´ t’éé’hadeeshghaazh ‘When he was just about to kick it, I shouted’ (Impf A . . . Perf A) ˛´ (11d) dk’ad˛e´˛e dah diisháahgo nikihonílta ‘Just as I was about to leave it started to rain’ (Impf A. . . Perf A) The situations presented in these verb bases are semantically instantaneous: {I catch sight of it}, {I burst it}, etc. The bases all have group A VLCs.10 Inceptives and terminatives: These morphemes are superlexical in function. They take a narrow view of an event, presenting the initial or final endpoint. They have predictable differences of compatibility and interpretation according to whether the event in question is durative or instantaneous. In English, the inceptive and terminative morphemes are verbs that take sentences as complements. The complement verb constellation presents the broad view of the event. The Navajo structures are different: narrow views are presented by prefixes within a single verb base. To investigate the broad-view forms associated with inceptives and terminatives, it is necessary to consider related verb bases, and to deal with cases in which there is no verb base presenting the broad view of the event. The inceptive and terminative morphemes of Navajo are optional prefixes; they appear with certain bases of the Momentaneous and Transitional VLCs. The broad view of an event is realized in a different verb base. For certain event concepts, the broad-view presentation—the verb and its arguments simpliciter—are of the same VLC. Others belong to a different VLC; in some cases the broad-view form does not exist. Therefore, it is necessary to abstract away from a given narrow-view base to a broad view of the event in question. The diagnostic criteria are applied to Navajo sentences with inceptive and terminative morphemes in the following manner. I ask whether a given sentence is semantically well formed, and whether the event of the verb base—abstracting away from the form with the superlexical morpheme—presents an event that is semantically durative or instantaneous. (12) gives a set of well-formed sentences with inceptive and terminative morphemes. The sentences have endpoint interpretations: they present the initial or final endpoint of the event in question. The superlexical prefixes are noted with each example.
10 Although
imperfective forms are available for group A bases, they tend to be used in marked contexts. In fact, there is a tendency to think of group A verb bases as taking only a perfective viewpoint. This tendency is very strong in the consciousness of speakers of the Navajo language. People are sometimes unable to give the imperfective forms of group A VLCs in isolation, although they produce the forms without difficulty in the appropriate contexts (Sally Midgette, personal communication).
Aspectual Categories in Navajo
39
(12a) niki: tsinaa’ eełgóó niki’ níłkó˛˛ó’ ‘I started to swim to the boat’ (Perf A:Mom) (12b) di: hooghan bidishníísh ‘I started to work on a hogan’ (Impf A:Mom) ˛´a ˛´’ (12c) nii: bilasáana bi’niiya ‘I started to eat the apple’ (Perf A:Mom) (12d) ni: diyógí ninítł’ó˛ ‘I finished weaving it (a blanket)’ (Perf A:Mom) (12e) ni: nihonítáál ‘I finished singing’ (Perf A:Mom) These situations are semantically durative from the broad point of view: swimming to the boat, eating an apple, etc., are events that take time. Consider how the broad view of such situations is conveyed. Often there is a broad-view verb base, but not always. For the examples above, for instance, there is a broad-view verb base for (12b) with the same VLC and without the inceptive morpheme; for (12c) the broad-view base has a different VLC. There are no broadview bases for (12a) and (12d).11 This variation is typical. See Section 4.3 for a discussion of verb bases and superlexical morphemes. Now consider some other examples with inceptives and terminatives. The sentences of (13) are good on the interpretation that they present preliminaries of a situation; those of (14) are semantically ill formed.12 (13a) nii: ??yi’niiltsa ‘I started to catch sight of, see it’ (Perf A:Mom) (13b) nii: bi’niiłdo ˛o ˛h ‘I started to burst it’ (Perf A:Mom) (13c) nii: bi’niitáál ‘I started to kick him’ (Perf A:Mom) (14a) ni: ∗ niiníłtsán ‘he stopped looking at it’ (Perf A:Mom) (14b) ni: ∗ niinítáál ‘he stopped kicking’ (Perf A:Mom) 11 The
(a) (c) (d) (e)
related broad-view forms of the sentences in (12) are these:
No broad-view verb base (cf. 22) Broad-view base, different VLC: bilasáana yíy˛a´˛´a ‘I eat the apple’ (Perf B:Dur) No broad-view base Broad view base, different VLC: diyógí sétło’ó ˛ ‘I wove the blanket’ (Perf B:Rep)
12 According to a native speaker, these forms cannot be used and in a sense they do not really exist: in other words, they are ungrammatical. With the ‘shoot’ verb, the form niiníłka’ conflicts with another verb entirely, niiníłkáá’ ‘he stopped tracking it’. I would like to thank Ken Hale and Paul Platero for judgments and comments on these and related examples.
40
C.S. Smith
(14c) ni: ∗ niiníłka’ ‘he stopped shooting at it’ (Perf A:Mom) These examples all present events which are instantaneous on the broad view, as we have seen above in (9) and (11). The differences found here are just those noted in Section 3.1 for inceptives and terminative morphemes with sentences of instantaneous events. This result patterns with two of the preceding three other diagnostics; the first, compatibility with direct durative forms, is irrelevant here since it excludes bases of group A. We have looked at four grammatical correlates of duration in Navajo: cooccurrence of a verb base with forms of direct duration and of indirect duration, occurrence and interpretation with inceptives and terminatives, and focus of the imperfective viewpoint. Together, they show a distributional pattern that differentiates verb bases according to the semantic features [Durative] or [Instantaneous]. There is some relation between VLC and the semantic feature of duration. All bases of group B VLCs appear to be durative, but the association is not exclusive. There are durative bases of group A by the criteria developed above. These bases appear with adverbials of indirect duration on the internal interpretation; the imperfective viewpoint has internal focus, they have related inceptive and/or terminative forms. Therefore, the semantic feature [Durative] in Navajo is not morphologically based. Instantaneous verb bases have type A VLCs. If they appear with indirect duratives, the imperfective viewpoint, and/or the inceptive, what is presented is an interval preliminary to the event. They do not appear with terminatives. But since the type A VLCs include both instantaneous and durative verb bases, the semantic feature [Instantaneous] in Navajo is not primarily morphological either. 3.2.3 The Feature ±Telic The difference between telic and atelic events seems to be a fundamental one for humans, from a cognitive point of view. Telicity has grammatical correlates in English, and similar correlates appear in other languages (Smith 1991). Recall that the tests are quite indirect: they are based on the grammaticized distinction between completion and simple duration. In Navajo, there is little evidence of a grammaticized feature of telicity. Navajo does not distinguish completion and direct duration in either verbs or adverbials. There are relatively few direct durative adverbials in Navajo, and no translation equivalent of the contrast in an hour/for an hour. The terminative forms of Navajo are verb prefixes which have the meaning of either completion and simple cessation: there is no contrast like that between finish and stop. The same form, terminative prefix ni-, has both meanings: it indicates that “verbal action reaches a stopping point, halts, or is finished” (Young and Morgan 1987:53). Looking elsewhere, we might expect that the morphological categories known as aspectual, the VLCs, would code the semantic feature of telicity. But the VLCs do not distinguish verb bases according to this feature. All categories except one
Aspectual Categories in Navajo
41
include both telic and atelic verb bases; the exception is the Transitional category, which has exclusively telic bases. But although all Transitional bases are telic, there are many other telic bases in Navajo. The verb bases of each VLC category are discussed, with examples, in Section 5 below. The Imperfective and Perfective modes are each expressed with different conjugational morphemes (cf. n. 6). Some scholars have suggested that there are dependencies between particular conjugational morphemes and telic verb bases (Young and Morgan 1987 and Rice 1995). There may be such dependencies for particular subclasses of verb bases, e.g., the Transitional VLCs. But it is difficult to find a pattern at the global level, partly because there are several types of dependencies between certain lexical prefixes, VLCs, and the viewpoint morphemes.13 The almost ambiguities mentioned in Section 3.1 do not distinguish telic and atelic event sentences in Navajo. Such ambiguities do not occur because the scope of the adverb is always determinate in a Navajo sentence. To summarize, I have presented evidence for grammatical correlates in Navajo of the temporal semantic features ±Dynamic and ±Durative. I suggest later that the lexicalization pattern of the Navajo language somewhat neutralizes the distinction between telic and atelic events (see Section 4 below).
4 The Situation Types of Navajo The grammaticized situation types of Navajo are States, Durative events, and Instantaneous events. Sentences associated with each situation type have the grammatical correlates of the distinguishing temporal features. Situation type and VLCs are related variously: Stative verb bases lack VLCs; nonstative verb bases require VLCs; all VLCs of group B, and some of group A, present Durative events; some group A bases present Instantaneous events.
4.1 The Stative Situation Type Verb bases of the Stative situation type have the distinguishing temporal feature [Static]. The Stative is a salient category in Navajo. I sketch below the properties of stative sentences within the framework of this analysis. Stative verb bases are known as “neuter” in the Navajo and Athapaskan literature. Stative verb bases are unlike nonstatives in their co-occurrence with viewpoint morphemes and the interpretation of those morphemes. Each Stative verb base is
13 The
pattern of dependencies is intricate, with many exceptions, and its force is much debated. Krauss (1969:82) and Young and Morgan (1987:104) argue that the perfective morphemes have consistent semantic force. I was unable to find a consistent semantics at the level either of situation types or the feature of telicity (Smith 1991). Axelrod takes a position similar to mine in discussing the perfective morphemes of Koyukon, another Athapaskan language (1993:35–40).
42
C.S. Smith
associated with one viewpoint morpheme which is not interpreted as either perfective or imperfective. I suggest elsewhere a value of partial information, the Neutral viewpoint, for these cases (Smith 1991). The internal structure of Stative verb bases is different from that of event verb bases: Statives do not have VLC morphology. The range of Stative verb bases includes physical and mental attributes, locations, beliefs. The examples illustrate; I notate the viewpoint morphemes as Neutral(Neut) in order to emphasize that they do not have their standard values: (15a) tucsondi naashá ‘I am in Tucson’ (Perf-Neut) (15b) nisneez ‘I’m tall’ (Impf-Neut) ˛´ (15c) yistsa ‘I’m pregnant’ (Perf-Neut) (15d) nástáán taah yí’áá n´ t’éé’ ‘The log used to extend into the water’ (Impf-Neut) Statives fall into three main formal classes, according to the mode morpheme they require. One class of statives requires an Imperfective morpheme, another class requires a Perfective, and the third requires forms close to the Progressive (Young and Morgan 1987). The classes have some semantic coherence: verb bases taking an imperfective or progressive morpheme are often attributive, while those taking a perfective morpheme are often positional (see Kari 1979). Sentences with Stative verb bases allow either an open or a closed interpretation. They are thus more flexible in interpretation than nonstatives. We can show this flexibility with conjunctions. It is felicitous to conjoin a stative sentence with an assertion that the state continues, or with an assertion that it no longer obtains (the morpheme n´ t’éé’ indicates past time). Conjunctions of each type are presented in (16). ˛d’a ˛´a ˛´ násta´an (16a) da˛a ´ taah yí’a´a´ nt’éé’ ´ do´o´ t’ahdii taah yí’a´ ‘Last spring the log extended into the water and it still does’ (16b) nástáán taah yí’áá n´ t’éé’ ‘The log extended into the water (but it doesn’t now)’ These examples show that the stative sentence has a Neutral viewpoint, since it is reasonable in the context of either type of continuation.14 This discussion does 14 In contrast, conjunctions like this with nonstative (event) verb bases and the Perfective viewpoint
are semantically ill formed. ∗ hosélbj’ dóó t’ahdii hashb˛i’ ‘I built a hogan and I’m still building it’ (Perf . . .Impf)
Since the Imperfective presents an event without endpoints, it is compatible with an assertion that the event continues.
Aspectual Categories in Navajo
43
not consider several important classes of statives, including derived statives such as generics and habituals (see Smith 1991 for discussion).
4.2 The Instantaneous Event Situation Type Sentences of the Instantaneous situation type in Navajo appear exclusively in verb bases of group A. These bases indicate events that are instantaneous in principle. Such events may actually take more than a single instant of measurable time—crossing the finish line in a race, for instance, probably uses up several milliseconds—but this does not disturb the concept of an instantaneous event. Instantaneous bases have the distributional correlates of dynamism and duration. Like other dynamic bases, they are formed with VLCs and allow a choice of viewpoint, Perfective or Imperfective. The Instantaneous verb bases are distinguishable by the criteria for durative events developed above. Instantaneous bases are not compatible with forms of direct duration or with terminatives. They may be compatible with inceptives, adverbials of indirect duration, and the imperfective viewpoint, with the consistent interpretation of stages that are preliminary to the event, rather than the event itself. Some examples of sentences with Instantaneous verb bases are given in (17). (17a) níyá ‘I arrive’ (Perf A:Mom) (17b) yiiłtsá ‘I catch sight of it’ (Perf A:Tr) (17c) séłbał ‘I wave it one time (e.g., a blanket)’ (Perf A:Sem) (17d) séłkah ‘I shoot at it’ (Perf A:Sem) The examples include telic and atelic events: the events of (17a) and (17b) are telic, while those of (17c) and (17d) are atelic and do not involve changes of state. I will not restate the examples for the correlates of durativity given above but will make an additional point. There is a difference in entailment between imperfective Instantaneous sentences and their durative counterparts. Durative imperfectives entail that a given event (or process associated with it, if telic) was going on; but an Instantaneous imperfective has no such entailment.15 We cannot infer from (18a) and (18b), for instance, that I left or that I fell asleep.
15 Characteristic
patterns of entailments for the different situation types have been recognized in the literature (see the references in n. 2).
44
C.S. Smith
(18a) dah diishááhgo, . . . ‘I was leaving . . . (when . . .)’ (Impf A:Mom) (18b) (k’ad鲲e ) ’iishháásh ‘I’m on the verge of falling asleep’ (Impf A:Mom) The morpheme k’ad˛e´˛e “almost, about to” often appears in sentences like these. The Instantaneous verb bases of Navajo include bases with inceptive and terminative morphemes. These narrow-view morphemes present the endpoints of durative events. As such they are events in themselves, events of an instantaneous nature.
4.3 The Durative Event Situation Type Durative verb bases are distinguished by the distributional properties of dynamism and duration. Dynamism: Durative verb bases appear with VLCs and allow a choice of aspectual viewpoints. Duration: Durative bases meet the four criteria of duration: appearance with direct durative forms, with inceptives and terminatives, and indirect verbs and adverbials; the imperfective viewpoints focus internal stages of the event. Durative verb bases of group B occur with direct duratives; those of group A appear with the other features. These points have been made above. I turn to another topic here, the presentation of durative events from more than one view. The discussion will add to our knowledge about the function of certain VLCs. Durative events may be presented from a broad or narrow view, as we have seen. These possibilities are realized in Navajo by related verb bases. The following array of Navajo sentences illustrates, with different presentations of the same situation, skinning it (an animal) in bases of the root [’ah]. (19a) násh’ah ‘I skin it’ (Impf B:Concl) (19b) dish’ááh ‘I start to skin it’ (Impf A:Mom) (19c) n´ dí’níísh’ááh ‘I just got a good start skinning it, when. . . ’ (Impf A:Mom) (19d) yish’ah ‘I am in the process of skinning it’ (Prog B:Cur) Compare the views taken in these bases: (19a) presents the event as a whole, (19b) and (19c) have the inceptive prefixes [di-] and [´ndií-]; (19d) presents an internal interval of the event with the Cursive VLC. Recall that the Cursive VLC requires the Progressive viewpoint (prefix ghi/yi). Since the Cursive VLC differentiates (19a) and (19d), I attribute the internal view to the Cursive category—and to the Progressive viewpoint, since the two are mutually dependent.16 16 In
a sense, two different ways of presenting situations are neutralized here. Lexical and superlexical morphemes present broad or narrow views of a situation: the Cursive VLC is superlexical,
Aspectual Categories in Navajo
45
Arrays like this show broad and narrow views of a situation in Navajo. The narrow-view verb bases have superlexical morphemes, which take the form of a specific prefix or a particular VLC; the broad-view bases have only the prefixes and stem that present the particular situation. Such arrays are an important feature of the language. They allow speakers to present a situation in complementary ways. Before discussing arrays further, I would like to comment on the prefix [ná-], which appears in (19a). This is a thematic prefix, associated with the root under the meaning “to skin” (Young, Morgan, and Midgette 1992:8). Thematic prefixes are often required for well-formedness but do not always contribute identifiable meaning to the verb base. The contribution of a thematic prefix to a verb base is formal, in the sense given in Section 2 above. In complex bases thematic prefixes are often overridden by other prefixes, as [ná-] is in (19b)–(19d) by the inceptives and the Cursive/Progressive prefix. I use the term “array” for related verb bases that differ only in taking a broad or narrow view. Their distinguishing feature is that they present the same situation, in the sense developed above. Therefore, the bases of an array must all have the same lexical morphemes. To make the notion of array precise, I use the distinction between lexical, superlexical, and formal morphemes: the bases of an array may differ only in superlexical and formal morphemes, and must have the same lexical morphemes. This requirement will distinguish array-related bases from verb bases that are related in other ways, for instance, verb bases with the same root but different lexical prefixes. To illustrate, I give another array of bases in (20). They present different views of singing a song or ceremony, root [táál]2 . The thematic prefix ho appears in all the examples; it is notated as “th,” for thematic: (20a) ’aho’niishtaał ’a - ho- ’nii - 0 - sh - taał it - th - start - impf - 1p - stem ‘I start to sing it (song, ceremony) ’ (Impf A:Mom) (20b) nihonishtaał ni - ho - ni -sh- taał stop - th- impf - 1p - stem ‘I stop/finish singing a song, ceremony’ (Impf A:Mom) (20c) hashtaal ho - 0 - sh - taał th - impf - 1p - stem ‘I sing a song, ceremony’ (Impf B:Dur)
presenting a narrow view of a situation, namely, an internal interval. The viewpoints of a language focus all or part of a situation. The Progressive viewpoint focuses on part of a situation, namely, an internal interval. Recall that the Cursive VLC and the Progressive viewpoint are mutually dependent.
46
C.S. Smith
(20a) and (20b) have superlexical morphemes; (20c) does not. These bases form an array, like those of (19): they are related in the intended sense. Verb bases with different lexical prefixes do not form arrays. (21) illustrates the difference between bases that are array related and bases that are not. (21a) and (21b) present different views of an event {I laugh}; (21c) presents a different event, {I smile, chuckle}; all three have the same root [dlo’] ‘laugh’. (21a) yishdloh yi - sh- dloh Curs/prog - 1p - stem ‘I’m laughing along’ (Prog B:Cur) (21b) ninishlóóh ni - ni - sh - lóóh stop - impf - 1p - stem ‘I stop laughing’ (Impf A:Mom) (21c) ch’ídinishdlóóh ch’ídi - ni - sh - lóóh lex pref - impf - 1p - stem ‘I smile, chuckle’ (Impf:Mom) The first two examples differ only in superlexical morphemes. (21a) presents an internal interval of the event with the Cursive VLC; (21b) has the terminative prefix ni and the VLC with which it is associated. Unlike the first two, (21c) has the prefix ch’ídi, which is a lexical morpheme and changes the meaning of the verb base; (21c) therefore belongs to another array. (Statives may also have array-related bases, but I discuss only arrays of durative events in this article.) There are arrays which consist only of narrow-view verb bases, as in (22) and (23): (22a) tsinaa’eełgóó niki’níłkó˛ó ˛’ ‘I started to swim to the boat’ (Perf A:Mom) (22b) tsinaa’eełdi ’aníłkó ˛ó ˛’ ‘I arrived swimming at the boat (= finished)’ (Perf A:Mom) (22c) tsinaa’eelgóó ’eeshkó˛ó ˛ł ‘I’m swimming along toward the boat’ (Prog B:Cur) (23a) hooghan bidishníísh ‘I start to work on a hogan’ (Impf A:Mom) (23b) hooghan nibinishníísh ‘I stop working on a hogan’ (Impf A:Mom) (23c) hooghan binaashnish ‘I am working on a hogan’ (Impf B:Ct) In (22c) the Cursive functions as a superlexical morpheme, presenting an internal interval of the event; in (23c) the Continuative VLC functions in the same way. In
Aspectual Categories in Navajo
47
these arrays no base exists which takes a broad view, presenting the event in its entirety. Such arrays were introduced in the discussion of superlexical morphemes (cf. the examples of 8). They offer an interesting problem of analysis. Since no verb base presents the events simpliciter, the narrow view forms cannot be derived from bases with the broad view.17 In these arrays the superlexical morphemes are the Cursive and Continuative VLCs and the inceptive and terminative prefixes. The latter require a base of the Momentaneous or Transitional VLCs. These VLCs cannot be said to function as superlexical morphemes, however, because there are many Momentaneous or Transitional bases without superlexical morphemes; several have been presented above. Durative events are also realized by lone verb bases or minimal arrays; (24) illustrates: (24a) shidá’ák’eh binaagóó béésh ’adishahí nísełt’i’ ‘I put a barbed wire fence around my field’ (Perf A:Rev) (24b) bighá’nígééd ‘I tunneled through it (a mountain)’ (Perf A:Mom) (24c) chidí’ánáshdlééh ‘I repaired the car’ (Impf A:Mom) (24d) yaaziid ‘I poured it’ (Perf A:Tr) These verb bases run counter to the multiple lexicalization pattern of Navajo. They can hardly be considered exceptional, since there are a great many of them.18 Minimal array verb bases tend to have prefixes which block other, related verb bases; morphological analysis may bring out an underlying relation between certain prefixes and VLCs (Ken Hale, personal communication). This is an interesting topic for future research. To speculate for a moment, the arrays of Navajo suggest that situations tend to be realized in complementary, relatively narrow views by Navajo speakers. Indeed, the standard way of talking about an event may be narrow rather than broad. Then the various narrow views, or multiple bases, would together realize the concept of a situation for Navajo speakers. The narrow bases of many arrays view either the
17 The
different bases of such arrays can be seen as multiple realizations of an event at a more abstract level. To develop such an analysis, it will be necessary to work out detailed lexical rules for Navajo. I hope to provide such rules in future work. 18 In an informal count of verb roots in Young, Morgan, and Midgette (1992), about 125 roots have only nonstative verb bases of group A, while about 275 roots have nonstative bases of both group A and group B. The count indicates a significant number of minimal arrays. It does not give a sense of how many minimal or arrayed bases there are, however, because roots and their bases vary widely. Some roots are extremely productive, with many related verb bases; others have relatively few bases. The verb bases related to a given root may include one or more arrays, or minimal arrays, or single bases of the Instantaneous situation type.
48
C.S. Smith
bounds of an event or an internal interval. For instantaneous events, the viewpoint subsystem allows more than one focus. The imperfective and perfective viewpoints focus on the preliminary stages or the single stage of the event itself. The main distinctions, then, seem to be between bounds and intervals.19 If this is correct, the arrays may reflect something of how a Navajo speaker views situations. Continuing to speculate, we might say that the multiple views of Navajo contrast with the standard way of viewing situations in English. I have suggested that for English speakers the broad view of a situation is basic, and the narrower views are shifts from that basic level (Smith 1991; 1993/1995). There is morphological support for this approach to English, since the superlexical forms always occur with a basiclevel verb constellation. But the morphological facts of Navajo support the multiple view. The arrays also suggest something about telicity in Navajo. I have noted that the language does not have grammatical correlates of this feature. This may seem curious, given the importance of telicity in apprehending events. I suggest that the Navajo system somewhat neutralizes the feature of telicity, because of its emphasis on the bounds, or endpoints, of events. The multiple lexicalization pattern provides many durative situation types with a verb base which presents the final endpoint of the event. (If there is no terminative form, convention allows the use of a related form to convey a narrowed view on the final endpoint; cf. 27–29 below.) Most durative events have a telic aspect, in effect, with verb bases that present the final endpoint of the event. The notions telic and atelic are available to Navajo speakers, of course. The telicity of an event may be conveyed by certain morphemes in the verb base, as in ‘ołta’ di yíyá (school-to perf-lp-walk). The combination of NP and the postposition [di] with a locomotion verb stem functions like a directional complement, giving a destination (not unlike the English I walk and I walk to school). Again, an adverbial ˛h hasísis’na’ (tree by-it up-lp-cfg-crawl) ‘I prefix can indicate a goal, as in tsin ba˛a climbed the tree’. There are many differences between telic and atelic forms, even though we do not find consistent distributional correlates of telicity. 4.3.1 Pragmatic Conventions of Use The choices made by speakers of a language are often conventionalized, conveying particular meanings which follow from the pattern of the language. I note several conventions for talking about durative events in Navajo. Strikingly, the Imperfective is often used to talk about events that are closed, whereas the Perfective is used to talk about events that are open. These conventions presuppose an array with group A and group B bases.
19 Another way of putting this is to say that the bounds of events are salient in Navajo; this point is
made by Midgette (1989b) in a somewhat different connection.
Aspectual Categories in Navajo
49
One of the most idiomatic ways to say you are in the midst of doing something is to use the Perfective viewpoint with an inceptive verb base, as Sally Midgette has informed me; (25) illustrates: (25a) ’ółta’góó déyá ‘I’m in a state of having started to go to school (= I’m on the way)’ (Perf A:Mom) (25b) bi’niiłgaii ‘I finished starting to heat it (= I’m heating it)’ (Perf A:Mom) These sentences invite the inference that the speaker is now in the process of going or heating it. The inference can be canceled by explicit information to the contrary (. . . but I didn’t actually continue). Such sentences can also mean that the speaker is definitely planning to go. This use of Perfective viewpoint to present a situation in progress may be partly due to the association of the Imperfective with duration, which in effect makes it less available to convey other meanings. To talk about closed situations speakers often use one of the imperfective viewpoints, especially if they wish to emphasize duration. For instance, Midgette reports that the following sentences are preferred by Navajo speakers for talking about a long journey (1987:100): ˛´a ˛´’dii na’nizhozhígóó shił’oolwoł (26a) shiida ‘for a long time I was driving toward Gallup’ (Prog B:Cur) ‘It took me a long time to get to Gallup’ ˛´a ˛´’ shá bíighah na’nízhoozhígóó shił oolwoł (26b) j˛í˛ída ‘all day toward Gallup with me vehicle was going’ (Prog B:Cur) ‘I drove all day to get to Gallup’ These sentences may be used to convey that the event is closed. Recall that direct duratives cannot occur with the Perfective viewpoint. There are also conventions of use for arrays that do not have a full set of choices. If a limited convention clashes with a general one, the limited convention overrides. One such convention provides a pragmatic view of the final endpoint of an event. Consider arrays with only two members, one presenting a broad view and the other a narrow, inceptive one, as in (27–29). (27a) yisdiz ‘I twist it’ (Impf B:Con) (27b) disgéés ‘I start to twist it’ (Impf A:Mom) (28a) yit’éés ‘it (meat, etc.) has cooked’ (Impf A:Mom) (28b) bi’niit’ees ‘it (meat, etc.) starts to cook’ (Impf A:Mom)
50
C.S. Smith
˛´ (29a) neest’a ‘(wheat, etc.) ripen’ (Perf B:Dur) (29b) bi’niit’˛í˛íh ‘it (wheat, etc.) starts to ripen’ (Perf A:Mom) (29c) noot’˛í˛íł ‘it (wheat, etc.) ripening along’ (Prog B:Cur) For such arrays the verb base without a superlexical morpheme may be used to convey a terminative view of the event. The Imperfective viewpoint is used this way for bases of group A (27a and 28a) and the Perfective viewpoint for bases of group B (29a). Minimal arrays (single bases) are even more limited, of course. For these verb bases the only choice is between Perfective and Imperfective viewpoints; the Perfective may be used to convey a pragmatically terminative view. The language has many more group A verb bases than group B bases. Group A bases tend to present events that are quite precise lexically, while group B bases tend to be general. Due partly to this asymmetry, there are numerous group A verb bases without array-related group B forms. Speakers can compensate for the asymmetry by using a group B form that is close in meaning. In other words, if no array-related group B base exists, speakers can use a general group B base to talk about an internal interval of a durative event. For instance, the root [zhee’] ‘hunt’ has sixteen bases with VLCs of group A, many of them with specific lexical prefixes, and four bases of group B. Among the group A bases are [’ahéé . . . zhééh] ‘make a circuit hunting’ (Mom), [’ałnáá . . . zhééh] ‘hunt between two places’ (Mom), [ha . . . zhééh] ‘climb up while hunting’ (Mom). None of these forms has an array-related group B base. But the very general Cursive base of group B, [na . . . zheeh] ‘be hunting’, can be used to talk about ongoing processes of these group A bases.20 The conventions emphasize the importance of the narrow-view bases, and the flexibility of the language in cases where a full array is not available.
20 To indicate the breadth of this phenomenon, I give another example of the striking differences in number and generality between group A and group B VLCs. There are close to fifty group A verb bases involving the notion ‘drive’ from the root [lo’], among them the following: [’a . . . łeeh] ‘drive unspecified object’, [’a’a . . . łeeh] ‘drive vehicle away out of sight’, [’e’e . . . łeeh] [’ada . . . łeeh] ‘drive vehicle into, e.g., a wash’, [’ada’a . . . łeeh] ‘drive down from a height’, [’ahéná . . . łeeh] ‘make a circuit driving’, [ná . . . łeeh] ‘drive or carry something around, detour around’, [yisdá . . . łeeh] ‘drive to safety’. There are five group B bases concerning driving generally, from the same root, e.g., [’a . . . łoh] ‘drive along in a vehicle’. Other bases from this root have meanings such as ‘throw loops’, ‘carry by a handle’, ‘put on brakes’, ‘trick, lure’. There are also many other bases referring to driving: for instance, the root [b˛a´˛] a´ has a large number of such bases, which tend to involve rolling and circular motion.
Aspectual Categories in Navajo
51
5 The Function of VLCs in Navajo VLCs are morphological categories of the verb base which vary in function and consistency. I use the notions of lexical, superlexical, and formal to analyze their semantic functions in the aspectual system of Navajo. Three types of VLCs can be identified according to their contribution to the verb base: lexical VLCs, superlexical VLCs, and a third group, which I simply call formal. The formal VLCs are heterogeneous: they do not convey a consistent temporal feature or make an identifiable semantic contribution to the verb base.
5.1 Lexical VLCs The lexical VLCs contribute consistently to the lexical meaning of a verb base. There are two subgroups. The Distributive, the Diversative, the Repetitive, and the Reversative form one subgroup; the Transitional, a certain subset of Continuatives, the Conclusive, and the Durative form the other. The VLCs of the first subgroup have specific lexical meanings, roughly according to the semantic name of the VLC. Thus the Repetitive bases indicate multiple events consisting of a series of subevents; the Distributive bases have the property of distributivity, etc. Three VLCs of this type have lexical meanings that pertain to plurality in some way: they are the Repetitive, the Distributive, and the Diversative. I list some examples of bases of each VLC; they are intended to be representative of the VLC categories but are not exhaustive. I assume that the English gloss adequately indicates the event concept conveyed by a verb base. In cases of uncertainty, the distributional correlates developed above can decide. Repetitive bases present multiple, durative events, as in (30): (30) Verb Bases of the Repetitive VLC (Group B) yishch’id ‘scratch it’, yishhozh ‘tickle him’, yishkad ‘flap it, slap it—as, a dusty hat’, náshka ‘I’m sewing it’, béshtloh ‘smear on it—as grease’, sitsiits’iin táláwosh bésh’ááh ‘I’m rubbing something (a bar of soap) on my head’ The Distributive bases present multiple events that involve the property of distributivity; these bases appear with a plural morpheme and/or the prefix taa: (31) Verb Bases of the Distributive VLC (Group B) biih daash’a’ ‘to place each one of them 3 + into it’, ndaash’a’ ‘set 3 + object about’, tł’oh ’ałch’˛i’ be’estł’ ónígíí ł˛í˛í’ bá ndaash’a’/ ndasé’a˛´ ‘I setting/set 3 + bales of hay around in each of 3 + locations for the horses’ The Diversative presents multiple events that involve successive subevents; they often appear with the prefixes tádi, na (multiple), taa:
52
C.S. Smith
(32) Verb Bases of the Diversative VLC (Group B) ˛´a ˛´’ shinaada˛´a˛´ bitaasha ‘visit them, go among them’, j˛í˛ída bitaaséyá ‘I visited my corn(field) today’, yitaadaabaah ‘they raid among them, make war on them’ The Reversative is less consistent than the other lexical VLCs. The signature prefix of the category, ná, means ‘to turn an object around or over’. It is used quite generally for events involving any type of semicircular motion, as illustrated: (33) Verb Bases of the Reversative VLC (Group A) náyiiłtééh ‘he turns an animate being over/around’, náyiilééh ‘he turns around/over, a ropelike object’, nashgo˛sh ‘he’s making sausage’ náháj˛ísh ‘. . . turning himself so as to lie on his side’, náyiidééh ‘move in a circle back to a starting point’, n´ deesht’ih ‘I will set something in a circle, e.g., rocks’ These VLCs make a fairly specific lexical contribution to the verb base; they may combine with other lexical morphemes (cf. the brief comment about subaspects below). Bases of the second group of lexical VLCs are consistent in a single temporal nature; the VLCs themselves make no specific, identifiable contribution to lexical meaning, however. Transitional bases, and certain Continuatives under certain circumstances, present events with the feature of telicity; the Conclusive and Durative present durative events. The Transitional category presents telic events. They vary in durativity, as the glosses suggest. For instance: (34) Verb Bases of the Transitional VLC (Group A) yiiłtool ‘clear up, as murky water’, yiibááh ‘become gray’, diilko˛o˛h ‘become smooth’, yiiz˛i˛ih ‘he’s coming to a standing position’, diishch’ééh ‘open one’s mouth’, yiishsh˛í˛íh ‘blacken something’ There is a subset of Continuatives which have the consistent telic meaning of a round-trip, only in sentences with the Perfective viewpoint. These Continuatives have a locomotive verb stem and the postpositional enclitic -góo, as illustrated: (35) tsinaa’eelgóó ni’séłkó˛˛ó’ ‘I swam to the boat and back’ (Perf B:Cont) The Conclusive and the Durative, both of group B, are consistent in the property of durativity, and thus pattern with the Durative situation type. They make no specific or identifiable contribution to lexical meaning. The events are telic and atelic:
Aspectual Categories in Navajo
53
(36) Verb Bases of the Conclusive VLC (Group B) yishbéézh’ ‘to boil, cook it’, nish’˛i˛ih ‘to steal something’, yishjeeh ‘to grease’, nishk’ah ‘to get fat’, yooshnééh ‘to forget’, yishchí ‘to give birth to’, yisht’eesh ‘to paint something black’, yisdiz ‘to spin it’, ntséskees ‘to think’ (37) Verb Bases of the Durative VLC (Group B) yashti ‘I’m talking’, łídíshchí ‘I’m flattering him ’, yishtlo’ ‘weave something’, yishkeed ‘to eat a roundish, chunky thing’, yishbéézh ‘to cook something’, náshkad ‘to sew it’, yishdlá ‘to drink it’ These VLCs may occur with a set of lexical prefixes known as sub-aspects (Young and Morgan 1987). They are seriative hi, semeliterative náá, reversionary náá, prolongative hi, and the superlexical inceptive and terminative morphemes discussed above. The reversionary, seriative, and semeliterative appear with a variety of VLCs; the inceptive, terminative, and prolongative prefixes require Momentaneous or Transitional bases. Bases with the subaspect prefixes do not have distinctive stem sets, so they do not form separate VLCs. Several subaspectual prefixes can appear together in a verb base, with various constraints and combinatory meanings (cf. Young et al. 1992:874–78). The lexical VLCs indicate aspectual features, which determine certain features of the situation specified in a given verb base. Lexical VLCs pertain to the internal temporal features of situations, e.g., duration, telicity, and different types of plurality. Thus the lexical VLCs are part of the aspectual system of Navajo at the lexical level, though not at the global level of situation types.
5.2 Superlexical VLCs Superlexical morphemes present part of a situation, and their only contribution to the verb base is this narrow view. Two VLCs function as superlexical morphemes. The Continuative and the Cursive present an internal interval of a durative event. The Cursive VLC appears obligatorily with the Progressive viewpoint, marked by the conjugational prefix ghi/yi ‘along’. The Cursive appears in many verb bases involving motion in a literal or extended sense, as in (38): (38) Motion Verb Bases of the Cursive VLC (Group B) (38a) ‘atiingóó yishááł ‘I’m walking along the road’ (Pg B:Cur) (38b) hooghangóó yish’nah ‘I’m crawling along toward the house’ (Pg B:Cur) The members of this category are actually quite diverse, as (39) makes clear: (39) Varied Verb Bases of the Cursive VLC yish’ah ‘I’m in the process of butchering (obj)’, dah yiibah ‘camping on a raid’, béésdził ‘strain at, e.g., childbirth, defecation, pulling a heavy object’, yishgoł ‘digging, spading along in a line’, yishgish ‘cutting along
54
C.S. Smith
in a line’, yishtł’óół ‘weaving along’, nááshbał ‘whirling along rapidly pirouetting’ Some of these verb bases involve motion only in a very metaphorical sense. The Continuative also presents an internal interval of an event, for verb stems which are not of the motion class; this VLC appears with the prefix na ‘around about’:21 (40) Varied Bases of the Continuative VLC (Group B) naakaad ‘spreads around—as a plant’, naash’na’ ‘be crawling around’, nahashniih ‘he’s buying, selling something’, ’ahaa nahiilniih ‘trading with each other, e.g., horses, sheep’, naada’a’ ‘he’s lounging around’, ndi’ni’ ‘he’s going about moaning’ Quite different are the subset of Continuatives, mentioned above, which indicate a round-trip when they appear with the Perfective viewpoint. The Continuative and the Cursive are complementary as superlexical categories: they appear with different verb bases. The inceptive and the terminative superlexical morphemes do not have the morphological status of separate VLCs, as noted above, but belong to the class of aspectual prefixes known as subaspects. If VLCs are classified according to whether they present a broad or narrow view of a situation, we might say that those categories not specified as narrow present a broad view of a situation or, alternatively, that they are not specified for this feature.
5.3 Formal VLCs The formal VLCs are the Momentaneous and Semelfactive, both of group A. These VLCs do not have the consistency of the other VLCs. They are not lexical or superlexical in function nor directly associated with a viewpoint. I give examples of each category; the glosses suggest their diversity of meaning. The Momentaneous is much the largest and semantically the most varied of all the VLCs, “the most pervasive and productive in the Navajo verb system” (Young et al. 1992:868). The category includes bases of instantaneous and durative events; and of telic and atelic events. For instance: (41) Verb Bases of the Momentaneous VLC (Group A) diitááh ‘it shatters’, haashłé ‘take out something’, bik’íhásh’ááh ‘put blame on someone’, áshłééh ‘make something’, disht’ééh ‘extend in straight line (e.g., wire)’, ’anáháshgeed ‘fill a hole back up by
21 The motion and nonmotion verb bases of Navajo differ in many ways (see the references above for discussion).
Aspectual Categories in Navajo
55
shoveling’, bigh’á’níshgééd ‘dig or tunnel through’, ndíshgééd ‘cut through— as a dam’ The Semelfactive is also a formal category. The term “Semelfactive,” which comes from Slavic linguistics, refers to a suffix which indicates a single event, usually atelic, such as flap a wing, cough. However, the members of this class vary in both telicity and durativity, as the examples show; Semelfactive bases have the prefix yi: (42) Verb Bases of the Semelfactive VLC (Group A) ˛sh ‘bird gives it a peck’, yiyiiłt’ish ‘hit once with blunt object’, yiyiighas yiyiiłta ‘give it a scratch’, yiisdis/sédis ‘roll an object up in a tarpaulin’, ’iisgis ‘to do laundry’, ’iishnih ‘to do the milking’ Although from the point of view of this article the Semelfactive is a formal category, it has a discernible semantic structure, with core and peripheral instances (cf. Midgette 1989a). The formal VLCs are heterogeneous and appear to be purely morphological categories. They may have been semantically coherent at an early stage of the language. To summarize, I have argued that the VLCs of Navajo are diverse in meaning and function. Those with consistent meanings contribute aspectual information of different types to a verb base. Lexical VLCs specifically concern the temporal features of a situation, and thus contribute to specifying a given event. Superlexical VLCs, like other superlexical morphemes, function at a different level to indicate a narrow view of a situation. The formal VLCs are varied, with no consistent meaning or function.
6 Conclusion This article discusses aspectual categories in Navajo. I have claimed that there are three global aspectual situation types in the language: Durative events, Instantaneous events, and Statives. The situation types are based on temporal features of dynamism and duration, which are grammaticized in the language. The morphological categories which are traditionally known as aspectual, the VLCs, are involved in the distinction of situation types. They constitute an important correlate of the Static/Dynamic distinction. There is an association between VLCs and the distinction Durative/Instantaneous: all bases with group B VLCs, and some bases of group A VLCs, are associated with Durative events. All bases associated with Instantaneous events are of group A VLCs. The VLCs also function at other levels of the aspectual system. As part of the superlexical subsystem, certain VLCs convey narrow views of a situation; others contribute specific lexical meanings of an aspectual nature to the verb base. The VLC categories, although they may be similar from the viewpoint of morphological structure, make diverse contributions to the aspectual system of Navajo.
56
C.S. Smith
Acknowledgement I wish to acknowledge much generous advice and help given to me in this work. I would like to thank Sally Midgette for sharing with me much of her knowledge, as well as for her patience and suggestions. I thank Robert Young for his encouragement, comments, and meticulous corrections of Navajo examples. I thank Ken Hale for stimulating discussions of much of this material. I would also like to thank the editor of this Journal for a penetrating reading of an earlier draft of this article. All errors of fact and interpretation are mine, of course. Works on Navajo which I consulted include Young and Morgan (1987), Young et al. (1992), Kari (1976; 1979; 1990), Midgette (1987), Speas (1986); also Hardy (1979) and Sapir and Hoijer (1967). The examples presented here are for the most part drawn from Midgette (1987) and Young, Morgan, and Midgette (1992); some were constructed in consultation with speakers of Navajo. [Editors’ note: All linguistic examples have been reproduced verbatim, with the exception of (13b) in which Smith used a different transcription system than elsewhere in this paper. In that single example, she used different transcription conventions for the voiceless lateral and for nasalized vowels. Our thanks to Ted Fernald, Paul Platero, and Tony Woodbury for discussion.]
Appendix: Discussion of ± Durative, ± Telic Grammatical Correlates 1 The Durative/Instantaneous Contrast 1.1 Sentences that present durative situations are compatible with direct durative adverbials, such as for an hour, in an hour, as (1) illustrates: (1a) Mary walked in the park for an hour. (1b) Mary built the sandcastle in an hour. The adverbials give the duration of the events. Now consider the sentences of (2), which present instantaneous events and have the same adverbials. (2b) is semantically ill formed; the adverbials of (2a) and (2c) have different interpretations. (2a) He coughed for an hour. (2b) ∗ The bomb exploded for an hour. (2c) The bomb exploded in an hour. In (2a), the interpretation is multiple: the event consists of a series of coughs. The adverbial triggers a shift to a durative interpretation (Smith 1993/1995). In (2c), the adverbial has an ingressive interpretation: it pertains to an interval before the event takes place. Thus in and for adverbials are interpreted differently for verb constellations of events that are semantically durative and instantaneous. 1.2 Sentences of durative situations are compatible with inceptive and terminative morphemes; such morphemes imply that a situation has duration. (3) illustrates: (3a) Mary began to build the sandcastle. (3b) Mary stopped building the sandcastle. (3c) Mary finished building the sandcastle. But sentences of instantaneous situations behave differently, as in (4): (4a) ?The bomb started to explode. (4b) ??The balloon started to burst.
Aspectual Categories in Navajo
(4c) (4d)
∗ The ∗ The
57
bomb stopped exploding. bomb finished exploding.
The inceptive sentences are more or less well formed according to whether the event can be seen as having preliminary stages: thus (4a) is slightly better than (4b) because of the possible interpretation that the fuse of the bomb was sputtering, presumably just prior to the expected explosion. The terminative sentences are simply ill formed (ignoring special slow-motion presentations in which no events are instantaneous). 1.3 Sentences of durative situations are compatible with indirect durative adverbials, which imply duration (slowly, quickly). But these adverbials are not compatible with sentences that present instantaneous events. The contrast is illustrated in (5): (5a) presents a durative event, (5b) and (5c) present instantaneous events. Only the first is well formed. (5a) The door opened slowly. (5b) ∗ The balloon burst slowly. (5c) ??The bomb exploded slowly. Some speakers allow sentences like (5c) on the ingressive interpretation, in which the sentence presents preliminary stages of an event rather than the event itself. 1.4 Imperfective viewpoints focus internal intervals of durative situations, but preliminary stages of instantaneous situations. The examples illustrate: (6a) presents a durative event, (6b) an instantaneous event. (6a) The door was opening. (6b) Mary was reaching the top. Both sentences are well formed, but the interpretations are different as stated. Many languages allow the imperfective viewpoint to focus preliminary stages of instantaneous events; but not all: in Mandarin Chinese, sentences like (6b) are ungrammatical (Smith 1991). 2 The Telic/Atelic contrast There are distributional correlates for telicity in English and many other languages. Telic verb constellations are compatible with adverbials of completion, such as in an hour; with forms of simple duration they are odd or require a special interpretation. In contrast, atelic verb constellations are compatible with verbs and adverbials of simple duration, such as stop and for an hour, and are odd with forms of completion. (7) illustrates: (7a) Mary walked to school in an hour. (telic) (7b) ??Mary walked in the park in an hour. (atelic) There is a reasonable interpretation of (7b): Mary has a certain amount of walking to do and completed it in an hour. This is an unusual telic interpretation of a verb constellation [Mary walk in the park] which is atelic at the basic level. With adverbials of simple duration the situation is reversed. In (8) the atelic verb constellation is fine but the telic is odd:
58
C.S. Smith
(8a) ??Mary walked to school for an hour. (telic) (8b) Mary walked in the park for an hour. (atelic) There is an interpretation of (8a): that Mary engaged in an activity—walking to school—but without any natural endpoint. This is of course an atelic reinterpretation of a verb constellation [Mary walk to school] which is usually telic. The notion of completion is part of the meaning of some verbs but not of others: compare finish and stop. The verb finish implies completion, whereas stop does not. Both verbs are compatible with telic sentences as complements, as the examples illustrate: (9a) Mary finished building a sandcastle. (9b) Mary stopped building a sandcastle. These sentences are both well formed, but with different meanings. But only stop is good with an atelic sentence, as (10) shows: (10a) Mary stopped walking in the park. (10b) ∗ ?Mary finished walking in the park. There is of course a telic interpretation of (10b), like the telic interpretation of (7b) noted above. All these contrasts are semantically based: the notion of completion is intrinsic to a telic event, irrelevant to an atelic event. These contrasts are found in French, Italian, Spanish, and Russian, among other languages, as well as in English. But not all languages have them. In Mandarin Chinese, for instance, there are no verbs or adverbials like those in the examples above. However, Mandarin marks the distinction between telic and atelic verb constellations with different classes of verb suffixes, known as Resultative Verb Complements (see Smith 1991).
References Axelrod, Melissa. 1993. The Semantics of Time: Aspectual Categorization in Koyukon Athabaskan. Lincoln: University of Nebraska Press. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Dahl, Östen. 1985. Tense and Aspect Systems. Oxford: Blackwell. Dowty, David. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Forsyth, John. 1970. A Grammar of Aspect: Usage and Meaning in the Russian Verb. Cambridge: Cambridge University Press. Freed, Alice. 1979. The Semantics of Aspectual Verb Complementation. Dordrecht: Reidel. Hardy, Fred. 1979. Navajo aspectual verb stem variation. Ph.D. dissertation, University of New Mexico. Houer, Harry. 1949. The Apachean verb B: the theme and prefix complex. IJAL 15:12–22. Kari, James. 1976. Navajo Verb Prefix Phonology. New York: Garland Press. Kari, James. 1979. Athabaskan Verb Theme Categories: Ahtna. Alaska Native Language Center Research Papers, no. 2. Fairbanks: University of Alaska.
Aspectual Categories in Navajo
59
Kari, James. 1990. Ahtna Athabaskan Dictionary. Fairbanks: Alaska Native Language Center, University of Alaska. Kenney, Anthony. 1963. Action, Emotion, and Will. London: Routledge & Kegan Paul. Krauss, Michael. 1969. On the Classification in the Athapaskan, Eyak, and the Tlingit Verb. IJAL Memoir 24. McDonough, Joyce. 1996. Epenthesis in Navajo. Essays in Honor of Robert Young, ed. E. Jelinek, L. Saxon, and K. Rice. Albuquerque: University of New Mexico Press. Midgette, Sarah. 1987. The Navajo progressive in discourse context: a study in temporal semantics. Ph.D. dissertation, University of New Mexico. Midgette, Sarah. 1989a. The semelfactive/repetitive contrast: aspect in Navajo. Paper read at the Athabaskan Languages Conference, Tucson, Arizona. Midgette, Sarah. 1989b. The weighting of lexical and inflectional aspect in Navajo. Ms. Mittwoch, Anita. 1980. The grammar of duration. Studies in Language 4:201–27. Rice, Keren. 1995. The morphology-syntax interface in Athapaskan languages: an overview. Paper presented at the Athapaskan Morphosyntax Workshop, Linguistics Institute, University of New Mexico. Ryle, Gilbert. 1949. The Concept of Mind. London: Barnes & Noble. Sapir, Edward, and Harry Hoijer. 1967. The Phonology and Morphology of the Navajo Language. UCPL 59. Berkeley and Los Angeles: University of California Press. Smith, Carlota S. 1983. A theory of aspectual choice. Language 59:479–501. Smith, Carlota. 1991. The Parameter of Aspect. Dordrecht: Kluwer. Smith, Carlota. 1993/1995. The range of aspectual situation types: derived categories and a bounding paradox. Proceedings of the Workshop on Tense and Aspect, ed. P. M. Bertinetti. Turin: Rosenberg & Sellier. Speas, Margaret. 1986. Adjunctions and predictions in syntax. Ph.D. dissertation, Massachusetts Institute of Technology. Vendler, Zeno. 1957. Verbs and times. The Philosophical Review 66:143–60. [Reprinted in Zeno Vendler, Linguistics in Philosophy. Ithaca: Cornell University Press.] Whorf, Benjamin. 1956. Language, Thought and Reality: Selected Writings of Benjamin Lee Whorf. New York: Wiley. Young, Robert, and William Morgan. 1980. The Navajo Language. Albuquerque: University of New Mexico Press. Young, Robert, and William Morgan. 1987. The Navajo Language, Rev. ed. Albuquerque: University of New Mexico Press. Young, Robert; William Morgan; and Sarah Midgette. 1992. An Analytical Lexicon of Navajo. Albuquerque: University of New Mexico Press.
Activities: States or Events? Carlota S. Smith
1 Introduction The general concepts of event and state are important to human beings as they manage in the world: they are helpful in recognizing dangerous and benign motion, situations that may result in changes of interest, etc. In the sentences of natural language these concepts are conveyed by four semantic categories proposed in Vendler (1957), known as Accomplishments, Achievements, States, and Activities. It is clear how three of these categories are related to the general concepts: the first two are events, the third are states. The status of the Activity category is less clear. Activities have something in common with each concept. I will argue here that Activities form a natural class with events, however. The argument is based on the behavior of sentences associated with the Activity concept in narrative and other sequential contexts. The keys to the argument are the property of dynamism and the contribution of the perfective viewpoint. Section 2 discusses the concepts and their linguistic expression; Section 3 examines Activity sentences in narrative and other sequential contexts; Section 4 presents an analysis of the Activity situation type, and discusses the modelling of narrative time; Section 5 concludes.
2 Situation Types The Vendlerian categories are semantic concepts: idealizations of types of situations characterized by temporal features. As realized in the sentences of a language, each category has unique distributional and interpretive properties. These properties are indirect syntactic correlates of the semantic concepts. The categories thus constitute covert linguistic categories in the sense of Whorf.1 I will refer to the categories as
1 Whorf (1956) argues persuasively that language has covert categories which have grammatical correlates and are thus indirectly grammaticized. The situation types are covert categories in this sense. Not all languages have all of the categories: in Navajo, for instance, certain distinctions are neutralized.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_3,
61
62
C.S. Smith
‘situation types’, using the term ‘situation’ as neutral between them. The two-valued temporal features Static-Dynamic, Telic-Atelic, Durative-Instantaneous distinguish the categories. I understand the features as follows: Consider first durative dynamic situations: they have initial and final endpoints and internal stages. The endpoints constitute change to and from a state of rest into the situation and out of it. The internal stages of a telic situation constitute progression of some kind toward the final endpoint (see the paragraph below). The internal stages of an atelic situation are motions of different types: external or internal, literal or metaphorical. There are also dynamic situations which consist only of a single stage: they are instantaneous in principle (see the comments on duration below). Such situations consist of an idealized single moment at which the dynamic event takes place, with a state of rest at the moments before and after.2 In contrast, situations with the feature [Static] do not change over time: they hold over an undifferentiated interval, with no structure: a state is the same at all times for which it holds. The features [Telic] and [Atelic] pertain to the final endpoint of a situation. Telic events have a natural final endpoint that is intrinsic to the notion of the event. This intrinsic endpoint involves a change of state, a new state over and above the situation itself. When this new state is reached the situation is completed and cannot continue. Telic situations have a heterogeneous part structure, so that there is no entailment from part to whole (Vendler 1957; Kenny 1963; Krifka 1989). 2 The way situations unfold in time can be seen in a temporal profile which relates the situation to times, or moments of time. The temporal profile of a dynamic situation differs at successive stages of time. Non-durative, single-stage, situations have only one stage at which something happens: the event takes place. Therefore the temporal profile shows occurrence of the situation at a given moment, with an idealized state of rest at the times before and after. The diagram illustrates (S indicates the single-stage situation): t2 t3 t1 rest S rest
In the temporal profile of a durative event times map to successive stages (I indicates the initial endpoint, F the final endpoint): t1 t2 rest I
t3 St1
t4 St2
t5 St3
t6 St4
t7 F
t8 rest
The endpoints can be taken as events in themselves, e.g., in inchoatives (Smith 1997). States do not take place in time, although they hold for a moment or interval of time. What this means in terms of a temporal profile is that there is no mapping of times to stages of the state: states have no internal structure. The diagram illustrates: (Ei indicates the change into a state, Eo indicates the change out of the state.) t1 rest
t2 t3 t4 t5 Ei - - - - - - - - - - - - -Eo
t6 rest
The initial and final endpoints of a state are changes which map to times; there is no such mapping for the interval over which the state holds.
Activities: States or Events?
63
The feature [Atelic] holds of situations that have an arbitrary final endpoint: they may terminate at any time, after an initial stage that is large enough to count as part of the activity (Taylor 1977). Atelic situations have a homogeneous part structure and thus an entailment from part to whole. Situations with the feature [Durative] hold or take place for an interval; situations with the feature [Instantaneous] hold or take place at an idealized moment (Vlach 1993). The actual duration of the latter type, e.g., reach the top, win a race, etc., may involve some unit of measurable time, but this is irrelevant to the notion of an instantaneous event (Dowty 1979). I assume that time is dense so that ‘moment’ is taken to represent some small meaningful unit, perhaps an idealized point. In Section 4.3 below I suggest that the successive stages of dynamic situations map to units of conventional time. The situation type categories with their distinguishing temporal features are summarized in (1); the names of the categories can be thought of as shorthand for each cluster of features. (1) Situation types3 State: static, durative (know the answer, love Mary, believe in ghosts); Activity: dynamic, durative, atelic (listen to music, push a cart, walk in the park); Accomplishment: dynamic, durative, telic (walk to school, draw a circle); Achievement: dynamic, instantaneous, telic (win a race, reach the top). The concepts associated with each situation type follow from their temporal features. States are static, homogenous situations. The concept of a state does not include endpoints: the beginnings and endings of a state are dynamic since they involve changes into and out of the state. Thus although they hold in time, states do not take time (Taylor 1977: 206). Activities are dynamic and homogenous, with arbitrary final endpoints. Accomplishments and Achievements are dynamic and heterogenous, with intrinsic final endpoints; they differ in duration. The situation type concepts are conveyed by verbs and their arguments, or verb constellations, as they appear in sentences. I assume that situations of all types are entities, following Kamp and Reyle (1993: 504); for a different view see Herweg (1991: 972). I use capital letters for the situation type names. For simplicity I will sometimes call a verb constellation or sentence by the situation type or
3 The full category of States includes ‘derived statives’ such as habitual and generic sentences and others (Smith 1991). There are also derived members of the other situation type categories, but the topic is beyond the scope of this article. The category of Achievements includes single-stage events that do not involve a change of state, e.g., flap a wing, cough. I have proposed elsewhere that these form a fifth situation type, the Semelfactive, with the features Dynamic, Atelic, Instantaneous (Smith 1991; see also Mittwoch 1991). I use the more standard classification here since questions about the fifth situation type are not relevant to the discussion of Activities.
64
C.S. Smith
temporal feature with which it is associated, for instance, an ‘Accomplishment’ or ‘dynamic’ sentence. There is a set of linguistic properties which are characteristic of the sentences with the defining temporal properties of the situation types. A cluster of properties realizes the semantic feature of dynamism (Vendler 1957; Dowty 1979; Smith 1991). Distributionally, dynamic verb constellations are compatible with the progressive viewpoint, with pseudo-cleft do, with forms of agency and volition; with verbs relating a situation to the passing of time (spend, take); with adverbials of motion (slowly, quickly, etc.). There are also interpretive properties: in sentences about the present time, dynamic constellations have an habitual interpretation, whereas state constellations do not. Modal verbs have different interpretations with dynamic and state verb constellations: with the former they may be taken as epistemic or deontic, but with the latter they have only an epistemic interpretation (Piñon 1995: 21). Accomplishment, Achievement, and Activity verb constellations have the linguistic properties associated with dynamism. The feature of telicity also has linguistic correlates. Distributionally, telic verb constellations are compatible with forms associated with completion, e.g., verbs like finish and in adverbials.4 Further, durative telic sentences are ambiguous with almost; with telic sentences there is no entailment from progressive or imperfective to perfective sentences. Accomplishments and Achievements have these properties. Atelic verb constellations are compatible with forms of termination, e.g., verbs like stop, and forms of simple duration, e.g., for adverbials.5 Atelic situations are homogeneous so that there is an entailment from progressive to perfective sentences. State verb constellations do not have the linguistic correlates of dynamism. States are compatible with forms of simple duration.
2.1 Classes of Situation Types The situation types can be divided into global classes, in addition to the four Vendler categories. There is a well-known division into discrete and non-discrete situations; these two categories correspond to the widely-accepted notions of event and state. Another division recognizes dynamic and non-dynamic situations as significant global categories. Scholars have differed as to the importance of 4 Completive adverbials (in. . .) have different interpretations with durative and instantaneous telic events. For Accomplishments, completive adverbials refer to the whole event (He built the house in a summer); for Achievements, such adverbials are ingressive, referring to the coming about of the event (He reached the top in an hour). 5 For Activities the possibility of occurrence with durative adverbials is a distinguishing property; but the actual occurrence of such adverbials is not. Paradoxically, sentences that have durative adverbials pattern with heterogeneous events, as noted above. More generally, the presence of in and for adverbials can shift the interpretation of a verb constellation. Shifts occur when an adverbial appears with a verb constellation that clashes in an essential temporal feature (Smith 1993/5, 1997).
Activities: States or Events?
65
dynamism. For Hinrichs 1981; Partee 1984; Dowty 1986; Herweg 1991, only mereological structure characterizes events. But for Vendler 1957; Mourelatos 1978; Bach 1981; Parsons 1989, both discreteness and dynamism characterize events. Events are dynamic, involving development in time (explicated in Krifka 1989; Verkuyl 1993; also Pustejovsky 1991). Each approach brings out certain properties of situations. I will argue that the second is useful in understanding how situation types are realized in sentences, and how sentences of different situation types function in narrative. I will show that Activities pattern with events in narrative. 2.1.1 Discrete and Non-Discrete Situations: The Strong Mereological View In this view the essential property of events is that they are discrete, specific entities. As such, they are bounded by their endpoints: “Events have a beginning and an end. . . a number of familiar properties of events. . . follow” (Bach 1981: 71). Events are heterogeneous: their part-whole structure is such that they cannot be subdivided into events of the same kind. Accomplishments and Achievements, the telic situation types, are event concepts. Accomplishment and Achievement sentences denote events. Recent work has shown that the class of events is not limited to telic situations, however (Smith 1993/5: Depraetere 1995). Activity verb constellations with explicit temporal bounds denote events: such situations are discrete, with non-uniform part structures. Explicitly bounded States may also belong to this category. (2) illustrates; the adverbials give the explicit limits. (2) a. Mary worked for 2 hours. b. Mary slept from 2 to 4 this afternoon. c. Mary was here for 2 hours. d. Mary was nervous from 2 to 4 this afternoon.
Activity Activity State State
These situations are heterogeneous in that there is no entailment from part to whole. The explicit limits make the difference. Note that these limits need not coincide with the beginning or termination of the event. In Mary worked for 2 hours, the span of three hours delimits the event presented in the sentence. But the working could go on for longer than that: one could continue. . . and she went on working for awhile.6 Sentences like (2a–b) pattern syntactically and semantically with telic sentences.7
6 An anonymous reviewer notes an account of durative adverbials which corroborates this point: M. Herweg, ‘Temporale Konjunktion und Aspekt’, Kognitionwissenschaft 2, 1991, 51–90. I have not seen the article. According to the reviewer, Herweg argues that durative adverbials specify the minimum duration of their arguments; they are often taken to specify exact duration, but such interpretation is due to pragmatically based inference. 7 Explicitly bounded Activities are discussed in detail in Smith 1993/5. Strikingly, they allow completive verbs and adverbials as telic sentences do. (i) presents a bounded Activity sentence with finish and take, and an in-adverbial.
66
C.S. Smith
The status of State sentences like (2c–d) is less clear.8 Though they have the feature of boundedness and can be identified as individuals, we may not want to classify temporally limited States as events; I shall not do so here. What is common to telic events and the situations of (2a, b) is that they are dynamic and bounded, with heterogeneous part-whole structures. They differ in the type of bound involved. While telic events have intrinsic bounds, temporally bounded situations have bounds which are explicit and independent (Heinämäki 1984; Depraetere 1995). Thus the class of events must be said to consist of bounded situations. Telic events are intrinsically bounded; other events are independently bounded. States are homogeneous situations, with a uniform part structure. They do not have discrete boundaries (unless explicitly stated, as noted above). States have the sub-interval property: when a state holds for an interval, it holds equally for any smaller interval of that interval (Bennett and Partee 1972). The situation type State uncontroversially belongs to the class of states. In the strong mereological view Activities also belong to this global class, due to their homogeneity. Some scholars claim that they have the subinterval property (Hinrichs 1981; Partee 1984; Dowty 1986; Herweg 1991). Situations presented with the progressive viewpoint also have the subinterval property (Vlach 1981; Dowty 1986). If mereological properties are primary in determining events and states (i)a. b. c.
Mary finished playing her violin for an hour. It took Mary 3 hours to play her violin for an hour. Mary played her violin for an hour in 3 hours.
(ib) and (ic) convey that within a three hour period, Mary did an hour of violin playing. Such sentences are ambiguous with almost, like Accomplishment sentences. (ii)
Mary almost played her violin for 3 hours. . . .but stopped after only 21/2 hours. . . .but decided not to because she had too much work.
The pattern of entailment of an explicitly bounded Activity is like that of a telic sentence: the part does not entail the whole. The reverse is true for simple Activities. The relation between durative adverbials and homogeneous situations, for instance as discussed in Vlach 1993, remains undisturbed by this result. 8 Explicitly bounded States are often ambiguous with almost but do not appear with forms of completion: (i)a. b. (ii)a.
#Michael finished being tired/here for an hour. #It took Michael an hour to be tired/here. Michael was almost here for an hour . . .but he didn’t come after all. . . .but he left after 45 minutes.
Explicitly bounded States have the same pattern of entailment as do explicitly bounded Activities.
Activities: States or Events?
67
(Herweg 1991), all situations that have the sub-interval property belong to the class of states. I will call this the strong mereological view. 2.1.2 Energeia: Dynamic vs Static Situations There is also an approach which takes situations with dynamism, or energeia, as a natural class which contrasts with the class of states.9 Dynamic situations, or nonstatives, take place in time. They occur in successive stages which are temporally located at different moments. Dynamic situations are “continually subject to a new input of energy”, as Comrie puts it (1976: 49). Since energy requires a source, we expect that when the energy ceases, so will the event. Thus dynamism brings with it both the assumption of an initial endpoint, and the possibility of an eventual final endpoint. The dynamic situation types are Accomplishments, Achievements, and Activities. The first two type are heterogeneous. Situations of the Activity type have a homogeneous part-structure, but with a certain restriction. Taylor (1977) notes that Activities require an initial stage of a certain size, as mentioned above. Bach makes a similar point. Discussing a possible entailment of a progressive to a perfect Activity sentence (If John is running, then John has run) Bach says that the entailment is “not literally. . . semantically valid . . . Consider the very beginning of John’s running. . . ” (1981: 71). There is generally a lower limit for certain Activities, imposed by the nature of the Activity (Parsons 1990). Activities “are much more divisive with respect to time than events, but generally still fall far short of states” (Piñon 1995: 23). Evidently, Activities differ from States. In some ways Activities are closer to events than they are to States. Activities take place in time. And like other situations involving change, Activities may terminate. In contrast, States are not dynamic. They consist of an undifferentiated period, continuing unless something happens to change them.10 This may be the basis for Taylor’s comment that although states are in time, they do not take time (1977: 206).
2.2 Aspectual Viewpoints Sentences have an aspectual viewpoint which affects the information conveyed about a situation. Before looking at sentences in narrative contexts, I consider briefly the contribution of viewpoint. Viewpoints make available for semantic interpretation
9 This
notion differs somewhat from Aristotle’s energeia. Cf. the discussion in Mourelatos 1993. that the change out of a state is itself a change of state. When a state is talked about as a discrete entity, e.g., Last month he was sick, the sentence refers to the state itself. The endpoints are understood, by inference. If the state no longer obtains the inferred final endpoint constitutes the ending of the state, Last month he was sick but he’s well now. If the state is ongoing, e.g., Last month he was sick and he still is sick, in such cases the endpoints constitute the beginning and ending of the stated temporal interval.
10 Recall
68
C.S. Smith
all or part of a situation; I shall say that they ‘present’ a situation. Aspectual viewpoints are imperfective, perfective or neutral; I focus on the perfective here.11 Sentences with imperfective viewpoints present an unbounded interval of a situation: the interval does not include endpoints. The English imperfective viewpoint is the progressive, which appears in standard speech only with non-State sentences.12 As noted, situations presented imperfectively have the sub-interval property. The English perfective viewpoint occurs with sentences of all situation types.13 Telic sentences with the perfective viewpoint present complete events with initial and final endpoints. The interpretation can be demonstrated by conjunction. If they present complete events, perfective telic sentences should be incompatible with assertions of non-completion. And they are incompatible with such assertions, as (3) and (4) show; infelicitous conjunction is marked by #. (3)a. #Donald fixed the clock and he is still fixing it. (Accomplishment) (3)b. #Mary opened the door, but she didn’t get it open. (Accomplishment) (4) #Bright Star won the race, but she didn’t win it. (Achievement) The English perfective presents a State as an ongoing, unbounded situation. Therefore conjoining a perfective State sentence with an assertion of continuation is felicitous: (5)a. Sam owned 3 peach orchards last year, and he still owns them. (State) b. The children knew the answer yesterday and they still know it. (State) Pragmatically, however, it is natural to infer from a State sentence in the past that the state no longer obtains, unless there is additional information as in (5b) that makes this interpretation untenable. The examples show that perfective event sentences present Achievements and Accomplishments in their entirety, as discrete individuals. Now consider Activity sentences with the perfective viewpoint.
11 The neutral viewpoint appears in certain languages in sentences that lack a viewpoint morpheme.
It allows a wider range of interpretation than either perfective or imperfective viewpoints. I ignore the neutral viewpoint here (Smith 1991/1997; Chapter 4). 12 In fact progressive stative sentences occur quite often, especially in informal speech. They are marked, presenting states as dynamic: I’m really loving this walk, The river is smelling particularly bad today. Progressive statives are discussed in Smith 1983, 1991. 13 Perfective viewpoints differ across languages; as far as I know they all appear with dynamic verb constellations and present them as bounded in some way. See Section 4.2.
Activities: States or Events?
69
Activity sentences with the perfective viewpoint are flexible. They may but need not convey a situation with endpoints, that is, a terminated situation. Conjunction with assertions of continuation gives variable results depending on what is asserted. Perfective Activity sentences are not compatible with assertions of simple continuation (6); but if a new unit of activity is asserted, the conjunctions are good (7). (6)a. #Lily worked and she may still be working. (Activity) b. #The dancers rehearsed and they may still be rehearsing. (Activity) (7)a. Lily worked and continued working after that without a break. (Activity) b. The dancers rehearsed and kept on rehearsing. (Activity) Apparently sentences with the perfective viewpoint present an Activity as a temporal unit, but not necessarily as terminated. I shall suggest later that such sentences present a segment of the Activity with an implicit temporal bound. State sentences and sentences with the progressive viewpoint present unbounded situations that have the sub-interval property. State sentences and progressives differ in other respects, however (Smith 1991, 1995). The next section begins with a general discussion of aspectual information in narrative, and then goes on to look closely at perfective Activity sentences.
3 Situations in Narrative 3.1 Discourse Dynamics To talk about situations in time, speakers use the complementary linguistic subsystems of temporal location and aspect. Temporal location allows a speaker to locate a situation in time from an external point of view, ignoring its internal structure, while aspect specifies the internal temporal structure of the situation (Comrie 1976: 5). The two sub-systems come together in narrative discourse. Narrative presents situations which are bound by a unifying thread of some kind. The temporal relations between these situations constitute the dynamics of narrative. Recent work has shown that aspectual information helps to determine the advancement of narrative time. The essence of a narrative is that its situations are taken to occur in sequence. According to Labov and Waletzky “strict temporal sequence . . . is the defining characteristic of narrative” (1966: 22). Sequential interpretations are due to linguistic forms which convey that the initial endpoint of one situation follows the final endpoint of another (Heinämäki 1984). Since perfective event sentences present discrete, bounded situations, they contribute to the sequential interpretation of a narrative. Sequence is also conveyed directly by temporal adverbials, e.g., then, on May 31, after, for 3 hours, etc.
70
C.S. Smith
Narrative time advances with perfective event sentences, and fails to advance otherwise, ignoring adverbials and other cues. This is a rough, general statement of the basic finding of discourse dynamics (Hinrichs 1981, Kamp and Rohrer 1983, Partee 1984). The examples below present illustrative fragments: in each, perfective sentences of Accomplishments and Achievements advance the narrative, while perfective State and progressive sentences do not. The fragments from actual texts are identified in the Reference section at the end of this article. (8) Vite docteur, dépêchez-vous. Mon mari a pris deux cachets d’aspirine, il a avalé sa lotion contres les aigreurs d’estomac, il s’est mis un suppositoire contre la grippe, il a pris un comprimé à cause de son asthme, il s’est mis des gouttes dans le nez, et puis il a allumé une cigarette. Et alors, il y a eu une énorme explosion. (Kamp and Rohrer 1983) (9) She put on her apron, took a lump of clay from the bin and weighed off enough for a small vase. The clay was wet. Frowning, she cut the lump in half with a cheese-wire to check for air bubbles, then slammed the pieces together much harder than usual. A fleck of clay spun off and hit her forehead, just above her right eye. (Peter Robinson) (10) I slipped outside into a shock of cool air and ran down the pier. Several small boats were rocking lazily to and fro in the water. I unfastened the rope to one, paddled out toward the “Republic”, then hauled myself hand over hand up a rope ladder to the topgallant bulwark, and over onto a broad empty deck. (Charles Johnson) These examples illustrate a basic narrative pattern in which events of comparable granularity are presented either as sequential or overlapping. Of course, narrative does not consist only of sequence. There are several other possibilities. Simultaneous, overlapping, and unordered readings are available for most situations, assuming pragmatic plausibility. Moreover, situations may be presented in the reverse order of their occurrence;14 there may be a change of granularity, a flashback, or an interpolation. I shall concentrate here on sequence and how it is conveyed without adverbials in narrative. In many accounts, Reichenbach’s notion of Reference Time (1947) has been extended to account for the dynamics of narrative. The basic idea is that narrative time advances with successive Reference Times. The clauses and sentences of a narrative either introduce a new, updating Reference Time, or have a Reference Time which overlaps the current time.15
14 The
possibilities depend partly on a given language. In French, for instance, the passé simple, a perfective past tense, cannot be used to present events out of the order in which they appear. 15 The approach has been worked out within Kamp’s Discourse Representation Theory by Hinrichs (1981), Kamp and Rohrer (1983), Partee (1984), Kamp and Reyle (1993) and others. Dowty (1986) gives a somewhat different account of the advancement of Reference Time.
Activities: States or Events?
71
Aspectual information determines how a situation relates to a given Reference Time. Each perfectively presented event clause introduces a new Reference Time interval, for events on the main line of a narrative. An event falls entirely within this Reference Time (RT) interval. The relation between an event E and the RT interval it introduces can be stated in terms of times that fall within the interval and the endpoints or temporal bounds of E, notated as I and F, BI and BF respectively. (11) Relation between an event and its RT interval For a given event E, RT interval I, and times t and t which fall within I. a. For event E with intrinsic bounds (telic), t and t coincide with I and F. b. For event E with independent bounds, t and t coincide with BI and BF. The values of BI and BF are given by the adverbial which states the independent bound. Whether BI and BF coincide with the beginning and end of the event is indeterminate. State and imperfective sentences do not introduce a new Reference Time and thus do not advance a narrative.16 The situations are presented in such sentences as unbounded: they include the current Reference Time rather than falling entirely within it. This account is widely accepted as it applies to perfective sentences of telic events and of States, and to progressives and other imperfective sentences.
3.2 Activity Sentences in Narrative We now ask how perfective Activity sentences function in narrative discourse. There are competing predictions. If Activities are primarily a type of state, Activity sentences in narrative would present unbounded situations whose endpoints do not fall within a given Reference Time. They would not be expected to advance narrative time without additional information (Hinrichs 1981; Partee 1984; Dowty 1986; Herweg 1991). But other factors suggest a different prediction. Activities are dynamic situations, with arbitrary final endpoints. Perfective Activity sentences might advance a narrative. More specifically, the prediction would be that the bounds or endpoints of an Activity presented perfectively fall within a given Reference Time (Aristar Dry 1983; Hatav 1989; Smith 1983, 1991). In this section I present the results of an informal empirical study of Activity sentences in narrative contexts. Recall that sentences associated with the Activity
16 This
nicely formalizes Jespersen’s (1931) observation that the English progressive provides a temporal frame for another event. Actually the facts are more complex. Sometimes a narrative does advance with a stative or a progressive; but such cases involve additional factors and require additional inferences. Note that bounds alone do not advance narrative time: neither bounded statives, habituals, or perfects do so, although they can invite an inference of a subsequent state, which contributes to advancement.
72
C.S. Smith
concept have unique distributional and interpretive properties. They can be shown to belong to the class of Activities by testing for these properties (Dowty 1979, Smith 1991). I will not carry out such tests here, but the reader is invited to do so. (12) gives constructed examples with telic and atelic clauses. The situations chosen might plausibly occur in sequence on pragmatic grounds; some are telic events, others are Activities. The question is whether the Activity clauses (underlined) advance the sequence. (12)a. He got up, played the piano, ate breakfast, and strolled in the park. b. She ate breakfast, strolled in the park, and listened to music. c. They rehearsed, strolled in the park, and went to the movies. These sentences do convey a sequence of situations which follow each other in time. But the effect might be due to the telic clauses in the sequences. To avoid this possibility, the sequences in (13) have only Activity clauses. (13)a. She played the piano, strolled in the park, and slept. b. They rehearsed, played with the children, and strolled in the park. c. He watched television, spoke to his mother, and slept. Here too each clause contributes to temporal advancement: the natural interpretation is that they present a sequence of situations, each bounded. (Although other interpretations are available as is generally the case if they are pragmatically plausible.) I now present some natural examples from texts which show that perfective Activity sentences can contribute to the movement of narrative time. The relevant clauses are underlined. (14)a. In the Manchester airport, Clinton spoke to Hillary from a pay phone. When he hung up, he was serene and unclouded. He began campaigning with a new resolve. b. Sperling finally jumped in during a rare opening in the conversation. “Well, I think the governor is busy, and we’ve taken enough of his time”, he said. Clinton laughed. The experts kept at each other. Sperling tried again. “Let’s let the governor get back to his golf”, he said. (Bob Woodward) c. . . .And then something happened, the two dogs arrived, to lick our hands and whine and jump around us . . . We had not remembered the dogs . . . we fled into our bedroom and into bed. We giggled and shrieked with relief, and the dogs went quietly back to lie in their places in the lamplight. (Doris Lessing) d. There was a small ivory push button beside the door marked ‘405’. I pushed (. . .) and waited. The door opened noiselessly about a foot. (variant in Depraetere 1995) e. A few days later I called on Dr P. and his wife at home . . . Mrs. P showed me into a lofty apartment, which recalled fin-de-siècle Berlin. A magnificent old
Activities: States or Events?
73
Bösendorfer stood in state in the middle of the room . . . Dr. P. came in, a little bowed, and advanced with outstretched hand toward the grandfather clock, but, hearing my voice, corrected himself, and shook hands with me. We exchanged greetings and chatted of current concerts and performances. Diffidently, I asked him if he would sing. (Oliver Sacks) f. He stared at me morosely. He stood up slowly, graceful as a panther. He walked across the room and looked into my office. He jerked his head at me and went in. He was a guy who owned the place where he happened to be. I went in after him and shut the door. He stood by the desk looking round, amused. ‘You’re small-time’, he said. ‘Very small-time’. (Raymond Chandler) g. Elmer came hurrying from the accounting office and he and Mr. Renehan – hastily summoned from next door – between them carried Mary Lou upstairs. She came to on the way and struggled to her feet on the first-floor landing. She wept, in front of them at first, before turning her back on everyone and hurrying up the second flight of stairs. (William Trevor) h. wayyomer ya‘aqov le’e¸tav liq¸t’avanim wayiqh¸u ’avanim say + WAY Jacob to-brother-his gatherImpf stones take + WAY stones. Then he told his kinsmen to gather stones and they took stones wayya‘asugal; wayyoxlu šam ‘al hagal make + WAY heap; eat + WAY there on the-heap and built a cairn; and there beside the cairn they ate together (Biblical Hebrew, cited Hatav 1989)17 In most of these passages, Activities are preceded and/or followed by telic events. The advancement might be due to the telic events, so that any homogenous situation in such contexts would advance narrative time. But this is not the case. As evidence, consider the following constructed examples based on (14a–b), in which states are followed by telic events. I have replaced the relevant Activity clauses of (14a–b) with State and progressive clauses (underlined): (15)a. In the Manchester airport, Clinton was restless. When he spoke to Hillary from a pay phone, he talked furiously. b. In the Manchester airport, Clinton was worrying about the campaign. When he spoke to Hillary from a pay phone, he talked furiously. (16)a. Sperling finally jumped in during a rare opening in the conversation. “Well, I think the governor is busy, and we’ve taken enough of his time”, he said. Clinton was restless. The experts kept at each other.
17 Hatav
discusses sentences presenting events on the main story line as opposed to backgrounded situations. In these examples, Activity clauses function as events do, advancing the narrative line. The verbs have endings of the way- conjugation, a type of inflection that appears only with events on the main story line.
74
C.S. Smith
b. Sperling finally jumped in during a rare opening in the conversation. “Well, I think the governor is busy, and we’ve taken enough of his time”, he said. Clinton was sitting beside him. The experts kept at each other. In these fragments the replacement clauses do not advance narrative time, whereas their counterparts in (14a–b) do advance it. The difference is due to the dynamism of Activities and, I shall argue, the fact that perfective Activity sentences present discrete units of the situation; see Aristar-Dry 1983 for other examples. The examples show that perfective Activity sentences can contribute to narrative advancement. In contrast, perfective State sentences and progressives do not do so. This result is not surprising since narrative advancement has to do with change, and the progression of time. I return to the discussion of advancement after a detour to consider an important objection.
3.3 Another Interpretation: Activity Sentences as Inchoatives We cannot immediately conclude that the conclusion given above is correct, however. One might claim that when sentences with Activity verb constellations contribute to narrative advancement, they function as inchoatives. Inchoative clauses or sentences denote the coming about of a situation; such comings-about are telic events.18 The final endpoints of Activities in narrative sequences, in this view, are not conveyed linguistically but are supplied by pragmatic inference. The objection is very difficult to prove or disprove. Since narrative provides a very strong cue to sequential interpretation, it is difficult to show directly that a given interpretation is semantically rather than pragmatically based. We cannot distinguish between a final endpoint that is part of the semantic information conveyed by the linguistic forms, and a final endpoint supplied by inference. There is probably no conclusive answer to the inchoative objection from consideration of the narrative sequences themselves. But there is relevant evidence elsewhere: we can examine perfective Activity sentences in other contexts which lend themselves to more than one interpretation. To do this we need contexts which clearly allow an inchoative or a terminative interpretation. In such contexts, we consider Activity clauses. Inchoative readings would present an instantaneous coming about of the Activity situation; terminative readings would present a durative, bounded Activity situation that does not continue beyond the interval presented in the sentence. The finding of terminative readings for Activities would support the claim that they can be terminative in narrative contexts.
18 Inchoatives may have an explicit verb such as begin or start: the inchoative interpretation also arises for perfective State sentences and sentences of other situation types in certain triggering contexts (Moens and Steedman 1987, Smith 1993/5).
Activities: States or Events?
75
Temporal clauses are helpful as diagnostics. Before- and after-clauses require a sequential interpretation of the situations in both main and embedded clauses (Heinämäki 1974). Either inchoative or terminative interpretations would be appropriate: both contribute to a sequential interpretation, as (17) illustrates. (17)a. Mary became angry before we rehearsed. b. Sam did his homework before we played in the park. The main clauses of (17a) and (17b) are an inchoative and a completed Accomplishment, respectively; the latter has a natural final endpoint. I now ask whether perfective Activity clauses are taken as inchoative or terminative in the context of such clauses. (18)a. b. c. d.
We rehearsed before/after Mary left. Sam rehearsed after we played in the park. Mary walked in the park after she did her homework. Mary left before/after we rehearsed.
The interpretation of both clauses seem terminative in these sentences. We can test further with conjunction. It should be reasonable to conjoin such sentences with assertions of continuation, if the Activity clauses are inchoative; but if they are terminative the conjunctions should be strange. (19) and (20) present several such conjunctions with (18a) as the first conjunct: (19)a. b. c. d. (20)a. b. c. d.
#We rehearsed before Mary left but we weren’t finished when she returned. #We rehearsed before Mary left and we were still playing when she returned. #We rehearsed when Mary left but we weren’t finished when she returned. #We rehearsed when Mary left and we were still playing when she returned. #We rehearsed after Mary left but we weren’t finished when she returned. #We rehearsed after Mary left and we were still playing when she returned. #We rehearsed after Mary left but we weren’t finished when she returned. #We rehearsed after Mary left and we were still playing when she returned.
The conjunctions are strange, although the situations they present are quite plausible. The terminative reading is strong in both cases. This suggests that durative, terminative readings are natural for perfective Activity sentences. Temporal conjunctions are strong diagnostic contexts for sequence. They afford an interesting comparison between Activity clauses and State and progressive clauses. All three present homogeneous situations. If homogeneity were the only relevant feature we would expect the three types to have the same interpretation in the context of temporal conjunctions. But in fact perfective Activity clauses differ from progressive and State clauses with these and other temporal conjunctions. Progressive clauses have only an overlapping interpretation with the temporal clause event. State clauses are taken as overlapping the temporal clause event, or are
76
C.S. Smith
ambiguous between an overlapping and sequential interpretation.19 This difference suggests that Activities do not make the same contribution to temporal sequence that States and progressives do. Another kind of evidence for a terminative interpretation comes from cases where additional cues suggest that the Activity holds for an interval. In (21), for instance, there is a sentence in which the main clause and a gerundive clause present Activities (underlined). The gerundive conveys an ongoing durative situation which is simultaneous with the situation of the main clause. (21) “Pull up a chair.” Val nodded and waved hospitably, and I sat beside him, spread out the papers, and explained about the wholly level-headed artist I had spent three hours with two weeks earlier . . . “Hm”, He pondered, looking again at the drawings, which still seemed funny to me after a fortnight’s close acquaintance. “Well, I don’t know. It’s too like aiming at the moon with a bow and arrow.” (Dick Francis) Thus the gerundive shows that He pondered is not an inchoative in this example. In the contexts above, we have seen that perfective Activity sentences have a terminative interpretation. But in other cases perfective Activity sentences present Activity situations as discrete segments with implicit temporal bounds.
19 To
see this, compare the examples below, which have progressive or State main clauses and temporal conjunctions, with (18), (19) and (20). a. Mary was walking in the park before Bill left. b. Mary was singing after Bill left. c. Mary was angry before John broke the glass. In (a) and (c) we interpret the situation of the main clause as overlapping with that of the temporal clause, not as terminative; (b) conveys that the main clause situation was in progress after the event of the temporal clause, due to the lexical meaning of after. There is also a difference between Activity, State, and progressives with when-clauses: d. e. f.
We rehearsed when Mary left. We were rehearsing when Mary left. Mary was angry when John broke the glass.
In (d) the main clause situation can be inchoative or terminative; in (e) it must overlap with the event of the main clause; in (f) it may be inchoative, or overlap the event of the main clause (Smith 1983). This data suggests that the property of homogeneity is not sufficient to determine whether a situation is an event or not, pace Herweg, who claims that temporal conjunctions have the effect of turning clauses of all homogeneous situations into events. According to Herweg temporal conjunctions are “semantically restricted to event-type expressions as arguments” (1991: 976). Herweg uses German examples; differences between German and English may be responsible for the discrepancy (Schilder 1997).
Activities: States or Events?
77
3.4 The Contribution of Activities Perfective Activity sentences do not necessarily present terminated situations. The implicit bounds need not coincide with the endpoints of the Activity. Evidence comes from other narrative passages for which the terminative interpretation of an Activity is implausible, as in the following examples; the Activity clauses are underlined: (22) She turned on the light and looked at Ernest lying beside her. He was sound asleep. He snored. But even though he snored, his nose remained perfectly still. (Virginia Woolf) (23) I sipped my drink and nodded. The pulse in his lean grey throat throbbed visibly.. . . An old man two-thirds dead and still determined to believe he could take it. “Your conclusions?” he asked suddenly. (Raymond Chandler) (24) John entered the president’s office. The clock ticked loudly. (from Dowty 1986) In (22), the only plausible interpretation is that Ernest snored before and after the event of looking. Similarly, pragmatic knowledge suggests that the Activities of (22), (23) and (24) have gone on for some time and will continue. We assume that these Activities will last as long as usual, however long that may be, in the absence of information to the contrary. The point has already been made by Dowty, who notes that “. . . an activity asserted to take place at interval I could perfectly well have begun before I or continued beyond I” (1986: 53). This analysis differs from Dowty’s in claiming that perfective Activity sentences present implicitly bounded segments and that they contribute to narrative advancement. The possibility of continuation is quite compatible with a segment of a situation that is implicitly bounded. The segment is a temporal unit: there is no information about beginning or ending. If the situation continues after the implicit bound, the continuation constitutes another temporal unit. The point is delicate but important for this analysis. We saw that assertions of continuation may felicitously be made in (5c–d), repeated here as (25a–b), with an additional example (25c). (25)a. Mary worked and continued working without a break. b. Sam rehearsed and kept on rehearsing. c. Alice thought about the math problem while Chris drove her to the station, and she kept on thinking about it all morning. The second conjunct of each sentence asserts that another segment, or temporal unit, follows the unit presented in the first conjunct. These examples show that temporal boundedness does not require termination. I noted above that the endpoints of an explicitly bounded event need not coincide
78
C.S. Smith
with the stated bound: in example (2) the events do not necessarily terminate. The same should hold for Activities with implicit bounds, and it does. Not surprisingly, interpretations differ for sentences in isolation and in context. Recall that in isolation, perfective Activity sentences tend to be interpreted as presenting terminated events. This is the force of the infelicitous conjunctions which assert simple continuation (5a–b). Yet we have seen that in narrative contexts the same type of sentence is not taken as terminative. The difference is due to a general pattern for how incomplete stimuli of all kinds are interpreted (Kanizsa 1976). In isolation, sentences are often semantically and pragmatically incomplete. People tend to give as complete as possible interpretations of such sentences (Smith 1977), so that termination is preferred. Continuing to explore the contribution of perfective Activity sentences to narrative, I have suggested that pragmatic knowledge determines the implicit bound of an Activity in a perfective sentence. To give content to this claim, I note some typical patterns. Whether the Activity is terminated, and the temporal extent of the unit presented, is often clear from information in the context in which a sentence appears. In one pattern an Activity is followed by a telic event. Several such examples appeared in (14), repeated here as (26). All have a strong sequential interpretation: the telic event in effect terminates the Activity. (26)a. There was a small ivory button beside the door marked ‘405’. I pushed it (. . .) and waited. The door opened noiselessly about a foot. b. In the Manchester airport, Clinton spoke to Hillary from a pay phone. When he hung up, he was serene and unclouded. He began campaigning with a new resolve. c. Dozing a little, Alleyn sat slumped forward in his seat. A violent jerk woke him. The sequence interpretation involves an inference that the Activity terminates. Pragmatic knowledge of how events normally relate to each other is essential here, of course. Other patterns consist of Activity situations in sequence, and Activities that overlap in time. (27) illustrates both patterns with an example from (14) in which three Activity situations occur (only one was noted above). (27) Sperling finally jumped in during a rare opening in the conversation. “Well, I think the governor is busy, and we’ve taken enough of his time”, he said. Clinton laughed. The experts kept at each other. Sperling tried again. “Let’s let the governor get back to his golf”, he said. (Bob Woodward) The sequential interpretation is that two situations occurred after Sperling’s first remark, Clinton’s laughing and Sperling’s second try. The overlapping interpretation is that the experts kept at each other during the period in which the two situations occurred.
Activities: States or Events?
79
Simultaneous Activities are also possible in narrative passages, as in this wellknown example from Hinrichs 1981. (28) Jaime was building another boat. He sang happily as he worked, the muscles of his brown arms rippled in the sun, and crispy wood shavings made a carpet between his bare feet and the sand. Recall too the constructed fragments presented in (12) and (13), in which Activities had a sequential and/or simultaneous interpretation. (The possibility is not limited to Activities: telic events may also allow non-sequential interpretations). Perceived situations, however, have a different pattern. They are not interpreted in terms of the situations themselves but according to the event of perception, as noted in Dowty 1986. The time required for the percept determines the temporal unit of a perfective Activity sentence, for instance in the examples (22) and (23) above, repeated here: (22) She turned on the light and looked at Ernest lying beside her. He was sound asleep. He snored. But even though he snored, his nose remained perfectly still. (Virginia Woolf) (23) I sipped my drink and nodded. The pulse in his lean grey throat throbbed visibly.. . . An old man two-thirds dead and still determined to believe he could take it. “Your conclusions?” he asked suddenly. (Raymond Chandler) We vicariously experience the narrator’s perceiving that Ernest is snoring and that the pulse is throbbing in the old man’s throat. For pragmatic reasons we infer that these situations continue before and after the unit of time presented in the sentences. Perceived events advance a narrative by the time in which the narrator grasps the situation. Such a time has nothing to do with the temporal duration of the situation itself, of course. Summarizing, perfective Activity sentences present a situation as a temporal unit, or segment. In discourse dynamic terms, such sentences introduce a Reference Time interval and the temporal unit of the Activity falls within the interval. The relation between such an Activity situation E and the Reference Time it introduces is like that given in (11b) for independently bounded events, except that in this case the initial and final temporal bounds are implicit, notated as IBI and IBF: (29) Relation between a perfectively presented Activity and RT For a given Activity E, RT interval I, and times t and t which fall within I: t and t coincide with IBI and IBF. The values of IBI and IBF are determined by context and world-knowledge, as we have seen. Whether or not IBI and IBF coincide with the beginning and ending of the Activity is indeterminate; this is similar to the indeterminacy of explicit temporal bounds as stated in (11b).
80
C.S. Smith
In this section I have shown that perfective Activity sentences contribute to narrative advancement, as do Achievements and Accomplishments. Perfective sentences present all three dynamic situation types as bounded. The bounds of telic and explicitly bounded situations are intrinsic or independent, while the bounds of Activities are implicit. But neither perfective State sentences nor progressives advance a narrative. This result suggests that both dynamism and boundedness are necessary for the advancement of narrative time.
4 The Semantic Analysis of Activities The discussion above has shown that perfective Activity sentences present implicitly bounded segments. The segments advance narrative time and thus function as discrete units. I now consider how to integrate this finding with the concept of an Activity and with the perfective viewpoint. Across languages the perfective viewpoint presents dynamic situations as bounded units. There are differences in the type of bound, according to the situation type and the language. The French and Russian perfective viewpoints present telic events as complete with intrinsic bounds. In Mandarin Chinese, however, the perfective viewpoint morpheme conveys implicit termination; there are separate verb complements which unequivocally convey completion (Smith 1991). Activities also vary with the perfective viewpoint. Russian Activity sentences must appear with certain perfective prefixes which lexically provide independent bounds.20 Polish has similar perfective Activity prefixes (Piñon 1993). The finding that the English perfective presents Activities with implicit temporal bounds is consistent with this picture. The Mandarin perfective morpheme also conveys implicit temporal bounds with Activity sentences. 21 20 There
are five such prefixes; they form perfective verb stems for Activity verb constellations. For instance, the ‘perdurative’ prefix pro- indicates limited duration; it appears with a time expression in the accusative case. The ‘delimitative’ prefix po- indicates a shorter period than expected, optionally with a time expression; a. b.
Ona prostojala na uglu celyj cˇ as. She stoodPerf on the corner for an entire hour. On porabotal (ˇcasok). He workedPerf a bit (for an hour).
Activity verb constellations require perfective prefixes of this class. Although the examples present situations that are specifically bounded, the situations are otherwise unchanged. As Flier puts it, “It is impossible to reconcile delimitation with any sort of net change or result” (1984: 45); the ‘pofective’ analysis of Galton (1984) makes the same point. 21 For State sentences and perfective viewpoints, languages vary considerably (Smith 1991). In English the only unmarked viewpoint for State sentences is perfective, which presents States as unbounded. In Russian states require the imperfective; in Mandarin Chinese and Navajo they require the neutral viewpoint. In French and other Romance languages, both the perfective and
Activities: States or Events?
81
4.1 The Activity Concept The concept of an Activity introduced above has the distinguishing features [Atelic], [Dynamic], [Durative]. Since they are atelic, Activities are homogenous. Suppose that Activities were classified with states. The interpretation of perfective Activity sentences as temporally bounded would be due entirely to the perfective viewpoint. That the bound of an Activity may be terminative would be due entirely to pragmatic considerations. This approach can account mechanically for the facts adduced above. But it has disadvantages. For one thing, there is too much disparity between the concept of an Activity and its realization in perfective sentences. The cases in which perfective Activity sentences present terminated situations are quite unexpected if the concept of an Activity has no endpoints. For another, on this view the contribution of perfective Activity sentences to narrative advancement is unexpected. Even in this view the concepts of the two situation types differ significantly; since Activities are dynamic and States static, the former would have more structure, namely successive stages; Schilder (1997) offers such an account in a slightly different connection. More satisfactory is an analysis in which the concept of an Activity includes potential endpoints. The feature [Atelic] characterizes the final endpoint as arbitrary. Note that the concept of a State lacks endpoints altogether due to the feature [Static]; the beginning and ending of a state are due to independent events, changes of state. The potential endpoints of Activities derive from the feature of dynamism. Dynamic situations are contingent upon energy. In the world that we know energy requires a source, and we expect dynamic situations to begin and end with the onset and cessation of energy. In this sense the dynamism of a situation brings with it the assumption of an initial endpoint, and the possibility of an eventual final endpoint. The expectation of a final endpoint is not always met – Activities may continue indefinitely, in fact or in imagination22 – but potential endpoints are part of the concept. This approach to dynamism and energy is reminiscent of some ideas in naïve physics, or commonsense knowledge, as Manfred Krifka has pointed out to me.23
imperfective viewpoints are available in the past tenses for all situation types; the perfective conveys termination for States. 22 Though the expectation of endpoints is not always met. For instance, as a reviewer points out, it is quite possible to say that the universe rotates eternally and mythical gods drink eternally. I would argue that the statements controvert standard default expectations. 23 Energy in this sense is close to the notion of dynamism, and is one of the deepest properties that humans recognize in the situations of the world. Not surprisingly, it is reflected in people’s concepts and their realization in language. There is evidence that infants have a cognitive correlate of energy, as manifested by mechanical forces in the world (Leslie 1994). Indeed, infants apparently use from early on a principle which Gelman characterizes as ‘Attend to the source of energy’. This enables them to differentiate between objects with internal sources of energy from those that are made to move by something else (Gelman 1990). Leslie, following Talmy (1988), argues that mechanical roles and relations are reflected directly in much of the verb-argument structure of natural language. The notion of energy cannot be reduced to spatiotemporal patterns. On precisely this point,
82
C.S. Smith
Recent experimental work has shown that energy is an essential factor in human perception and categorization. And there is evidence that adults and infants understand the mechanical properties of energy sources in a way that cannot be reduced to spatiotemporal patterns. The evidence comes from experiments in which subjects are indirectly shown to grasp the properties of launching events (Leslie 1994). The concept of an Activity as a dynamic event with potential endpoints is consistent with commonsense notions. The perfective viewpoint presents Activities with implicit temporal bounds which may but need not coincide with its endpoints. In earlier work I proposed that the perfective viewpoint presents sentences of all situation types with their characteristic endpoint features (Smith 1991), but the investigation here shows that this is incorrect. The Activity concept has potential endpoints only. In Activity sentences the perfective viewpoint actually adds information, conveying a temporally bounded unit. We have seen that the temporal bound may or may not coincide with the endpoints of the Activity. Other approaches to the concept of an Activity have recognized their potential for an extrinsic bound and/or arbitrary final endpoint. For instance, Kamp and Reyle analyze Activity verbs as “incomplete”; but note that such verbs allow a bound imposed by a complement or adverbial (1993: 564). If we recognize that the perfective viewpoint also has the effect of imposing a bound on Activities, the analysis of Kamp and Reyle is close to the one proposed here.
4.2 Dynamism, Conventional Time, and Narrative Time Dynamic situations unfold in time, with each successive stage involving a change of some kind. This can be modeled by mapping the stages of dynamic situations into successive times. To implement the mappings of stages to times, we must identify the stages of dynamic situations, including Activities. Some Activities have clearly recognizable stages: breathing, walking and chuckling, for instance, consist of individual breaths, steps, and chuckles. But other Activities, such as playing in the sand, dreaming, enjoying, are harder to organize into distinct stages. Still others vary with the participants. Swimming has clear stages for humans, depending on the type of stroke employed, but it’s not clear that stages can be identified if the creature swimming is a fish. The sub-parts of an Activity may not correspond to successive stages in any clear way, which makes the mapping problematic. The same is true for other situation types. The units of conventional time provide a way to handle this problem. They nicely allow a modelling of dynamism without complete dependence on the particulars of
Leslie claims that infants understand the mechanical, and not just the spatiotemporal properties of such events. His evidence comes from experiments concerning infants’ grasp of launching and the notion of agency (1994). Pascal Boyer kindly provided me with references to this material.
Activities: States or Events?
83
situations.24 The successive stages of dynamic situations can be identified with units of conventional time. In case where there is no clear relation between the structure of the event and conventional time, the successive stages are imposed entirely by conventional time. The units of a given mapping are determined by the granularity of the narrative in question. For instance, units of different sizes figure in the narrative passages of this article. There are rather delicate moments of perception, e.g., the throb of pulse in the old man’s throat; and larger units, e.g., a ride and a swim. The units of temporal mapping vary considerably, but provide the temporal beat of narrative through the successive stages of dynamic situations. In the same manner, conventional time can model narrative advancement. Recall that advancement takes place as Reference Time changes: new Reference Times are triggered by events or Activity segments. Each new Reference Time would be identified with a unit of conventional time. This would allow the simplifying assumption that narrative and conventional time are of the same kind (cf. Friedman 1990). The notion of conventional time is implied anyway in narrative advancement. Recall that Reference Time may advance due to a temporal adverbial; to an event that is independently or intrinsically bounded; to an implicitly bounded event (ignoring other factors). The first two by definition involve conventional time; the other two imply it. Narrative has a temporal beat which is based in the successive stages of its events. The notion of temporal beat may explain another kind of time found in narrative. The receiver is often aware of the passing of time, relative to knowledge of how time moves in the world. Consider (10) for instance, repeated here as (30). (30) (a) I slipped outside into a shock of cool air and ran down the pier. (b) Several small boats were rocking lazily to and fro in the water. (c) I unfastened the rope to one, paddled out toward the “Republic”, then hauled myself hand over hand up a rope ladder to the topgallant bulwark, and over onto a broad empty deck. In this short narrative there is a sense of time passing: from (a) we infer that the narrator runs along the pier toward the water; time continues to pass in (b) as he reports the percept of boats rocking in the water, and continues with the succession of events in (c). This sense of time passing is due in part to the beat of conventional time which underlies narrative. Conventional time continues to pass, even without the clear jumps of pragmatic advancement. The dynamism of events is always present in narrative at the time line.
5 Conclusion I have shown that perfective Activity sentences present temporally bounded units of Activity situations, contributing to the advancement of narrative time. At the beginning of this article I raised the question of how best to subdivide the situation 24 I
thank Nicolas Asher for helpful discussion of this point.
84
C.S. Smith
types, suggesting that narrative would provide evidence. The predictions of the strong mereological view are not borne out: Activities pattern with events rather than states in sequential contexts. Imperfective event and Activity sentences do not advance time; perfective event and Activity sentences do advance it. To restate the conclusion arrived at above, both boundedness and dynamism are necessary for the advancement of time. Activities cannot be assimilated to the class of events: they are homogenous rather than heterogeneous, and they have only the potential for endpoints. Yet Activity sentences form a natural class with full-fledged event sentences on the basis of their dynamism and their behavior in narrative and sequential contexts. The perfective viewpoint presents all dynamic situations as bounded. The bound may be intrinsically part of the situation concept, as with sentences of telic events. It may be independent and explicit, as in sentences with temporal adverbials. Or the bound may be implicit, as in perfective Activity sentences. Acknowledgments I would like to thank Sheila Glasbey and Manfred Krifka for comments on an earlier version of this article. I also thank three anonymous reviewers for their critiques.
References Aristar-Dry, H.: 1983, ‘The Movement of Narrative Time’, Journal of Literary Semantics 12, 19–53. Bach, E.: 1981, ‘On Time, Tense and Aspect: An Essay Concerning English Metaphysics’, in P. Cole (ed.), Radical Pragmatics, Academic Press, New York, pp. 63–81. Bach, E.: 1986, ‘The Algebra of Events’, Linguistics and Philosophy 9, 5–16. Bennett, M. and B. Partee: 1972, Toward the Logic of Tense and Aspect in English, System Development Corporation, Santa Monica, California. Comrie, B.: 1976, Aspect, Cambridge University Press, Cambridge. Depraetere, I.: 1995, ‘On the Necessity of Distinguishing Between (Un)boundedness and (A)telicity’, Linguistics and Philosophy 18, 1–19. Dowty, D.: 1979, Word Meaning and Montague Grammar, Reidel, Dordrecht. Dowty, D.: 1986, ‘The Effects of Aspectual Class on the Temporal Structure of Discourse: Semantics of Pragmatics?’, Linguistics and Philosophy 5, 23–33. Flier, M.: 1985, ‘The Scope of Prefixal Delimitation in Russian’, in M. Flier and A. Timberlake (eds.), The Scope of Slavic Aspect, UCLA Slavic Studies, Volume 12, Slavica Press, Columbus, Ohio. Friedman, W.: 1990, About Time, MIT Press, Cambridge, Mass. Galton, A.: 1984, The Logic of Aspect: An Axiomatic Approach, Clarendon Press, Oxford. Gelman, R.: 1990, ‘First Principles Organize Attention to and Learning about Relevant Data: Number and the Animate/Inanimate Distinction’, Cogmtive Science 14, 79–106. Hatav, G.: 1989, ‘Aspects, Aktionsarten, and the Time Line’, Linguistics 27, 487–516. Heinämäki, O.: 1974, The Semantics of English Temporal Connectives, Ph.D. dissertation, University of Texas. Heinämäki, O.: 1984, ‘Aspect in Finnish’, in C. deGroot and H. Tommalo (eds.), Aspect Bound, Foris, Dordrecht. Herweg, M.: 1991, ‘Perfective and Imperfective Aspect and the Theory of Events and States’, Linguistics 29, 969–1010. Hinrichs, E.: 1981/1986, ‘Temporal Anaphora in Discourse’, Linguistics and Philosophy 9, 63–82. Jesperson, O.: 1931, A Modern English Grammar on Historical Principles, Part IV, Allen and Unwin, London.
Activities: States or Events?
85
Kamp, H. and C. Rohrer: 1983, ‘Tense in Texts’, in R. Bauerle et al. (eds.), Meaning, Use, and Interpretation of Language, de Gruyter, Berlin. Kamp, H. and U. Reyle: 1993, From Discourse to Logic, Kluwer, Dordrecht. Kanizsa, G.: 1976, Subjective Contours. Scientific American, 234, 48–52. Kenny, A.: 1963, Action, Emotion, and Will, Humanities Press, New York. Krifka, M.: 1989, ‘Nominal Reference, Temporal Constitution and Quantification in Event Semantics’, in R. Bartsch et al. (eds.), Semantics and Contextual Expressions, Foris, Dordrecht. Krifka, M.: 1992, ‘Thematic Relations as Links between Nominal Reference and Temporal Constitution’, in I. Sag and A. Szabolcsi (eds.), Lexical Matters, Chicago University Press, Chicago. Labov, W. and J. Waletzky: 1966, ‘Narrative Analysis: Oral Versions of Personal Experience’, in J. Helm (ed.), Essays on the Verbal and Visual Arts. Proceedings of the 1966 Annual Meeting of the American Ethnological Association, Seattle, University of Washington Press. Leslie, A. M.: 1988, ‘The Necessity of Illusion: Perception and Thought in Infancy’, in L. Weiskrantz (ed.), Thought without Language, Oxford Science Publications, Oxford. Leslie, A. M.: 1994, ‘ToMM, ToBy, and Agency: Core architecture and Domain Specificity’, in L. Hirschfield and S. Gelman (eds.), Mapping the Mind: Domain Specificity in Cognition and Culture, Cambridge University Press, Cambridge. Mittwoch, A.: 1991, ‘In Defense of Vendler’s Achievements’, Belgian Journal of Linguistics 71–85. Moens, M. and M. Steedman: 1987, ‘Temporal Ontology in Natural Language’, in Proceedings of the 25th Annual Conference of the Association for Computational Linguistics, Stanford, California. Mourelatos, A.: 1978, ‘Events, Processes, and States’, Linguistics and Philosophy 2, 415–434. Mourelatos, A.P.D.: 1993, The Pre-Socratics. Princeton, NJ: Princeton University Press. Parson, T. 1989, ‘The Progressive in English: Events, States and Processes’, Linguistics and Philosophy, 12, 213–41. Parsons, T.: 1990, Events in the Semantics of English, MIT Press, Cambridge, Mass. Partee, B.: 1984, ‘Nominal and Temporal Anaphora’, Linguistics and Philosophy 7, 243–286. Piñon, C.: 1993, ‘Aspectual Composition and the “Pofective” in Polish’, unpublished paper, read at the Slavic Linguistics Conference, MIT. Piñon, C.: 1995, An Ontology for Event Semantics, Ph.D. dissertation, Stanford University. Pustejovsky, J.: 1991, ‘The Syntax of Event Structure’, Cognition 41, 47–81. Reichenbach, H.: 1947, Elements of Symbolic Logic, Macmillan, London. Schilder, F.: 1997, Temporal Relations in English and German Narrative Discourse, Ph.D. dissertation, The University of Edinburgh. Smith, C. S.: 1977, ‘The Vagueness of Sentences in Isolation’, in W. Beach et al. (eds.), Proceedings of the 13th Meeting of the Chicago Linguistic Society, Chicago Linguistic Society, University of Chicago. Smith, C. S.: 1983, ‘A Theory of Aspectual Choice’, Language 59, 479–501. Smith, C. S.: 1991/1997, The Parameter of Aspect. Dordrecht: Kluwer. 2nd edition, revised. Smith, C. S.: 1993/5, ‘The Range of Aspectual Situation Types: Shifts and a Bounding Paradox’, in P. Bertinetto et al. (eds.), Temporal Reference: Aspect and Actionality, Rosenberg and Sellier, Turin. Smith, C. S.: 1995, ‘The Relation between Aspectual Viewpoint and Situation Type’, Address to the Linguistic Society of America; Eric electronic database. Taylor, B.: 1977, ‘Tense and Continuity’, Linguistics and Philosophy 1, 199–220. Vendler, Z.: 1957, ‘Verbs and Times’, The Philosophical Review; reprinted 1967, Linguistics in Philosophy, Cornell University Press, Ithaca, New York. Verkuyl, H.: 1993, A Theory of Aspectuality, Cambridge University Press, Cambridge. Vlach, F.: 1981, ‘The Semantics of the Progressive’, in P. Tedeschi and A. Zaenen (eds.), Syntax and Semantics 14: Tense and Aspect, Academic Press, New York.
86
C.S. Smith
Vlach, F.: 1993, ‘Temporal Adverbials, Tenses, and the Perfect’, Linguistics and Philosophy 16, 231–283. Whorf, B.: 1956, ‘Grammatical Categories’, in J. Carroll (ed.), Language, Thought, and Reality, Selected Writings of Benjamin Lee Whorf, MIT Press and John Wiley, New York.
Example Sources: Raymond Chandler, The Long Good-bye Dick Francis, Banker Charles Johnson, Middle Passage James Joyce, Dubliners Doris Lessing, Under My Skin Ngaio Marsh, Vintage Murder Peter Robinson, A Necessary End Oliver Sacks, The Man Who Mistook His Wife for a Hat William Trevor, Two Lives Eudora Welty, Delta Wedding Robert Woodward, The Agenda Virginia Woolf, Lappin and Lapinova
The Syntax and Interpretation of Temporal Expressions in English Carlota S. Smith
Abstract The only obligatory temporal expression in English is tense, yet Hans Reichenbach (1947) has argued convincingly that the simplest sentence is understood in terms of three temporal notions. Additional possibilities for a simple sentence are limited: English sentences have one time adverbial each. It is not immediately clear how to resolve these matters, that is, how (if at all) Reichenbach’s account can be reconciled with the facts of English. This paper attempts to show that they can be reconciled, and presents an analysis of temporal specification that is based directly on Reichenbach’s account. Part I is devoted to a study of the way the three times – speech time, reference time, event time – are realized and interpreted. The relevant syntactic structures and their interaction and interpretation are examined in detail. Part II discusses how a grammar should deal with time specification, and proposes a set of interpretive rules. The study offers an analysis of simple sentences, sentences with complements, and habitual sentences. It is shown that tense and adverbials function differently, depending on the structure in which they appear. The temporal system is relational: the orientation and values of temporal expressions are not fixed, but their relational values are consistent. This consistency allows the statement of principles of interpretation. An interesting result of the study is that the domain of temporal specification is shown to be larger than a sentence. Sentences that are independent syntactically may be dependent on other sentences for a complete temporal interpretation; complements may be dependent on sentences other than their matrix sentences. Time adverbials and tense may be shared, in the sense that a temporal expression in one sentence may contribute to the interpretation of another sentence. These facts have important consequences: only a grammar with surface structure interpretation rules can account for temporal specification in a unified manner, because more than one sentence may be involved. Context is thus shown to be crucial for the temporal interpretation of sentences.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_4,
95
96
C.S. Smith
Part I – The Temporal System of English 1 Temporal Interpretation of Simple Sentences In this section I discuss how the notions essential to temporal interpretation are conveyed in simple sentences. In particular, I ask what elements of sentences correspond to speech time, reference time, and event time. The answer to this question shows that there is no contradiction between the traditional view that sentences have one time adverbial, and the scheme for temporal specification in which three separate times are distinguished. Both are correct, as I hope to demonstrate. The demonstration will bring out the fact that time specification is an area in which semantic and syntactic structures are quite different. In a sense, what follows constitutes an argument for a grammar with an autonomous syntax, although I will not be directly concerned with how a grammar can best deal with this material until the second part of the paper.1 According to Reichenbach,2 temporal specification involves three notions of time: Speech Time, Reference Time, and Event Time. Speech Time (ST) is the time at which a given sentence is uttered, that is, the moment of utterance. Reference Time (RT) is the time indicated by a sentence, which need not be the same as ST. Event Time (ET) refers to the moment at which the relevant event or state occurs, which need not be the same as RT. For instance, in (1) ET and RT are the same, and are prior to ST: (1) Marilyn won the prize last week. In (2) all three times are different: (2) Marilyn had already won the prize last week. For the second example ST is the moment of utterance, RT is last week, and ET is an unspecified time prior to last week. To understand the temporal specification of a sentence, one must know the values of the three times, and their relations to each other. The relations of sequence and simultaneity are basic to the system. Two times may be simultaneous, or one may precede the other: RT may but need not be simultaneous with ST, and ET may but need not be simultaneous with RT.
1 The discussion is limited to tense, time adverbials, frequency adverbials, and auxiliary have. Not included in the analysis at this time are conditional and contrary-to-fact sentences, aspect, modals, relative clauses, temporal NPs such as former president, and adverbial sentences. I would like to thank Lauri Karttunen, and the other members of the informal syntax discussion group of the University of Texas, Department of Linguistics, for helpful discussions of the topics covered in this paper. 2 Reichenbach outlines his system in Reichenbach (1947).
Temporal Expressions in English
97
Speech Time is the keystone of the system, in that Reference Time is oriented to it. When RT is simultaneous with ST, RT indicates Present time; when RT precedes ST, it indicates Past time; when RT follows ST, it indicates Future time. Only Reference Time is actually specified in independent sentences. Event Time is not specified if it differs from Reference Time, but the relation between the two is given. Since RT is oriented to ST, it need not be specified. In support of these statements, consider how the following English sentences are interpreted temporally: (3) (4) (5) (6) (7) (8)
Allan swam at midnight. Tim mowed the lawn yesterday. Joe wrecked the car before evening. We fixed the hammock before Todd left. The boys had already eaten dinner. Mary left the party before the guest of honor arrived after she had spilled coffee on her dress.
All of these sentences are interpreted as referring to the Past; the way the three reference periods are established will be discussed below. Close consideration of the examples shows that they specify RT, and the relation between ET and RT, but they do not specify an ET that differs from RT. To see this, consider the information one may draw from these sentences. In (5), for instance, one does not know when Joe actually wrecked the car: only that he did so before a specified time, evening (RT). In (8), although the time adverbial is complex, it indicates only one time as RT. Complex adverbials do not specify both ET and RT, but rather give a detailed specification of RT. These sentences are typical of non-habituals in English. Although only RT is specified, they are not felt to be incomplete. It is sufficient for interpretation to know RT and the relation between ET and RT; the relation between ST and RT is evident if one knows RT.
1.1 Relational Values The absolute values of temporal expressions change, and their functions differ, according to the syntactic configuration in which they occur. However, the relational values of the temporal expressions are consistent. This consistency is an essential part of the temporal system of English, it will be shown, and underlies the analysis presented here. The relational values of temporal expressions mirror the three relations possible among times, simultaneity and sequence. Temporal expressions fall into three classes according to whether they indicate simultaneity, anteriority, or posteriority. Consider first the relational values of adverbials. Certain adverbials have explicit relations with the moment of speech: thus right now is simultaneous with ST, yesterday precedes ST, tomorrow follows ST. These will be referred to as explicitly
98
C.S. Smith
Past, Present or Future, following current usage. Other adverbials are not anchored to a particular point and will be referred to as Unanchored: for instance, on Tuesday or in March may indicate a time that precedes or follows ST. These are the classes of temporal adverbials (frequency adverbs are not relevant to this classification; they are taken up in Section 3 below). Classification of Adverbials ANTERIOR (←) yesterday;—ago; last— Explicitly Past on Tuesday; in April; etc. Unanchored SIMULTANEOUS ( = ) now; right now; at this moment; Ø Explicitly Present POSTERIOR(→) tomorrow; next—; in— Explicitly Future on Tuesday; in April; etc. Unanchored Explicitly anchored adverbials are oriented to the present moment. For instance, the times of (9), (10) and (11) are computed in relation to ST, although all indicate a Past RT. (9) Arthur borrowed the canoe yesterday. (10) They sold the house 3 weeks ago. (11) Gwen passed the examination last month. For some speakers, anchored adverbials are always oriented to ST; for others, they can also occur in dependent sentences with a different orientation. See Section 2 for further discussion of anchored adverbials. The other temporal expressions of English are prepositions, tenses, and auxiliary have. The traditional meaning for auxiliary have is anteriority; I have argued elsewhere (Smith, 1976b) that this is its relational value in English, and that it indicates that ET is anterior to RT. Prepositions have the relational values that correspond to their lexical meanings: e.g. before indicates anteriority, at indicates simultaneity, after indicates posteriority. The relational values of temporal expressions in English are given below. ANTERIORITY (←) past tense Past adverbial Unanchored adverbial before, etc. auxiliary have Past RT
SIMULTANEITY ( = ) present tense Present adverbial zero (Ø) adverbial at, on, Ø, etc.
Future adverbial Unanchored adverbial after, etc.
Present RT
Future RT
POSTERIORITY (→)
In the analysis to be developed below, the relational meanings of temporal expressions are referred to frequently.
Temporal Expressions in English
99
1.2 Past, Present, Future Reference Time The question of how RT is established can now be taken up. I will show that it is the COMBINATION of tense and adverbial that establishes RT. Some combinations establish RT and others do not; however, all combinations occur somewhere in English. Sentences that do not have RT cannot be fully interpreted without additional information. The possible combinations of tense and adverbial are these: tense present present present present
adverbial Present Past Future Unanchored
tense past past past past
adverbial Present Past Future Unanchored
The combinations of tense and adverbial that establish RT have compatible relational values, whereas non-RT combinations have contradictory relational values. For instance, past tense and Future adverbials have the values ← and → and do not establish RT in combination with each other. I now give the RT combinations, with examples: tense present present present past past
adverb Present Future Unanchored Past Unanchored
RT Present Future Future Past Past
(12) (13) (14) (15) (16)
I am playing now.3 Chris is working tomorrow. Emily leaves on Thursday. Scott won the race a week ago. I won the race on Tuesday.
The following combinations do not establish RT: past past past have present
Future Present Unanchored Past
(17) (18) (19) (20)
Ross was leaving in 3 days. Emily was annoyed now. Ross had left on Tuesday.4 Last week, Todd accidentally stumbles on a snail.
Sentences (17), (18) and (19) cannot be interpreted without additional information, which shows that they do not establish RT. (17) tells us that Ross was going to leave 3 days from some time, but no more; (19) gives no anchor for the Tuesday on which Ross left; (18) tells us that Emily was annoyed at some time, but not more. What is needed in each case is a point of reference: that is, an RT. In the following section I will show that sentences like (17), (18) and (19) occur as complements dependent on sentences other than the matrix, and as independent sentences that depend semantically on other sentences. 3 In this section I discuss only time adverbials that are not habitual, so that frequency adverbials are excluded. Frequency adverbials are discussed in Section 3. 4 With a different word choice, the sentence would probably be taken as ambiguous. For instance, I left on Thursday or I was leaving on Thursday might be taken as a Future-in-Past, that lacked RT. See below for this second interpretation; it is facilitated by imperfect aspect.
100
C.S. Smith
The Historical Present is exemplified by (20). Although the Historical Present is not semantically dependent on other sentences as are (17), (18) and (19), it appears to be dependent on context in ways not yet understood. I will not deal with the Historical Present in this paper. Sentences with tense alone are incomplete semantically. The reference period is not unambiguous, and the relation between ET and RT is not specified. Since they specify so little, sentences with tense alone may be interpreted in more than one way, depending on the context in which they occur. For instance, consider the interpretation of (21): (21) Albert is playing tennis a Something unusual is scheduled for tomorrow: Albert is playing tennis. b We can’t discuss the problem now: Albert is playing tennis. Depending on the context, (21) is taken as having either Present or Future RT. The same kind of vagueness and flexibility can be seen with (22): (22) Albert was playing tennis a I saw him yesterday afternoon: Albert was playing tennis. b The plans for the following day were made: Albert was playing tennis. In (22a), (22) establishes a Past RT, whereas in (22b) the same sentence does not establish RT and represents what has been called a Future-in-Past. Detailed interpretations of sentences of this type will be given in Section 2; what I wish to establish here is that sentences with tense alone are incomplete semantically in that they do not establish RT. Time adverbials may be simple or complex. Complex time adverbials may have prepositional phrases, embedded sentences, or both; in principle, they may have infinitely many of each. [For discussion of adverbials, see Crystal (1966) or Leech (1969).] Complex time adverbials are single units in temporal interpretation, even though they need not occur as a unit in surface structure. Compare, for instance, the interpretation of (23), (24) and (25): (23) Bill arrived at 10 o’clock. (24) Bill arrived at 10 o’clock in the morning last Wednesday. (25) Last Wednesday, Bill arrived at 10 o’clock in the morning. There is only one RT for all of these sentences, that specified by the adverbial in combination with past tense. The complex adverbial of (25) specifies RT more precisely than does that of (24) or (23), but only RT. Adverbials have a double function in the sentences discussed above. The adverb contributes to the specification of RT, as indicated; the introductory preposition also plays an important role. The preposition gives the relation between ET and RT, according to its relational value. If the relational value of a preposition is ←, for instance, an adverbial introduced by that preposition indicates that ET precedes RT as well as contributing to the specification of RT. For instance: (26) Phyllis decorated the cake before midday.
Temporal Expressions in English
101
In this sentence, RT is Past, midday; ET precedes RT as indicated by before. Note that although an adverb and its introductory preposition form a constituent syntactically, they have different functions semantically. The adverb establishes RT in conjunction with tense, the preposition gives the relation between ET and RT. If an adverb lacks an introductory preposition, the relation between ET and RT is taken to be simultaneous. The reader has probably noticed that will is not treated as a tense in this analysis. There are several strong arguments against such a treatment. First, note that will can occur with Present and Past, as well as Future, sentences: (27) The store will have your book by now. (28) The documents will have arrived last week. If will were to be treated as a future, it would be necessary to set up at least one other will to account for sentences such as (27) and (28); but this would be undesirable, since all have the same predictive meaning. Moreover, will is not the only predictive form that appears in Future and other RTs; there is no reason to give it a status different from may, for instance. The syntactic complexities of will-deletion, a transformational rule which would be required if will were a future tense, also make the analysis of will as tense dubious. (For discussion of this question, see Braroe (1974), Jenkins (1972), Lakoff (1969), Smith (1975a, 1976b).) Will and other modals occur with present and past tense. With present tense, modals have Future or Present RT, depending on the adverb with which they occur. They are in this respect like other verbs: (29) Allan will be in Colorado now. (30) Mary will be in the Valley tomorrow. However, the distribution of modals with past tense is more complicated: modals do not appear with past tense in independent sentences. (31), for instance, is grammatical only if taken as semantically incomplete and dependent on another sentence. (31) John would work tomorrow. (I ignore here the contrary-to-fact uses of these forms.) The fact that (31) is incomplete semantically is predicted by the analysis here: according to the analysis, the combination of tense and adverbial of a sentence establish RT, and combinations that do not establish it are incomplete. In (31) the adverbial cannot specify RT; the sentence can be interpreted only if RT is established in another sentence. In other words, the sentence is incomplete not because of the modal but because of the combination of tense and adverbial. So far, I have established that reference times are indicated by combinations of tense and adverbial. In such combinations, the adverbial contributes to the specification of RT and gives the relation between ET and RT.
102
C.S. Smith
1.3 Event Time I turn now to the specification of Event Time. So far, no examples have specified ET unless it is simultaneous with RT. If sentences with one time adverbial specify RT, sentences with two adverbials might be expected to specify both RT and ET. Consider some examples; the adverbials are given in two positions to facilitate the interpretation. (32) (33) (34) (35) (36) (37) (38) (39)
Bill wrecked the car last night 3 weeks ago. 3 weeks ago, Bill wrecked the car last night. Tom broke his leg on Wednesday as soon as he was released from the hospital. On Wednesday, Bill broke his leg as soon as he was released from the hospital. John arrived yesterday on Monday. On Monday, John arrived yesterday. Mary takes the train in a month tomorrow. Tomorrow, Mary takes the train in a month.
Most native speakers that I have consulted find these sentences ungrammatical and almost unintelligible;5 asterisks are omitted so as not to prejudice the reader. The examples in which one adverb is fronted are somewhat easier to interpret than the others; the fronted adverb is taken to specify RT, so that a sentence like (39) for instance may be interpreted, roughly, to mean “It will be the case tomorrow that Mary takes the train in a month.” (Sentences like (33) and (39) must be distinguished from sentences in which part of a time adverbial is fronted, as in (40): (40) On Tuesday Bill played squash at 2 pm. Complex adverbials of this type constitute one adverbial in the present analysis.) The ungrammaticality of the examples above shows that ET cannot be specified by simple adjunction of a time adverbial to a sentence specifying RT. The examples also show that English sentences can have only one time adverbial. A more fruitful approach will be to look for sentences that indicate an ET different from RT. There are such sentences. For instance: (41) They told me yesterday that the play had closed 3 weeks ago. (42) I heard last night that the show was opening in a few days. The complement sentences of both these examples indicate an ET different from RT. Note that the adverbials in the complements specify ET, and the relation between ET and RT is given by the relational value of the adverbial. The interesting point about these examples is that RT for the complements is partly established in the matrix sentence. The complements have as RT a time of 5 Some
speakers accept such sentences; for instance, see Hornstein (1975).
Temporal Expressions in English
103
the matrix, and can be said to SHARE that time. I have shown above that adverbials contribute to the specification of RT; in these sentences RT has been specified so that the complement adverbial is free, as it were, to specify ET. In other words, ET can be specified only if – for a given sentence – RT is also specified. There are syntactically independent sentences that correspond in their temporal interpretation to the complements of (41) and (42). For instance: (43) John had read the article three weeks ago.6 (44) Harry was arriving tomorrow. Both of these sentences are somewhat odd in isolation: they are felt to be incomplete in some way. Within Reichenbach’s scheme for temporal specification, it is clear what is odd about these sentences; they lack a reference time. The examples do not have combinations of tense and adverbial that establish RT. A nearby sentence can give an RT for sentences like this, and in fact such sentences cannot be fully interpreted except in a domain larger than a sentence. The semantic dependence of sentences like this have far-reaching implications for the question of how a grammar can best account for time specification. It can be shown that the information completing these incomplete sentences is just that which can appear in matrix sentences for the corresponding complements. Therefore the same principles are involved in relating syntactically and semantically dependent sentences to other sentences that establish RT. There are some sentences with two time adverbials that seem to be counterexamples to the analysis developed here. These sentences are not grammatical for all speakers; they are a subset of the type of sentence exemplified in (32), (33), (34), (35), (36), (37), (38) and (39). (45) Last night, Mary had disappeared 3 months ago. (46) Next June, Todd will be graduating in a month. The interpretation of these examples is that the first adverbial establishes RT and the second gives ET. The first adverbial is interpreted with the tense, and the second separately, on the pattern of complements with shared RT discussed above. That (45) and (46) follow the pattern of shared RT sentences suggests that they are in fact reductions of sentences that share RT. This is the analysis that will be suggested here. I suggest that sentences like (45) and (46) have a 2-sentence source, the first containing the matrix RT and a proform for the second sentence. For instance, a plausible source for (45) would be something like (47) or (48): (47) Bill told me last night: Mary had disappeared 3 months ago. (48) It was the case last night: Mary had disappeared 3 months ago. 6 The ‘already’ reading, in which John read the article some time previous to three weeks ago, is not intended here.
104
C.S. Smith
(48) is preferable to (47) as a source since it does not involve irrecoverable deletion. A sequence such as (48) is artificial in isolation, but can be imagined as part of a discourse. A more natural sequence is (49): (49) Bill finally told me the terrible thing that has been bothering him. Last night, Mary had disappeared 3 weeks ago. A detailed account of such sequences must be undertaken before a precise source for sentences that are reductions of 2 sentences can be suggested. However, I would like now to give some arguments in favor of the general approach. The first adverbial in sentences like (45) is rigid in position, unlike normal time adverbials. It can occur only to the left of the sentence, preceding and set off by comma intonation. A precedent for analyzing (45) as a reduction of two sentences may be found in Emonds’ analysis of parentheticals, discussed in Emonds (1974). Emonds suggests that sequences like (50) underlie sentences with parentheticals such as (51): (50) Mary is a secret agent; you know it. (51) Mary, you know, is a secret agent. A transformation inserts the non-anaphoric material of the second sentence into the first. The source suggested here for sentences with two time adverbials is similar to the sequence in (51): there is one full sentence, and one sentence consisting of a predicate of existence, a pro-form, and some new material. The transformation that reduces the two sentences to one inserts the new material of one sentence into the other and sets it off with comma intonation. Sentences with two time adverbials differ, of course, from the parentheticals that concern Emonds. In Emonds’ examples the inserted material is subject and verb (you know), whereas subject and verb introduce the new material in the sentences discussed here. The fact that such different types of material may be inserted into other sentences is suggestive: perhaps reduction of two sentences to one is a more general process in English than has been heretofore appreciated. A final piece of evidence for the 2-sentence analysis comes from examination of contexts in which sentences like (45) and (46) occur. Whenever I have found such sentences in actual discourse, or tried to construct a plausible context for them, there have been other sentences in the neighborhood that shared the RT and had a verb of saying or a related verb. Such sentences are like those of the suggested source sequence. To summarize, sentences that specify ET are not independent sentences. They are dependent semantically on another sentence for the specification of RT, and are similar in this respect to complement sentences. Further discussion appears in Section 2, which is devoted to the analysis of complement sentences.
Temporal Expressions in English
105
1.4 Event Time and Auxiliary have There is one independent structure that I know of in which ET can be specified, or at least indicated: Present sentences with have.7 Have is a relational element semantically, indicating that ET precedes RT. Its semantic function is similar to that of relational adverbials, but it differs syntactically. Consider, for instance, the interpretation of (52): (52) They have eaten all the fudge while you were out. This sentence seems to be a counter-example to the claim that ET can be specified only when RT is specified: its interpretation is that the adverbial indicates ET, yet there is no specification of RT. However RT is, in effect, specified in (52). Consider the role of have in the sentence: have indicates anteriority of ET from RT, and therefore implies a specific RT. Since the sentence has Present RT, a specific RT is available, namely ST. In other words, the appearance of have in a Present sentence directly implies that the RT of the sentence is ST. Therefore (52) is not a counter-example to the claim that RT must be specified if ET is specified. Present sentences with a modal and have also have an implied RT of ST. For instance: (53) They will have planted all the roses while you were away. (54) They will have arrived last week. (55) Bill may have won the race on Tuesday.8 (53), (54) and (55) all make predictions about the Past, but from a Present RT. As in (52), the adverbials are past and indicate an ET that precedes ST. Sentences similar to these but with Past or Future RT cannot be constructed. The combinations of tense and adverb are ungrammatical, or are interpretable only as specifications of RT. Sentences with Past RT that correspond to (53), (54) and (55) would have past tense, have, and a Past or Unanchored adverbial. For instance: (56) They had eaten all the cookies while we were away. (57) Bill had already arrived last week. (58) They would have left on Tuesday. In (56) and (57) the combination of tense and adverbial is interpreted as indicating Past RT. This interpretation occurs because Past sentences have no specific 7 Adverbials that can occur in this construction are strictly limited; the matter is discussed in Smith (1976a). 8 See Hofmann (1966) for a different analysis of such sentences.
106
C.S. Smith
orientation point, as Present sentences have ST in the Present. Therefore the adverbials are taken to contribute to RT. Sentences like (58) have already been discussed; dependent on other sentences, they do not specify RT. An attempt to construct Future sentences corresponding to (52) runs into difficulty. What makes (52) unusual is that it specifies ET and makes the specification of RT unnecessary. Recall that Future RT is indicated by present tense and Future or Unanchored adverbial. Sentences may have only one time adverbial, and Future sentences have no specific orientation in the Future as Present sentences have ST in the Present. This means that there is no way to specify a separate ET. In fact the combination of present tense, have and Future or Unanchored adverbials is ungrammatical and must be blocked, as the following sentences show. (59) ∗ They have eaten all the cookies tomorrow. (60) ∗ They have arrived on Tuesday. The reason for the ungrammaticality of these sentences is that have and the adverbials are not compatible elements. Both indicate relations between times: have indicates anteriority, while Future adverbials indicate posteriority (to ST). The forms are mutually exclusive and cannot occur together. The possibility that remains is a Future sentence with a modal, have, and an adverbial, for instance: (61) They will have eaten all the cookies tomorrow. (62) They will have eaten all the cookies on Tuesday. In this case the sentences are grammatical but they specify RT rather than ET. These are the only adverbs that could occur in such sentences: with Past or Present adverbials the sentences would not have Future RT. As in the Past sentences, there is no possibility of interpreting the adverbial as ET. That the Present examples are interpreted as specifying ET is due partly to the fact that the combination of present tense and Past adverbial does not function to specify RT in independent sentences. Just as important, however, is the interpretation that RT = ST for Present sentences with have. This interpretation, or implication, frees the adverbial to specify ET. The implication is possible because ST has a central position in the system of time specification, and need not be given explicitly.
1.5 Summary I have shown that sentences specify ET only when RT is already specified, and that this situation occurs only under particular circumstances. The reason for this is syntactic: English sentences are limited to containing one time adverbial, although the adverbial may be complex and distributed in surface structure. Therefore RT can be specified when a sentence shares RT with another sentence, or when RT is implied in Present sentences with auxiliary have. That sentences with one time adverbial are interpreted as specifying RT shows that RT is essential for interpretation.
Temporal Expressions in English
107
2 The Temporal Interpretation of Complement Sentences This section deals with the vexed question of how complements in English sentences are interpreted temporally. I present an analysis that explains the interpretation of well-known cases, and of others not usually mentioned in the literature. The interpretation of complements is somewhat problematic.9 One fact that makes analysis difficult is that temporal expressions may have different values in complements than they do in independent sentences. Further, the distribution of temporal expressions is apparently different for complements and independent sentences. Presumably for these and other reasons, a unified account of complements has so far not been given. I suggest that complements have resisted satisfactory analysis because they have not been approached as part of the temporal system of English. It is necessary to consider the entire system in order to deal with part of it. The account of the temporal system developed above provides a basis from which complements can be analyzed without difficulty. I give first a general account of the detailed analysis to follow. Before embarking on the discussion, I present a set of examples that indicates the various types of complements to be provided for. a. (63) The boy said that he was eager to enter the debate. (64) Stuart will announce tomorrow that he will enter the debate in a week. (65) The spokesman assured us a week ago that the candidate was leaving 3 days earlier. (66) I remembered in the morning that Ed had left the party before midnight. b. (67) The leaders claim that the tribes were betrayed. (68) The President will say next month that Congress resisted him weeks earlier. c. (69) The narrator says that the heroine was worried now. (70) Mrs. Dalloway will murmur that the party had been a success. (71) The Egyptians realized that the world is round. In group a, matrix and complement have the same tense, exhibiting what is traditionally known as Sequence of Tense.10 However, the sentences in this group are not all interpreted in the same way. In (63) for instance the matrix and complement are taken to be simultaneous, but in (64) the complement is not taken to be simultaneous with the matrix. Sequence of Tense is not a strong enough notion to account for this difference. In group b, matrix and complement do not have the same tense. The complement of (67) has the normal temporal interpretation, in which past tense is anterior to ST; but that of (68) has an interpretation in which the past tense is not anterior to ST. 9 See
Braroe (1974) for discussion of some of the problems involved.
10 Sentences in which matrix and complement have the same tense, and are taken as simultaneous,
exhibit Sequence of Tense. The phenomenon is discussed by Jespersen, among many others. Ross (1967) proposes a transformational rule to account for it; Smith (1976b) and Riddle (1976) argue against such an analysis (from different points of view).
108
C.S. Smith
The complements of both groups are dependent syntactically and semantically on their matrix sentences. The sentences of c, however, are different: the complements cannot be interpreted temporally without more information. They depend syntactically on their matrix sentences, but semantically they are dependent on other sentences. A fairly satisfactory account of these examples – which exhaust the possible types of matrix and complement – can be given by extending the account of temporal specification to include relations between sentences. What is needed is an understanding of how sentences may be temporally related to each other, and principles stating how the relationships are realized. I will argue that complements of English are dependent on other sentences in two different ways, First, a complement may have as a point of reference – that is, RT – a time established in the matrix or another sentence. In this case the sentences will be said to SHARE the time established in the matrix, since it holds for both. The principle of interpretation for such sentences will be called the Sharing Principle. A complement may also be anchored, or ORIENTED, to a time in the matrix rather than to ST. Recall that RT in independent sentences is oriented to ST: in other words, ST provides the point of orientation to which a reference time is simultaneous or sequential. In a dependent sentence RT has its usual relational value, but is related to a time established in the matrix rather than to ST. The principle of interpretation for such sentences will be called the Orientation Principle. The principles of Sharing and Orientation account for the interpretation of the various examples above. In the sentences of a, matrix and complement have the same tense and are explained by the Sharing Principle: a time established in S1 acts as RT for S2 . In the sentences of b, the Orientation Principle applies: the RT of S2 is oriented to a time established in S1 . The sentences of c are interpreted as sharing a time with a sentence other than the matrix; the other sentence establishes RT for the complement. For these sentences, the Sharing Principle must have a domain larger than a sentence.
2.1 Same Tense in Matrix and Complement I now discuss sentences in which matrix and complement have the same tense; such sentences are interpreted by the Sharing Principle, and are exemplified in a of Section 2. Consider the interpretation of sentence (72): (72) They told us yesterday that Tom had arrived 3 days earlier. The matrix sentence establishes an RT in the Past, namely yesterday; the complement is dependent on the matrix, in a way to be precisely determined. The complement has a combination of tense and adverbial that does not establish RT, according to the principles of Section 1. Yet (72) is fully interpretable: informally, we can say that S2 indicates an arrival 3 days anterior to yesterday. Therefore, S1 establishes
Temporal Expressions in English
109
RT for S2 . The relationship between the two sentences is particularly evident in this case, because S2 lacks RT. Since S2 is interpretable it must be getting RT from some temporal expression outside S2 , and S1 is the only possibility. There are three important points to notice here. First, a time established in S1 is essential to the interpretation of both sentences. (In (72) only one time is established in S1 , because ET = RT; see Section 3.2 for a discussion of what time in S1 acts as RT for S2 .) S1 and S2 can be said to share the time in question, in that it holds formally and semantically for both: the sentences have the same tense, and the adverbial of S1 has both sentences in its domain. Recall that RT is established by the combination of tense and adverbial. Second, the complement specifies a time other than RT – namely, ET. The interpretation of S2 involves two specified times, RT and ET. To see this, consider the roles of the adverbials in matrix and complement: the adverbial of S1 contributes to RT, whereas the adverbial of S2 specifies ET. A rough schematic interpretation of (72) brings out the difference: (73) S1 : RT: Past, yesterday ET = RT S2 : RT = RT1 ET ← RT ET: 3 days earlier. The arrow in the interpretation above indicates anteriority; similarly, → will be used in such interpretations to indicate posteriority, and = to indicate simultaneity. The interpretation shows that complements such as that of (72) differ from independent sentences in specifying ET. Independent sentences can indicate ET but not specify it, if ET differs from RT. For instance, take the interpretation of (74): (74) Harry ate before noon. The time specified by this sentence is noon in the Past, which serves as RT; we know that ET precedes RT because of the introductory preposition before, but we do not know the actual time of ET. A sentence must establish RT, and English allows only one time adverbial per sentence, as noted in Section 1. Since an adverbial is necessary to establish RT, it follows that a single independent sentence cannot specify ET. However, a complement differs from an independent sentence in having, as it were, two adverbials, because the matrix adverbial is available to the complement. Since the matrix adverbial contributes to the specification of RT for the complement, the complement adverbial is free to specify ET. The third point of interest, in the interpretation of (72), is how the adverbial in S2 indicates the relation between ET and RT. In the complement of (72) ET is taken as anterior to RT. Recall that the adverbial of the complement is explicitly Past, and that its relational value is anteriority. It is the relational value of the adverbial in S2 that gives the relation between ET and RT; the adverbial itself specifies ET. For instance, in the examples below the relational values of the adverbials in S2 differ, and the relations between ET and RT of S2 differ accordingly:
110
C.S. Smith
(75) I realized at midnight that Sam had left the party earlier. ET ← RT (76) The nurse explained that the doctor was busy now. ET = RT (77) The office announced last week that the chairman was resigning in two days. ET → RT Unanchored adverbials have the values of both anteriority and posteriority: they are ambiguous, out of context, between the values of ← and →. Therefore (78) has two readings, as the disambiguating contexts show. (78) The nurse explained that the doctor was working on Tuesday a . . . so he couldn’t have committed the crime. ET ← RT b . . . so he couldn’t come to the charity bazaar. ET → RT If a sentence has in the complement auxiliary have and an Unanchored adverbial, it specifies unambiguously that ET is anterior to RT, since have indicates anteriority: (79) The nurse explained that the doctor had been working on Tuesday.
2.1.1 Matrix Event Time as Complement Reference Time I turn now to two essential details of the interpretation. I first look more closely at the time in S1 which S2 shares, and then consider the interpretation of embedded anchored adverbials. The Sharing Principle says that a time established in S1 acts as RT for S2 . In the relevant examples above, it is RT of the matrix that acts as RT for the complement: both sentences have the same RT. But in those examples, ET is simultaneous to RT, so that only one time is indicated in the matrix. Sentences in which the matrix ET is not simultaneous to RT exhibit a rather different dependence, for instance: (80) They announced before noon that the fugitive had been caught 3 hours earlier. (81) I told the hotel clerk several hours after midnight that I was leaving in 2 hours. In these examples, the complement depends on ET of the matrix rather than RT. To see this, consider how to compute the time specified as ET by the complement. In (80), when was the fugutive caught? If RT2 = RT1 , the fugitive was caught 3 hours before noon; but if RT2 = ET1 , he was caught 3 hours before a time before noon. The latter corresponds unambiguously to the interpretation of (80). Again, in (81), if RT1 = RT2 then I planned to leave two hours after midnight; but the sentence can only mean that I was to leave 2 hours later than some time several hours after midnight. All of the examples can be accounted for with the generalization that ET1 acts as reference time for S2 ; included of course are sentences where ET = RT in S1 . This same dependency occurs in more complicated embeddings. Generally, then, if an embedded sentence has the same tense as the sentence above it, ET of the higher
Temporal Expressions in English
111
sentence acts as RT for the lower. This relation holds even if the higher sentence does not have a combination of tense and adverbial that independently establishes RT: (82) Sam announced that Bill had told him 3 days earlier that Charlie was arriving in a week. (83) Mary says now that Sue will announce next week that Amy is threatening to disclose in 5 days that Jane leaves in March.
2.1.2 Embedded Anchored Adverbials There is a complicating factor in the interpretation of many sentences that share a time between matrix and complement. The adverbial of S2 specifies ET2 , and the relational value of the adverbial gives the relation between ET and RT. Complications arise if the adverbial in S2 is explicitly Past or explicitly Future, because these adverbials are usually oriented to ST. If they appear in a complement that has an RT different from ST, they may be interpreted with respect to ST or to RT of the complement. This means that sentences with embedded ST-anchored adverbials are ambiguous as to the value of ET2 . For instance: (84) Sally told me on Tuesday that Bill had arrived 3 weeks ago. (85) I explained on Tuesday that Bill had left last week. (86) The radio reported on Tuesday that Bill had disappeared last week. In (84), Bill may have arrived 3 weeks before ST, or 3 weeks before Tuesday; in (85), he left either a week before ST or a week before Tuesday; and so on. The same ambiguities hold for sentences with Future RT and explicitly Future adverbials: (87) Bill will announce on Tuesday that he is leaving next week. There appears to be some difference among anchored adverbials, and among speakers, as to whether the adverbials are flexible in orientation.11 For some speakers these adverbials are inflexible, and can be anchored only to ST; for them, (84), (85), (86) and (87) are not ambiguous. A number of people find that some adverbials are more tightly anchored than others: yesterday, for instance, is for many less flexible than a week ago; tomorrow is less flexible than in a week. When embedded adverbials such as these are taken as anchored to ST, their interpretation is rather complicated. The adverbial has two functions, one with respect to RT2 and the other to ST. For instance, consider the interpretation of (87). The adverbial of S2 has the relational value of posteriority, and gives the information that ET2 11 For
some speakers and in some contexts, anchored adverbials may be dependent on a time not mentioned in a sentence at all, but understood or appearing elsewhere in the discourse.
112
C.S. Smith
follows RT2 . The adverbial also specifies ET2 , and this specification is calculated from ST. Such a calculation involves a coding by the speaker of the time referred to in the complement. This is why the ST-anchored interpretation is particularly plausible for sentences in which S1 and S2 have the same subject, and rather implausible where the embedded sentence is not easily available to the speaker. Compare, for instance: (88) I admitted yesterday that I found the letter a week ago. (89) Fred knew on Tuesday that the gang had stolen the jewels last week. Adverbials that are oriented to ST can occur in deeply embedded sentences, producing multiple ambiguities, for instance: (90) The reporters found out on Tuesday that the investigators had been told 3 days earlier that the count had left last week. In summary, the Sharing Principle works in the following way. It applies to syntactically dependent sentences that have the same tense, and interprets ET1 as RT2 . If S2 has no adverbial, the sentences are taken as simultaneous; if S2 has an adverbial, it specifies ET2 which is not simultaneous with RT2 .12 The relational value of the adverbial gives the relation between ET and RT, and the adverbial specifies ET. Adverbials in S2 that are usually anchored to ST may be taken as oriented either to ST or to RT2 .
2.2 Different Tenses in Matrix and Complement I turn now to sentences in which matrix and complement do not have the same tense. There are two possibilities: either S1 has past tense and S2 present tense, or S1 has present tense and S2 past. The different cases will be handled by the Orientation Principle and by an extended version of the Sharing Principle. Consider first sentences such as (91) and (92). In both, S1 has present tense and S2 past tense, and S2 has a combination of tense and adverbial that establishes RT. (91) The report states that the spy was denounced last month. (92) The investigator will insist next month that he talked to the suspects 3 weeks earlier. In (91) the complement has the same temporal interpretation as it would in an independent sentence: RT is anterior to ST. But in (92) the complement does not receive 12 Certain
sentences must be understood as having a complex ET: these cases where the complement refers to more than one point in time, e.g. Bill said yesterday that Tom had already left 3 days earlier.
Temporal Expressions in English
113
its ‘normal’ interpretation. The time referred to is prior to S1 but not necessarily prior to ST, whereas normally Past RT is anterior to ST. These interpretations are predicted by the Orientation Principle, which says that the RT of the complement is oriented to a time established in S1 rather than to ST. The relational value of RT to its point of orientation is unchanged; what is different is that the point of orientation is not necessarily ST. For instance, in (91) S1 has a Present RT and RT = ST; the complement is oriented to Present RT and automatically receives its normal interpretation. In (92) the matrix has a Future RT and RT = ET; the complement has Past RT, and is taken as preceding RT1 . But RT1 is posterior to ST, so that RT2 need not be anterior to ST. In contrast to the Sharing Principle, the Orientation Principle says that S2 establishes its own RT. One might object that sentences such as those just considered do not differ in general interpretation from the examples of the preceding section – that is, that the complements share a time with the matrix in all cases. The objection can be tested by asking whether, in sentences where S1 and S2 have different tenses, S1 establishes RT for S2 . This is the essential point of the Sharing analysis. Crucial examples will have complements that do not establish RT, and that have a different tense from the matrix. For instance: (93) William will insist next week that Mary was returning to London in three days. (94) Bill will say next week that Mary had left 3 days ago. (95) Bill says that Mary was leaving in 3 days. If some form of the Sharing Principle is applicable to these sentences, they should be fully interpretable temporally: S1 should provide RT for S2 . But none of the examples can be fully interpreted: more information is needed. In (95) for instance, one knows that Mary was leaving 3 days after some point, but not what point. These examples show conclusively that S1 does not establish RT for S2 in the sentences under consideration. No form of the Sharing Principle could account for them, therefore, and the Orientation Principle is shown to be necessary. The matrix time to which complements such as (92) are oriented is ET. To see this, consider the interpretation of (96), where ET1 is not simultaneous with RT1 . (96) Sam will announce before midnight that Sue left 3 hours earlier. The complement RT is anterior to its orientation point, so one can ask whether Sam will announce that Sue left at 9 (3 hours earlier than midnight), or that she left earlier than this. The sentence can only be interpreted to mean that the point of reference for 3 hours earlier is before midnight, not midnight. For both the Sharing Principle and the Orientation Principle, then, ET1 is the time in S1 that is crucial for the interpretation of S2 . According to the Sharing Principle, ET1 is RT2 ; according to the Orientation Principle, ET1 is the point of orientation for RT2 .
114
C.S. Smith
2.2.1 Sentences to Which Sharing and Orientation Principles Both Apply There are a number of sentences to which both the Sharing Principle and the Orientation Principle might apply, according to the preceding discussion. These are sentences in which matrix and complement have the same tense, and in which the complement has an RT combination of tense and adverbial. In the examples that supported the Sharing Principle, the complements do not have RT combinations; in the examples supporting the Orientation Principle, the complements establish an independent RT. I now ask whether the two principles give different interpretations for sentences to which both apply; and if so, which one is correct. Consider, for instance: (97) Bill will say tomorrow that the committee rules on the problem in three days. (98) Sam believes that he is justified now. (99) Sharon admitted that she had already arrived on Tuesday. The Sharing Principle can apply to these examples, because S1 and S2 have the same tense; the Orientation Principle can apply, because S2 establishes RT. In many cases a sentence receives the same interpretation from both principles; this is true of (97), (98) and (99). For instance, both interpretations of (97) are given in (100) (100) Interpretations of (97) Sharing S1 : Future RT, tomorrow
Orientation S1 : Future RT, tomorrow
ET = RT S2 : RT = ET1 ET → RT ET: in 3 days (from tmw) ET = RT S2 : RT → ET1 RT: in 3 days (from tmw) ET = RT
According to the Sharing interpretation, matrix and complement have the same RT because ET = RT in S1 ; the Future adverbial in S2 indicates that ET2 is posterior to RT, and it specifies ET2 . According to the Orientation interpretation, S2 establishes its own RT, in which ET and RT are simultaneous; RT2 is posterior to RT1 because of the relational meaning of Future RT. However, there are sentences for which the principles give different interpretations. Consider, for instance, (101) and its interpretations (102). (101) Bill said yesterday that Tom was sick. (102) Sharing S1 : Past RT, yesterday ET = RT S2 : RT = ET1 ET = RT Orientation S1 : Past RT, yesterday ET = RT S2 : RT ← ET1 ET = RT
Temporal Expressions in English
115
(101) can mean only that matrix and complement are simultaneous. The Sharing Principle correctly makes this interpretation, but the Orientation Principle predicts that S2 be anterior to S1 . Therefore, only the Sharing Principle must be allowed to apply to sentences like (101). There are two ways to accomplish this. The rules of interpretation might be ordered, with the Sharing Principle applying first. This would ensure that the Orientation Principle would never apply to sentences to which the Sharing Principle was applicable. Another possibility would be to state the Orientation Principle so that it applied only to sentences in which the complement did not establish RT. The complement of (101) has only tense and therefore does not establish RT. The first of these alternatives is preferable because there is another case (see below) which requires that the Sharing Principle precede the Orientation Principle. 2.2.2 Present-Tense Matrix and Past-Tense Complement I now turn to sentences such as (103) and (104), in which S1 has present tense and S2 past tense, and S2 does not establish RT. According to the analysis developed above, neither principle of interpretation applies to them. (103) The prosecuting attorney claims that the nurse was tired now. (104) The public will learn next week that Smith had already withdrawn his offer of open negotiations. Note, however, that neither of these sentences can be fully interpreted. A point of reference is needed for now in (103) and for have in (104). In the terms of this analysis, both complements require an RT. What is needed to interpret these examples are other sentences that supply RT for the complements. In fact, sequences of the type required occur fairly frequently, especially perhaps in rather formal discourse of the narrative type. The following sequences allow an interpretation of (103) and (104): (105) It was 3 o’clock in the morning when the old lady rang for the nurse on duty. The prosecuting attorney claims that the nurse was tired now, and didn’t pay much attention to the old lady. (106) The conference took place before March, ostensibly to arrive at a peaceful solution. But the public will learn next week that Smith had already withdrawn his offer of open negotiations. In these sequences the sentences preceding (103) and (104) establish reference times for the complements of the sentences. Thus, 3 o’clock in the morning is the point of reference for now in (105); before March is the point of reference for had already in (106). These reference points are related to the complements of the following sentences just as the time established in a matrix sentence relates to a complement that has the same tense. The sentences involved have the same tense, and ET of the
116
C.S. Smith
preceding sentence acts as RT of the complement. In terms of the previous discussion, the complement sentences share a time established in a preceding sentence. Indeed, except for the fact that S1 and S2 are not syntactically dependent on each other, the situation is the same as in the case of sentences to which the Sharing Principle applies. An extension of the Sharing Principle can account for the interpretation of (103) and (104), in contexts that allow for such an interpretation. In the examples discussed in Section 3, syntactically dependent sentences share a time; in the examples discussed here, syntatically independent sentences share a time. If the Sharing Principle were extended so that its domain were larger than a sentence, the different cases could be accounted for in the same way. The extended Sharing Principle would work roughly as follows: from the complement of a sentence, it would look for a sentence with the same tense. If the matrix has a different tense, another sentence would be sought that had the same tense, and could provide RT. The exact conditions under which a sentence may provide RT for a neighboring sentence is an interesting and important topic that requires investigation in its own right. It is interesting to note that an extended Sharing Principle is needed elsewhere in the grammar of English. The relevant sentences are syntactically independent, unlike the complements just examined, but semantically dependent in exactly the same way on other sentences. Such sentences have tense + adverb combinations that do not establish RT, as mentioned briefly in Section 1. For instance: (107) Ross was leaving in 3 days. (108) Ross had left on Tuesday. (109) Ross was annoyed now. Exactly the same principles that apply to complements without RT predict the interpretation of (107), (108) and (109). In isolation they cannot be fully interpreted; in the neighborhood of a sentence that has the same tense and that establishes RT, they can be interpreted according to the Sharing Principle. I have discussed sentences like this at some length elsewhere. Although they seem odd in isolation, they are natural and far from infrequent in fictional and non-fictional narratives; see the examples in Smith (1976b). All the cases in which S1 has present tense and S2 past tense have now been discussed. If S2 establishes RT, the complements are accounted for by the Orientation Principle. If S2 does not establish RT, the extended Sharing Principle applies. It was shown that the extended Sharing Principle is needed independently to account for sentences that are semantically incomplete but syntactically complete. 2.2.3 Past-Tense Matrix and Present-Tense Complement One type of sentence has yet to be considered: that in which the matrix has past tense and the complement has present tense. The following sentences illustrate:
Temporal Expressions in English
117
(110) The Egyptians knew that the earth is round. (111) Sam told me that Mary is leaving next week. (112) I heard last night that Whitney is sick. I will argue that an interpretation of these sentences can be made by using the extended Sharing Principle, and by postulating an abstract performative sentence in which RT = ST. First, however, consider how the Orientation Principle might apply to them (since the matrix and complement have different tenses, the Sharing Principle is not applicable).
2.3 Summary According to the Orientation Principle, the relational value of the complement RT gives the relation between RT1 and RT2 . In these examples S1 has Past RT and S2 has Present RT. Since the relational value of Present RT is simultaneity, the Orientation Principle predicts that the complements of (110), (111) and (112) are simultaneous with the matrix sentences. But this is incorrect: the complements of all three examples refer unambiguously to ST and not to the matrix. The correct interpretation of such sentences is that the speaker is responsible, as it were, for the complement’s being true or relevant at ST. More precisely, they indicate that the same event or state referred to holds at the time referred to in the matrix and at ST. For instance, (112) means that Whitney was sick last night, and that he is sick at ST with the same illness. It cannot mean that he is sick again. Not all matrix verbs allow speaker-oriented complements. It appears to be a valid generalization that factive verbs and verbs of saying do allow them.13 The following sentences, which do not contain such verbs, are distinctly odd: (113) Mary feared that Bill is sick. (114) The family thought that the money is safe. Since the Orientation Principle does not make correct predictions for sentences like these, one might seek to revise the relational system to allow for such predictions. The system might be revised so that Present RT would have a value other than that of simultaneity: rather, Present RT might be given the value of a time between Past and Future, anterior to Future RT and posterior to Past RT. If Present RT were given such a value, (110), (111), (112) would be correctly interpreted by the Orientation Principle. The complements of the sentences, oriented to the matrix RT, would indicate a time posterior to the Past, namely the Present. But this revision will not work for the full range of relevant cases. It fails for sentences like (115), in which both S1 and S2 have present tense: (115) Gwendolyn will say tomorrow that the Abbess of Crewes is dangerous. 13 These facts are noted by Braroe (1974); Kiparsky and Kiparsky (1970) argue that factivity is the
relevant property but do not consider a wide range of examples.
118
C.S. Smith
Sentences like (115) have only the interpretation that matrix and complement are simultaneous. But if the Orientation Principle were revised so that Present RT did not indicate simultaneity, the wrong prediction would be made for (115). In fact, no one principle can explain all the relevant sentences. In some cases complements with present tense are taken as simultaneous with the matrix, and in other cases such complements are taken as simultaneous with a sentence other than the matrix. What is needed is a way to allow the two different principles of interpretation to apply. If the Sharing Principle were applicable to some of the cases, and ordered before the Orientation Principle, it would be possible to provide for the interpretation of sentences like (112) and (115). Speaker-oriented complements pose a problem for analysis because there is no sentence to which the complement appears to be related. This problem can be solved by positing an abstract performative sentence, associated with the main sentence, in which RT = ST.14 Such a sentence would be available to the extended Sharing Principle. The Sharing Principle relates sentences with the same tense; in its extended form it relates a complement to a sentence other than the matrix. For sentences with speaker-oriented complements, the performative sentence and the complement – both of which have Present RT and ST – would be related by the extended Sharing Principle. The extended form of the principle was introduced in the preceding section to account for certain complements, and for independent sentences that are semantically related to other sentences. In the cases discussed here, a complement is dependent on its associated performative (S0 ) rather than on the sentence that directly dominates it (S1 ). The interpretation of adverbials in speaker-oriented complements is straightforward, according to the principles developed above. The relational value of the adverbial gives the relation between ET and RT, and the adverbial specifies ET; if there is no adverbial, ET is simultaneous with RT. In (111) for instance the complement adverbial is explicitly Future. This means that ET2 follows RT0 , or ST, and is future with respect to ST; the complement has just this interpretation. By positing an abstract performative sentence one can also account for the interpretation of embedded speaker-oriented adverbials such as yesterday. In the discussion of such adverbials, it was noted that they may be taken as oriented to ST rather than to RT of the sentence in which they occur. For instance, (116) has such an interpretation for many speakers. (116) The butler reported yesterday that the count had vanished a week ago. If the adverbial is oriented to ST, then ET2 is a week prior to ST rather than to yesterday, which is RT1 . If a performative with RT = ST is associated with each sentence, the speaker-oriented interpretation of (116) can be accounted for in a natural manner. One more adjustment must be made to account for the interpretation of speaker oriented complements. The principles of interpretation must be ordered so that Sharing precedes Orientation, and applies obligatorily if application is possible. This 14 The notion of an underlying performative sentence is due to Sadock (1969) and (1974). A super-
ordinate performative associated with each sentence would also be quite plausible.
Temporal Expressions in English
119
ordering will prevent the Orientation Principle from applying to such sentences, and yet allow it to apply in other cases. For instance, consider (117) and (118). (117) Myrna reported that the bomb will go off next week. (118) The papers will report next week that the bomb went off on Tuesday. The extended Sharing Principle will apply to (117), relating the complement to S0 , the performative associated with the sentence. Since (117) is interpreted, the Orientation Principle will not apply to it. (118) does not meet the conditions for the Sharing Principle, since the complement has Past RT and neither S0 nor S1 have past tense. Therefore the Orientation Principle applies, relating RT of the complement to RT1 rather than to ST. This ordering of the principles of interpretation is necessary because of an asymmetry in the way complements are interpreted. Present is always related to a sentence with Present or Future RT: a complement with present shares RT with a sentence other than the matrix, if the matrix is Past. Past, however, need not be related to a Past RT. It is because of the flexibility of the Past that the Orientation Principle is needed, and because of the formal similarity of the crucial cases that an ordering of the principles is needed. Summary. I have developed in this section an account of the temporal interpretation of complement sentences in English. Relating, by principles of interpretation, matrix sentences and their complements, and complements to sentences other than the matrix, all the possibilities are provided for. The extended Sharing Principle applies to most of the cases; but the Orientation Principle is necessary for an important small group. The next step in this research will be to investigate the circumstances, syntactic and perhaps semantic, in which a sentence may contribute to the interpretation of a neighboring sentence. This step is necessary for the formulation in precise terms of the principles presented here.
3 Habitual Sentences This section characterizes habitual sentences, and shows how they fit into Reichenbach’s scheme for temporal specification.15 No special extension of the system is needed to account for habituals, although Reichenbach did not discuss them particularly. Habitual sentences are particularly interesting for the study of syntax and semantics, because adverbial forms may function quite differently in habitual and non-habitual sentences. Habitual sentences indicate that an event or state recurs, with a given frequency, during a given interval. Consider some typical habitual sentences:
15 See
Smith (1976a).
120
C.S. Smith
(119) Lee was often in love last summer. (120) Bill swam 3 times a week in March. (121) Scott got up early every morning that year. None of these sentences refers to a particular event or state: rather, they indicate the recurrence of an event or state. Since habituals do not refer to particulars, they do not specify a particular moment or interval. This is the crucial fact about habituals from the point of view of time specification. Habituals are thus fundamentally different from non-habituals, which pick out a particular moment or interval for ET even though it may be specified only relationally. An adverbial of frequency signals the habitual sentence. As the examples show, the adverbial may be general or quite specific. Frequency adverbials may indicate the number of times per unit that an event or state recurs; they may also indicate the typical time of recurrence. The unit (day, week, etc.) must have an indefinite determiner. If the determiner is definite the result is ungrammatical as (123) below; if the determiner is deictic the result is grammatical but not habitual. (124), for instance, gives number of repetitions rather than rate of recurrence and does not contain a frequency adverbial. (122) They went to the movies 3 times a week. (123) ∗ They went to the movies 3 times the week. (124) They went to the movies 3 times that week. (125) has a complex frequency adverbial which indicates the time of recurrence as well as its rate: (125) They went to the movies every afternoon at 3 o’clock last year. Completely specified habitual sentences indicate the interval during which the recurrence takes places as well as its frequency. However, habitual sentences often have less than complete temporal specification. The interval may be unspecified, as in (126); the frequency may be omitted, as in (127): (126) They went to the movies 3 times. (127) Fido chased cats.16 Another possibility is that a sentence that does not have an habitual reading in isolation, receives one in the context of a frequency adverbial. For instance: (128) Mary ate an apple. (129) At lunchtime every day, the same thing happened: Mary ate an apple. 16 Lawler
(1972) discusses this type of sentence.
Temporal Expressions in English
121
Many sentences can take on an habitual reading in the appropriate context; sentences that specify a number of repetitions, with a deictic (as in (124)) cannot be interpreted as habitual.17 Temporal expressions play different roles in habitual and non-habitual sentences. Consider, for instance, the interpretation of the following examples: (130) John got up at noon. (131) John got up at noon every day last summer. In (130), a non-habitual, the adverbial specifies RT, which is simultaneous with ET. However, in the habitual (131) the same adverbial (at noon) is part of the frequency adverbial and does not specify RT. Since non-habitual sentences pick out a particular moment or interval, it will be convenient to refer to them as ‘specifying’. Habituals can now be fitted into the scheme for temporal specification that underlies the analysis of this paper. Habituals specify RT, the interval during which the recurring event or state appears. Habituals do not have a particular ET: rather than specifying one or several events, they indicate recurrence. A natural analysis, then, will interpret the frequency adverbial that is characteristic of habituals as ET. It is the indication of recurrence for ET that precludes a specific interpretation, so that this account gives exactly the correct results. The analysis of (131) for instance, would be something like (132): (132) s[RT: past, last summer ET = RT ET: every day at noon]. While specifying sentences specify ET under certain circumstances only, it is normal for habituals to specify ET; as pointed out at the beginning of this section, a frequency adverbial signals that a sentence is habitual. Since specifying sentences have the same adverbial forms with different functions, one would expect to find structurally ambiguous sentences in which a time adverbial might be part of ET or part of RT. The possibility is even greater since sentences may not be fully specified temporally. Structurally ambiguous sentences can be constructed with adverbials that refer to intervals: (133) Janet swam in the mountain lake from June to September. This sentence might be taken as specifying the interval during which the separate events of swimming recurred, the frequency and time of recurrence being unspecified; on this interpretation the time adverbial specifies RT. It might also be taken as specifying the time (though vaguely), but neither the interval nor the
17 ‘Incomplete’
sentences are frequently uttered and written, and almost always are understood without difficulty. This is due to the contribution of context, I believe, and to principles of interpretation according to which speakers choose that interpretation requiring the fewest assumptions or additional information. I plan to state these principles - which are strategies, essentially - in a forthcoming paper.
122
C.S. Smith
frequency during which the swimming took place. A fuller version of this second interpretation might be (134): (134) Janet swam in the mountain lake from June to September every year during her childhood. If a sentence has two time adverbials and a frequency specification and one adverbial has been moved e.g. to the front, the sentence is structurally ambiguous. The displaced adverbial might be part of the frequency specification, or it might indicate RT. Actually such sentences are rarely ambiguous when the details of the adverbials are considered: adverbials associated with frequency specifications tend not to specify intervals, and only such adverbials can lead to structural ambiguity of this particular type. The analysis of habitual sentences shows that time adverbials may function as part of the frequency specification, as well as to specify RT or a particular ET. Semantically, the relevant characteristic of habituals is that they do not indicate a particular moment or interval; syntactically, habituals characteristically contain a frequency adverbial. The crucial fact about the interpretation of habituals is that the frequency specification functions as ET.
4 Conclusions Regarding Temporal Interpretation The analysis developed above is based on consistent relational values for the temporal expressions of English. I have emphasized the interaction of adverbials and other elements in temporal specification. It is the combination of adverbial and tense that establishes RT; adverbials specify ET, when syntactic conditions allow; adverbs may contribute to the specification of an habitual. In order to interpret an adverbial, then, the syntactic configuration in which it occurs, and the relational values of the relevant expressions, must be considered together. An important result of this study is that sentences that are dependent on others for semantic interpretation need not be syntactically dependent. The same rules of interpretation apply, moreover, regardless of syntactic dependency, to all temporal expressions. When semantically dependent sentences are considered as well as complements and independent sentences, it is evident that all combinations of temporal expressions occur in English, so that there is no need for selectional restrictions such as those sketched in e.g. Aspects. Literary forms do not need special treatment in this analysis, it should be noted. There are no rules for shifting into e.g. the style indirecte libre, or the more widelyused reportive style. Rather, these forms are generated as combinations of temporal expressions like any others, and are interpreted by the basic principles set up for the sentences of the language. It has been suggested that some languages have special syntactic forms for literary discourse; Benveniste (1966) has made this suggestion for French, and Hamburger (1968) has developed a similar approach for German. For English, as the discussion above makes clear, no such special approach
Temporal Expressions in English
123
is needed. One can say that the forms taken as literary tend to be those in which syntactically independent sentences are dependent on other sentences for full semantic interpretation.
Part II – The Treatment of Temporal Expressions in Generative Grammar Temporal expressions and their interpretations have been discussed in Part I of this paper. Part II is concerned with the question of how temporal specification should be handled in a generative grammar. There are three possible approaches, I believe, to this issue. Generative rules might be constructed that would relate semantic representations to surface structures, following the general outline suggested by McCawley and others as ‘generative semantics.’ Alternatively, syntax might be autonomous in a grammar, with semantic rules relating underlying structures to semantic representations. This is the approach of the ‘standard theory’, of course. A third possibility is that syntax be autonomous, and rules of interpretation relate surface structures to semantic representations. I will consider all three approaches, and will argue that only the third can deal adequately with temporal specification. In Section 3 below I discuss rules of interpretation, and present a set of rules for time specification.
1 Unified Generative Rules: Semantax I begin with a brief discussion of a proposal that tense and adverb be treated as a single category in underlying structure. Kiparsky (1968), McCawley (1971), and Gallagher (1970) have all supported this proposal, although without much detail.18 The proposal hardly represents a full-fledged account of time specification, but it seems worth considering since it is one of the few explicit proposals that have actually been made in this area. It seems reasonable to interpret the proposal as referring to the reference time of a sentence. An underlying adverbial indicating Past, Present, or Future RT would be the source for a copy that would have the form of surface tense. This would account for the fact that tense does not given enough information to establish RT. For sentences with Present RT, then, an underlying Present adverbial would be the source for present tense. This might work for Present sentences, but it is unable to account for sentences with either Future or Past RT. Consider, for instance, sentences with Past RT. The adverbials that occur to establish Past RT are either explicitly Past, or Unanchored: (1) They hired Carol last week. (2) They hired Carol on Tuesday. 18 Following
a suggestion of Kiparsky (1968), McCawley (1971) and Gallagher (1970) suggest a generative semantic treatment.
124
C.S. Smith
There is nothing intrinsically Past about adverbials that are Unanchored – how then could they be sources for past tense? The same difficulty arises for Future sentences, which have either Future or Unanchored adverbials and present tense: there is no way that an Unanchored adverbial could function as the source for present tense in Future RT sentences. The analysis of tense as a copy of an underlying adverbial is dependent, perhaps, on the analysis of will as a Future tense. If English has a Future tense, then the problem noted above does not arise for sentences with Future RT (although it remains for sentences with Past RT). However, there are good arguments against the analysis of will as a future tense; see Section 1 of Part I, and the references mentioned there. Since the proposal that tense and adverbial are a single category in underlying structure is feasible for only one RT of English, it must be rejected. A more abstract semantic representation would be needed for a grammar that related semantics directly to surface structures. One possibility would be an abstract temporal constituent, which would generate the temporal elements in direct relation to their semantic function; a transformational rule could distribute the temporal elements into sentences before the application of cyclic syntactic rules. Such a possibility was suggested in Smith (1975b). I outline how such a constituent might be organized. The temporal constituent, if keyed to a Reichenbach analysis of time specification, might have two parts, termed Reference and Relation. The Reference constituent would specify RT and its relation to ST. The Relation constituent would specify the relation between ET and RT, and ET where appropriate. The following generative and transformational rules account roughly for temporal specification in independent sentences (adverb movement is not included). Selectional details are omitted except for the essential adverbial selection. Generative rules: Temporal constituent Reference
Past Present Future Relation Rel ET Habitual
→ Reference+Relation ⎧ ⎫ ⎨ Precedes ST (Past) ⎬ → Simultaneous with ST (Present) ⎩ ⎭ Follows ⎧ ST(Future) ⎫ ⎨ Past ⎬ adverb → past tense ⎩ ⎭ Unanchored → present tense+Present adverb ⎧ ⎫ ⎨ Future ⎬ adverb → present tense + ⎩ ⎭ Unanchored → Rel ⎧ + (ET) ⎫ ⎨ Precedes ET ( ← ) ⎬ → Simultaneous with ET (=) ⎩ ⎭ Follows ET ( → ) → Habitual → Frequency adv + (Rel+Unanchored adv)
Temporal Expressions in English
125
→ last—, —ago; etc. → this—; now; etc. → next—;
in—; etc. NP Unanchored adv → + (PP) S Precedes ET → before, (aux) have, etc. Simultaneous with ET → at, on, while, Ø, etc. Follows ET → after, as soon as, etc. Frequency → Number + Unit Past adv Present adv Future adv
Transformational rules for distributing temporal elements (pre-cyclic): S [ . . . Aux [X] . . . ]VP . . . ]Temp [tense + Adverb + Rel + (Habitual)]
1 2 34 5 6 7 Structural Change: 1 6 + 2 3 4 + 8 # 7 + 9 5 Condition: 8 is not aux have S [ . . . Aux [tense (modal) × ]y]Temp [have] 1 2 3 4 Structural Change: 1 2 3 + 5 4
8
9
5
The first, portmanteau, transformation moves temporal elements from the abstract temporal constituent into the sentence. Tense becomes the first member of Aux; the relational element and Adverb are adjoined by Chomsky adjunction to the VP, creating a time adverbial; the habitual is sister-adjoined to the VP. The second transformation places have appropriately within Aux. The rules given above would have to be extended if they were to include dependent sentences. There are three extensions that would be necessary: first, two combinations of tense and adverb occur only in dependent sentences, and are not provided for in the rules above. The additional combinations are the non-RT pairs, past tense and future adverb, and past tense and present adverb.19 They are exemplified in (3) and (4): (3) Kelly left tomorrow. (4) Kelly was satisfied now. A second extension would be required by the fact that ET may be specified in dependent sentences if different from RT; no provision for a non-habitual ET is made in the rules above, since it is not specified in independent sentences. Although the same temporal elements indicate ET and RT, the rules would have to be changed for ET to account for the difference in semantic interpretation. The third change involves the interpretation of tense in dependent sentences, which differs significantly from the interpretation of tense in independent sentences. 19 I
omit discussion of the Historical Present, as in the preceding sections.
126
C.S. Smith
All three extensions of the rules would require conditions that are incompatible with generative rules of this type. It would be necessary to look ahead to a higher sentence in generating the temporal forms of the complement. This might be possible for the tenses of complement sentences with shared RT if, looking ahead to see that a sentence was a complement, a temporal constituent could be generated with an empty RT. A transformational rule could then copy the tense of the higher sentence into the complement RT. This would produce the correct combinations of tense and adverbials for shared RT complements. It would not, however, work for other complements, or provide for the generation of a separate ET. These problems might be dealt with if drastic formal changes were made in the way semantic representations were generated. (Actually no one has given any careful account of how underlying structures are to be generated in this framework.) Structures involving more than one underlying sentence might begin with the generation of the highest rather than the most deeply embedded sentence (one has no clear idea of what constitutes a sentence here). If such a way of generating could be specified, provisions for temporal elements in dependent complements could be made. Context-sensitive rules might be keyed to the RT of a matrix sentence, for instance, to allow for the interaction between matrix and complement. Even such drastic changes – if they could be worked out – would not save the approach of semantax, however. There are difficulties of a totally different kind that arise when the full range of sentences with temporal elements is considered. These difficulties could not be handled by allowing embedded sentences access to information in higher sentences, or by any other revision of the rules under discussion. Semantic representations give all the information needed to complete a given sentence, and the rules proposed above generate structures that are complete semantically and syntactically. However, some sentences are not complete semantically, as shown at length in Part I of this paper. They present an insuperable problem for rules that interpret underlying structures. If sentences with incomplete time specifications were somehow generated from underlying semantic representations, other rules would be needed in the grammar to account for their interpretation. Such sentences may be semantically completed with material from other sentences; but rules of interpretation with a domain larger than a sentence would have to apply to surface structures. The conclusion seems unavoidable that this approach to time specification cannot be successful.
2 Autonomous Syntax and Underlying Structure Interpretation I turn next to a treatment in which semantics and syntax are separate: rules of semantic interpretation would apply to the underlying structures generated by syntactic phrase structure rules. This approach is preferable to the one explored above. The semantic interpretation rules would have access to full structures, so that no difficulties arise in accounting for the relationships between matrix and complement sentences. The fact that temporal elements have different functions and values in different configurations would not be problematic, because the full configurations
Temporal Expressions in English
127
would be available to the rules. Another point in favor of this approach is that the phrase structure rules of an autonomous syntax are simpler than generative rules that relate semantic and surface structures to each other. Compare, for instance, the relevant rules of Aspects or a standard text to the generative rules of the preceding section. The generative rules are cumbersome because they relate elements that function differently from a syntactic point of view than from the semantic point of view. For instance: the prepositions that introduce time adverbs are syntactically a unit with the adverb; but semantically they indicate the relation of ET and RT, while the adverb contributes to RT. Auxiliary have is a verbal auxiliary in terms of syntax, but semantically it is a relational element. Autonomous phrase structure syntactic rules would generate sentences with tense, have and a time adverbial. The latter two are optional, of course. Two transformations would be needed to account for surface structure variation: a rule or rules of Adverb Movement, and a Reduction transformation that inserts the time adverbial of one sentence to the left of another.20 The question arises as to whether any restrictions should be imposed on the combinations of temporal expressions generated by the grammar. I have pointed out that all combinations of tense and adverbial occur in English. Certain combinations, however, are not independent. The traditional way to deal with the facts is to take as basic the independent combinations, and produce the dependent forms with additional operations (in generative grammar, with transformations such as Sequence of Tense). Another approach is to generate all the combinations with a basically unrestricted set of phrase structure rules. Certain combinations, of course, must be interpreted as dependent and/or semantically incomplete. This approach is both simpler and more satisfactory than the other: no transformations are needed to produce the so-called shifted forms, and restrictions need not be imposed on the generative rules. I have argued in favor of doing away with Sequence of Tense rules in Smith (1976b); see also Riddle (1976). (A crippling argument against Sequence of Tense rules is that they change meaning, since the dependency of a complement is indicated by tense.) Certain restrictions on phrase structure rules are required, however, to block combinations of elements that are ungrammatical rather than dependent. Certain combinations with have are ungrammatical, for instance: (5) ∗ Mary has left at noon. (6) ∗ Jane has arrived tomorrow. Sentences like (5) are discussed by Hofmann and McCawley, and in Smith (1976b); (5) is intelligible although clearly ungrammatical. (6), on the other hand, is almost impossible to interpret. I will discuss the two cases separately. There are perfectly good English sentences that have the same temporal elements as (5):
20 A
third rule would also be needed to account for tense in adverbials with embedded sentences, perhaps as suggested in Smith (1975a).
128
C.S. Smith
(7) Mary has often left at noon. (8) Mary’s having left at noon didn’t surprise me. (9) Mary may have left at noon. In (7), a frequency adverbial makes the sentence habitual; in (8) and (9), the temporal expression of which have is a part does not dominate the entire sentence. What must blocked is just the configuration exemplified in (5), in an independent Specifying Present sentence. The combination of relational elements in sentence (6) – have and a Future adverb – must be blocked throughout the grammar. The expressions are contradictory semantically, but since they play different syntactic roles a special syntactic restriction is needed to block their simultaneous occurrence. Note that both restrictions involve auxiliary have, a form with several idiosyncratic properties. Difficulties arise for this approach, as for the preceding, with the interpretation of sentences that are semantically incomplete. Such sentences will be generated by the syntax, which allows tense as the only temporal element and all combinations of tense and adverbial. Since the interpretive rules operate on underlying structures, they would not be able to do anything with incomplete sentences except to recognize their incompleteness. Dependent sentences would have to be handled differently, with semantic rules that interpreted sequences of sentences and that applied to surface structures. These rules would duplicate in many ways the rules dealing with underlying structures: dependent sentences relate to other sentences just as complements relate to matrix sentences. Since rules relating to underlying structures cannot deal with sequences of sentences, they are not adequate to account for temporal specification. It is preferable to adopt the third alternative suggested at the beginning of this discussion: a grammar with an autonomous syntax and semantic interpretation rules that apply to surface structure. One should perhaps not be very surprised that a surface structure approach is called for in this domain. Surface structure considerations have already proved to be important for other aspects of sentences that involve deixis.
3 Autonomous Syntax and Surface Interpretation This approach has the advantages of autonomous syntax, without the disadvantages of underlying structure semantic rules. I will propose a set of semantic interpretation rules that apply to the surface structure of sentences and of domains larger than a sentence. These rules, and the relevant syntactic rules, are stated in Section 3.2 below. In Section 3.1 I give a general account of how the rules are constructed. The semantic interpretation rules relate surface structures to semantic representations. Surface structures are interpreted via model configurations that represent temporal elements as they occur in sentences. Associated with each configuration is a semantic representation that constitutes its interpretation. To interpret a particular sentence, one matches its surface configuration with the appropriate model, and then consults the associated semantic representation.
Temporal Expressions in English
129
The semantic representations are organized according to Reichenbach’s scheme: for each sentence, RT and its relation to ST, the relation of ET to RT, and ET (where appropriate), will be given. The actual values of these notions will be given rather crudely, in terms of the adverbials and other forms that occur in sentences; a more sophisticated treatment will no doubt be desirable at a later stage.
3.1 Principles of Interpretation Surface structure configurations will be interpreted according to the principles developed in the first part of this paper. The principles are simple, and can be applied by referring to the information given in surface structure. Recall what information is necessary for the interpretation of a sentence. One must know the relational values of the tense and adverbial(s); whether the sentence has a frequency adverbial; whether the sentence is syntactically independent or not. With this information, a sentence can be interpreted. I will show, in discussions of the main types of configurations that appear, just how the principles of interpretation are applied. Independent sentences have tense and a time adverbial.21 A decision must be made as to whether these constitute an RT combination: if they do, tense and adverbial have the same relational value, or the relational value of tense is simultaneous. The combinations of tense and adverbial are repeated here for convenience: RT Present Past Past Future Future — — —
Tense present past past present present present past past
Adverbial Present Past Unanchored Future Unanchored Past22 Present Future
The principles of interpretation compute the value of RT, and the relation between ET and RT, by considering the relational values of the time expressions in each configuration. As an example, consider sentence (10) below. The surface (temporal) configuration of (10) is (11). The configuration includes only temporal elements, and indicates their relational value; this information is crucial, as shown above, for interpretation. The configurations also indicate syntactic embedding where relevant. 21 Sentences
with tense alone are not complete semantically and therefore are not independent. combination is that of the Historical Present and recurs, of course, throughout the rules. It will be noted but not discussed. 22 This
130
C.S. Smith
(10) Sue boarded the ferry before noon. (11) S [pastTemp [P Unanch]]. (11) is to be read as follows: tense is represented as ‘present’ or ‘past’; temporal adverbials are labelled ‘Temp’, to distinguish them from frequency adverbials, labelled ‘Freq’. ‘p’ indicates the preposition that introduces the adverbial; ‘Unanch’, ‘Past’, ‘Present’, and ‘Future’, indicate the four types of adverbials. The configuration (11) is matched with its associated semantic representation, (12). (12)
S [ RT
← ST: Unanch ET ← RT ].
(12) indicates the relational value of RT, and the relation between ET and RT.23 The actual semantic representation for (10), in which the particular material of the sentence occurs rather than the general symbols←and Unanch, is (13). (13)
S [ RT:
Past, noon ET before RT ].
The interpretation of other sentences with RT combinations of tense and adverbial is determined in the same manner. Adverbials specify RT no matter how complex they are, so that the internal structure of adverbials need not be included in the model configurations. The actual position of the adverbial in surface structure is irrelevant for this interpretation (although it may well be important for determining other properties such as those associated with emphasis and presupposition). Habituals, that is, sentences with frequency adverbials, are interpreted in the same manner also, except that the frequency value functions as ET: (14) Steve will play tennis every day next summer. (15) S [ present [Freq] Temp [Future] ]. (16) S [ RT → ST: Future ET = RT ET: Freq ]. If no introductory preposition appears in a temporal adverbial, ET = RT. Habitual sentences such as (17), that lack specification of RT, will not be included in the rules given here. (17) Steve played tennis every day. For the interpretation of sentences with complements, it is essential to determine how the complement is related to the matrix. If the complement shares a time with the matrix RT, its RT is ET1 and its adverbial specifies ET and the relation between ET and RT; if the complement does not share the matrix RT, it establishes RT as usual but is oriented to another sentence. Complements that share a time with the 23 This
notation, introduced in Part I, represents that Past RT is anterior to ST, Present RT simultaneous with ST, and Future RT posterior to ST. Arrows will be used throughout the rules to indicate anteriority or posteriority, and the equals sign will represent simultaneity.
Temporal Expressions in English
131
matrix have the same tense as the matrix sentence.24 An example of such a sentence is (18); its temporal configuration and associated semantic representation are (19) and (20), respectively. (18) They told Rachel last week that Pete left Texas in 3 days. (19) S [ past Temp [ Past ] s [ past Temp [ Future ] ] ].25 (20) S [ RT ← ST: Past ET = RT s [ RT = ET1 ET → RT ET: Future ] ]. Complements that do not share the RT of the matrix are oriented to the matrix RT rather than to ST; the complement is interpreted as simultaneous or sequential to the matrix RT. This difference in orientation is indicated in the semantic representation of such complements. For example: (21) Reuben will announce next week that he conferred with Jack on Tuesday (22) S [ present Temp [ Future ] s [ past Temp [ P Unanch ] ] ] (23) S [ RT → ST: Future ET = RT s [RT ← RT1 : Unanch ET = RT ] ]. Sentences that have two adverbials as those above (resulting from the Reduction transformation) are analyzed in the same manner as those above, except that RT is established with the leftmost adverbial and tense. For example: (24) Last week, Bill had won the race 3 days ago. (25) S [ Temp [ Past1 ] past have Temp2 [Past2 ] ]. (26) S [ RT ← ST: Past1 ET ← RT ET: Past2 ]. Just as the rules assign a relational interpretation to adverbials, auxiliary have is assigned the interpretation of anteriority. Sentences with nonRT combinations of tense and adverbial do not have a complete semantic representation. The model configurations that correspond to such sentences will be associated with semantic representations marked accordingly: (27) Mary was amused now. (28) S [ past Temp [ Present ] ]. (29) S [ INCOMPLETE ]. To interpret a sentence like (27) one looks for an appropriate larger configuration. What is needed is a sentence that establishes RT, and that has the same tense as. The two sentences can be analyzed as sharing RT, and the adverbial of the incomplete, or dependent sentence specifies ET. A configuration that would provide an interpretation for (27), for instance, is (30): 24 Complements will be interpreted by the Sharing Principle when this interpretation is possible, or
by the Extended Sharing Principle, according to the rules of application developed in Part I. adverbials play two roles in sentences like this: they indicate the relation between ET and RT, and specify ET.
25 Anchored
132
(30) [S [ past Temp [ Unanch ] ] S [ past
C.S. Smith Temp [ Present ] ] ].
and the semantic representation associated with this configuration is (31): (31) [S [ RT ←ST: Unanch ET = RT ] S [ RT = RT1 , ET = RT ] ]. Since I have been concerned in this paper only with establishing single complete semantic interpretations, no configuration will be associated with a semantic representation in which more than one RT occurs in an independent sentence. A set of model configurations, their associated semantic representations, and examples of each is presented in the following section. They correspond to the types of sentences discussed here: sentences with tense and adverbial that establish RT, habitual sentences, complement sentences, sentences with tense and adverbial that do not establish RT. These rules are far from complete: a number of problems remain to be solved or investigated before a less tentative set of rules can be offered. However, the rules that are given indicate clearly that this approach is both desirable and feasible. I mention briefly some of the problems, ignored in the rules, that would have to be handled in a complete set of interpretive rules. The semantic representations interpret sentences as complete whenever possible. This means that they do not deal with potential ambiguity. Sentences are potentially ambiguous when they have an interpretation in isolation, but may receive a different interpretation(s) in different context(s). Example (32), for instance, is potentially ambiguous: (32) Jonathan played croquet on Tuesday. On one reading it is complete, indicating a Past RT and ET simultaneous with RT. There are two incomplete readings, dependent on context: in the neighborhood of a frequency adverbial (32) may be taken as habitual; in the neighborhood of a sentence with which it might share RT, it might be taken as a dependent sentence that specified ET (this would involve the → interpretation of the Unanchored adverbial). Only the first of these readings will be accounted for in the rules that follow, although a full set of interpretive rules should deal with potential ambiguity in some way. I plan further research on this topic, which includes not only potentially ambiguous sentences but also the conditions that force one or another reading of such sentences. The configurations for interpreting incomplete sentences are little more than schematic: they indicate adjacent sentences with the appropriate temporal elements. At this point it is not possible to give a fuller account; what is needed is, again study of the conditions for semantic dependency between sentences. The rules are stated only for single embeddings; however, multiple embeddings are interpreted by the same principles, so that new situations would not arise if the rules were to be extended to cover sentences in which more than one sentence were embedded. Several difficulties arise in connection with the recognition of adverbials in surface structure. Since the Adverb Movement transformation may break up complex adverbials, it is necessary to amalgamate them into one unit for
Temporal Expressions in English
133
interpretation. No mechanism has been provided for doing this. It is also necessary to distinguish between distributed adverbials (the result of Adverb Movement) and sentences with two adverbials; both have similar surface structures. Compare, for instance, (33) and (34): (33) On Tuesday, Tom had arrived at noon. (34) Last week, Tom had arrived 3 days ago. Both sentences have two temporal prepositional phrases, one initial and one final. Yet their interpretations are entirely different: (33) contains one complex adverbial that contributes to the specification of a Past RT; (34) contains two adverbials, one specifying RT and the other ET. (34) results from the reduction of two sentences, and its adverbials cannot be amalgamated into one. Thus it contrasts with (33), whose adverbials can be amalgamated. Compare for instance: (35) Tom had arrived at noon on Tuesday. (36) ∗ Tom had arrived 3 days ago last week. Another difficulty involves the interpretation of habitual sentences. Frequency adverbials may contain temporal adverbials (e.g. every day at noon). In habitual sentences, then, a temporal adverbial may be part of the frequency adverbial or part of the specification of RT. If an habitual sentence has two temporal adverbials, it may not be clear which one is associated with the frequency adverbial. Actually, the difficulty only arises with temporal adverbials that indicate an interval, as in the examples below: (37) Jeff rode the rollercoaster from morning till night from June till August. (38) From June till August, Jeff rode the rollercoaster from morning till night. (39) From morning till night, Jeff rode the rollercoaster from June till August. Such sentences may have to be regarded as structurally ambiguous. They are rarely ambiguous, however, when the individual lexical items are considered. Such considerations may have to be provided for in complete rules for sentences of this type. Another omission that I wish to mention is the serious consideration of incomplete sentences. Although I have attempted to account for their interpretation above, informally, I am not yet certain of the principles involved. Incomplete habituals are not treated at all, nor are sentences with complements in which the matrix sentence does not establish RT. To complete this list (although not, I am sure, the list of omissions in the rules), the absence of complex adverbials that have embedded sentences, such as (40), should be noted; (40) Mrs Bogen left before Mrs Mackay had arrived. Interpretation of these will be integrated into the rules at a later stage; their syntactic analysis is given in Smith, 1975a.
134
C.S. Smith
3.2 Interpretive Rules In this section I present a list of interpretive rules. Each rule has the form of a pair, the pair consisting of a model surface structure configuration and a semantic interpretation. The pairs are numbered; following the rules is a list of example sentences, numbered to correspond with the rules. The reader will thus be able to find an example for every rule given. The first group of rules, list A, accounts for simple sentences with tense and one time adverbial, or have, or both, or neither. All combinations are included. Rules with a domain larger than a sentence, which allow interpretation of those configurations marked Incomplete, are given at the end of the list. List B accounts for sentences with complements; list C, for sentences with two time adverbials. Model configurations are given to the left and semantic representations to the right. All the examples and configurations deal with specifying sentences: frequency adverbials are not included in these rules.26 3.2.1 Sentences with One Time Adverbial
26 Frequency
adverbials would present no problem for the rules; they have been discussed in the preceding part of this paper. They are omitted here because they present problems of recognition rather than analysis, and do not require separate treatment in terms of RT from the sentences analyzed here. 27 Historical Present. 28 Only a very few prepositions appear with explicitly Present adverbials. 29 No other adverbials will appear in such sentences because of restrictions on have.
Temporal Expressions in English
135
Configurations with complete interpretations of the incomplete sentences above are given in this list. The incomplete sentences appear as S2 in the configurations. The facts are undoubtedly more complicated and more fluid, allowing more than one order and a certain amount of intervening material. As noted above, these particular configurations are merely schematic since not enough is known about configurations of this type to allow a complete presentation.
In the interest of brevity and readability, the first S of configurations such as 1820 will be represented simply as Past RT, Present RT, Future RT; the relation of ET to RT in the first S will be omitted in the semantic representation since it is not relevant to the interpretation of the incomplete sentence, S2 in these configurations. The interpretation of these incomplete, or dependent, sentences is consistently that RT is ET of a preceding sentence; recall that sentences share RT when the dependent sentence has the same tense as its matrix or independent anchoring sentence.
30 See
preceding note.
136
C.S. Smith
3.2.2 Sentences with Complements These rules account for the semantic interpretation of complements, in sentences where the matrix S establishes RT, and there is one embedding. Complement sentences are interpreted as sharing the matrix RT when possible; when the two have different tenses and this interpretation is not possible, the complement is oriented to the matrix or to another sentence; when the two have different tenses and the complement has present tense, however, the orientation interpretation is not possible and the complement shares the tense of another sentence (a performative or a preceding sentence). Matrix sentences that fail to establish RT are not included. Their interpretation will involve another sentence; the patterns of embedding and dependency at this level of complexity are the same as those of multiple embedding, which is not different from single embedding in this respect. As in the preceding rules involving more than one sentence, the matrix S will be represented simply as having Past RT, Present RT, or Future RT. Larger configurations required for interpreting the incomplete sentences are given after the rules for matrix sentences of each time period. To conserve space, temporal adverbials are labelled ‘T’, except when they are referred to in a semantic representation.
Temporal Expressions in English
137
Configurations with a larger domain than those above, which include the configurations marked Incomplete, are given below. They all involve past tense complements of sentences with Present RTs; those that do not have RT combinations are dependent on separate Past sentences that establish RT for them. Configurations with a larger domain than the above, which include the configurations marked Incomplete, are given below. They all involve present tense complements of sentences with Past RTs; those that do not have RT combinations
31 See
preceding note. Present. 33 This sentence is odd semantically because the deictics are contradictory. 32 Historical
138
34 Historical
C.S. Smith
Present.
Temporal Expressions in English
139
are dependent on separate (or superordinate performative) sentences that establish RT for them. The two possibilities will not be distinguished structurally here because the structure of a performative sentence and its relation to the sentence it dominates is not entirely clear; however, the immediately preceding sentences with Present RTs in these configurations could be performatives.
35 Historical
Present.
140
C.S. Smith
Configurations with a larger domain than those above, which include the configurations marked Incomplete, are given below. They all involve past tense complements of sentences with Future RT; those that do not have RT combinations are dependent on separate Past sentences that establish RT for them.
This completes the list of semantic interpretation rules for sentences with complements.
3.2.3 Sentences with Two Adverbials The sentences interpreted below result from the Reduction Transformation, which inserts a time adverbial from one sentence into another. As noted above, the interpretation of such sentences is similar but not identical to that for complement sentences. RT is established by the first adverbial and tense; the second adverbial gives ET and
Temporal Expressions in English
141
its relation to RT. The construction is less flexible than the complement construction. See the brief discussion of the Reduction Transformation in the following section; I list here only the combinations that are clearly grammatical, although even such judgments as that are difficult to make in this shadowy area. 80. 81. 82. 83. 84. 85. 86.
87.
88.
Future [T [Pres] presT [P ]] S [RT=ST ET → RT ET: Temp2 ] S 1 1 Unanch
Past [T [Pres] presT [P ]] s [RT=ST ET ← RT ET: Temp2 ] S 1 2 Unanch S [T1 [P Past]pastT2 [P Fut]] s [RT←ST: Temp1 ET → RT ET: Temp2 ]
Past [T [P Past] past (have)T [P ]] s [RT←ST: Temp1 ET←RT Temp2 ] S 1 2 Unanch S [T1 [P Fut] presT [P Pres]] S [RT → ST: Temp1 ET=RT]
Fut ]] s [RT→ST: Temp1 ET → RT ET: Temp2 ] S [T1 [P Fut] presT2 [P Unanch
Fut ]] s [RT → ST: Temp1 ET → RT ET: S [T1 [P Unanch] presT2 [P Unanch Temp2 ]
Fut ]] s [RT ← ST: Temp1 ET → RT ET: S [T1 [P Unanch] past T2 [P Unanch Temp2 ]
Past ]] s [RT ← ST: Temp1 ET ← RT ET: S [T1 [P Unanch] past have T2 [P Unanch Temp2 ]
I now give a list of sentences to go with the rules of semantic interpretation. Only one version of each sentence is given here; that is, the schematic configurations for domains larger than a sentence do not have example sentences. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
Jerry is repairing the spinnaker Jerry has repaired the spinnaker Jerry is repairing the spinnaker now Hart won the election last year Hart is leaving tomorrow Hart is leaving on Wednesday Hart has repaired the spinnaker now They have eaten all the cookies while you were out Greta Garbo starred in the film “Anna Karenina” Burt Lancaster had starred in the film “The Killers” Judy Garland starred in the film “A Star is Born” years ago Della sent the proofs to the printer on Wednesday Della was leaving tomorrow Della was optimistic now
142
C.S. Smith
15. 16. 17. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39.
Della has sent the proofs now Della had sent the proofs to the printer a week ago Della had sent the proofs on Wednesday Allan believes that the committee is supporting him now Allan believes that the secretary has resigned Allan believes that the committee has decided now Historical Present (not covered in this paper) Shirley thinks that the party starts in two hours Shirley thinks that the semester begins on Thursday Shirley thinks that the volcano erupted Shirley thinks that it snowed yesterday Shirley thinks that Mark was getting suspicious now Shirley thinks that the semester began on Thursday Shirley thinks that Tom was leaving in 3 days The narrator says that the Fat Lady had left The narrator says that the circus had closed now
on Tuesday The narrator says that the elephants had arrived yesterday The office announced yesterday that Professor Thrum is available now The office announced yesterday that Professor Thrum has arrived The office announced yesterday that Professor Thrum has retired now Historical Present The office announced yesterday that Professor Thrum leaves in 3 months The office announced yesterday that Professor Thrum retired The office announced yesterday that Professor Thrum retired last week The office announced yesterday that Professor Thrum retired in April The office announced yesterday that Professor Thrum was retiring in 3 months The office announced yesterday that Professor Thrum was away now The office announced yesterday that Professor Thrum had retired The office announced yesterday that Professor Thrum had retired now The office announced yesterday that Professor Thrum had retired
last week on Wednesday The candidate will say next week that he is available (now) The candidate will say next week that his opponent has been unreliable The candidate will say next week that he has selected his running-mate now Historical Present in 3 days The candidate will say next week that he announces his choice on Thursday The candidate will say next week that he won the election last month The candidate will say next week that he saw the documents in April The candidate will say next week that he realized the difficulties now The candidate will say next week that he expected to leave in 3 days The candidate will say next week that he had overestimated the demand for hotdogs
40. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.
63. 64. 65. 66. 67. 68. 69. 70. 71. 72.
Temporal Expressions in English
143
73. The candidate will say next week that he had reserved judgment at this time 74. The candidate
will say next week that he had seen the documents last summer in July
in 3 weeks 80. Today, Harry leaves on Tuesday
3 weeks ago 81. Right now, Harry is leaving on Tuesday 82. Last week, Harry was leaving in 3 days
2 weeks ago 83. Last week, Harry had left on Tuesday 84. Next week, Harry leaves now
in a month 85. Next week, Harry leaves on Tuesday
on Sunday 86. On Tuesday, Harry leaves in 5 days
in 3 days 87. On Friday, Harry was leaving on Tuesday
3 days ago 88. On Monday, Harry had left on Saturday
3.3 Syntactic Rules for Temporal Elements The syntactic rules relevant to this paper are few. They include phrase structure rules for time adverbials and the auxiliary, with the constraints on co-occurrence discussed above; the transformation(s) of Adverb Movement; the Reduction Transformation; a Tense Copying Transformation for embedded adverbial sentences. Neither the auxiliary PS rule not the Tense Copying Transformation will be discussed in this paper, but I shall give a brief account of the others. The time adverbial rules allow for a temporal adverbial and an habitual, both of unbounded complexity. As shown in Part I, only one time adverbial appears in a sentence. Time adverbials consist of prepositional phrases or embedded sentences, or both; in principle there is no upper bound to the possibilities, as the examples suggest: (41) Lee arrived on Tuesday on Tuesday at 9 am in the morning at 9 on Tuesday in the morning at 9 on Tuesday in May in the morning at 9 on Tuesday in the first week of May. . . (42) Lee arrived before Dennis left before Dennis left last week
144
C.S. Smith
before Dennis left after Steve called for help before Dennis left after Steve called for help when he saw the burglars steal in as soon as the fire started. Embedded adverbial sentences do not have independent tense, as noted above; the Tense Copying Transformation copies the tense of the main sentences into the adverbial sentence. Frequency adverbials consist of a frequency and a temporal constituent. The frequency constituent indicates the rate of recurrence per unit of the relevant event or state; the unit must have an indefinite determiner: (43) 3 times a day ∗ 3 times the day 3 times that day (not habitual) The temporal constituent of a frequency adverbial indicates the particular time of recurrence, and may in principle have all the complexity of independent time adverbials, for instance: (44) John swims every Friday before he goes home after he has played tennis. The following phrase structure rules will generate the appropriate structures. Time adverbials are optional, of course, and their separate constituents need not appear together. Time adverbial Temporal PP Habitual Frequency Unit
→ (Temporal) + (Habitual) → PP
NP →P (PP) S → (Frequency) + (Temporal) → Number + Unit → Indef Det + Noun
The Adverb Movement Transformation moves all or part of a temporal adverbial to another position in the sentence. Time adverbials appear in a number of positions, so that more than one rule may be necessary; other types of adverbials also move, so that the rule should perhaps not be restricted to time adverbials. Generally, if part of an adverbial is moved and part left behind, it is the rightmost (in standard order the most general semantically) constituent, for example: (45) Pauline called at 2 pm on Sunday. (46) On Sunday Pauline called at 2 pm. (47) At 2 pm Pauline called on Sunday.
Temporal Expressions in English
145
In spite of the oddity of (47), I think that all combinations should be allowed; one can find an appropriate environment – emphasis, or a preceding question – for combinations that are odd in isolation. The Reduction Transformation applies to two sentences and results in one sentence. Schematically, it looks like this: S1 :[ProS] V TempAdv 1 2 3 → S2 :NP V X TempAdv 4
4,1
Since this is a syntactic transformation, the temporal adverbials are specified syntactically, allowing all possible combinations of tense and adverbials. Recall that in sentences resulting from this transformation the first adverbial and the tense establish RT, and the second adverbial specifies ET. The analysis is the same as for sentences with complements that share RT with their matrix sentences; not all combinations, then, are interpretable. For instance, sentences in which the tense and first adverbial do not form an RT combination must be blocked; (48) and (49) exemplify: (48) ∗ Last week, Bill is tired now. (49) ∗ Next month, Alma left yesterday. The semantic interpretation rules can block sentences such as (48) and (49), since the tense and first adverbial are not RT combinations. There are other uninterpretable combinations also, however: (50) ∗ Next week, Bill is tired now. The formulation of principles to block sentences like (50) awaits further understanding of the construction, and of temporal specification across sentences in English.
References Benveniste, E.: 1966, Problèmes de linguistique générale, Gallimand, Paris. Boyd, I. and Thorne, J.P: 1969, ‘The deep grammar of modal verbs’, Journal of Linguistics 5, 57–74. Braroe, E.: 1974, The Syntax and Semantics of English Tense Markers. Monographs from the Institute of Linguistics, University of Stockholm. Monograph 1. Crystal, D.: 1966, ‘Specification and English tense’, Journal of Linguistics 2, 1–34. Emonds, J.: 1974, ‘Parenthetical clauses. You take the high node and I’ll take the low node’, Papers from the Comparative Syntax Festival, Chicago Linguistics Society, Chicago, Ill. Gallagher, M.: 1970, ‘Adverbs of time and tense’, Papers from the 6th Regional Meeting, Chicago Linguistics Society, Chicago, Ill. Hamburger, K.: 1968, Die Logik der Dichtung, E. Klett, Stuttgart. Hofmann, T.: 1966, ‘Past tense replacement and the modal system’, Harvard Computation Laboratory, National Science Foundation Report No. 17, Cambridge, Mass.
146
C.S. Smith
Hornstein, N.: 1975, ‘As time goes by: a small step towards a theory of tense’, Montreal working papers in linguistics 5. Jenkins, L.: 1972, Modality in English Syntax, Ph.D dissertation, M.I.T. Cambridge, Mass. Jespersen, O.: 1931, Modern English Grammar on Historical Principles, George Allen and Church, London. Kiparsky, P.: 1968, ‘Tense and mood in Indo-European syntax’, Foundation of Language 4, 30–57. Kiparsky, P., and Kiparsky, C.: 1970, ‘Fact’ in Bierwisch and Heidolph, eds. Progress in Linguistics, Mouton, The Hague. Lakoff, G.: 1969, ‘Presuppositions and relative grammaticality’, in W. Todd, ed., Studies in Philosophical Linguistics, Great Expectations Press, Cincinnatti, Ohio. Leech, G.: 1969, Toward a Semantic Description of English, Indiana Univ. Press, Bloomington. Lawler, J.: 1972, ‘Generic to a fault’, Proceedings of the 8th Regional Meeting. Chicago Linguistics Society, Chicago, Ill. McCawley, J.: 1971, ‘Tense and time reference in English’, in Fillmore and Langendoen, eds., Studies in Linguistic Semantics, Holt, Rinehart, Winston, New York. Reichenbach. H.: 1947, Elements of symbolic logic. University of California Berkeley, Calif. Riddle, E.: 1976, ‘A new look at sequence of tense’, unpublished paper. Ross, J.: 1967, Constraints on Variables in Syntax. Ph.D dissertion, M.I.T. Cambridge, Mass. Sadock, J.: 1969, ‘Hypersentences’, Papers in Linguistics 1, 283–370. Sadock, J.: 1974, Towards a Linguistic Theory of Speech Acts, Academic Press, New York. Smith, C.: 1975a, ‘The analysis of tense in English’, Texas Linguistic Forum 1, 71–89. Smith, C.: 1975b, ‘A new look at auxiliary have in English’, Proceedings of the Sixth Annual Meeting of the Northeastern Linguistic Society, McGill University, Montreal. Smith, C.: 1976a, ‘Present curiosities’, Proceedings of the 12th Regional Meeting, Chicago Linguistics Society, Chicago, Ill. Smith, C.: 1976b, ‘A theory of auxiliary have in English’, Indiana University Linguistics Club. Verkuyl, H.: 1971, On the Compositional Nature of the Aspects, Foundations of language supplementary series 15.
The Temporal Reference of the English Futurate Carlota S. Smith
There are two closely-related ways of talking about the future in English, the willfuture and the futurate. The constructions are not interchangeable, because the futurate involves some kind of plan, schedule, control, or pattern of events, while the will-future is not so restricted. This restriction on the futurate is familiar and has been discussed by grammarians at least since Jespersen 1931. Less familiarly, the futurate does not fit neatly into a general syntactic-semantic account of temporal reference in English: neither adverbial nor aspectual forms play their characteristic roles in this construction. Because it is problematic, the futurate is of some interest for the study of temporal reference. Particularly interesting are the truth-conditional and presentational variations of aspect. The futurate in a general account of temporal reference is the specific topic of this paper. The futurate serves also as an area in which to discuss and demonstrate an approach to the general question of how syntactic and semantic structures should be related in a grammar. I will argue for intermediate semantic structures that represent abstract scope relations yet are relatively close to the syntax. The rationale for this type of structure is that it would serve as input for different types of semantic constructs, such as model-theoretic and text structures. As an introduction, consider the following futurate sentences and two questions about their temporal interpretation. Present tense and a future time adverbial are the hallmarks of the futurate construction. (1) a. b. c. d. e. f.
Roger is playing squash tomorrow. Roger plays squash tomorrow. The sun is setting at 5 this afternoon. The sun sets at 5 this afternoon. Emily is leaving tomorrow. Emily leaves tomorrow.
The first question concerns the time that a futurate talks about, and the role of tense and adverbial in specifying it. Intuitively it is clear that the futurate spans two times, the present moment and the future time indicated by the adverbial. It also seems clear that the futurate is “about” the present. In semantic interpretation, to evaluate a sentence for truth or falsity we need to know what time a sentence is R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_5,
147
148
C.S. Smith
about; to adopt the useful term of Reichenbach 1947, we need to know the reference time of the sentence.1 Generally the time adverbial contributes to the specification of reference time, but this does not seem to be the case with the futurate. The matter is discussed in section 1 below. Now consider the difference in interpretation between the aspectual pairs of the sentences in (1). The differences are, oddly, rather elusive. For example, a and b do not differ – as one might expect – in that the progressive talks about an ongoing activity of squash-playing whereas the simple form does not. Nor do the other pairs differ in this way. Certain features can be found that differentiate between particular aspectual pairs, but it is difficult to generalize them. For instance in some futurates the progressive form seems to make a slightly weaker prediction than the simple form; but the pair c and d show that this is not always the case. Indeed, it seems most unlikely that one interpretive notion such as uncertainty would be appropriate for the entire range of futurate sentences. Rather, what is wanted is a semantic account of the futurate that allows the pragmatic interpretations that occur. I present such an account in the following sections.
1
Generally in English the temporal location of a sentence is the time specified by the tense and time adverbial, in combination.2 To evaluate the truth or falsity of a sentence one finds the time specified and asks whether the sentence is true at that time. For example, take the sentence of (2): (2) a. b. c. d.
Mary worked yesterday. We were swimming at 6 o’clock. He is sleeping now. They will arrive tomorrow.
To evaluate these sentences, we look at the time specified by the combination of tense and time adverbial and ask whether the event in question occurred at that time. Thus for (2a) one asks whether, at the time yesterday, the sentence Mary work is true; for (2b) one asks whether at the time 6 o’clock past, the sentence We swimming is true. This is not the pattern on which futurate sentences are understood and evaluated, however. Futurates are not evaluated at the time specified by the future adverbial but at the moment of speech, or Speech Time (ST).3 To see this, consider the examples of (3): (3) a. John rehearses tomorrow. b. John is rehearsing tomorrow. 1 Reichenbach has presented an important analysis of temporal reference, which I will draw on below. 2 Arguments and rules for this analysis are presented in Smith (1978), (1980). 3 Reichenbach argues that the temporal reference of a sentence involves three times, Speech Time, Reference Time, and Event Time.
The English futurate
149
To evaluate these sentences we do not find tomorrow and ask whether John rehearse is true. Nor do we ask the question at ST; whether John is actually rehearsing when these sentences are uttered is irrelevant. Rather, to evaluate the sentences of (3) one looks at the situation at ST, to see whether it licenses the prediction. Thus one might ask whether John’s name is on the rehearsal schedule for the following day, whether he will be in town then, etc. If the situation at ST licenses the prediction John rehearse tomorrow, the sentences of (4) are true. Thus the role of the future adverbial is not to specify the time of evaluation, but rather another time – here, the time of the predicted situation. As noted above, the time of evaluation of a sentence is Reference Time (RT) in Reichenbach’s terminology; the future time specified in a futurate is Event Time (ET). I have suggested that the Reference Time of a futurate sentence is the moment of speech, ST. There is some evidence for this from time adverbials. Futurates can always have a second time adverbial that indicates present time: (4) Now At this moment John rehearses tomorrow Today In these examples the standard pattern of interpretation of tense and adverbial occurs. In this pattern the first (present time) adverbial specifies RT in combination with the tense, and the second adverbial specifies ET. I have shown (Smith 1978) that when RT is specified and another adverbial is present, that adverbial specifies ET. Unfortunately examples like (4) do not give conclusive evidence for the structure of futurates, because there are other constructions that also seem to allow a present adverbial. The other constructions are not of the same type, however. They have special restrictions and are analyzed as reductions of sentences with sentential complements. For extensive discussion see Smith (1978). Stronger evidence for the claim that futurates are evaluated at ST comes from a comparison of futurates and will-futures. Compare the sentences of (5): (5) a. Mary will rehearse at 3 o’clock. b. Mary rehearses at 3 o’clock. The will-future is evaluated at the time specified by the future adverbial: one looks at a future 3 o’clock and asks whether Mary rehearse is true. But for the futurate, one looks at ST and asks whether Mary rehearse at 3 o’clock is true. And, tellingly, futurates are compatible with a question about what will actually happen in the future whereas will-futures are not. Thus (6a) is good, (6b) distinctly odd. (6) a. Mary leaves tomorrow but I won’t be surprised if she changes her mind. b. Mary will leave tomorrow but I won’t be surprised if she changes her mind. This is just what one would expect if the first conjunct of (6a) is evaluated at ST whereas that of (6b) is evaluated at tomorrow. (The second conjunct of both is evaluated at tomorrow.)
150
C.S. Smith
These points show that the reference time of the futurate is indeed ST, and that the future adverb specifies a time that has another function in the sentence. If a present adverbial is taken as underlying all futurate construction, their temporal interpretation can be handled with general rules needed anyway for the interpretation of English. I suggest in Smith (1978) a set of such rules, that map combinations of tense and time adverbials into times in a framework based on Reichenbach’s account of temporal reference. In the system a combination of present tense and a present time adverbial indicates a present RT. I now suggest that the interpretive rules be written so as to map surface structures, or sentences, onto abstract semantic structures. These structures would represent the semantic scope of temporal expressions as well as their interpretation in times.4 Such structures mediate between the syntactic and semantic structures of a language, by giving information relevant for semantic structures of different types; they would not be limited to temporal expressions, of course. For the constructions under discussion here, I propose that abstract structures roughly along the lines of (7) below, be set up to model the semantic scope of temporal (and other) expression. Structures like (9) would be the input to interpretive rules for e.g. model-theoretic or situational semantics, for constructing text structures, etc. The higher nodes represent general notions (such as reference time, event time, aspect) and the lower nodes represent notions and expressions particular to a given language. (7) Temporal Time Tense
Present Past
S
Frequency, Aspect Duration Adverb
NP
VP
Progressive Simple
In this structure temporal expressions appear under the Temporal and Aspectual nodes; temporal expressions may also be located in the inner sentence, as shown in (9). (For the purposes of this paper, I will consider only time adverbials of location, ignoring frequency and duration; the analysis of modals, including will, is also ignored.) Tense and time adverbial, giving the reference time of a sentence, are dominated by the Temporal constituent. So, for example, (8) represents a temporal structure associated with sentence (2a).
4 See the discussion of the perfect toward the end of the paper. The perfect, like the futurate, involves two times other than Speech Time.
The English futurate
151
(8)
Temporal
S’
Time
Aspect
tense
adverb
past
yesterday
S
Simple
NP
VP
Mary
work
I propose that in the representation of the futurate a present adverbial occurs in the Temporal constituent, together with present tense. The future adverbial is part of the inner sentence. (9) below presents a representation of a temporal structure associated with sentence (1a). (9) S’’ S’
Temporal Aspect
S
Time Progressive tense
adverb
present
present
NP
VP
Adverb
Roger play squash tomorrow
It should be noted that structures like (9) are not incompatible with the insights of Reichenbach about temporal reference. The structure allows two places for a time adverbial (the Temporal constituent and the inner sentence) and thus for the specification of two times. Reichenbach’s system has three times: Event Time, Reference Time, and Speech Time. Speech Time is the keystone of the system, since past, present and future, the general reference times, are determined by their relation to ST. Past time precedes ST, future time follows ST, present time is simultaneous with ST. Thus although ST and its relation to RT is not directly specified in structures such as (8) and (9), it is an essential part of their interpretation.
2 I now turn to the aspectual interpretation of futurate sentences.5 I will show that the progressive does not have its usual interpretation in the futurate, and then discuss proposals for handling it in the futurate construction. 5 I shall discuss only perfective and imperfective aspect in this paper; references to the simple form will be to cases where it indicates perfective aspect.
152
C.S. Smith
As a preliminary, consider the more or less standard interpretation of the progressive and simple verb forms in English. I will assume without discussion that the progressive and simple forms signal the aspect of a sentence, the progressive indicating imperfective and the simple perfective aspect.6 Following traditional accounts and recent model-theoretic work on aspect, I take the meaning of the perfectivity contrast to be roughly as follows. The perfective presents an event as a whole, including its initial and final endpoints and without regard to internal structure; the imperfective presents an event as activity or process, ignoring endpoints7 . For events with natural endpoints, process is associated with a final outcome or result. I do not consider states in this paper; for discussion, see Smith (1981) and (1983). It is important to note that perfectivity crucially affects the semantic interpretation of a sentence. Whether a sentence is true may depend on its aspect, for sentences dealing with certain types of events. Thus for a given situation John was drawing a circle may be true, but John drew a circle may be false; perhaps John changed his mind in media res and drew an eggplant instead. As Vendler, Dowty, and others have emphasized, differences like this can arise because certain events involve completion in their temporal schemata. Such events have at least two essentially different stages (before completion and completion itself) and are said to be heterogenous. Some events do not involve completion – Vendler’s example is pushing a cart – and they are homogenous.8 Unlike the aforementioned cases, sentences about homogenous events do not differ in truth value if they differ aspectually. So, They were running is true when They ran is true because there is no outcome or result with which the process of running is associated. The standard interpretation of the progressive does not hold for futurates as Dowty (1977) notes. Consider the interpretation of the following, for instance: (10) a. He is rehearsing tomorrow. b. She is writing the report in the morning. We do not understand these sentences to say that a process of rehearsing or reportwriting is actually going on at ST; yet ST is the reference time of the futurate, the time being talked about. What we do understand is that certain preliminaries are under way, and these preliminaries are associated with the outcomes of rehearsing and report-writing. The general notion of process that holds for the progressive clearly needs some revision if it is to account for futurate progressives.
6 Adverbials also may contribute to the aspectual interpretation of a sentence, as emphasized in the work of Vendler and in the references of the following note; however I shall not be concerned with adverbials and aspect in this paper. 7 These matters are discussed in Dowty et al. (1981), Vlach (1981), Bennett and Partee (1972). 8 Heterogenous events are, in the terminology of Vendler, achievements and accomplishments; homogenous events are activities.
The English futurate
153
There is a natural extension of the way a process is associated with an outcome, that might include futurate progressives in a general account of the progressive.9 We can allow a process associated with an outcome to include the preliminaries of a situation – those stages at which plans are made, patterns noted, schedules organized. The futurate progressive, according to this view, talks about the early stages of a process associated with an outcome; later stages might include other processes that are taken as actions rather than preliminaries. Note that the line between preliminary and actual stages of an activity or process can be drawn, although there are some unclear cases. To adapt an example of Vlach’s: if I am lying in bed thinking about my morning schedule – which involves getting dressed and walking to work – it would be appropriate for me to say I am walking to work this morning because plans and intentions can be taken as preliminaries; but I cannot say I am walking to work (now) because I am not actually walking to work yet. Consider in this regard the interpretation of (11): (11) We are building a summer house next year. The situation at ST might involve plans, an architect, shopping expeditions; the actual activity would be building; the outcome of the process would be the completion of the summer house. However, the extended process approach is difficult to generalize to sentences about homogenous events, events that do not involve completion. For such events there is no outcome associated with a process or activity: the event consists only of process or activity. To see the difficulty compare the sentences of (12), where rehearse is a homogenous event. (12) a. Mary was rehearsing yesterday. b. Mary is rehearsing tomorrow. (12) says that at the time yesterday the process Mary rehearse was going on, and this is a standard progressive. But for an event such as rehearsing no goal or outcome is associated with the process. Yet (12b) cannot be interpreted as saying that the process of Mary rehearsing is actually going on; rather, a process associated with Mary rehearse tomorrow is going on. The problem with the notion of extended process and events such as rehearsing is this: in the futurate perspective all events are heterogenous. An extended process involves an event with stages that are essentially different. Applying this notion to a homogenous event would require a radical change in the structure of the event. The appealing simplicity of the extended process approach cannot be maintained, therefore. Since the notion of extended process enables a general account of the progressive, there might be some interest in revising it appropriately. Before attempting a revision, however, it will be worth while to look more closely at the approach. 9 This
approach is discussed in Dowty (1977).
154
C.S. Smith
3
Further investigating the extended process approach to the futurate, we now consider a question about the temporal range of the futurate: how many situations does it include? With the notion of extended process we have made an unstated assumption about the futurate. We have assumed that the temporal range of the futurate includes two situations, the present and a future situation. Futurate sentences are taken to present both situations: the progressive, for instance, would focus on a process that is expected to continue into the future. In this interpretation, both preliminary processes and the actual processes they are associated with are presented in futurate sentences. Consider in this light a sentence like (13), a standard example. (13) The Dodgers are playing the Yankees tomorrow. In the suggested reading, this sentence focuses at ST on a process at the preliminary stage; and it is expected that the actual process will follow. I will now argue that the assumption of two situations is incorrect. In the next section I sketch an analysis in which the futurate presents only a present situation. If the range of the futurate stretches into a future situation, as suggested by the extended process analysis, a question arises about the scope of the single aspectual marker. The question is whether the aspectual applies to both present and future situations. For the futurate as sketched here we might propose an abstract semantic structure that represents two situations, each with an aspectual indication of some kind. I will assume that a situation always has an aspectual, although for a wider range of material this may not be the case.10 Now, how would an underlying structure involving two aspectuals be realized in a futurate sentence? The syntax of English does not allow more than one verbal auxiliary to a verbphrase, and aspect is indicated only in the main verbphrase of a futurate. This suggests that the aspect of the second situation, the inner sentence of a futurate, cannot be indicated on the surface. There is another possibility, an analysis in which a futurate sentence would realize an underlying structure with two situations. This analysis exploits the fact that perfective aspect is indicated by the simple verb form (the absence of an auxiliary). We might claim that the future situation is always presented with perfective aspect in a futurate sentence. Such a claim is plausible just because the marker of perfective aspect has no phonological shape. Note that there seems to be nothing wrong semantically with a structure that has two situations and two aspects. The problem, in this view, is simply that such a structure cannot be realized with the English futurate if the future situation is imperfective. The reason is syntactic: the inner sentence of a future cannot have the progressive auxiliary. (Of course, a different construction could realize a semantic structure with two situations and two aspectuals.)
10 It
is not clear that situations are always presented with an aspect; nominalizations for example may not have aspect underlyingly. The question deserves investigation.
The English futurate
155
We now have two competing analyses of aspect in futurate sentences, both maintaining the idea that futurates involve two situations. The analyses make different predictions about the aspect of the inner sentence, or future situation. One analysis predicts that the aspect of the inner sentence is always perfective, because there is no syntactic place for the form that marks imperfective. The other analysis predicts that the aspect of the inner sentence varies with that of the main sentence – that the aspect of the main verb has the inner sentence (that presents the future situation) in its scope. These predictions can be tested with diagnostics of the aspect of the inner sentence. I suggest two such diagnostics, one using adverbials and the other conjunction. Completive adverbials (in an hour, etc.) are compatible with perfective but not imperfective aspect, as noted in Vendler 1967. Compare for example the sentences of (14): (14) a. John walked to school in an hour. b. ∗ John was walking to school in an hour. On the intended completive reading, (14b) is ungrammatical: comparison with (14a) shows that the ungrammaticality is due to the combination of completive adverbial and imperfective aspect. Following this pattern, completive adverbials can be used as diagnostic of the aspectual reading we are interested in. What is needed is a test sentence: a futurate progressive with a completive adverbial. If the inner sentence of a futurate progressive is indeed imperfective, the test sentence should be ungrammatical; it is has perfective aspect or is ambiguous, with a perfective reading, the test sentence should be grammatical. Consider the following examples: (15) Mary is writing a report tomorrow in an hour. (16) Tonight I’m cooking in an hour a dinner that took Julia Child all day. (17) Next semester we’re grading all the papers in three days. These examples suggest that the inner sentence of a futurate has perfective aspects: all three are perfectly grammatical. One might object that these examples are not convincing, because adverbials with in are not necessarily completives. They can be taken as part of an event-description, as in the dialogue of (18): (18) A sees B frantically scribbling at a very short manuscript A : What are you doing? B : C bet me that I couldn’t write a sonnet in 5 minutes, but I’m doing just that: I’m writing a sonnet in 5 minutes.
156
C.S. Smith
Here we might say that B was writing-a-sonnet-in-5-minutes, rather than writing a sonnet; on this reading the adverbial is not a sentence operator semantically.11 However examples like (18) merely show that adverbials with in can appear in more than one structure; the completive readings for the other cases remain undisturbed. The adverbial diagnostic, then, supports the analysis of perfective aspect for the inner sentence of a futurate progressive. Another test for imperfective aspect is whether an imperfective sentence can felicitously be conjoined with a sentence about completion of the activity in question. Imperfectives can be conjoined felicitously with an assertion that the activity was broken off. Perfectives, which involve completion cannot be so conjoined with felicity. The examples illustrate: (19) a. Sarah was making a quilt but she didn’t actually make one. b. Sarah made a quilt but she didn’t actually make one. (19a) is intelligible and felicitous, whereas (19b) is contradictory. Now consider conjunctions of this type, where the first conjunct is a futurate progressive. If the inner sentence has perfective aspect, we would expect these conjunctions to be contradictory. (20) Sarah is making a quilt tomorrow, but she may not actually make one. (21) Martin is building a model airplane next week but he may not actually build it. Both these examples are contradictory, indicating that the inner sentences of the first conjuncts have perfective aspect. (The judgment is difficult to make because there is a different, more natural reading, in which the future event will not occur at all; this is the reading alluded to in the discussion of example (6b) above.) Note that the results of both tests go in the same direction. We now have a fairly reasonable account of the analysis in which a futurate involves two underlying situations. We have seen that the aspect of the present situation may vary, but that of the future situation is semantically perfective — for surface syntactic reasons. In this account, it is interesting to note, syntactic possibilities constrain semantic ones. The analysis explains quite nicely one’s intuition, noted in the introduction to this paper, about aspect in the futurate construction. The intuition is that there is little difference in the interpretation between simple and progressive futurates. The explanation of this account is that sentences of both types present the future event in the same way, with perfective aspect. For example (22a and b) both present the future situation with perfective aspect; they differ as to how the situation at ST is presented, of course. (22) a. John plays a tennis match with Stuart tomorrow. b. John is playing a tennis match with Stuart tomorrow. 11 This
use of adverbials is not limited to in, e.g. i A: What are you doing? B: I’m running as fast as I can; it’s good for the cardiovascular system ii: I went to Paris for three days, and ended up staying for three months.
The English futurate
157
There is, unfortunately, a rather serious disadvantage to this analysis. It allows semantic structures with any number of situations and aspectual markers, that are related in a grammar to surface structures with only one aspectual marker. And if we can analyze futurates as involving two situations, why not 3? – the preliminary situation, a situation associated with an outcome (for heterogenous events), and the outcome. But then, why not 4 or 5 situations, including onsets? But this is a reductio ad absurdum. In short, the analysis allows indeterminate relations between semantic and syntactic structures and therefore cannot seriously be maintained.
4
There is another way of looking at the futurate, which avoids these difficulties while preserving the positive points made above. The basic idea is that the futurate presents a single situation, that obtains at ST and is associated with a future time and a future situation. The role of the future situation is as final endpoint of the present situation. This interpretation says that aspectually the future situation functions completively, and it leads to a simple and satisfying analysis of the futurate. Simple futurates on this approach present a predictive situation, that is, a situation at ST that constitutes preliminaries and/or prediction. The situation extends into the future and its final endpoint is the occurrence of what is predicted. Predictive situations are heterogenous: there is a change of state when the outcome of the prediction is reached. The future situation has no internal structure in the futurate, then: its role is only as final endpoint of the predictive situation. The progressive futurate presents a predictive situation as a process (that has begun before ST) that is associated with an outcome, which is the predicted situation. The process consists in preliminaries: the schedule, plan, arrangements, patterns, etc. that license at ST predictions about the future. Presented in the focus of imperfective aspect, the preliminary situation is part of a dynamic process that is associated with the actual occurrence of the event or state in question. Here too the future situation has no internal structure: it is presented only as the outcome associated with the predictive situation. This view of the futurate is consonant with the results of the preceding section. In both simple and progressive futurates, the future situation is presented without regard to internal structure. The difference between the extended process notion and that of preliminary process offered here is this: in the former, preliminaries are part of a process that may have later, active stages. Here, preliminaries are the whole process. We now ask how to derive the preliminary process interpretation from a temporal structure such as (9). The structure has one aspectual and two adverbials, one dominated by the temporal constituent and the other by the inner sentence. There are two possibilities, one structural and the other specifying, associated with specification of the notion of futurity. On the structural approach a special rule would associate the predictive interpretation with a configuration such as (9). The approach will work if the futurate is the only construction that has the temporal configuration of (9). However, there may be another construction with the same configuration – the English perfect, which also involves two times. At this point it is not clear to me whether the perfect has
158
C.S. Smith
a structure like (9) or a structure in which the temporal adverbials all occur within one underlying constituent. It is also not clear to me how much systematic weight should be given to the differences between the concepts “past” and “future”, in rules that deal with temporal matters. Another approach would appeal directly to the notion of futurity in some way to be determined. For instance, rules might allow for the predictive interpretation be made for the relevant configuration, when the inner sentence specifies the future. This is the specifying approach mentioned above. Analyses of temporal reference tend to emphasize the symmetry of pastness and futurity; but in language their asymmetries may be just as important, if not more so. Syntactically past and future differ in English, of course; and there are other differences. Of considerable importance here is the area of prediction, which includes the futurate of course. Sentences about the past are not taken as predictions in English unless so marked with a modal, as in They will have left before last Sunday. Sentences about the future are always predictive, whether they have modal or futurate form.
5
Two aspectual puzzles about the futurate were noted at the beginning of this paper. On the one hand, futurate progressives do not indicate events that are “in process” in the usual sense; on the other, the difference between simple and progressive futurate is difficult to pin down. The first puzzle is explained by the notion of preliminary process. We are now able to clear up the second. The aspectual difference between futurates is elusive for two reasons. The first reason is that, as we have seen, both perfective and imperfective present the future situation in the same way — as the final endpoint of a predictive situation. The second reason is that the aspectual difference in futurates is not based on a truth-conditional difference. The same truth conditions hold for both perfective and imperfective sentences. In other words, the situations in which a perfective futurate are true, are just those in which an imperfective futurate are true. Consider for example the sort of situation that would license futurate predictions such as (1a and b), repeated here: (1) a. Roger is playing squash tomorrow. b. Roger plays squash tomorrow. For the perfective prediction to be made, there must be a situation in which preliminaries are recognizeable. This is exactly the situation that would license (1a), the imperfective futurate. The difference between the sentences is the presentation, or aspect: in the imperfective presentation the preliminaries are seen as part of an ongoing process. Of course, the two aspects do not lose their characteristic semantic meanings when they appear in futurate sentences. The perfective focuses on eventas-a-whole, the imperfective focuses on process. This difference in focus can be reflected in the articulation of the truth-conditional relation between sentences and actual situations. Futurates with different aspects have the same truth conditions, but they do not have the same truth-condition articulations.
The English futurate
159
Although aspect can reflect truth-conditional differences, the futurate construction shows that it does not always do so. Speakers always have a certain amount of choice as to how to present a situation. In the case of the predictive situations that license the futurate, the choice is particularly open. One can focus on a predictive situation as a process to present it in a particular way. The point that I wish to make here is that this kind of speaker’s choice does not depend on whether the situation in the world is “actually” a process but rather on what aspect of the situation the situation the speaker wishes to present.12 The futurate is not unique in offering this type of aspectual choice to the speaker. Another example can be found in statives, that is sentences about situations generally thought of as stative.13 Although in standard English statives do not occur with the progressive, sentences like (23b) are fairly common in colloquial English: (23) a. Maurice likes his little brother this week. b. Maurice is liking his little brother this week. In (23b), what is commonly thought of as a state is presented as a process or activity – that is, as an event. As far as I can tell the sentences of (23) have the same truth conditions, but they differ as to how the situation in question is presented. In this paper, using the futurate construction as an example, I have suggested that the relation between syntax and semantics be analyzed through the use of mediating structures. The mediating structures presented for temporal reference give the scope of temporal expressions within a sentence. Such structures can serve as input to interpretation of different kinds. I have also shown that in the analysis of aspect it is fruitful to consider factors that are not truth-conditional. Acknowledgments I would like to thank the Cognitive Sciences Seminar at the University of Texas, and most particularly Mark Steedman and Frank Vlach, for stimulating discussions of much of the material presented here.
References Bennett, M. & Partee, B. (1972). Toward the logic of tense and aspect in English. Santa Monica, CA: System Development Corporation. Dowty, D. (1977). Word meaning and Montague grammar. Dordrecht: Kluwer. Dowty, D., Wall, R. E., & Peters, S., Jr. (1981). Introduction to Montague semantics. Dordrecht: Reidel. Jespersen, O. (1931). A new science: Interlinguistics. Psyche, 11, 57–67. Reichenbach, H. (1947). Elements of symbolic logic. New York: MacMillan. Smith, C. S. (1978). The syntax and interpretation of temporal expressions in English. Linguistics and Philosophy, 2, 43–100.
12 This
important point was suggested to me in discussions with Frank Vlach.
13 A stative situation is, very roughly, one that is stable and homogenous. For discussion see Smith
(1983).
160
C.S. Smith
Smith, C. S. (1980). Temporal structures in discourse. In C. Rohrer (Ed.), Time, tense, and quantifiers (pp. 355–375). Tübingen, Germany: Niemeyer. Smith, C. (1981). The Futurate Progressive: Not simply Future + Progressive: Proceedings of Chicago Linguistic Society, 17, 369–382. Smith, C. S. (1983). A theory of aspectual choice. Language, 59, 479–501. Vendler, Z. (1967). Linguistics and philosophy. Ithaca, NY: Cornell University Press. Vlach, F. (1981). The semantics of the progressive. In P. J. Tedeschi & A. E. Zaenen, (Eds.), Syntax and Semantics, Vol.14: Tense and aspect (pp. 271–292). New York: Academic Press.
The Domain of Tense Carlota S. Smith
1 Introduction The syntactic domain of tense is the clause: tense appears in some form in every clause of a tensed language. Semantic interpretation of tense requires information from context, however. This has been clear at least since Partee’s (1984) demonstration of the anaphoric properties of tense. In this chapter, I will show that the facts about context are quite complex, perhaps more so than has been appreciated. There are three patterns of tense interpretation according to the type of discourse context in which a clause appears. I will introduce the notion of discourse mode to account for the different types of context. I offer an interpretation of tense in Discourse Representation Theory, a framework that is organized to deal with information from the context. I also show that a syntactically based theory can handle contextually based tense interpretation. I first set out the basic analysis of tense and show how it applies to sentences in isolation (Section 2). I then discuss types of discourse context (Section 3), patterns of tense interpretation (Section 4), and the formal analysis of tense (Section 5). I conclude with a summary and a prediction about temporal interpretation in tenseless languages (Section 6).
2 Tense in Single Sentences 2.1 Syntax I assume that a tense morpheme is generated in each clause. In Principlesand-Parameters Theory (now the Minimalist Program), tense heads a functional category. In Discourse Representation Theory, the tense morpheme appears under the Auxiliary node in surface structure.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_6,
161
162
C.S. Smith
2.2 Semantics My approach to the analysis of tense is based on Reichenbach (1947). Each tense involves three times and conveys information about two relations between them: a relation between speech time and reference time (SpT and RT), and a relation between reference time and event time (RT and ET). In the simple tenses (past, present, future), ET and RT are the same; for the relative tenses, they differ. The notion of reference time has been controversial.1 RT makes the information associated with tense relatively complex, and some scholars have argued that it is unnecessary for simple tenses. However, I think that the arguments in favor of RT are decisive. The examples are from English; languages differ so that the details of each system must be worked out separately. The classic argument for RT, due to Reichenbach, concerns the analysis of relative tenses such as the perfect. The argument is still good. The perfect differs conceptually but not truth-conditionally from the simple tenses. Consider these examples: (1) a. Mary arrived. b. Mary has arrived. c. On Sunday, Mary had (already) arrived. The difference between (1a) and (1b) is that of temporal standpoint or perspective. In (1a), the event is set squarely in the past: RT is the same as ET. In (1b), the event is presented from the standpoint of the present, so that RT is the same as SpT. (1c) is a past perfect requiring three different times for semantic interpretation: SpT, RT (the Sunday before), and ET, which precedes that time. (2) states schematic meanings for the tenses of (1). (2) Present: RT = SpT, ET = RT Past: RT < SpT, ET = RT Past perfect: RT < SpT, ET < RT If simple and perfect tenses are analyzed in this way, we capture both conceptual and truth-conditional meanings. A second argument, also due to Reichenbach, concerns the temporal relation between events and states in multiclause sentences. In (3a), the arrival occurs during the interval of the smile; in (3b), it precedes the smile. (3) a. Mary was smiling when John arrived. b. Mary smiled when John arrived. 1 There are some well-known difficulties with Reichenbach’s theory: in the abstract, it predicts more temporal relations than are found in natural language; and the relations between times are not always clear. These difficulties are not crippling, and various proposals have been made to overcome them (Comrie 1985; Hornstein 1990). I will use the Reichenbach-based approach developed by Kamp and Reyle (1993)—with some differences, which I point out below.
Domain of Tense
163
The notion of RT provides a locus for relating the events in a principled manner (Hinrichs 1986). Similar contrasts occur with events expressed in independent sentences. Another argument for RT concerns the phenomena of shifted deixis. As is well known, deictic adverbials such as now and in three days, normally oriented to the moment of speech, can orient to a past (or future) time, as in (4). (4) Mary sat down at the desk. Now she was ready to start work. In such contexts, the shifted now suggests Mary’s perspective. The notion of RT is the anchoring point for this perspective. Tense interpretation interacts with aspectual information, as the sentences of (3) show. I will briefly introduce the aspectual notions we need. Aspect: Aspectual systems have two components, viewpoint and situation type (Smith 1991). Situation type indirectly classifies a sentence as expressing an eventuality, a state or an event. The information is conveyed by the verb and its arguments (the verb constellation). Aspectual viewpoint, conveyed morphologically, focuses all or part of the eventuality. Sentences with the perfective viewpoint have the simple verb form and focus events with endpoints. Imperfective (progressive) sentences have the verb auxiliary be+ing; they focus an internal interval of an event, without endpoints. (5) illustrates. (5) a. Mary walked to school. (Perfective) b. Mary was walking to school. (Imperfective) The perfective focuses events as bounded; the progressive focuses them as unbounded. States are unbounded; in English, they are expressed in sentences with the perfective viewpoint.2 The property of boundedness is crucial for the interaction of tense and aspect. Bounded events are included in the ET interval (ET ⊆ e); unbounded events and states overlap or surround the ET interval (ET O e). Example (6) illustrates. (6) a. Lee built a sandcastle. John left. b. John was working. Lee was at school. We understand the events of (6a) as taking place within the interval of ET—unstated in these examples. The ongoing event and state of (6b) overlap ET: that is, they hold during the ET interval and are understood to hold before and after it as well.
2 The temporal schema of a state does not include its endpoints: changes in and out of a state are not part of the state itself. The perfective viewpoint focuses the temporal schema of a state and thus expresses an unbounded eventuality (Smith 1991). The progressive (imperfective) viewpoint appears neutrally only with events. There are some marked stative progressives, for example, I’m really loving this walk.
164
C.S. Smith
Discourse Representation Theory (DRT): The theory introduces entities for individuals, eventualities, and times with each clause of a sentence. Tense information is conveyed by features associated with a given tense. These features trigger conditions in the discourse representation structure (DRS) that relate the times. The time entities are distinguished as t1 , t2 , t3 and associated with an interpretation in the DRS. Aspectual information is interpreted by rules that construct the DRS. The construction rules interpret a given verb constellation as expressing a state, telic event, and so on. The information appears as a condition on the eventuality introduced for the clause.
2.3 Tense Interpretation of Single Sentences The interpretation of tense is deictic for single sentences: all simple tenses are oriented to SpT. This point is not controversial. As we have seen in the examples above, past tense conveys that RT precedes SpT, while present tense conveys that RT and SpT are the same.3 For concreteness, I give a semiformal statement for two sentences. The first has past tense and expresses a bounded event, the second has present tense and expresses an ongoing, unbounded event; t1 = SpT, t2 = RT, t3 = ET. (7) a. John left. E: bounded event t2 < t1 , t3 = t2 ET ⊆ e b. John is working. E: unbounded event t2 = t1 , t3 = t2 E O ET The construction of detailed DRT representations is discussed in Section 5.
3 Sentences in context I now turn to the interpretation of tense in sentences in discourse. The question immediately arises of how to investigate this topic: what does one look at? My first thought was to use genre as an organizing principle. Discourse is traditionally classified into genres according to purpose and other criteria; thus, novels, newspaper articles, and business letters constitute separate genres. Texts of different genres might function as different types of context. I use the term discourse for spoken and written material, text for written. Only written texts will be discussed here. The standard genre categories function as different types of activity; they are rooted in context and culture. The notion of activity type is introduced in an 3 For
such cases, Partee (1984) suggests anaphoric reference to a time established in the context.
Domain of Tense
165
important article on discourse structure by Levinson (1979). An activity type is “any culturally recognized activity. . .in which language plays a significant role. . .[They are] goal-defined, socially constituted, bounded, events with constraints on participants, setting. . .and allowable contributions” (1979, 368). Levinson develops a strong case for the pragmatic basis of activity types. The pragmatic basis of genre makes it inappropriate for close linguistic analysis. Levinson’s argument applies to genre categories.4 Pragmatic factors, including detailed expectations, predominate in the understanding and interpretation of genre. People recognize the structure of a discourse by using their cultural and contextual knowledge. This is another take on the well-established point that structure is not “in” the text but is constructed by the receiver. Genre is also too global a category: texts of a given genre tend to be quite varied when examined carefully. In narratives, for instance, events and states occur in sequence, bound together by a unifying theme. But narrative episodes rarely consist entirely of such sequences. There are descriptive passages, and sometimes commentary. Similarly in the expository genres, narrative sequences that depart from the main argument line are often found.
3.1 Discourse Mode At the local level of the passage, one recognizes stretches of text that are intuitively of different types—for example, narrative, description, argument. These stretches tend to have a particular force and a characteristic cluster of linguistic features and interpretations. Narrative, description, and argument make different contributions to a text; they have different linguistic features and interpretations, including tense interpretations. I introduce the notion of discourse mode for such stretches of text (Smith 2000). I posit five modes: Narrative, Report, Description, Information, and Argument. The list is not exhaustive (it ignores conversation and procedural discourse, for instance), but it includes types that appear in many texts. The list of modes should be relatively short if it is to make significant generalizations. One way to keep it short is to allow for variation within a mode. I will assume that the modes vary in point of view, or authorial stance, and in level of formality. Discourse modes cut across genre categories. The modes can be characterized with two interpreted linguistic features, both relating to temporality. One feature is the type of entity introduced into the universe of discourse; the second is the principle of advancement. One way of getting at what a text is about is to ask what sorts of things the text deals with. More formally, we ask what kinds of entities the text licenses in semantic 4 Levinson is interested in relating the structure of a discourse to its function. “Wherever possible,” he writes, “I would like to view structural elements of an activity as rationally and functionally adapted to the point or goal of the activity—functions members of the society see the activity as having” (1979, 369). Levinson recognizes divisions of an activity into subunits: court case, seminar, and so on. Within each unit, there may be prestructured sequences required by conventions and other parts.
166
C.S. Smith
representation. The most familiar entities are eventualities: events and states and a larger category of statives, sentences that express generalizations due to pattern or regularity. There is another important category, that of abstract entities, which consists of facts and propositions. Texts also introduce entities for individuals and times, not relevant here.
3.2 Types of Entities Entities are conceptual categories, expressed linguistically at the level of the clause. They are realized by verb constellations and nominal forms. The linguistic forms have distinct distributional properties and are covert categories of grammar in the sense of Whorf (1956). Therefore, the types of entities invoked by a text are interpreted linguistic features. The main types of entity are eventualities, general statives, and abstract entities. They differ in temporality: eventualities and general statives are temporally located in the world, whereas abstract entities are not. (8) and (9) illustrate the first two categories. (8) Eventualities a. Mary won the race. John opened the door. Lee rehearsed. (Event) b. The cat is on the table. Sam is tired. Mary likes ice cream. (State) (9) Stative (general) a. The lion has a bushy tail. (Generic) b. John often fed the cats last year. (Generalizing-habitual) The linguistic properties that distinguish events and states are quite well known (Vendler 1957; Dowty 1979; Smith 1991). General statives are exemplified in (9). (9a) is a generic sentence, denoting a kind; (9b) makes a generalization over patterns of situations; both are discussed in Krifka et al. 1995. Generics and generalizations are semantically stative, though the latter have the syntactic characteristics of events (Smith 1991). Linguistic expressions of abstract entities—facts and propositions—also have distinct characteristics when they occur as complements of particular predicates. The point was first made by Vendler (1967,1972); it has been elaborated by Asher (1993) and Peterson (1997). Abstract entities are important for this study because they predominate in certain text modes. The following examples illustrate: Abstract entities: not temporally located (10) Facts: objects of knowledge a. I know that Mary refused the offer. b. Mary’s refusal of the offer was significant.
Domain of Tense
167
(11) Propositions: objects of belief a. I believe that Mary refused the offer. b. Mary’s refusing the offer was unlikely. Not all expressions of abstract entities have distinct linguistic characteristics. Sentences that directly express facts and propositions cannot be identified linguistically, although they can be so identified when they appear as clausal complements. There is a strong correlation between the discourse modes and types of entity. Entities of different types predominate in each of the discourse modes. They fall into two groups on the basis of temporality, as summarized in (12) and (13). (12) Temporal Narrative: events, states Report: events, states, general statives Description: states, statives, ongoing events (13) Atemporal Information: facts, statives Argument: abstract entities, general statives This summary ignores the question of how to decide when a given type of entity predominates in a text. The question will be addressed in further work.
3.3 Advancement All texts advance through a structured domain, but not in the same manner: texts of different modes have different principles of advancement. Advancement is a linguistic feature in the sense that information in the text gives rise to interpretation. There is also a literal sense of advancement, in which a hearer or reader processes the text unit by unit. In the temporally organized modes, a text advances as location changes—time or space. The text modes of Argument and Information are not temporally organized, though they may include eventualities that are temporally located. To understand text advancement in the atemporal modes, we need something other than dynamism. We find it in the notions of metaphorical location and motion, from the spatial domain. The semantic domain of an atemporal discourse can be seen as terrain to be traversed: a metaphorical space. The discourse advances as key reference moves metaphorically from one part of the domain to another.5 Such motion is closer to spatial than to temporal location. The spatial domain is basic to our understanding of the world, and it underlies our notion of time (Clark 1973). Notions associated with space underlie the organization of atemporal texts. Space is 5 I identify key reference as the referent of the NP that plays a focal role in a clause, usually the role of theme (for discussion, see Smith 2003).
168
C.S. Smith
not unidimensional, like time: rather, it allows directions of various kinds. Similarly, direction in a text domain can be hierarchically up or down, lateral, and so on. We need the complexity of space to model metaphorical motion. I suggest, then, that the notion of discourse mode accounts nicely for the different types of passages that appear in discourse. The discourse modes cut across genre lines. For further discussion of the discourse modes, see Smith 2003.
4 Patterns of Tense Interpretation Tense is interpreted differently in the discourse modes of Narrative, Report, and Description. There are patterns of interpretation for passages of each type. I discuss here only the discourse modes in which advancement depends on temporal factors.
4.1 Narrative Narrative has been studied extensively and is fairly well understood. The main unit is the episode, a sequence of consequentially related events and states; the order in which they occur is crucial for understanding (Moens 1987). The essence of Narrative is dynamism, in which events and states are related to each other. After the first sentence of a narrative, the times are sequential or simultaneous with previous times in the text. They are not related to SpT. The narrative fragments in (14) illustrate. (14) Narrative: states, events related to each other a. When I got to Harry’s Waldorf Towers apartment, they were winding up the meeting downstairs. Harry appeared about a half-hour later, greeted me warmly, went immediately to the telephone. (Lillian Hellman, Scoundrel Time, Little, Brown, Boston, 1976, 55) b. One night in November 1961, Alice went into the tub room to put some clothes in her old wringer washing machine. When she turned on the light, there was a rat the size of a small cat sitting on the machine. (J. Anthony lukas, Common Ground, Knopf, New York, 1985, 149) c. 1 She put on her apron, took a lump of clay from the bin and weighed off enough for a small vase. 2 The clay was wet. 3 Frowning, she cut the lump in half with a cheese-wire to check for air bubbles, then slammed the pieces together much harder than usual. 4 A fleck of clay spun off and hit her forehead, just above her right eye. (Peter Rcbinson, A Necessary End, Avon Books, New York, 1989, 182) In these fragments, narrative time advances with perfective event sentences and with explicit temporal adverbials, and fails to advance otherwise. This is the basic finding of discourse dynamics (Hinrichs 1986; Kamp and Rohrer 1983; Partee 1984). I take
Domain of Tense
169
it that the basic case of narrative is sequence; I ignore flashbacks, changes of scale, and the like.6 To see just how narrative dynamism works, let us consider (14c) in more detail. The narrative conveys a series of events that follow each other at a time prior to SpT. There is one state (sentence 2), which is simultaneous with the time of the preceding event. The time line in (15) illustrates; E indicates and event, S a state. (15) Time line for (14c) . . .t1 . . .t2 . . .t3 . . .t4 . . .t5 . . .t6 . . .t7 . . . < SpT E1 E2 E3 E4 E5 E6 E7 S1 Aspectual information provides the dynamism or lack of it in (14c). Bounded events occur in sequence; states are simultaneous with the time of the preceding event. Tense conveys continuity in these fragments. Past tense is not interpreted deictically: if it were, the events would be related to SpT rather than to each other. Nor do we interpret (14) as expressing a series of events successively prior to one another. The same pattern holds for narratives in the present or future tense. There are two principles for tense interpretation in Narrative. If a sentence expresses a bounded event, RT advances. If the eventuality is not a bounded event, RT does not change and tense is anaphoric. The two patterns are set out in (16). (16) a. Continuity pattern, narrative advancement: bounded events E1 . . .. . .. . .. . .. . .E2 . . .. . .. . ....E3 . . .. . .. . .. . . RT2 > RT1 RT3 > RT2 RT1 < SpT b. Anaphoric pattern, simultaneous with preceding RT: states, progressives E3 . . .. . .. . ... S1 . . .. . .. . . RT2 = RT1 RT1 Similar principles are stated by Kamp and Reyle (1993) using a Reichenbach-based approach. These principles do not account for advancement in the other two temporal discourse modes, as I will now show.
4.2 Report Reports give an account of eventualities and their significance from the temporal standpoint of the reporter. They are, like narratives, mainly concerned with events and states. It is not primarily the dynamism of events but the position of the reporter 6 There are many departures from the sequential norm. Events may appear in reverse order; there may be a change of level from the general to the particular; events may be simultaneous. For instance, many simultaneous events are presented in E.L. Doctorow’s novel Ragtime, as James Higginbotham has pointed out to me.
170
C.S. Smith
that advances the text. Eventualities are related to SpT, rather than to each other. In reports there is “. . . an immediate link between [SpT] and the reported events” (Caenepeel 1995, 231). I have found report passages in texts of various genres. Reports are discussed in Caenepeel 1995 as a genre. This is not surprising: there is a certain correspondence between genres and modes. For each mode, there is a genre in which that mode predominates: narratives predominate in fiction, reports predominate in (certain) newspaper articles, and so on. In reports, eventualities are ordered with respect to an advancing SpT and to changing location. Thus, tense is deictic in reports. Order of presentation, a key factor in narratives, is less significant in reports. Caenepeel observes that the order of presentation could be changed in a report without changing the information given. The deictic interpretation of tense is like that of independent sentences, as in examples (7) and (8). (17) illustrates; the first fragment appears in Caenepeel 1995. (17) Report: eventualities related to SpT a. 1 The war to free Kuwait began a few minutes before midnight last night, as squadrons of American fighter-bombers blasted Baghdad. 2 And carly today the Ministry of Defense confirmed that British forces were also involved “in military action.” 3 The American F-15Es took off at 9:50 pm GMT from the largest US air base in central Saudi Arabia. 4 “There were loud explosions, obviously bombs, in three parts of the city,” the American Cable News Network, CNN, quoted one of its correspondents as saying. . . . 5 The White House said the US-led attack was aimed at Iraqi troops in both Iraq and Kuwait. 6 The American broadcasting networks in the Iraqi capital reported brilliant flashes of light, thunderous explosions and heavy anti-aircraft fire in the sky. 7 The attack began about 3am on a moonless night. 8 A squadron of British Tornado ground attack aircraft also took off from Bahrain just after lam. (Daily Express 1991) b. 1 A week that began in violence ended violently here, with bloody clashes in the West Bank and Gaza and intensified fighting in Southern Lebanon. 2 Despite the violence, back-channel talks continued in Sweden. 3 Israeli, Palestinian and American officials have characterized them as a serious and constructive dialogue on the process itself and on the final status issues. 4 News accounts here say that Israel is offering as much as 90 percent of the West Bank to Palestinians, although it is difficult to assess what is really happening by the bargaining moves that are leaked. (New York Times 2000) The past tenses and adverbs in these fragments are deictic, relating to SpT: last night and early today in (17a) indicate the position of the reporter, as does here in (17b). There are also changes of tense. (18) gives a time line for (17b).
Domain of Tense
171
(18) Time line for (17b) E1 . . . . . . . . .E2 . . . . . . . . .S1 . . . . . . . . .S2 . . . . . . . . . RT: < SpT < SpT = SpT = SpT Sentences 3 and 4 are in the present tense, so that the eventualities are simultaneous with SpT; sentence 3, in the present perfect, expresses the result state of a past event. Passages in the deictic pattern often have sentences with different tenses, as in the fragments above. As usual, aspectual information interacts with tense interpretation. Eventualities that are unbounded overlap with the present or a past RT. Bounded events are included in a past RT. Bounded events are not expressed in present tense sentences.7 This is true of language generally. Thus, in English all present tense sentences express unbounded situations, owing to the interaction of present tense and aspectual viewpoint. There are exceptions, marked cases such as the “sports announcer present” or the historical present. Present progressive sentences express ongoing events (19a); present simple (perfective) sentences express states (19b) or generalizations—also semantically stative (19c). (19) a. Mary is eating an apple. b. Mary is tall/pleased. c. Mary feeds the cat. The deictic pattern of past tense interpretation is given schematically in (20). (20) Deictic tense interpretation E1 . . . . . . . . .E2 . . . . . . . . .S1 . . . . . . . . .S2 . . . . . . . . . RT: <SpT <SpT < SpT < SpT In the deictic pattern, eventualities and deictics are oriented to SpT.
4.3 Description Description is portrayal of a scene or state of affairs by means of language. In description, pictures are drawn of particular objects, characters, and mental states, as Cairns notes (1902, 114). Description is indispensable to narration, also common in works of other modes, and predominant in travel books. 7 Kamp and Rohrer note that “no event whose duration is properly included within that of the entire discourse may be reported in the present tense” (1989, 72). They suggest that a presupposition leading to it underlies verbal communication. The presupposition is that verbal expression takes place only after the thought it expresses has been conceived. Moreover, communication is in principle instantaneous; communication of a bounded event, even if instantaneous, requires at least one instant after the completion of the event.
172
C.S. Smith
In descriptive passages, time is stable or suspended, without dynamism. If there is motion, it does not involve significant changes of state and there is no sense that time advances. Tense is anaphoric: all the sentences of a given passage have the same RT. Advancement is spatial: the text advances as the reader goes from one part of the scene to another. Usually a locative adverbial appears at the beginning of a description, with scope over the material that follows; such a phrase appears in both examples in (21). (21) Description a. 1 In the passenger car every window was propped open with a stick of kindling wood. 2 A breeze blew through, hot and then cool, fragrant of the woods and yellow flowers and of the train. 3 The yellow butterflies flew in at any window, out at any other, and outdoors one of them could keep up with the train, which then seemed to be racing with a butterfly. 4 Overhead a black lamp in which a circle of flowers had been cut out swung round and round on a chain as the car rocked from side to side, sending down dainty drifts of kerosene smell. (Eudora Welty, Delta Wedding, Harcourt, Brace, New York, 1945, 1) b. 1 On the big land below the house a man was ploughing and shouting admonitions to the oxen who dragged the ploughshares squeaking through the heavy red soil. 2 On the track to the station the loaded wagon with its team of sixteen oxen creaked and groaned while the leader cracked his whip that reached to the horns of the leader oxen and yelled on a note only they understood.. . . 3 On the telephone wires the birds twittered and sang. 4 The wind sang not only in the wires, but through the grasses, and the wires vibrated and twanged. (Doris Lessing, Under My Skin, HarperCollins, New York, 1994, 107) To account for the temporal stability of descriptions like this, I suggest that they be modeled with a durative time adverbial that has scope over the entire passage. The eventualities expressed in these fragments include states and ongoing events (progressives), which fit the anaphoric pattern discussed above. There are also perfective events, in sentences 2, 3, and 4 of both (21a) and (21b). All are activities. In this context, one interprets them as continuous and/or iterative—for example, the breeze blowing through, the lamp swinging, the birds twittering. The perfective viewpoint of an activity focuses a bounded unit that may but need not coincide with the endpoints of the activity. In descriptive contexts, the activities are expected to continue more or less indefinitely; the contribution of the perfective viewpoint to activities is discussed in Smith 1999. Perfective events do not fit the anaphoric pattern, which is limited to unbounded events. To reach a better understanding of description, consider (22), a variant of (21b). I have changed the example slightly; here, sentence 3 has a telic verb constellation
Domain of Tense
173
(walk to school). The descriptive mode is undisturbed and, strikingly, sentence 3 does not have dynamic force. (22) 1 On the big land below the house a man was ploughing and shouting admonitions to the oxen who dragged the ploughshares squeaking through the heavy red soil. 2 On the track to the station the loaded wagon with its team of sixteen oxen creaked and groaned. 3 A group of children walked to school. 4 On the telephone wires the birds twittered and sang. In this context, the event of sentence 3 is taken as atelic, an instance of coercion. The descriptive mode overrides the potential dynamism of the event. The coercion effect can be attributed to the tacit time adverbial of description posited above. Within the scope of a durative time adverbial, telic sentences regularly undergo a shift of situation type to atelic. This is a general phenomenon (Moens and Steedman 1987; Smith 1995). One well-known set of examples involves durative time adverbials and telic sentences with telic verb constellations. (23) a. The children walked to school for an hour. b. Mary read a book for an hour. The durative time adverbial overrides the telicity of the internal verb constellation. These sentences can only be taken as atelic: there is no sense that the children got to school, or that Mary finished the book—on the contrary. Similarly, A group of children walked to school in (22) is reinterpreted as atelic, under the scope of the tacit durative adverbial of description. The actual duration of such an adverbial is determined by context. The assumed time is often either the time during which the situation is expected to hold—in (21a), perhaps the duration of the train trip—or the time it takes for the perceiver to scan a scene and become aware of its properties. In Description, the anaphoric pattern of tense interpretation holds for all eventualities. There are coercion effects due to the tacit adverbial of duration.
4.4 Summary Three broad patterns of tense interpretation have been demonstrated for nonfirst clauses. Tense conveys continuity, anaphora, or deixis. Each discourse mode has a slightly different pattern, as set out in (24). (24) Patterns of tense interpretation in the temporal discourse modes Continuity: nonfirst clause, bounded events, narrative mode Anaphora: nonfirst clause, unbounded events and states, narrative mode, all eventualities, descriptive mode Deictic: default—all other cases
174
C.S. Smith
Notice that the narrative pattern departs from the deictic interpretation that is traditionally said to hold of tense. Equally important, other patterns must be recognized to account for the interpretation of tense in discourse. In the next section, I consider how to account formally for these patterns.
5 Formal Accounts of Tense Interpretation I offer an account of tense interpretation in DRT, assuming an extended version that recognizes and records discourse mode. I then show that a syntactically based theory can also deal with the dependence of tense on discourse mode.
5.1 Discourse Representation Theory DRT was developed to account for conceptual as well as truth-conditional meaning, and to deal with contextual dependencies. Tense interpretation can be nicely accounted for within this theory. Indeed, an analysis of Narrative in the DRT framework is given by Kamp and Reyle (1993). Here I sketch a somewhat different analysis of narrative advancement and the other patterns of interpretation. In DRT, syntactic surface structure is the basis for construction of a discourse representation structure (DRS). Construction rules interpret the information of surface structure in terms of entities and conditions, which are encoded in the DRS. When available, linguistic context also provides information for the construction rules. 5.1.1 Independent Sentences The DRT treatment of independent sentences, outlined above, is presented here in some detail. Syntactic surface structures are the input to construction rules; the output of the rules appears in the DRS. In this sketch, I provide a surface structure and a DRS; I will not state the construction rules. See Kamp and Reyle 1993 for detailed discussion of the construction algorithm. Surface structures are relatively simple, as illustrated schematically in (25b); IP = Infl Phrase; AuxP = Auxiliary Phrase. AuxP has the progressive auxiliary, or the zero morpheme with the simple verb form. Information percolates up the tree so that it is available at the IP node: tense, viewpoint (Ø in this case); the event entity e; the individual entities x and y. (25) a. Mara put on her apron. Tense, Ø, e, x, y trigger entities and conditions that are introduced into the DRS. Information about them arises as follows: the VP is associated with an event or state
Domain of Tense
175 IP: tense, ø; e, x, y
NP
Mara
I’
I
AuxP: perfective (simple verb form)
tense
Aux’
Aux
VP: e
ø
put on her apron
entity; the participants appear in the VP; AuxP is associated with the perfective or imperfective viewpoint. The tense morpheme triggers the introduction of three temporal entities into the DRS, and two features that relate them. The times will be notated simply as t1 ,t2 ,t3 ; they are associated with the temporal concepts SpT, RT, ET. The features, A and B, relate the times to each other, providing what Kamp and Reyle (1993) call the “two-dimensional theory” of tense. I shall say that the A feature relates t1 and t2 , and the B feature relates t2 and t3 . The conditions interpret t1 as SpT (the default case),8 t2 as RT, t3 as ET. Tense is deictic in (25a), since it is an independent sentence. The viewpoint is perfective, so that the event is included in ET, t3 . The past tense has the A and B features introduced above: feature A t2 < t1 ; feature B t2 = t3 . This information is introduced into the DRS for (25a), roughly as in (26): the entities appear at the top, the conditions that specify them below. The A and B features appear on lines 2 and 3 of the DRS; t2 and t3 are specified only relationally because no time adverbial or context gives other information for interpreting the sentence. Line 4 specifies the aspectual information: the event is included in ET. The other lines specify the event and individual entities.
8 In some cases, t
1 is an orientation time other than SpT—for instance, in complex sentences where a complement clause is oriented to a time established in the main clause, as in The senator will predict next week that in a year he will have resigned. Here the complement RT is a year from the main clause RT (next week), and auxiliary have in the complement indicates that ET is prior to RT. The main clause provides the orientation time to which will is related.
176
C.S. Smith
(26) DRS for Mara put on her apron t1 t2 t3 e x y 1. 2. 3. 4. 5. 6. 7.
t1 =SpT t2 < t 1 t 2 = t3 t3 ⊆ e x = Mara y = her apron e: Mara put on her apron
When constructed for a discourse, the DRS is dynamic: each sentence is interpreted with information already in the DRS, so that the DRS is a cumulative representation. The DRS in (26) illustrates the deictic interpretation of tense, which holds for independent sentences and for passages in the Report mode. In this pattern, a sentence is related to SpT according to the content of the A and B features associated with it. (27) Deictic Principle of tense interpretation A sentence S involves three times. t1 = SpT. Feature A: t2 is related to t1 according to the tense of S. Feature B: t3 is related to t2 according to the tense of S. The principle applies to independent sentences and as a default to other sentences unless another principle applies. In the perfect tenses, not discussed here, the B feature provides that t3 either precedes or follows t2 . 5.1.2 Continuity and Anaphora Tense Patterns The interpretation of tense for nonfirst sentences of passages in the Narrative and Description modes requires a departure from the deictic pattern. Moreover, there are differences between the Narrative and Description modes. In this section, I offer an account of all three patterns in the DRT framework. For the nondeictic patterns, I use a technique of analysis that in effect neutralizes the A feature of tense. Information about the mode of a given passage must be available to the construction rules; each mode will trigger the appropriate rule(s) of tense interpretation. I will assume that the mode can be determined by considering types of entities and other factors. Mode information will be stated in an extended DRT along lines suggested by Asher (1993) for handling discourse coherence relations.9 Since it 9 Asher (1993) proposes an extension of DRT, Segmented DRT. The extended theory includes information about discourse coherence relations. I assume that a similar extension would be an appropriate place for information about discourse mode.
Domain of Tense
177
depends on mode, tense may be interpreted at a relatively late stage in the construction process. SpT plays no role in the continuity and anaphora tense patterns, but the notions of ET and RT are needed. Recall that RT is relevant to related eventualities, shifted deixis, and tenses where ET = RT (the perfect and “future in the past”). The notion of ET is needed for the relative tenses (not discussed here) and to relate aspectual information to time. Accordingly, the DRS for these patterns will omit the temporal entity t1 (SpT). This will have the effect of neutralizing the A feature associated with tense, the feature that gives the relation between t1 and t2 . Without a temporal entity t1 , no information can be stated concerning the relation between t1 and t2 . The rules will provide that tenses interpreted in the continuity and anaphora patterns introduce two times, t2 and t3 (RT and ET), into the DRS. This neutralizing technique does not require a change in the information associated with tense. Either two or three times will be associated with a tense, according to the context. If three times are introduced, both the A and B features will be active; if two, only the B feature will be active. Assuming this technique, we need principles for calculating continuity and anaphora interpretations of tense for narrative and descriptive passages. For Narrative, I adapt the ideas of Kamp and Rohrer (1983) and Kamp and Reyle (1993). Advancement occurs with a bounded event sentence: in such cases, RT advances, following the preceding RT. This is the Continuity Principle. The actual time of advancement may be specified adverbially; otherwise, it is taken as the standard run time for the type of event expressed. The Limited Anaphora Principle provides that RT does not advance with sentences expressing unbounded eventualities: ongoing events, states, and general statives are simultaneous with the preceding RT. (28) Narrative Continuity: Advancement Principle If S is a nonfirst sentence of a narrative passage, RTx immediately preceding, and if S expresses a bounded event: RTy > RTx (29) Narrative: Limited Anaphora Principle If S is a nonfirst sentence of a narrative passage, RTx immediately preceding, and if S expresses an unbounded eventuality: RTy = RTx For simplicity, I have avoided sentences where ET = RT. They include perfect tenses and the “future in the past,” as in On Sunday, Mary had arrived three weeks ago and Mary said that she would leave. Sentences like this do not advance narrative; they are stative and fall under the Anaphora Principle. I illustrate the Continuity and Anaphora Principles in a DRS for a narrative fragment of three sentences; the first two introduce events, the third introduces a state. The first sentence introduces three times (subscripted a) in the DRS. The others introduce two times (subscripted b, c). The events (e) are numbered 1, 2; s is the state. The entities are all introduced at the top of the DRS.
178
C.S. Smith
This DRS provides two RTs (t2 ) for the three sentences: the first is specified as prior to SpT on the Deictic interpretation, line 2; the second follows the first by the Advancement Principle, line 8. The third RT is equal to the second by the Anaphora Principle, line 15. The events are included in the intervals of t3 (ET), lines 4 and 10; the state overlaps t3 , line 17. (30) 1 Mara put on her apron. 2 She took out a lump of clay. 3 The clay was wet. t1a t2a t3a e1 x y t2b t3b e2 z w t2c t3c s u 1. t1a = SpT (Deictic) 2. t2a < t1a 3. t2a = t3a 4. e1 ⊆ t3a 5. x = Mara 6. y = her apron 7. e1 : Mara put on her apron 8. t2b > t2a (Advancement) 9. t2b = t3b 10. e2 ⊆ t3b 11. z = x 12. w = a lump of clay 13. e2 : She took out a lump of clay 14. t2c = t2b (Anaphora) 15. t2c = t3c 16. s O t3c 17. u = w 18. s: The clay was wet
For descriptive passages, the Advancement Principle must be suppressed, and the Anaphora Principle extended to apply to all types of eventuality. (31) Description: Full Anaphora Principle If S is a nonfirst sentence of a descriptive passage, RTx immediately preceding: RTy = RTx For contexts not identified as Narrative or Description, the Deictic Principle of tense interpretation applies. The principles use the technique of introducing two temporal entities in the DRS for tenses interpreted in the continuity and anaphoric patterns. This technique neutralizes the A feature associated with tense. It decouples tense from SpT in a given sentence and allows the statement of narrative advancement in terms of RT.10 10 A
different approach to the interpretation of tense in narrative is presented by Kamp and Reyle (1993). In their version of the two-dimensional analysis, tense introduces two times for all clauses and a third time is calculated, the temporal perspective point (similar to Reichenbach’s RT). Kamp
Domain of Tense
179
5.2 The Syntax-Based Approach Tense interpretation with a Reichenbach flavor has also been considered extensively in the generative framework of Principles-and-Parameters Theory, now the Minimalist Program (Hornstein 1990; Zagona 1990). This framework provides that temporal and aspectual information appears in syntactic structure, including Logical Form (LF). Tense either originates syntactically as a verbal affix and moves to a functional tense projection at LF (Stowell 1996) or is generated under TP (Tense Phrase; Thompson 1999). Semantically, tenses are dyadic predicates, relational expressions that directly encode relations between temporal entities (Giorgi and Pianesi 1997). Deictic tense interpretation is not problematic in this framework because it requires no information from outside the clause. I therefore proceed directly to a discussion of the other patterns. The key factor in nondeictic tense interpretation is dependence on context. The RT of a nonfirst sentence either follows the preceding RT or is simultaneous with it (assuming no adverbial information). The dependence can be handled in a syntaxbased theory by providing that a temporal element appear in a topiclike position high in the syntactic tree (Stowell (1996) makes a similar proposal, though for a different purpose). Then a dependent RT would be bound by the RT element in the preceding sentence, also high in the syntactic tree. Assuming that this notion can be implemented, I will adapt the neutralizing technique to the syntactic framework. The analysis must provide that t1 is neutralized in certain contexts. Assume that three times are generated in all sentences. In Narrative and Description contexts, the SpT entity of a nonfirst sentence will be bound by the preceding SpT entity. Then it will not be available for the relation between SpT and RT, so that this relation will be neutralized (however it is expressed). The structure that I assume is essentially the one proposed by Thompson (1999), except for the final position of SpT. There are three temporal entities and an eventuality entity. The latter is associated with the head of VP. ET is also associated with VP (Thompson gives persuasive syntactic evidence for this). RT is associated with the head of AspP (Aspect Phrase). SpT is generated under TP and moved to Spec, CP, where it is in a position to be bound by a form in a preceding sentence.11 The tree structure in (32) has the NP subject raised to Spec, IP; TP is omitted for simplicity.
and Reyle use these times to interpret independent and first sentences on the deictic pattern. They add another time, the Rpoint, for nonfirst sentences. They state principles using the four times for advancement and the limited anaphora of narrative. Their account could easily be augmented to deal with the other patterns presented here. They too would have to identify and note Narrative and Description contexts. For Description, the Advancement Principle would be suppressed and the Anaphora Principle extended to all eventualities. 11 In Thompson’s analysis, SpT is associated with the head of TP, located above AspP.
180
C.S. Smith
(32)
CP SpT
C’
IP:RT
NP
I’
I tense RT
AspP: imperfective/perfective
Asp’
Asp
VP: e, ET
The Continuity and Anaphora Principles can be adapted to a syntactically based representation, given the possibility for dependence on context. I do not attempt to state the principles here, but note simply the information that they would require. Like the principles stated above for a DRT approach, they would relate the RTs of different sentences. The principles would need access to context for information about discourse mode and whether a temporal entity in the preceding sentence binds the highest temporal entity in the tree. Thompson (1999) presents an analysis that is similar to the continuity pattern.12 She uses “linking rules” that relate the RTs of clauses and independent sentences with the adverb then. The principles also need aspectual information, available as shown in (32). Recall that in the continuity pattern, RTy follows RTx for a sentence expressing a bounded event. In the anaphora pattern, RTs are simultaneous for unbounded eventualities (as in Narrative) or for all eventualities (as in Description). In the deictic pattern, adjacent sentences are not temporally related to each other.
6 Conclusion 6.1 Summary Tense interpretation depends on the discourse context in which a sentence or clause appears. The relevant information about context is determined at the level of the passage. I introduce the notion of discourse mode to account for passages of different
12 Thompson
claims that there is no need to posit independent principles to account for the discourse behavior of tense, a point that is called into question by the material presented here.
Domain of Tense
181
types. There are five discourse modes, each with defining linguistic characteristics of a temporal nature. They differ in the types of entities they introduce and their principle of advancement. Three of the discourse modes are temporally organized, with different tense interpretation principles for each one. In the Report mode, tense is deictic; in the Narrative mode, tense conveys either continuity or limited anaphora; in the Description mode, tense is fully anaphoric. Aspectual information is an essential factor in tense interpretation. I have shown how the dependency of tense on context can be handled in dynamic semantic theories such as Discourse Representation Theory, and I have sketched an approach in a syntactically based theory.
6.2 Tenseless Languages The different patterns of tense interpretation have implications for tenseless languages. The natural question to ask about such languages is how the information of tense is conveyed. Let us assume that the deictic pattern is basic for languages generally, including tenseless languages. We might expect that analogues to the other patterns can be found and that they are triggered by additional information in the context. In the English fragments presented above, this information comes from discourse mode. The default for tenseless languages follows the deictic pattern for present tense: unbounded eventualities are taken as present, bounded events are taken as past (e.g., in Mandarin Chinese and Navajo; Smith 1991). The study of discourse modes suggests that nondefault interpretations are triggered by additional information in the context. More specifically, we would expect that in tenseless languages, explicit cues can be found for the nondefault cases. I list the main cases in (33), using examples in English (for simplicity, I ignore the interpretation of future time). (33) a. Unbounded events in the past Mary was drawing a circle. b. Unbounded states in the past Kim was rich. c. Bounded states in the past John was here yesterday. (but isn’t now) I am currently exploring the predictions about tenseless languages for Mandarin Chinese in joint work with Mary Erbaugh (Smith and Erbaugh 2001). Acknowledgements I would like to thank the members of the audience at the International Round Table on the Syntax of Tense and Aspect for helpful comments and questions.
References Asher, Nicholas. 1993. Reference to abstract objects in discourse. Dordrecht: Kluwer. Caenepeel, Mimo. 1995. Aspect and text structure. Linguistics 33, 213–253.
182
C.S. Smith
Cairns, William B. 1902. The forms of discourse. Boston: Ginn. Clark, Herbert. 1973. Space, time, semantics and the child. In Timothy Moore, ed., Congnitive development and the acquisition of language. New York: Academic Press. Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. Dowty, David. 1979. Word meaning and Montague Grammar. Dordrecht: Kluwer. Giorgi, Alessandra, and Fabio Pianesi. 1997. Tense and aspect. Oxford: Oxford University Press. Hinrichs, Erhard. 1986. Temporal anaphora in discourses of English. Linguistics and Philosophy 9, 63–82. Hornstein, Norbert. 1990. As time goes by. Cambridge, Mass: MIT Press. Kamp, Hans, and Uwe Reyle. 1993. From discourse to logic. Dordrecht: Kluwer. Kamp, Hans, and Christian Rohrer. 1983. Tense in texts. In Rainer Bäuerle, Christophe Schwarze, and Arnim von Stechow, eds., Meaning, use and interpretation of language. Berlin: de Gruyter. Kamp, Hans, and Christian Rohrer. 1989. A Discourse Representation Theory account of tense in texts. Manuscript, University of Stuttgart. Krifka, Manfred, Francis Jeffry Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro Chierchia, and Godehard Link. 1995. Genericity: An introduction. In Gregory Carlson and Francis Jeffry Pelletier, eds., The generic book. Chicago: University of Chicago Press. Levinson, Stephen. 1979. Activity types in language. Linguistics 17, 356–399. Reprinted in Paul Drew and Andrew Heritage, eds., Talk at work. Cambridge: Cambridge University Press (1992). Moens, Mark. 1987. Tense, aspect and temporal reference. Doctoral dissertation, Centre for Cognitive Science, University of Edinburgh. Moens, Mark, and Mark Steedman. 1987. Temporal ontology in natural language. In Proceedings of the 25th Annual Meeting of the Association for Computational Linguistics, Stanford, Calif. Partee, Barbara. 1984. Nominal and temporal anaphora. Linguistics and Philosophy 7, 243–286. Peterson, Philip. 1997. Fact, proposition, event. Dordrecht: Kluwer. Reichenbach, Hans. 1947. Elements of symbolic logic. New York: Macmillan. Smith, Carlota S. 1991. The parameter of aspect. Dordrecht: Kluwer. Smith, Carlota S. 1995. The range of aspectual situation types: Shifts and a bounding paradox. In Pier Marco Bertinetto, ed., Temporal reference: Aspect and actionality. Turin: Rosenberg and Sellier. Smith, Carlota S. 1999. Activities: States or events? Linguistics and Philosophy 22, 479–508. Smith, Carlota S. 2000. Characterizing discourse modes with linguistic tools. Paper presented at the Annual Meeting of the Society for Text and Discourse, Lyon, France. Smith, Carlota S. 2003. Modes of discourse. Cambridge: Cambridge University Press. Smith, Carlota S., and Mary Erbaugh. 2001. Temporal information in sentences of Mandarin. In Xu Liejiong and Shao Jingmin, eds., New views in Chinese syntactic research: International Symposium on Chinese Grammar for the New Millennium. Hangzhou: Zhejiang Jiaoyu Chuban she. Stowell, Tim. 1996. The phrase structure of tense. In Johan Rooryck and Laurie Zaring, eds., Phrase structure and the lexicon. Dordrecht: Kluwer. Thompson, Ellen. 1999. The temporal structure of discourse: The syntax and semantics of temporal then. Natural Language and Linguistic Theory 17, 123–160. Vendler, Zeno. 1957. Verbs and times. The Philosophical Review 66, 143–160. Reprinted in Zeno Vendler, Linguistics in philosophy. Ithaca, NY.: Cornell University Press (1967). Vendler, Zeno. 1967. Linguistics in philosophy. Ithaca, NY.: Cornell University Press. Vendler, Zeno. 1972. Res cogitans: An essay in rational psychology. Ithaca, NY.: Cornell University Press. Whorf, Benjamin L. 1956. Language, thought, and reality. Cambridge, Mass.: MIT Press. Zagona, Karen. 1990. Times as temporal argument structure. Manuscript, University of Washington. Paper presented at the conference “Time in Language,” Cambridge, Mass., March 1990.
Tense and Context in French Carlota S. Smith
In this article I discuss the tense system of French from a cross-linguistic perspective. Taking the characterization of tense as a point of departure, I consider certain tenses in French. The main thrust of the discussion is semantic and pragmatic. The role of context is a recurrent theme: the interpretation of tense depends crucially on information in context, a point often not sufficiently appreciated for tense. Section 1 gives a general discussion of tense and its properties across languages; Section 2 discusses the atemporal meanings of past tense; Section 3 addresses the question of a Future tense: whether it ever makes sense to posit such a tense, or whether Future always belong to a separate, modal category; Section 4 shows that a tensed clause may have one of three different temporal interpretations, depending on the type of discourse context in which the clause appears; Section 5 concludes.
1 Introduction: Tense as a Cross-linguistic Category For languages that are tensed in the traditional sense–such as French and English– tense is a verbal inflection that gives information about temporal location. At one level, then, it is a morphological category. Syntactic theory of the past twenty years accords tense an important role in syntax as well. In Principles and Parameters and related theories, tense is associated with abstract syntactic structure (cf. Pollock 1989). Tense heads its own syntactic projection and participates in case assignment, agreement, and subjectivity; the distinction between finite and non-finite clauses is made within the category of tense. Thus the inflectional morphemes of tense are part of, or form, the syntactic ‘spine’ of a sentence. Tense in different languages can be considered in terms of what kind of system they form: how many degrees of remoteness they offer in past and in future, for instance Comrie (1986). There are interesting questions about the semantic meaning of tense, and the pragmatic interpretations that may arise. The semantic interpretation that may arise. The semantic interpretation of tense is deictic, with Speech Time at the center. The approach assumes the category of tense and tries to understand its contribution to sentence interpretation. R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_7,
183
184
C.S. Smith
Different questions arise when one takes a more general cross-linguistic perspective. The most basic questions concern the definition of tense as a category. It is not always obvious in analyzing a language whether a particular form is a tense or not, as I have found in recent work on Navajo. Thus one finds oneself asking whether all morphemes that convey temporal information should be taken as tenses. The matter is discussed in Smith, Perkins, Fernald 2003. At the same time, I would maintain the distinction between tense and adverb, which is well-motivated both syntactically and semantically. All languages including Hopi (cf. Malotki 1983) seem to have temporal adverbs; it is not clear that all languages have tense (cf. Dahl 1985). The grammatical category of tense can be defined in broad or narrow terms. Several positions can be identified. In current work, Dahl & Velupillai (2005), posit a broad category that enables them to deal with many different languages. For them, tense, mood and aspect (TMA) form a systematic whole in which individual ‘grams’ may express such notions as past, perfective, perfect. Members of this category vary widely in morphological and syntactic characteristics. A very different, equally broad view, posits that all languages have ‘abstract tense’ which may or may not be realized as morphological tense on the surface (cf. Huang 1984). In contrast, I will take a relatively narrow approach here. I suggest that ‘tense’ is a morphological category with syntactic ramifications, and certain general semantic meanings. I consider the traditional category of tense and try to give a characterization that is useful both in understanding the system of a tensed language and in dealing with languages where the category of tense is not so clear-cut. The result allows us to identify the morphemes of a language as more or less tense-like if they have all or some of the properties of tense. I take French as an example of a language with tense par excellence. I propose morphological, syntactic, and semantic properties. Morphology and syntax: The relatively narrow approach assumes that tenses are grammatical morphemes that express location in time (cf. Comrie 1986). Further, I take it that tense is an inflectional category that plays an important role in the abstract grammatical structure of a sentence. This differentiates tense from adverbial morphemes that indicate location in time. As part of the sentence spine, so to speak, tense is obligatory. Semantics: I understand the semantics of temporal meaning along the lines of an extended Reichenbach approach. The ideas of Reichenbach (1947) have been developed by Partee (1984), Hinrichs (1986), Smith (1991/7), Kamp and Reyle (1993). Three times are needed to account for temporal interpretation: Tense codes the relation between Reference Time, the time talked about in a sentence, and Speech Time; and the relation between Reference Time and Situation Time.1 The present tense 1 ‘Reichenbach’s term for ‘Situation Time’ is ‘Event Time’. Klein (1994) proposes a slightly different approach that also involves three times: for him, tense codes the relation between Topic Time and Speech Time, and aspect codes the relation between Topic Time and Situation Time.
Tense and Context in French
185
conveys that Reference Time is simultaneous with Speech Time, past and future tenses convey that Reference Time precedes or follows Speech Time, respectively. In simple tenses Reference Time and Situation Time are simultaneous; in complex tenses Situation Time may precede or follow Reference Time (as in the French plusque-parfait, for instance). In addition, I suggest below that the atemporal meanings often associated with non-present tense give us another way of identifying tense morphems. There are thus two facets of tense, the morpho-syntactic and the semantic. I posit two features for each: together, I suggest, the four features characterize tense in many languages. (1) Features of tense Morpho-syntactic (i) Inflectional: verbal morpheme or auxiliary (ii) Syntactic projection; obligatory Semantic (iii) Temporal meaning basic: tense codes 2 relations between Speech Time and Reference Time; Reference Time and Situation Time (iv) May have atemporal meanings in certain contexts I assume that if all the features apply to a morpheme, it is a tense; and that the more features apply to a given morpheme, the more tense-like it is. The morpho-syntactic feature that tense is inflectional means that it has ramifications in the grammar of a language, as noted above. The feature that tense is an obligatory syntactic projection means that a tense morpheme is required to appear in all full clauses. The third feature, that the morpheme convey temporal information, is semantic. As stated within a Reichenbach-inspired framework, this feature requires that tense code the relation between three times, and differentiates tense from other forms that convey information (cf. Smith and Erbaugh 2005). The fourth feature, participation in atemporal interpretation, is less familiar than the other three. It is based partly on Iatridou’s (2000) discussion of the meanings of the past tense. (I should note that Iatridou herself does not propose atemporal meaning as a feature or criterion for tense generally). Cross-linguistic studies show that in many unrelated languages, tense has both temporal and atemporal and meanings. The atemporal meanings-non-actual, conditional, hypothetical, contrary-to-fact-are often associated with the past tense in particular. Iatridou’s discussion focuses on Modern Greek and English; essentially the same pattern is found in other Indo-European languages, and in many other languages, including Papago (cf. Hale 1969), Proto-Uto-Aztecan (cf. Steele 1975), Japanese, Korean (cf. Cho 1997), Hebrew, Turkish (cf. Enç 1996), Basque and others (cf. James 1982). The atemporal feature is relevant as a criterion for tense only if the resources of a given language allow. Some languages, for instance, have
186
C.S. Smith
specialized morphemes for conditionals and counter-factuals and do not use past tense to indicate atemporality; Iatridou notes that Tagalog is of this type. In Burmese, a language without surface tense, dedicated irrealis forms are used (cf. Nichols 2003). The French past tenses have the four features of tense listed above. Tense is an obligatory, inflectional morpheme; it plays a key role in the grammatical articulation of the clause (cf. Pollock 1984); tense has a primary temporal meaning, and certain tenses have atemporal meanings in conditional and counter-factual contexts.
2 Atemporal Meanings in Certain Contexts Atemporal meanings—conditional, hypothetical, contrary-to-fact-are associated with certain past tenses in French. These meanings arise in the context of conditional and counter-factual clauses. Thus French fits the portrant of a tensed language. This section explores the fourth feature in French. I do not deal with subjunctives, a related phenomenon. The discussion will focus on the interpretation of dependent si clauses. The first set of examples is from Fleischman (1989): it presents parallel instances from English, French, and Spanish. The French tenses are the présent, the imparfait, and the plus-que-parfait, In the dependent si clauses the tenses do not have temporal meaning. (2) a. If I have time, I’ll write to you. a.’ Si j’ai le temps, je t’écrirai. a”. Si tengo tiempo, te escribo. b. If I had time, I would write to you. b’. Si j’avais le temps, je t’écrirais. b”. Si tuviera tiempo, te escribiría. c. If I had had time, I would have written to you. c’. Si j’avais eu le temps, je t’aurais écrit. c”. Si hubiera tenido tiempo, te habría escrito. In each case, the tense the if-clause of a conditional has a non-actual meaning. The présent tense in the ‘a’ examples conveys some degree of uncertainty; the past tense in the ‘b’ examples conveys a greater degree of uncertainty; the pluperfect tense in the ‘c’ examples conveys a hypothetical or contrary to fact meaning. S. Fleischman uses the metaphor of temporal distance to explain the interpretation of tense in these examples. “Temporal distance is pressed into service to express model distance, particularly the speaker’s assessment of the certainty / reality / actuality status of a predicated situation” (cf. Fleichman 1989: 4–5).
Tense and Context in French
187
I add other examples of conditional clauses from my own investigations with native speakers of French (all the examples given below, and others along the same lines, have been checked with several speakers). The a sentence have the présent, the b sentences the imparfait, and the c Sentences the plus-que-parfait. (3) a. Si Jean passe l’examen, il y réussira. b. Si Jean passait l’examen, il y réussiait. c. Si Jean avait passé l’examen, il y aurait réussi. (4) a. Si tu achetais ce costume, tu ressemblerais à Humphrey Bogart. b. Si (un jour) tu achetais ce costume, tu ressemblerais à Humphrey Bogart. c. Si tu avais acheté ce costume, tu aurais ressemblé à Humphrey Bogart. The a sentences are Conditional, expressing a possibility in the future. The b sentences are also about the future, but suggest that the possibility is not very likely to occur. These sentences exemplify the Future Less Vivid. They are about the future, and still realizeble, therefore distinct in meaning from past conditionals, usually known as counter-factuals (cf. Iatridou 2000: 234). The c examples are usually interpreted as Counterfactual, in the absence of specific contextual information that suggests otherwise. The counter-factual interpretation of past conditionals is pragmatic rather than semantic, as many have noticed. To see this, consider the following scenario: You are a doctor. You arrive at the hospital in the morning to find that one of your patients manifests disturbing symptoms. You wonder if the night nurse, who is no longer on duty, gave the medicine that you had prescribed. You suggest looking at the nurse’s records to find out: (5) Si l’infirmière la lui avait donnée, elle l’aurait noté. This sentence is a past conditional, with the same forms as in (3c) and (4c) above. But it is hypothetical rather than counter-factual in the scenario given: neither speaker nor audience actually knows what has happened. Other natural scenarios for hypothetical interpretation are criminal investigations, and reconstructions of historical events. See Karttunen and Peters (1979) for a discussion of the contextdependent nature of counter-factual interpretation. The term ‘counter-factual’ is therefore somewhat misleading; I retain it here because it is commonly used for this construction. I will use the general term ‘conditional clause’ for clauses introduced by si. The preferred interpretation is often counterfactual for past conditional clauses, because people often know whether the past event they are taking about has taken place. The conditional and hypothetical interpretations can be explained with the Gricean maxim of Quantity. The two clauses of the maxim allow for two competing notions of quantity, as Levinson (1983) emphasizes. The first clause says (a) Give only as much information as you need to; the second says, (b) Give as
188
C.S. Smith
much information as you can. Since people usually know what happened in a given situation, it’s natural to use clause (a) in interpreting a conditional sentence. If one doesn’t know what happened, one uses clause (b) to arrive at an interpretation. French has other past tenses besides the imparfait and the plus-que-parfait, notably the passé composé and the passé simple; I do not consider here the passé antérieur or the past subjunctive forms. Strikingly, neither the passé composé nor the passé simple are fully acceptable in conditional clauses: (6) a. ??? Si demain son état de santé a empiré, contactez-moi tout de suite. b. ∗ Si demain son état de santé ∗ empira, contactez-moi tout de suite. c. Si demain son état santé empirait, contactez-moi tout de suite. The imparfait appear in (6c), for contrast. According to native speakers, the passé composé is strange in this construction, whereas the passé simple is absolutely impossible. To account for the different meanings of past tense, Iatridou introduces a semantic ‘exclusion feature’. She proposes that the past tense morpheme “provides a skeletal meaning of the form ‘T(x) excludes C(x)’. [· · · · ] C(x) stands for the x that for all we know is the x of the speaker. [· · · · ] The variable x can range over times or worlds” (cf. Iatridou 2000: 246). When the exclusion feature ranges over times, it has past meaning. When in ranges over worlds, it excludes the current reference world, as in the interpretation of the past tenses above. A clause with atemporal meaning is not taken as an assertion that a proposition is true in the current world. In Iatridou’s account the temporal meaning of the past tense excludes the present Topic Time. The term Topic Time, due to Klein (1994), denotes the time for which an assertion is made in a sentence. The notions of Topic Time and Reference Time appear to be very close. However, Bohnemeyer (2003) argues convincingly that there are important differences between Klein’s system and the modified Reichenbach approach. The exclusion feature is part of the meaning of a given tense; in a Reichenbach-based account, it affects the statement of the relation between Reference Time and Speech Time. The French imparfait and plus-que-parfait have a provision that does the work of Iatridou’s semantic exclusion feature that ranges over times or worlds. However, the passé composé and the passé simple do not have it. One can say in Iatridou’s terms that the exclusion feature for these tenses ranges over times only. The French past tenses code aspectual as well as temporal information. The imparfait normally conveys an imperfective aspectual viewpoint, while the plusque-parfait has the perfective viewpoint. In her discussion of the atemporal past tense, Iatridou (2000) shows that past tenses in atemporal contexts do not have their usual aspectual meaning in Modern Greek or in English. The same is true for French. The aspectual meaning of these tenses is neutralized in the context of a si-clause, neither imparfait nor plus-que-parfait has their standard aspectual value. Usually, of course, the imparfait conveys the imperfective viewpoint and the plus-que-parfait conveys the perfective viewpoint. (There is individual variation in languages on this
Tense and Context in French
189
point: Navajo viewpoints have their consistent aspectual value in conditional contexts (cf. Smith et al. 2003). Co-occurrence and interpretation provide evidence for the claim that the French tenses do not have their usual aspectual values in conditional clause contexts. Consider the following examples, in which the tenses appear in si-clauses with forms that are not usually compatible with their aspectual value. Imperfectives are not normally compatible with completive verbs or adverbials (cf. Smith 1991/7). In (7) the si-clauses with the imparfait have completive meaning. One might claim that the completive interpretation of (7a) is pragmatic, due to a natural extrapolation from the information semantically conveyed in the clause. But in (7b) the completive meaning is conventional, conveyed by the completive adverbial dans une semaine. (7) a. Si vous repeigniez cette maison, vous pourriez la vendre avant l’été. b. Si vous repeigniez cette maison dans une semaine, vous pourriez la vendre avant l’été. The examples show that the imparfait has neither its temporal or aspectual value in the context of a si-clause. The same is true for the plus-que-parfait: the si-clause in (8) is ambiguous. It allows either a bounded or unbounded interpretation. In the former Jean began to talk when Marie arrived; in the latter he was already talking when she arrived. (8) Si Jean n’avait pas parlé quand Marie était arrivée, il aurait vu qu’elle était dérangée. This kind of ambiguity is not found in standard uses of the plus-que-parfait. I will assume that the aspectual viewpoints of the imparfait and the plus-que-parfait are neutral in these contexts. The neutral viewpoint allows open or closed interpretations, depending on information in the context. To provide for this, we posit a neutral viewpoint that makes visible for semantic interpretation an event entity with the initial endpoint, and at least one internal stage if it is durative; or a stretch of a state entity.2 This is indeterminate enough to allow the interpretations that arise. The neutral viewpoint is not entirely unconstrained, however; for discussion and examples see Smith (1991/7). I conclude that the French imparfait and plus-que-parfait have the properties of tense given above in (1). But the other past tenses, the passé simple and the passé composé have all but the last property, atemporal meanings in fixed contexts. The temporal meaning associated with these tenses must differ accordingly. I give schematic meanings for the four tenses below, assuming construction rules like those of Discourse Representation Theory, or DR Theory (cf. Smith 1991/7, 2 Since states are entirely homogenous, with the subinterval property, this is enough to guarantee that the state entity has the relevant properties.
190
C.S. Smith
Kamp and Reyle 1993). The representation explicitly specifies a different aspectual viewpoint, and alternative worlds, when an imparfait or plus-que-parfait morpheme appears in the context of a conditional si-clause, indicated by the notation ‘Ccontext’. The representations do not differentiate between the passé simple and the passé composé; the differences between these tenses arise from other features. (9) Past tense meanings a. imparfait: imperfective;
RT < SpT; SitT = RT in C-context, neutral; RTw = SitTw
b. plus-que-parfait: perfective;
RT = SpT, RT < SpT; RT < SitT in C-context; neutral; RTw = SitTw
c. passé composé: perfective;
RT < SpT; SitT = RT
d. passé simple: perfective;
RT < SpT; SitT = RT
These meanings are interpreted by the construction rules of DR Theory. The provision concerning the temporal relation between RT and SpT is vacuous here, as Iatridou 2000 points out. In the context of a conditional clause, the rules for the imparfait and plus-queparfait would choose the neutral viewpoint and the atemporal, alternative world meanings.
3 Can Future be a Tense? The status of the future as a tense has been controversial for quite some time. Semantically the future always has an element of modality. As J. Lyons puts it “Futurity is never a purely temporal concept; it necessarily includes an element of prediction or some related notion” (cf. Lyons 1977: 677). This point is not in question, I take it. But there is a question as to whether a future is ever a tense. Temporally the future is symmetrical to the past. Both are related to Speech Time: the past precedes Speech Time, the future follows it. Future morphemes have a modal property as well. What is at issue is the modal meaning of the future: for some scholars, modality precludes its membership in the grammatical category of tense. In one view, only past and present morphemes can be considered tenses. J. Bybee, for instance, claims that “the future does not belong in the same grammatical category [tense] as the past” because of its essentially modal nature (cf. Bybee 1985: 157). Supporting this view L. Stassen notes that across languages, future time expressions tend to have idiosyncratic formal behavior (adduced in Stassen 1997: 354 et seq; Stassen excludes future from the category of tense). On the other hand, however, Palmer (1985) notes that future morphemes may either pattern with modals, as they do in English; or they may pattern with past and present tenses, depending on the language (cf. Palmer 1985: 217). And O. Dahl and V. Vallupulai
Tense and Context in French
191
in a recent study, assume the category of future tense and find many instances of it across languages (cf. Dahl & Velupillai 2005). I take the view that a future morpheme may be part of a tense system, if it has the features of a tense posited above. Thus I assume that the question is independent of the semantic fact that the future has modal force. I will develop an account of the French future tense, the futur, that supports this view. (I will not consider other ways of conveying futurity in French). French has a morpheme known as the futur: it is inflectional, marked on the verb, and has a future temporal meaning. The futur is a prime candidate for a future tense. What might suggest that it is not a tense? according to the view sketched above, futures should be categorized as modals – presumably, both syntactically and semantically. To investigate this aspect of the status of the futur, I have developed a set of examples based on Enç’s 1996 study of will in English. Enç asks whether will is a tense. She notes that some linguists assume that will is ambiguous between a tense and a modal (Hornstein 1990), others have assumed that it is a modal (cf. Jespersen 1924, Smith 1978), and others find the evidence inconclusive (cf. Comrie 1986). Enç (1996) considers those cases where will is more or less temporal in meaning, rather than its clearly modal occurrences. She argues that future meaning is not enough for a morpheme to qualify as a tense, since futurity is common to several intensional expressions. Therefore, the future meaning of will doesn’t necessarily mean that it should be categorized as a tense. She then argues that will patterns with modals rather than with tenses in interpretation. To do this she considers the interpretation of clauses with will in sequence of tense and relative clause contexts. Enç shows that clauses with will do not have the sequence of tense interpretation that is typical of the past tense.3 They pattern with modals rather than with the past tense. Enç concludes that will is a modal not a tense. The arguments for the status of English will as a modal are quite convincing. The basic observation about sequence of tense is that a past tense in certain contexts does not necessarily have its usual relational value of pastness to an anchor time. One such context is a complement clause embedded under a past tense matrix clause. Past tense clauses in such contexts tend to convey a time that is simultaneous with that of the main clause, rather than anterior to it. There are some variations according to the situation type of the embedded clause.4 Adapting the tests used by Enç, I look at the French futur in sequence of tense and relative clause contexts. Of course, there is an important different between the languages in this regard: in French, modals are verbs and do not form a distinct syntactic category as English
3 Enç analyses English as having only one tense, the past. For her what is commonly known as present tense morphology is actually agreement. Therefore she compares modal interpretations only to past tense interpretations in modal contexts. This difference does not substantially affect the discussion here. 4 The term ‘situation type’ refers to the aspectual value of a verb and its argument, or verb constellation. The main classes are events and states.
192
C.S. Smith
modals do. Although the tests are not perfectly analogous, they do provide an additional source of evidence about the status of the French futur. The question that I ask is directly analogous to Enç’s: do clauses in the futur pattern interpretively with the French past tense? For the latter I use the passé composé and the imparfait. The past tense usually conveys that Reference Time is prior to an anchor time. The relational value is ‘anteriority’, which I note as ‘<-’. In these contexts the anchor time is that of the matrix clause, rather than the default Speech Time. The facts about the présent and past tense are well known. A past tense complement clause embedded under a past tense matrix clause may be taken as simultaneous with the matrix time, or as before the matrix time. In contrast, present tenses complements have their usual relational value. The complement clause examples are in 3 groups: passé composé and imparfait under passé composé, présent under présent, futur under futur. The présent conveys that Reference Time is simultaneous with Speech Time, noted as ‘=’; the futur conveys that Reference Time is posterior to Speech Time, noted as ‘->’. Compare the interpretations of the embedded clause in each group: (10) Past under past. Relational value [<-RT] a. Marie a dit qu’elle souffrait. state: = or
RT] a. Marie dira qu’elle souffrira. state: = or -> b. Marie dira qu’elle gagnera. event: = or -> The interpretations of past-under-past and future-under-future readings are symmetrical: both allow a reading in which the tense has its usual relational value, anteriority or posteriority, and a simultaneous reading as well. Event embedded clauses tend to have the anterior or posterior reading, whereas state embedded clauses tend to have the simultaneous reading. For completeness, I also give examples with a modal verb in the sequence of tense complement context. The interpretations are like those of sentences with the futur: (13) Modal verb under modal verb: a. Marie doit dire qu’elle doit souffrir en silence. b. Marie doit dire qu’elle doit gagner.
state: = or -> event: ->
This pattern suggests that the French futur is a tense like other tenses. Now consider the passé composé and the futur in relative clauses with the same tenses in the main clause. Here too one finds the sequence of tense reading, for stative clauses only. Thus in (13a) and (13c), relative clauses with both past and future
Tense and Context in French
193
tenses have symmetrical interpretations in which the time of the state is simultaneous to the time of the main clause situation. In this reading neither passé composé nor futur has its usual relational value. For completeness (14) shows that modal verbs in the futur have the same sequence-of-tense interpretation. (14) Relative clauses a. Marie a vu l’homme qui souffrait. b. Marie a dit qu’elle a gagné c. Marie verra l’homme qui souffrira. d. Marie verra l’homme qui parlera.
state: = or <event: <state: = or -> event: ->
(15) Relative clauses with modals e. Marie doit voir l’homme qui doit souffrir en silence. f. Marie doit voir l’homme qui doit gagner.
state: = or -> event: ->
There were differences among speakers with these and related examples. For some the modals are always forward-looking; for others, they may be simultaneous or forward-looking. The point for this discussion is that the interpretation of French futur is symmetrical to that of past tenses, in sequence of tense contexts. In relative clause contexts, past and future tenses have their consistent relational value in event sentence; in state sentences, they have either the sequence of tense simultaneity value, or the relational value of the tense.5 There is no justification here for not treating the futur as a tense; nor for treating modals differently from other tensed verbs. The futur has the morpho-syntactic features of tense given above, and the basic meaning of temporality. The fourth feature, atemporality, does not apply: the futur does not have atemporal uses. The reason may be simply that conditional morphemes are available for this purpose, like the modal-and-future morphemes of other European languages.6 I give a schematic meaning for the French futur, assuming it has a neutral aspectual viewpoint. (16) Future tense meaning futur: neutral; RT>SpT; SitT = RT Recall that the neutral viewpoint is informationally indeterminate allowing bounded or unbounded – essentially perfective or imperfective – interpretations. I now turn to the interpretation of tense in context.
5 The audience at the Colloque agreed with the judgements of the native speakers whose judgments I report here. 6 In Navajo, which does not have dedicated conditional morphemes or inflections, there are future forms – independent morphemes and inflection – that have atemporal meanings in conditional contexts (cf. Smith, Perkins & Fernald 2003).
194
C.S. Smith
4 Tense Interpretation in Discourse All discourse advance through a structure of some kind. I will show that there are three different principles of discourse advancement, each involving a distinct interpretation of tense. I demonstrate with discourse passages realizing different ‘Discourse Modes’. The notion of discourse mode, presented in Smith (2000, 2003), characterizes passages of written texts. Before introducing the Discourse Modes, I make some remarks about tense in discourse. For sentences in isolation, tense gives information that temporally locates the situation of a clause. But the information is quite vague. The semantic meaning of a non-present tense locates situations before or after Speech Time; present tense of course locates situations at Speech Time.7 Other information often specifies further. For instance, a past adverbial in a sentence or its context can specifies further. For instance, a past adverbial in a sentence or its context can specify the time conveyed by a past tense (e.g. Marie est arrivée hier). The point that contextual information can specify the interpretation of a tense is well-known, especially since the work of Partee (1984). However, the facts are quite complex for discourse, perhaps more than has been appreciated. In discourse passages, tense information is part of the input to an interpretive process. The temporal interpretation of a given tense in a given context varies according to the type of context and other information. In some cases the situation expressed by a clause is temporally related to other situations in the passage; in some cases it is temporally related to an anchor time. The key factor is the Discourse Mode of a text passage. Within a text one recognizes stretches that are intuitively of different types, e.g. narrative, description, argument, commentary. These stretches tend to have a particular force and a characteristic cluster of linguistic features and interpretations. They realize different Discourse Modes. I posit five modes: Narrative, Report, Description, Informative, and Argument. Each makes a different contribution to the text. The list is not exhaustive– it omits conversation and procedural discourse, for instance–but includes the modes that commonly appear in written texts. The list of modes should be relatively short if it is to capture significant generalizations. I will assume that the modes very in point of view. I do not consider persuasive discourse a separate mode or genre: persuasion appears in texts of many kinds. The modes cut across genre lines. Two linguistic features characterize the modes, both involving temporality in the larger sense. The first is the type of situation that a text passage introduces into the universe of discourse, the Discourse Representation Structure. Texts introduce individuals, concepts, and times into the universe of discourse. They also introduce situations such as events and states, which are familiar from aspectual studies. In formal accounts, individuals, times, and situations are
7 There are languages with more specific meanings for past and future tenses, expressing different degrees of remoteness (cf. Dahl 1985, Comrie 1986).
Tense and Context in French
195
represented as different types of entity in the representation. Each entity is licensed by information in a text. I work with a new, extended classification of situation entities that recognizes several types of non-dynamic, stative situations. This classification distinguishes the five discourse modes and is perhaps their most important feature. There are three main types of situation entity: Eventualities, General Statives, and Abstract entities.8 The three types are distinct on conceptual grounds, and have distinguishable linguistic characteristics. (17) Classes of situation entities Eventualities – specific events and states General statives – generic and generalizing states Abstract entities – facts and propositions Situation entities are conceptual categories, expressed linguistically at the level of the clause. They are realized by verb constellations and by nominal forms (cf. Smith 1991/7). The classes have linguistic correlates and thus constitute covert linguistic categories in the sense of Whorf (1956). In each discourse mode certain types of situation entity predominate, as follows: (18) Different situation entities predominate in passages of each mode Narrative Report Description Information Argument
– eventualities – eventualities, general statives – states, ongoing events – general statives – abstract entities, general statives
The linguistic correlates for eventualities are worked out for French and several other languages in Smith (1991/7); see Smith (2003), and Smith (in preparation), for study of general statives and abstract entities in French. The Discourse Modes are also distinguished by a second linguistic feature: the principle of advancement, or text progression, that holds for a given mode. Readers advance through the episodes of a story, the stages of an argument, the classifications of an informative text. Each of these involves a different type of advancement. Three of the modes are temporal: Narrative, Description, and Report. Passages of the temporal modes advance as location—temporal or spatial—changes. Narrative advances through narrative time, with situations related to each other. Description is static temporally; it advances by changes in spatial location, within the scene described. In Reports, situations are related to the time of report, often the presentSpeech Time, and advancement involves a change of time. The text modes of Argument and Information are not temporally organized, though they may include 8 Only abstract entities expressed by complement clauses can be recognized linguistically. Facts and propositions that are expressed by independent clauses do not have consistent linguistic correlates (cf. Smith 2003).
196
C.S. Smith
eventualities that have temporal location. Atemporal texts progress with metaphorical changes of location through the information space of the text. I shall focus on the temporal modes in this article; for discussion of the atemporal modes see Smith (2003). The temporal modes, Narrative, Description and Report, progress by three different temporal principles. Narrative primarily introduces event and state entities into a discourse. The situations are temporally related to each other according to their aspectual properties, and information specified by adverbs. In narrative tense conveys continuity. Description primarily introduces states and ongoing events into the universe of discourse, and temporally locates them at a single established time. Tense in description conveys a time anaphoric to a previously established time. Report primarily introduces events and states, and general statives; they are temporally related to Speech Time. In Report tense is deictic; this is the default, found in the atemporal modes as well. The principles of interpretation are modeled using an extended Reichenbach framework. In texts with the principle of tense continity, Reference Time advances according to bounded events or temporal adverbials. In texts with the principle of deicitic tense, Reference Time changes with different relations to Speech Time. In texts with the principle of tense anaphora, Reference Time is simultaneous with a previously established time. To calculate temporal location in a discourse passage one must have access either directly or indirectly to the discourse mode of the passage. I do not develop the formal analysis here; see Smith (2003) for a detailed account. Narrative: Narrative consists of consequentially related events, recounted in the sequence in the sequence in which they occur–the main story line (cf. Labov and Waletzky 1966, Kamp and Rohrer 1983, Moens 1987). The essence of a narrative is dynamism: as the events of the narrative unfold in sequence, we understand that narrative time advances. The dynamism that advances narrative time is due to aspectual information, and to explicit time adverbials and inference. Narrative advances with bounded, perfective events and with explicit temporal adverbials. It fails to advance with states and ongoing events, unless additional information warrants an inference of advancement. The information that tense conveys after the initial sentence of a narrative is continuity. Bounded events are conveyed by clauses with event verb constellations and the perfective viewpoint–the passé composé and the passé simple. The first example illustrates a standard past tense narrative. The clauses that advance RT are marked by an arrow and numbered subscripted Es (for Event). (19) Past tense narrative . . . ll parlait des personnes dont les propriétés bordaient la route. Au milieu de Toucques ->E1 il dit « En voilà une Mme Lehoussais. . . ». Félicité n’entendit pas le reste; les chevaux trottaient, l’âne galopait; ->E2 tous enfilèrent un sentier, ->E3 une barrière tourna, ->E4 deux garçons parurent, et ->E5 l’on descendit devant le purin, sur le seuil même de la porte. (G. Flaubert, Un Coeur Simple, p. 19)
Tense and Context in French
197
(20) Continuity interpretation of tense E1 . . .. . .. . .. . .. . .E2 . . .. . .. . .. . .. . .E3 . . .. . .. . .. . .. . .E4 . . .. . .. . .. . .. . .E5 RT1 < SpT RT2>RT1 RT3>RT2 RT4>RT3 RT5>RT4 Strikingly, almost the same principle holds for present tense narrative: event verb constellations and adverbs advance narrative time. (21) Present tense narrative . . .Dans son dos, ->E1 les mains jouent avec la carte de la lune qu’il a dans sa poche fessière. II pense aux rochers de la lune et à la fine poussière météorique, semblable à du charbon mulvérulent, qui la recouvre. ->E2 Puis elles la lâchent et ->E3 se prennent elles-mêmes comme tenon et mortaise jusqu’à ce qu’il ->E4 finissse par les balancer devant lui. (R. Detambel, Mésanges, 2003) The continuity interpretation is stated as a temporal principle of interpretation for narrative dynamism: (22) Narrative Dynamism Principle a. Bounded event clauses advance narrative time; State and unbounded event clauses do not advance narrative time. b. With RTx immediately preceding; if e is a bounded event: RTy > RTx An additional statement is needed for adverbial advancement. Like other pragmatic principles, this one can be overridden by explicit information. Description: Time is static in description: situations are temporally located, but the time does not change. In passages of Description, time is stable or suspended, without dynamism. Situtations are located at a time already established in a text so that all the clauses of a given passage have the same Reference Time. Description passages appear in fiction, travel writing, and accounts of states of affairs. Tense is anaphoric to a time already established in the discourse: in the fragment below the bounded event aperçut provides the RT for what follows. The entities in Description are unbounded events and states. Often a locative adverb or other anchoring information appears at the beginning of a description. I assume a tacit durative time adverb that has scope over the passage. Event verb constellations in passages of Description are atelic, sometimes by coercion due to the tacit time adverb. If motion is involved, it is without significant changes of state and there is no sense that time advances. (23) Description Monsieur Guillaume regarda la rue Saint-Denis. -> ll aperçut alors le passant en faction, qui de son côté contemplait le patriarche de la draperie. = Monsieur Guillaume portait de larges culottes de velours noir, des bas chinés et de souliers carrés à collet carré, enveloppait son corps légérement voûté d’un drap
198
C.S. Smith
verdâtre garni de grands boutons en métal blanc mais rougis par le usage. = Ses cheveux gris étaient si exactement aplatis et peignés surson crâne jaune, qui le faisaient ressembler à un champ sillonné. = Ses petits yeux verts, percés comme avec une vrille, flamboyaient sous deux ares marqués d’une faible rougeur à défaut de sourcils. (Balzac, La Maison duchat-qui-pelote, p. 31) (24) Anaphoric interpretation of tense E0 . . .. . .. . .. . .S1. . .. . .. . .E1. . .. . .. . . RT1 < SpT RT2 = RT1 RT2 = RT1 (25) Static Interpretation Principle a. States and unbounded events are located at an established time b. RTx immediately preceding: RTy = RTx The continuing RT may be past or present – or, in principle, future – according to the context of the passage. In a sense this pattern complements narrative dynamism: it applies to unbounded situations and locates them at a single time. Report: In Reports, eventualities and general statives are related to Speech Time and are not temporally related to each other. Time and space adverbials are common in reports, as are changes in tense. Significantly, the events of a report could appear in different order without making a real different –quite unlike narrative (cf. Caenepeel 1995) Tense and adverbials are Deictic, oriented to Speech Time. As before, an arrow before a clause indicates that it advances time; an = indicates that the previous RT is maintained. (26) Report -> Quatre employés de la Compagnie parisienne de chauffage urbain ont été tués et neuf autres blessés grièvement dans l’explosion, merecredi 15 novembre, d’une canalisation, porte de Clignancourt à Paris, dans le 18e arrondissement. -> 2 Le drame s’est produit lorsque les ouvriers réalisaient, à treize mètres sous terre, des tests dans une nouvelle canalisation de vapeur de chauffage urbain. -> 3 Les victimes s’apprêtaient à mettre en service cette nouvelle conduite. = 4 Les causes de l’explosion sont inconnues. -> 5 Les personnes blessées se trouvaient à trois mètres environ sous terre et coordonnaient l’opération. (Le Monde 11/17/00) The deictic interpretation maintains Speech Time as the center, and there are changes in time with changes in tense. Situations related to Speech Time. Tense and adverbials are deictic, related to Speech Time. (27) Deictic interpretation of tense E1 . . .. . .. . .E2 . . .. . .E3 . . .. . .S1. . .. . .. . . RT : < SpT <SpT <SpT = SpT No additional construction rules or principles for calculating advancement are needed for the Report mode. The deictic principle is applied anew for each clause.
Tense and Context in French
199
Conclusion: The interpretation of tense in discourse varies according to the discourse mode of a passage. To calculate temporal location in a discourse passage one must have access either directly or indirectly to the discourse mode of the passage; see for instance the DR theory rules in Smith (2003). The semantic meanings of tense must be compatible with different pragmatic interpretations.
5 General Conclusions Investigation of French tenses supports generalizations about tense, and enables us to arrive at a better understanding of tense as a cross-linguistic category. The discussion and example have established the following points. (a) The French imparfait and plus-que-parfait have a hypothetical, often counterfactual, meaning. in conditional contexts. They thus participate in the alternation between temporal and atemporal meanings that appears in many languages of the world. The passé simple do not have atemporal meanings, however. (b) The French futur functions both syntactically and semantically like the other tenses of French. This supports the view that the status of a future morpheme depends on the particulars of a language. The form marking future may be a real tense–as in French: it patterns with other tenses in various ways, including sequence of tense interpretations. In contrast, English will patterns with modals syntactically and interpretively and is not considered a tense. (c) Tense in discourse is interpreted according to three different patterns, depending on the Discourse Mode of a text passage. Discourse Modes distinguish text passages passage according to the classes of entity the passage introduces and the principle of advancement. The discussion and examples have shown that tense interpretation is pragmatic, varying according to the discourse mode of a passage. For each of these points about tense, access to context is essential: the conditional context of si or another form that triggers atemporal interpretation; the context of sequence of tense interpretations; the matrix sentence for a complement or relative clause; the Discourse Mode of a text. Acknowledgement I would like to thank the audience at the Colloque for interesting comments and discussion. I especially thank Jean-Pierre Cauvin, Judy Coffin, Pascal Denis, and Dina Sherzer for kindly providing native speaker judgments.
References Bohnemeyer, J. (2003). Relative tense vs. aspect: The case reopened, Presentation at SULA 2, University of British Columbia. Bybee, J.(1985). Morphology, Amsterdam: Benjamins. Caenepeel, M. (1995). Aspect and text structure, Linguistics 33: 213–253. Cho, E. (1997). Counterfactuals in Korean and Japanese: Interaction between verbal morphology and interpretation, Ms., Cornell University, Ithaca, NY. Comrie, B. (1986). Tense, Cambridge: Cambridge University Press. Dahl, O. (1985). Tense and Aspect Systems, Oxford: Blackwell.
200
C.S. Smith
Dahl, O., Velupillai, V. (2005). Tense and aspect: Introduction, in: B. Comrie, M. Dryer, D. Gil, B. Comrie, (eds), The World Atlas of Language Structures. Oxford: Oxford University Press, pp. 266–267. Dahl, O., Velupillai, V. (2005). Future tense, in: M. Hasplemath, M. Dryer, D. Gil, B. Comrie, (eds), The World Atlas of Language Structures. Oxford: Oxford University Press, pp. 270. Enc. M. (1996). Tense and modality, in: S. Lappin,(ed), Handbook of Contemporary Semantic Theory, Oxford University Press . Fleischman. S. (1989). Temporal distance: A basic linguistic metaphor, Studies in Language 3: 1–50. Hale, K. (1969). Papago /cim/, International Journal of American Linguistics 35: 203–212. Hinrichs, E. (1986). Temporal anaphora in discourses of English, Linguistics and Philosophy 9: 63–82. Hornstein, N. (1990). As Time Goes By, Cambridge, MA: MIT Press. Huang, C.T.J. (1984). On the distribution and reference of empty pronouns, Linguistic Inquiry 15: 531–574. Iatridou, S. (2000). The grammatical ingredients of counterfactuality, Linguistic Inquiry 31: 231–270. James, D. (1982). Past tense and the hypothetical: A cross-linguistic study, Studies in Language 6: 375–402. Jesperson, O. (1924). The Philosophy of Grammar, London: G. Allen & Unwin. Karttunen, L., Peters, S. (1979). Conventional implicature, in: Ch. K. Oh, P. A. Dineen (eds), Syntax & Semantics 11: Presupposition, New York: Academic Press, 1–56. Kamp, H., Reyle, U. (1993). From Discourse to Logic, Dordrecht: Kluwer. Kamp, H., Rohrer, C. (1983). Tense in texts, in: R. Bauerle, R. Schwarze, A. von Stechow (eds), Meaning, Use and Interpretation of Language, Berlin: de Gruyter. Klein, W. (1994). Time in Language, London: Routledge. Krifka, M., Pelletier, F., J. Carlson, G., ter Meulen, A., Chierchia, G., Link, G. (1995). Genericity: An introduction, in: G. Carlson, F. J. Pelletier, (eds), The Generic Book, Chicago, Illinois: University of Chicago Press. Labov, W., Waletzky, J. (1966). Narrative analysis: Oral versions of personal experience, Essays on the Verbal and Visual Arts: Proceedings of the 1966 Annual Meeting, American Ethnological Society, Seattle: University of Washington Press. Levinson, S. (1983). Pragmatics, Cambridge: Cambridge University Press. Lyons, J. (1977). Semantics, Cambridge: Cambridge University Press. Malotki, E. (1983). Hopi time: a linguistics analysis of the temporal concepts in the Hopi language, Berlin: New York: Mouton. Moens, M. (1987). Tense, aspect, and temporal reference, Ph.D thesis, Centre for Cognitive Science, University of Edinburgh. Nichols, L. (2003). Attitude meaning and form in the interpretation of counterfactuality, Presentation at SULA 2, University of British Columbia. Palmer, F. (1985). Mood and modality, Cambridge: Cambridge University Press. Partee, B. (1984). Nominal and temporal anaphora, Linguistics and Philosophy 7: 243–286. Pollock, J.-Y. (1989). Verb movement, UG, and the structure of IP, Linguistic Inquiry 20: 365–424. Reichenbach, H. (1947). Elements of Symbolic Logic, London: Macmillan. Smith, C. S. (1978). The syntax and interpretation of temporal expressions in english. Linguistics and Philosophy 2: 43–100. Smith, C. S. (1991/7). The Parameter of Aspect, Dordrecht: Kluwer. Smith, C. S. (2000). Discourse Mode: A linguistically interesting level of text structure, Proceedings, Third Workshop on Spoken and Written Texts. Austin, Texas: Department of Linguistics, University of Texas. Smith, C. S. (2003). Modes of Discourse, Cambridge: Cambridge University Press.
Tense and Context in French
201
Smith, C. S. (in preparation). The discourse modes of French. Smith, C. S.; Erbaugh, M. S. (2005). Temporal interpretation in Mandarin Chinese, Linguistics, 43: 713–756. Smith, C. S., Perkins, E., Fernald, T. (2003). Temporal Interpretation in Navajo. Proceedings of the SULA Conference. Amherst Working papers in Linguistics, Amherest, MA: University of Massachusetts. Stassen, L. (1997). Interansitive Predication, Oxford: Oxford University Press. Steele, S. (1975). Past and irrealis, International Journal of American Linguistics 41: 200–217. Whorf, B. L. (1956). Language, Thought, and Reality, Cambridge, MA: MIT Press.
The Acquisition of Time Talk: Relations Between Child and Adult Grammars Carlota S. Smith
Abstract A Reichenbach-type theory is proposed that accounts for a child’s early system of temporal reference and relates it directly to the adult system. This theory makes predictions different from those of a competing theory, a strong decentring hypothesis. Experimental results and data from spontaneous speech are examined, and the predictions of the first theory are borne out: young English-speaking children express notions of temporal ordering as well as aspectual notions.
1 Introduction The ability to include temporal reference in utterances develops gradually, even slowly. Apparently the slow pace is due to several factors: the syntactic complexity of full-fledged time talk (a phrase I owe to Emmon Bach), the semantic complexity of the notions involved, and the relatively advanced cognitive development that is required. The importance of the third factor has been strikingly demonstrated in naturalistic and experimental studies in this area. Children begin to use temporal adverbials well after they have developed the ability to produce other adverbials, and relatively complex temporal expressions are not acquired until about 5 years. The interpretation often given of these facts is a cognitive one: only after a certain stage of development, it is suggested, can children talk about time (cf. Cromer (1968) and Ferreiro (1971), who stress the importance of cognitive factors in the acquisition of temporal forms and temporal notions, Cromer using a naturalistic and Ferreiro an experimental approach). I will discuss in this paper the acquisition of temporal reference, with special attention to different stages of development, and to the relation between these stages and adult grammars. It is an open question whether children’s grammars are ‘part’ of adult grammars – whether there is a systematic relation between child and adult grammar. Time talk seems an area in which investigation of this question could be fruitful. The grammar of adult time talk is fairly well understood, although there are some competing analyses. This understanding is an indispensable prerequisite for the study of acquisition. Further, a good deal is known about the acquisition of temporal forms in English, the language I will focus on here. R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_8,
209
210
Acquisition of Tense
Children’s grammars might undergo shifts at crucial stages of cognitive re-organization; whether they actually do so is not now known. It would be of the utmost interest if a shift in grammars were demonstrated – one that constituted reorganization rather than a more gentle change. Such a shift would show that child grammars are not necessarily part of adult grammars. If this is the case, the structure of the former does not allow inference about the structure of the latter. Time order and aspect are both involved in temporal reference. By time order, I mean expression of how events are ordered relative to a reference point. Simultaneity, anteriority, and posteriority are the basic ordering relations. Aspect involves the temporal contour of events, in the happy phrase of Hockett (1958: 237); only perfective and imperfective aspect will be considered in this paper. In English, both time order and aspect are indicated by verb inflection, verbal auxiliaries, and time adverbials. I consider here the early stages of development, when children do not regularly use temporal adverbials. The discussion below focuses on the acquisition of past time reference and its integration with the expression of aspectual distinction. The relevant linguistic forms are the past and present tense inflections and the progressive auxiliary be+ing. I will not, however, deal with morphological irregularities in this paper.
1.1 Temporal Reference: Child and Adult There is almost general agreement in the acquisition literature that by roughly the end of the third year, most English-speaking children have acquired the progressive morpheme ing and the inflection for past tense. Roughly, progressive inflection indicates imperfective aspect and nonprogressive indicates perfective aspect; the situation changes with the presence of certain adverbials. Acquisition of ing and the past tense was proposed by Brown (1973), and documented more fully by de Villiers and de Villiers (1973), Kuczaj (1976), and Bloom et al. (1977). Brown’s criteria for acquisition require that a morpheme be used in 90% of the relevant obligatory contexts. In appropriate use of the past tense, the event referred to must have occurred prior to the time of speech. Appropriate use of the progressive is more difficult to pin down (cf. below); it is usually optional. Brown took as obligatory contexts for the progressive those cases where (1) the child was engaged at the time of speech in the activity mentioned or (2) where previous related utterance used the progressive. He took ing alone as evidence of acquisition of progressive aspect. Knowing that past and progressive forms are used appropriately is hardly sufficient as an account of children’s temporal reference. What is needed is a theory of such reference: as George Miller (1975) puts it, one would like to know what ‘pastness’ means for the child. In the absence of such a theory and its confirmation, children’s use of temporal forms is open to more than one interpretation. I have proposed a theory of adult temporal reference which provides a basis for understanding
The Acquisition of Time Talk
211
children’s temporal reference (Smith 1978). The theory deals with temporal ordering, and is based on the work of Hans Reichenbach. According to Reichenbach (1947), a full temporal system for English and other languages involves the notions of SPEECH TIME (ST), REFERENCE TIME (RT) and EVENT TIME (ET). The times may be simultaneous, or one may follow another. All three times are required for English because of sentences in the perfect, such as (1): (1) John had already talked to Mary last Friday. This sentence involves 3 times: the RT is last Friday, which precedes ST; the event of talking (ET) precedes RT. I have also given other evidence that English requires 3 times, in Smith 1978 and other recent papers. RT and ET may be simultaneous, as in (2)–(4): (2) Mike is sleeping. (3) Ann won the tennis match on Thursday. (4) Rich arrives tomorrow. The time indicated by a sentence can be understood in terms of its relation to ST. If RT is simultaneous with ST, the time referred to is present; if RT precedes or follows ST, the time referred to is past or future. The temporal system must be flexible enough to account for shifts in orientation, such as complements that orient to matrix sentences and fictional narratives. This can be done by giving temporal expressions consistent relational values and allowing them to shift in absolute value. In English, tense and time adverbials function together to establish temporal reference (see further, Smith 1978). I suggest that children’s early temporal ordering system differs from the adult system in two ways: only two times are involved, and orientation is fixed at ST. The system has two times, and the basic relations of simultaneity and sequence. If this approach is correct, it gives a precise meaning for pastness: pastness in the child’s system indicates a time anterior to ST. Only with the development of a certain complexity can the system allow for a third time and flexible orientation. In this view children’s ordering systems are related systematically to those of adults. The child system is simpler but not different in organization from the adult: both have the essential property of relating a time to an orientation time by simultaneity or sequence. The approach to the perfective-imperfective aspectual distinction that I will talk here is roughly that summarized in Comrie (1976). IMPERFECTIVE aspect in linguistic utterances takes an event without regard to its completion or termination – from within as it were – whereas PERFECTIVE aspect takes an event as a whole. (This notion of perfectivity and imperfectivity is somewhat different from the traditional account of these aspects in Slavic languages. It does however correspond to the central non-habitual meanings of the aspects, according to Cochrane 1977 and Newton 1979.) Events themselves may also be referred to as perfective or imperfective. Perfective events have natural endpoints, of result or completion (drawing a circle, climbing stairs); imperfec-
212
Acquisition of Tense
tive events are activities without natural endpoints, although of course they may be terminated (jumping rope, swimming).1 The basic aspectual notions of perfective and imperfective events are apparently understood early (Brown 1957).2 What develops gradually is the ability to refer to events from more than one aspectual perspective. It is also necessary for the two parts of the temporal reference system (time order and aspect) to integrate with each other, and this happens slowly, partly perhaps because the syntax of English is relatively complex in this area.
1.2 Temporal Ordering and Cognitive Development Another point of view suggests a very different interpretation of children’s early use of linguistic forms associated with temporal ordering. Studies of cognitive development, notably those of Piaget and his colleagues, show that an important stage is reached at about the age of 6 years. This is the concrete operational stage, when children are able to deploy a complex of abilities in life and in the laboratory. It has been suggested that children are limited, before this stage, to an egocentric view of the world and are unable to ‘decentre’. It is not entirely clear just what the ramifications of egocentricity are, but one possibility is that young children are unable to talk about time other than present time. This is of course a strong view of the relation between egocentricity and temporal reference. Cromer (1968) takes this view in his discussion of temporal reference in the records of Adam and Sarah, two of the children studied by Brown. Cromer suggests that their failure to use many temporal structures before 6 years was due to their inability to decentre. The fact that both children talked about the past and the future well before this stage is not explained, however. (Cromer mentions an intermediate stage of ‘actionizing’, suggested by Bruner, but does not discuss it in detail.) Ferreiro (1971) also adopts the strong view of decentring in her discussion of experimental studies with Frenchspeaking children. More recently the strong view of decentring has been taken by Bronckart and Sinclair (1973). They suggest that when temporal ordering forms (tense) appear in the speech of children under 6, the forms are used to code aspectual rather than temporal distinction. If they are correct, it has seemed that children were coding temporal order when they were really coding aspect; this mistake in interpretation was possible because perfective aspect is generally appropriate when referring to an event in the past. Not until children have acquired the cognitive abilities 1 Duration is another factor; it will be omitted here. There are two rationales for this omission. First, duration is not in principle related to the notion of perfectivity; see Vendler (1967) and Smith (1976). Second, Bronckart (1973) has shown that duration was not significant in children’s linguistic treatment of different types of events. 2 One cannot know, of course, just how young children interpret events or utterances. In the experiment reported in Brown (1957), children’s judgements corresponded to adult judgements. When events are referred to as perfective or imperfective in this paper, I will be using standard adult judgements (and standard events).
The Acquisition of Time Talk
213
associated with decentring do they use tense to code temporal order, Bronckart & Sinclair claim. The evidence for their claim comes from an experiment conducted with Frenchspeaking children (Bronckart and Sinclair 1973). In the experiment, 3- to 8-yearolds were invited to describe events enacted before them in the very recent past, following the familiar Genevan technique. The events occurred immediately before the children described them. The events varied, notably in perfectivity: some events had clear results and natural completions (perfective events) and others did not (imperfective events). The children’s task was to describe the events. Through the age of 6, they tended to use the présent for events without clear endpoints, and the passé composé for events with endpoints. Thus the present was associated with imperfective events and the past with perfective events. Bronckart & Sinclair interpret this result as showing that the children were using tense to code aspectual differences. According to them, young children do not understand or express temporal order: ‘the role of tenses as indicators of temporal relations only begins to be apprehended from 6 years’. If this interpretation is correct, it should hold generally, since cognitive development is said to be the crucial factor. Extending the interpretation to English, we would reinterpret the past tense forms found in children’s speech as signalling perfective aspect rather than anteriority. This interpretation is very interesting. It supports the strong view of the relation between decentring and temporal reference; and it directly implies that, in the area of temporal reference, children’s grammars are different from adult grammars. I will refer to this view as the STRONG DECENTRING HYPOTHESIS .
2 The Study: The Strong Decentring Hypothesis In the remainder of this paper I shall consider the question of whether the strong decentring hypothesis can be maintained in the light of data from English. Two kinds of data will be presented or referred to: records of spontaneous speech, examined for the predicted correlation between past tense and perfectivity of events mentioned; and an experiment similar to that of Bronckart & Sinclair, conducted with English-speaking children. The strong decentring hypothesis makes a specific prediction about the type of events referred to with past inflection: such events should be perfective.
2.1 Perfective and Imperfective A brief discussion of perfective and imperfective aspect, and how they are indicated in English, is a necessary preliminary since it affects the interpretation of the data. Consider first the distinction between perfective and imperfective activities or events. The former are activities that by their very nature involve completion – sometimes with an enduring result, as in building a cabin, sometimes not,
214
Acquisition of Tense
as in climbing the stairs. Unless the cabin is completed, or all the stairs climbed, one cannot be said to have done these things. The notion of completion is clear, although in some cases what actually constitutes completion may be somewhat imprecise. Imperfective activities, on the other hand, can be stopped at any moment without change in the notion of what sort of event has occurred. The difference between perfective and imperfective events is brought out by a well-known semantic test, involving entailment. The entailments are different for progressive sentences referring to perfective and imperfective events. For imperfective events, if ‘I was Xing’, then ‘I Xed’ (John was dancing; John danced). Bur for perfective events, ‘I was Xing’ does not entail that ‘I Xed’ (John was building a house; John built a house). To know whether the second sentence is true, additional information is necessary: whether or not John completed the house (cf. Dowty (1977), Vendler (1967)). The work of Verkuyl (1972) has shown that most linguistic expressions cannot be classified simply according to the type of event to which they refer. A verb may indicate a perfective or imperfective event, depending on the rest of the sentence (objects, subjects, adverbials are all relevant). For instance, run indicates a perfective event in (5), and an imperfective event in (6): (5) Joan ran to school. (6) Joan ran in the park. Certain adverbials, however, are associated with perfectivity or imperfectivity, as Vendler (1967) showed. Speakers usually have a choice, in talking about events, as to whether to consider them as wholes (perfective) or without regard to endpoints (imperfective). This choice is available to the speaker regardless of the type of event. Languages differ as to whether the choice is obligatory or not. The choice is obligatory in English, and is indicated without adverbials by the presence or absence of the progressive auxiliary be+ing. A sentence with auxiliary be+ing indicates imperfective aspect; nonprogressive or simple aspect indicates perfectivity. The presence of certain adverbials can change the interpretation of a sentence in the simple tense to perfective, but this is irrelevant here since the children we are concerned with did not use such adverbials. There is no paradox in referring to imperfective events with perfective aspect, or perfective events with imperfective aspect. Either type of event may be viewed from within, or taken as a whole. An imperfective event may be reported without regard to environment, as in (7), and so may an imperfective event, as in (8): (7) Bill was building a cabin. (8) Mary was laughing. However, the interpretation of perfective aspect depends on the type of event. Perfective aspect implies completion for perfective events and termination for imperfective events, as the following examples illustrate:
The Acquisition of Time Talk
215
(9) Mary climbed the stairs. (10) Michael swam in the lake. A particular aspect is rarely obligatory, even when, as in English, the choice itself must be made. The aspect chosen by a speaker depends largely on extra-linguistic factors, such as what information about the event is shared by speaker and hearer. (There are a very few obligatory contexts in English, for instance, embedding under adverbials such as before and after.) In English, the realization of imperfective aspect is more complex syntactically than is perfective. Perfective or simple aspect is indicated by a verb+tense form; imperfective by the auxiliary be+ing, which has several special properties. Young children often produce sentences with just progressive ing, and since the auxiliary be carries tense, these sentences are opaque as to tense. One cannot tell, without contextual information, whether a typical child form such as (11) should be interpreted as (12a) or (12b): (11) Mary working. (12a) Mary is working. (12b) Mary was working. This opacity may have led to underestimation of the production of past progressives – that is, utterances intended as past progressives – by young children learning English.
2.2 Data from Spontaneous Speech Data about the spontaneous speech of 17 English-speaking children were examined to test the predictions of the strong decentring hypothesis as made by Bronckart and Sinclair. They claim that past tense is associated with perfective events.3 This claim leads to the prediction that verbs with past tense indicate perfective events. If this prediction is borne out, the early past forms of English-speaking children are semantically perfective although they have a suffix that indicates temporal order in the adult language. The data come from summaries of two sets of transcripts of spontaneous speech. One summary lists all the utterances of Adam and Sarah that pertained to past time in the speech samples collected by Roger Brown and his colleagues. The summary was made by Richard Cromer and appears as an appendix to his dissertation; although Cromer listed some verbs that had present inflection or no inflection as referring to past time, I have considered only verbs with past tense inflection. The second 3 This is one of two possible interpretations of their claims about perfectivity. The other will be discussed below. The ambiguity arises because they make no distinction between the aspectual nature of an event and the aspect used in referring to it.
216
Acquisition of Tense
summary lists all the verbs used with past inflection in transcripts of spontaneous speech collected by Stanley Kuczaj; he studied his son Abe and 14 other children and presents the lists in an appendix to his dissertation. The data vary in quantity: the number of relevant utterances for Adam and Sarah are meagre; the data for Abe copious, those for the other 14 children of Kuczaj’s study relatively full. All of the children used past tense inflection to report imperfective events and perfective events. The perfectivity of an event is often not correctly indicated by a list of verbs alone, as noted above. Therefore the counts given here are only approximate. However, since an effort was made to count only clear cases, the percentages reported probably underestimate children’s actual use of past inflection for imperfective events. I give some examples of complete utterances with verbs reporting imperfective events, before summarizing the actual counts. (13) Know what the bear seed? (Abe, 2;11) A goose flewed (Abe, 2;6) They all danced. (Sarah, 4;6) The results of the transcript study are presented in Table 1. The percentages indicate, for each child, how many of the verbs with past inflection were imperfective (that is, verbs that usually refer to imperfective events). No child had more than 60% of such verbs, but there was a good deal of difference between the children. Individual differences, different situations in which the speech was recorded, and number of utterances available for each child, may explain the wide range. Adam and Sarah appear in both younger and older groups, since material is available from two stages and the percentage of imperfectives changed. Ages are given in parentheses; the initials indicate children studied by Kuczaj. The imperfective verbs listed for each child appear in the appendix to this paper. The interpretation of these findings is straightforward. The children used past tense with both perfective and imperfective verbs; therefore their use of this inflection cannot be associated with perfectivity in events. This means that past tense inflection cannot be analysed, for these children, as suggested by the strong decentring hypothesis. There is, however, another interpretation of the strong decentring hypothesis. In this interpretation past tense indicates perfective aspect but is not associated exclusively with perfective events. Records of spontaneous speech cannot be used to test this idea, because perfectivity and pastness coincide: events that are terminated or completed are past. Recall that perfective aspect is appropriate for reporting both perfective and imperfective events, depending to some extent on extra-linguistic factors. Since pastness and perfectivity are confounded, other information is required to give information as to whether children use past inflection to indicate pastness or perfectivity. To answer the question of whether young children have the notion of pastness, I would like to call upon some disparate results of already-published work. Halliday (1975), in a discussion of the development of narrative abilities, gives examples of exchanges in which a child tries to talk about a past event before he has the linguistic
The Acquisition of Time Talk
217
Table 1 Imperfective verbs with past tense inflection as a percentage of all verbs with past inflection Over 40%
25–40%
Under 25%
Younger group D. N. (3;0) H. K. (3;5) K. M. (4;0) G. D. (4;0)
Abe (2;7) I. B. (3;7) V. Q. (3;8)
Adam (3;6) Sarah (3;6) N. E. (2;5) M. Z. (2;5)
Older group Sarah (4;6) L. R. (4;5) H. L. (5;0)
Adam (4;6) F. Y. (4;8) C. P. (5;8) A. B. (5;0) J. W. (5;8)
means to do so. Halliday’s data were rich in contextual information. For example, he gives this dialogue between an adult and a child of 1;8:
(14) C: Try eat lid. A: What tried to eat the lid? C: Try eat lid. Goat . . . man said no . . . goat try eat lid. . .man said no.
The child was referring, out of the blue, to an incident that had taken place on a recent visit to the zoo. Halliday interprets this to mean that the child had a notion of pastness, however rudimentary – that he was trying to talk about events that had occurred in the past. Stoel-Gammon and Cabral (1977) discuss the development of the reportative function in a child learning Brazilian Portuguese. They give examples in which children try to report or discuss past events, and not merely to refer to them as the past-equals-perfective analysis suggests. In some current work, Sachs (1977) discusses the development of children’s ability to talk about the past. She is studying this topic, and reports that by the age of 3, children are able to talk about some past events, although there are limitations on their ability to do so. The studies mentioned above are all observational and draw on detailed contextual information. There are also some relevant experimental results that point in the same direction. Harner (1976) and Herriot (1969) studied comprehension of tense inflections with picture selection tasks. Both found that children had developed notions of temporal ordering: children of 3 and 4 were able to perform successfully in the experiments conducted (which were, incidentally, quite different). I take it as well established, then, that English-speaking children have the notion of pastness by the age of 4 years, and that they tend to indicate pastness with past tense inflection.
218
Acquisition of Tense
2.3 Experimental Data In another approach to the acquisition of time talk, an experimental study was conducted with English-speaking children (diPaolo and Smith (1978)). We elicited descriptions of events from 4- to 6-year-olds in an experiment similar to that of Bronckart & Sinclair. French and English have rather different systems for indicating tense and aspect; it seemed likely that in comparing the responses of children in the two languages we would find out what behaviour was general and what was language-bound. In particular, we were interested in whether English-speaking children – who indicate aspect in spontaneous speech very early – would behave differently from the French-speaking children in the aspect experiment. The experiment, reported in detail in diPaolo and Smith (1978), differed in one respect from that of the other experiment. Before the main task, the children in our experiment performed actions with the toys according to descriptions given by the experimenter. All the children were able to do this without difficulty. The stimuli consisted of eight actions, presented singly and in combination. Four had definite endpoints (climb stairs, leave garage, bump, close box) and four did not (push, dance, fly, ride). The main task was like that of the Bronckart & Sinclair study: children watched while an experimenter performed actions of different types with toys, and then recounted what they had seen (immediately afterwards). The subjects were 28 children, aged 4;7 to 6;6. The results were strikingly different from those of the French-speaking children. The English-speaking children in this study used the past tense almost exclusively in their descriptions: fewer than 7% of the total responses had the present tense. Both progressive and simple aspect appeared in the responses. Of particular interest here is the relation between the type of stimulus event and the aspectual forms produced by the children. The patterns of response differed for type of stimulus event and age, but three general comments can be made. (1) All children responded to perfective events with both simple and progressive aspect, but they responded most frequently with simple aspect. Similarly, children described imperfective events with both aspects, but their most frequent responses had imperfective aspect. (2) In descriptions of double events, the progressive form invariably referred to the imperfective event and the simple verb to the perfective event. (The stimulus actions included simultaneous events that varied in perfectivity. Bronckart and Sinclair state that double actions were presented to their subjects, but give no examples and do not discuss responses to these stimuli.) (3) Perfective events elicited simple aspect more frequently than imperfective events elicited the progressive. Taken together, these results show that English-speaking children command both progressive and simple aspect – that is, perfective and imperfective – with the past tense by the age of 5. They can, but need not, report events in the aspect that corresponds to their perfectivity. The youngest children showed less sensitivity than the others to different stimulus types. The mid group, as noted above, showed a high correlation between the aspect of an event and the aspect of its report. The oldest children had the most varied response patterns; this may indicate a developing sophistication in the use of aspect, especially in situations with two actions.
The Acquisition of Time Talk
219
It is difficult to give a detailed interpretation of these results. Since no context was given to the stimulus actions, no one aspect was appropriate for reporting them. The relatively high correlation in aspect between event type and report may be due to children’s general abilities and preferences. It may also be due to the fact that the experimental situation emphasized just these properties. These results contrast with those of the Bronckart and Sinclair aspect experiment, done with French-speaking children. The younger children in that experiment, who correspond to Groups 1 and 2 in the one reported here, used tense inflection to code aspectual differences, and there was a high correlation between aspect of stimulus action and report for all children under 6.
3 Discussion The data from transcripts of natural speech and from experiments show clearly that English-speaking children talk about the past relatively early, certainly by their fourth year. They are sensitive to aspectual differences. There is some evidence from naturalistic studies that in early linguistic expression of aspect children talk about perfective events with simple aspect and imperfective events with progressive aspect (see Bloom et al. (1977)). The results of the English aspect experiment suggest that this pattern becomes weaker for older children. This is hardly surprising: older children, presumably more aware of subtle differences, are freer to choose among different linguistic expressions. They are freer simply because they are better talkers. All in all, the strong decentring hypothesis is not supported by these results. It is undeniable that children’s time talk becomes more complex, notionally and syntactically, at or near the age of 6 years; but apparently they have the notion of pastness and the linguistic means for expressing it (at least in some languages) well before that age. It is difficult to know how much of the English data is relevant to interpreting the Bronckart & Sinclair experiment. That experiment presents a rather special situation (as do all experiments, of course) and can best be evaluated in conjunction with other material from French-speaking children. It does seem that the Bronckart & Sinclair results, interesting as they are, cannot be generalized to predict that young children have neither the notional not linguistic awareness of pastness. There are two rather different lines of argument that might explain the difference between the results obtained in the two experiments. One line is linguistic: the tense and aspect systems of French and English are quite different, and this may account for the children’s different use of tense and aspect. These differences, in the relevant areas, can be summarized very briefly as follows. In English, both tense and aspectual distinctions are obligatory, consistently. In the adult language, a verb must indicate tense and aspect; the possibilities for aspect are the same for all tenses. In French, on the other hand, tense is obligatory but aspect is not and most tenses are neutral aspectually. To indicate aspect, adverbials are used. The system is not entirely consistent because there is one tense, the imparfait, that is not neutral aspectually. This tense indicates imperfective aspect, in the relevant interpretation.
220
Acquisition of Tense
(I ignore the passé simple here, since it is not in the repertoire of many French speakers and is not used by young children.) To indicate imperfect aspect, then, one uses the imparfait or an adverbial (or other appropriate expression); to indicate perfect aspect, one uses adverbial or other expressions. The obligatory, systematic difference between perfective and imperfective aspect in English may mean that children learn to express aspectual notions with relative ease. French-speaking children may learn to express these notions later, because in French the system is relatively complex. Thus linguistic differences might explain the patterns of responses in the two experiments. If so, then the Bronckart and Sinclair interpretation of their findings is supported, although not their cognitive explanation. One would say that the French children used tense to express aspect because of the complexity of the French language, not because of a cognitive inability to deal with temporal notions.4 The different descriptions of the English-speaking children would be attributable to the fact that English is less complex in this area. I should like to suggest another line of explanation for the findings of Bronckart and Sinclair. This explanation preserves their focus on aspect, but not their claim that their French subjects were using tense to code aspect. I suggest that the children in the Bronckart & Sinclair experiment were describing the final stage of the actions they watched the experimenter perform. If this is correct, it explains the children’s use of tense in a natural way. Just at the end of the actions perfectives had their result and imperfectives were in progress. For instance, a turtle that performed the action of climbing stairs would be at the top of the stairs; the actual climbing would be over at the end of the action. But a turtle that was swimming at the beginning of the action would be swimming at the end. Thus at the end the climbing of the stairs would be past, while the swimming would be present. It should be recalled that there was one difference between the French and English aspect experiments. The experiment done with English-speaking children had two parts: first, children performed actions with the toys after the experimenter described an action, and then they described actions performed with the toys by the experimenter. The aspect experiment done with French-speaking children had only the second of these tasks. It may be that the children who both described and performed the actions took a more complete view of the actions they saw performed than did the children who described the actions only. The latter group may have described what they saw at the end of the experimenter’s actions with the toys. (There may have been other subtle differences between the experimental situations, and the children’s experience with narrative, that would be relevant here.) What I am suggesting is that the younger children in the French experiment took a viewpoint like that of the historical present. This should not be surprising: it is common to find the historical present in children’s early narratives (as noted, for instance, in Weinrich (1964)). If the children’s descriptions were of situations at 4 In diPaolo and Smith (1978) this analysis is suggested. I now think that it cannot be maintained. One piece of evidence against it comes from a recent dissertation, Virbel (1975). Virbel’s dissertation contains examples of the imparfait used spontaneously by a 4-year-old native speaker of French, something that the complexity hypothesis would not predict.
The Acquisition of Time Talk
221
the time the stimulus activities ended, their use of tense was consistent with adult notions of appropriate temporal ordering. This explanation brings the French results into accord with the general account of time talk given at the beginning of this paper. Antinucci and Miller (1976) discuss a similar use of tense, in connection with the data of spontaneous speech. Their data and analysis are not comparable with those of this paper, however. They worked with much younger children and considered as perfective only events with actual results. The ability to decentre may be acquired in stages. It is not implausible that at one stage of development children would talk about the past as if it were the present. This is just what happens in adult use of the historical present, in which a present air of immediacy is conveyed. Narratives in the past, on the other hand, take a past viewpoint while maintaining the relation to the present. The ability to maintain more than one time in this way is, I take it, typical of the abilities that develop with decentring.
4 Conclusion: Time Talk Reconsidered I return now to the topic of time talk. A close look at data from English and French has supported the claim that early time talk involves two times and the basic relational values of simultaneity and sequence. The orientation is fixed, I have suggested, at ST. In talking about the past according to this model, a child refers to a time preceding ST. This does not require a viewpoint other than present, but rather indicates a time that has a certain relation to the present. I shall refer to this rudimentary time talk as Stage I. At this same stage, children can signal aspectual differences, but linguistic complexity (and perhaps notional complexity as well) can interfere with the integration of aspect and temporal ordering. The syntax of Stage I is limited to verb inflections in French and English. Although relevant adverbials and modals sometimes are used, they are apparently occasional rather than consistent. An intermediate stage, II, seems to be evident in the French descriptions. This stage involves the same times and ordering relations as Stage I, but allows a focal point other than ST. There are more possibilities for temporal description than at Stage I. The wider range comes from flexible use of the times and relations rather than from additions to the system. At Stage I, children can refer to times other than the present, but always from the point of view of the present. At Stage II, children can narrate a sequence of times from a point of view other than the present. Probably this stage begins for most children by the fourth year; an example from spontaneous speech is that 4-year-olds often make rather detailed plans about things they will and/or want to do. There may be cognitive differences in the ability to talk about the past and the future that cannot be captured by an essentially semantic account such as this. The adult system has a third time that is related to the others by simultaneity or sequence. The adult system also has an orientation time that is not fixed, and temporal dependencies of more than one type. Combinations of tense and time adverbials
222
Acquisition of Tense
allow for these and other subtleties. Whether these additions to the system involve more than one intermediate stage of development is an interesting topic for future research.
Appendix: Imperfective Verbs Appearing in Transcript Summaries Adam at 4;6, 21 verbs: share, hear, peek, look, go, touch, hurt, draw; Sarah at 4;6, 22 verbs: dance, dig, see, scribble on, chew, look for, go ahead; Abe at 2;7, 57 verbs: cry, push, kiss, hold, see, help, like, rub, cough, roll, work, touch, say, draw; N.E. at 2;5, 44 verbs: kiss, need, move, want, jump, love, ?say, turn, jump; M.Z. at 2;5, 48 verbs: cry, help, play, sneeze, snow, show, use, want, carry, talk, go, see, say, hear; D.N. at 3;0, 8 verbs: cry, want, play, hear, squeeze (?); H.K. at 3;5, 56 verbs: belong, cry, look, scare, watch, pull, hear, sing, say, happen, splash, teach, stick, hurt; I.B. at 3;7, 62 verbs: splash, want, live, scare, play, look, watch, cry, try, push, shake, fight, go, see, read, feel; V.Q. at 3;8, 55 verbs: live, like, peek, pull, push, watch, help, scare, want, hold, keep, sit, bend, stand, tell, think, go; K.M. at 4;0, 31 verbs: crawl, crunch, fuss, play, push, tickle, wave, wiggle, walk, try, want, show, stand, think, see, blow, hurt, ride; G.D. at 4;0, 39 verbs: brush, play, look, pull, stay, share, show, watch, sit, read, hear, go, have, let, stick; L.R. at 4;5, 65 verbs: borrow, brush, comb, play, watch, push, wave, laugh, skate, float, pretend, move, try, share, help, scare, kiss, blow, see, fly, sit, go, run, grow, stick; F.Y. at 4;8, 34 verbs: crawl, like, love, listen, want, go, know, blow, see; H.L. at 5;0, 14 verbs: camp, help, dance, snow, talk, walk, push, blow, fly, think, dig, forget, read; A.B. at 5;0, 39 verbs: rip, push, snow, want, look, shine, go, know, stand, hide, grow, think, tell, read, see; C.P. at 5;8, 53 verbs: need, pull, rain, wiggle, want, help, look, play, rock, talk, wait, paint, hold, ride, swim, sit, fly, think, read, tell; J.W. at 5;8, 47 verbs: happen, kiss, push, pretend, survive, stay, want, try, use, crawl, flood, know, get, hear, read, see, think, go.
References Antinucci, F. and Miller, R. (1976). How children talk about what happened. J Ch Lang 3. 167–89. Bloom, L., Lifter, K. and Tanouye, E. (1977). Semantic organization of verbs in child language and the acquisition of grammatical morphemes. Unpublished paper.
The Acquisition of Time Talk
223
Bronckart, J. (1973). Aspect et temps. Rev Psychol Sci Ed 5. 147–76. Bronckart, J. and Sinclair, H. (1973). Time, tense, and aspect. Cognition 2. 107–30. Brown, R. (1957). Linguistic determinism and the part of speech. J Ab Soc Psychol 55. 1–5. Brown, R. (1973). A first language. Cambridge, Mass: Harvard University Press. Cochrane, S. (1977). Verbal aspect and the semantic classification of verbs in Serbo-Croatian. Unpublished Ph.D dissertation, University of Texas. Comrie, B. (1976). Aspect. Cambridge: C.U.P. Cromer, R. (1968). The development of temporal reference during the acquisition of language. Unpublished Ph.D. dissertation, Harvard University. de Villiers, J. and de Villiers, P. (1973). A cross-sectional study of the acquisition of grammatical morphemes in child speech. J Psycholing Res 2. 267–78. diPaolo, M. and Smith, C. S. (1978). Cognitive and linguistic factors in the acquisition of temporal and aspectual expressions. In P. French (ed.), The development of meaning. Tokyo: Bunka Hyoron. Dowty, D. (1977). Toward a semantic analysis of verb aspect and the English imperfective progressive. Linguistics & Philosophy 1. 45–77. Ferreiro, E. (1971). Les relations temporelles dans le langage de l’enfant. Geneva: Librairie Droz. Halliday, M. A. K. (1975). Learning how to mean. London: Edward Arnold. Harner, L. (1976). Children’s understanding of linguistic references to past and future. J Psycholing Res 5. 65–84. Herriot, P. (1969). The comprehension of tense by young children. Ch Dev 40. 103–10. Hockett, C. (1958). A course in modern linguistics. New York: Macmillan. Kuczaj, S. (1976). A study of the acquisition of certain verbal inflections. Unpublished dissertation, University of Minnesota. Miller, G. (1975). Pastness. Paper presented at the conference, Psychology and biology of language and thought, Cornell University. Newton, B. (1979). Scenarios, modality and verbal aspect in modern Greek. Lg 55. 139–68. Reichenbach, H. (1947). Symbolic logic. Berkeley: University of California. Sachs, J. (1977). Talking about the there and the that. PRCLD 13. 56–64. Smith, C. S. (1976). Present curiosities. Papers from the twelfth annual meeting of the Chicago Linguistics Society, 568–81. Smith, C.S. (1978). The syntax and interpretation of temporal expressions in English. Linguistics & Philosophy 2. 43–100. Stoel-Gammon, C. and Cabral, L. (1977). Learning to tell it like it is: the development of the reportative function in the child’s speech. PRCLD 13. 64–72. Vendler, Z. (1967). Linguistics in philosophy. Ithaca, New York: Cornell University Press. Verkuyl, H. (1972). On the compositional nature of the aspects. Dordrecht: Reidel. Virbel, M. (1975). Étude du détachement et de l’extraction dans le langage de l’enfant. Université de Provence. Centre d’Aix-Marseille. Weinrich, H. (1964). Tempus. Stuttgart: Wohlhammer.
The Acquisition of Tense: Bootstrapping into Syntax Carlota S. Smith
In this talk I will suggest that the linguistic category of tense plays an important role in language acquisition. There is an interesting relation between the representations underlying tense and the representations of syntax. Both require the property of scope. Scope is essential in the semantic representation of tense, and is also essential for syntactic representation. Thus, when they have acquired the category of tense children have all they need to construct full syntactic representations. Full-fledged syntactic representation involves both hierarchy and scope. I assume that children develop the basis for hierarchical representations at the first stage of active language acquisition – the lexical stage with which children begin to produce utterances.1 At the next stage, I propose, the conceptual structure of tense gives children a leg up into syntax. Tense conveys information about temporal location. In developing my argument, I will begin with the semantics of temporal location, and will show the role of scope in the concepts expressed by tense. Next, I’ll discuss how temporal location is conveyed in early child language, focusing on English. I’ll give evidence that children express temporal location well before they acquire the grammatical forms for doing so. I comment on early uses of tense, and show that the full syntax of tense mirrors its conceptual structure. My proposal is closely related to a controversial idea in language acquisition, the Stages Hypothesis. According to this hypothesis, functor categories are acquired later than lexical categories. The final section of the talk discusses some difficulties that arise in interpreting the evidence for and against the Stages Hypothesis. One type of difficulty is due to the fact that tense – like most other categories – is acquired gradually. Accounting for gradual acquisition is a problem which bedevils many discussions of syntax acquisition and especially the Stages Hypothesis.
1 At the lexical stage children are learning the basic lexical categories, e.g. nouns and verbs; and I assume that they gradually learn the hierarchical nature of these categories.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_9,
225
226
C.S. Smith
1 The Concepts of Temporal Location In a recent book called About Time, which gives an interesting overview of the temporal domain, William Friedman (1990) distinguishes three different kinds of time which are important for human beings: experiential time, logical time, and conventional time. I would suggest that linguistic time is an indispensable fourth. The temporal concepts which are expressed in language allow a certain complexity in expression and thought. These concepts include temporal orientation, and with it the possibility of shifted and embedded orientations and temporal perspective. Both are used in the linguistic expression of temporal location. Tense and time adverbials express temporal location in many languages (but not all: Chinese, Indonesian, Thai, Dyirbal, Classical Hebrew, and Malay are all languages that do not have tense).2 To understand the acquisition of a given category, we need to know its role in the target system. Let us consider briefly what tense contributes to systems of temporal location. I discuss tensed languages here, using English, but the analysis is a general one. Time, like space, requires an orientation point for mapping and locating. By relating it to an orientation point, one locates an event or state. Events or states – I use the term “situations” for both – are located before, after, or simultaneous with such a point. Temporal location thus involves the notion of orientation, and the ordering relations of sequences, simultaneity and overlap. The moment of speech, or Speech Time, is the primary orientation point in language. Situations that are located in the present are simultaneous with Speech Time; situations located in the past precede Speech Time; situations in the future follow it. Present and past tense are temporal expressions, which locate a situation by indicating the relation of the situation to Speech Time. Consider, for instance, the simple tense in (1):
(1)(a) Mary is in Panama. (b) Mary was in Panama. (1a) locates the situation talked about in the present: the present tense expresses simultaneity with Speech Time. (1b) locates the situation in the past: the past tense expresses anteriority to Speech Time.3 Similarly, expressions of futurity code posteriority to Speech Time. These are the simplest cases. English, like many other 2 Aspect, which I will not consider here, adds another important dimension to the temporal concepts which are expressed in language. In languages without tense, aspect plays a particularly important role. Aspect is the semantic domain concerning the internal temporal structure of situations. It deals with situation type, e.g. event and state; and with aspectual viewpoint, e.g. perfective, imperfective, progressive. See Smith (1991) and Bonevac and Smith (1993). 3 Tense is not always oriented to Speech Time. In many languages, one or more tenses are flexible and may be understood relative to another time. This is the distinction in linguistic between ‘absolute’ and ‘relative’ tense: the former is oriented only to a given time, usually Speech Time, the latter is flexible in orientation; see Comrie (1985) for discussion. In English, both past and present tense function in either way: they may orient to a present or future time but not to a past time (cf. Bonevac & Smith, 1993).
The Acquisition of Tense: Bootstrapping into Syntax
227
languages, also allows orientation points other than Speech Time. In (2), for instance, the time of the lower clause is temporally dependent on the main clause. (2) Mary said last Tuesday that Jane was leaving in 3 days. Here the lower clause is temporally oriented to the main clause. To see this, consider the interpretation of the adverbial in 3 days which appears in the lower clause: it is calculated from the time established in the main clause, last Tuesday. The combination of tense and adverbials allows this type of dependency. Adverbials make times more specific than do tenses alone. The two types of expressions function together in specifying temporal location (Smith, 1991; Bonevac & Smith, 1993). Temporal perspective is another essential factor in the linguistic expression of temporal location. English, like many other languages, has expressions which can only be understood with the notion of a temporal perspective, or Reference Time. For instance, consider the difference between the examples in (3). They have the same truth conditions, but differ subtly in conceptual meaning. (3)(a) John has (already) left Panama. (b) John left Panama. Both sentences present situations which occurred prior to Speech Time, but they do so from different perspectives: the perspective of (3a) is the present, while that of (3b) is the past.4 All sentences present a situation from a particular temporal perspective. The full temporal location system of a language may be quite complex. In English, for instance, it involves verbal auxiliaries as well as tense and adverbials. The future is expressed either with present tense and an adverbial, or with a modal auxiliary (Mary arrives tomorrow, Mary will arrive tomorrow). Mastery of the system does not occur all at once, due to the complexity of both conceptual and linguistic factors. Tense is apparently the first temporal expression that appears in children’s spontaneous speech.5 Tense is a grammatical category which codes temporal orientation and location. The category of tense is the functor type, with a few fixed members and meanings. Functor categories include tense, determiners, prepositions, and complementizers. Arguably, they differ from open lexical categories such as nouns, verbs, and adjectives. The syntactic distinction between lexical and functional categories plays an important role in traditional and some current syntactic theories (cf. Fukui & Speas, 1986; Speas, 1991). 4 The notion of Reference Time is due to Reichenbach (1947). It plays a role in most current work on the temporal information conveyed in sentences and discourse. 5 Many adverbials that are simple syntactically are relatively difficult conceptually, e.g. yesterday. Adverbials may also have relatively complex syntactic structures, e.g. before and after occur with NPs and with embedded sentences.
228
C.S. Smith
There is an interesting disparity between the surface and underlying structures associated with tense. Tenses are verbal affixes which appear relatively low in the surface syntactic structure of a sentence. Yet semantically, tense is high in the representation of a sentence: tense relates a situation to a time and thus has scope over the verb and its associated arguments. (The verb and its arguments, or verb constellation, linguistically indicate a given situation.) The difference between the surface and the semantic structure of tense is shown schematically in (4) for the sentence Mary fed the cat. A more complete surface structure for this sentence is given in (9) below. (4) Structures underlying Mary fed the cat.
(a) Simplified surface structure S NP
VP V
NP
Mary feed+Tns
Det
N'
the
cat
(b) Schematic semantic structure S
tense
Verb constellation
past
Mary feed the cat
I assume that syntactic structures are associated with semantic structures, for a given language, as in these representations. Semantic structures realize conceptual structure, although not all conceptual structures have a linguistic realization.
2 Acquisition of Temporal Expressions: The Early Stages 2.1 Young children understand and make references to events in the past, and perhaps the future, before the age of two years. They do so well before the category of tense appears systematically in spontaneous speech. Slightly later,
The Acquisition of Tense: Bootstrapping into Syntax
229
the linguistic category of tense itself appears – when the notion of scope is already available to the child through the semantic representations underlying temporal notions. In a very real sense, tense plays a bootstrapping role in the abstract representations which underlie conceptual and syntactic knowledge. With the acquisition of tense the child is in position to handle other abstract syntactic representations. When children first begin to talk about situations in the past and future, they do not use tense or adverbs to indicate time. Rather, they mention salient points about the situation. Talking about an event in the past, say a trip to the zoo, can be successfully managed by using nouns and verbs, especially when a child is addressing an adult who knows of the event. Children who talk in this way tend to have an MLU (Mean Length of Utterance) of more than one; I take it that one-word stage is pre-syntactic. The following is an exchange between a child at 1;8 (1 year and 8 months old) and his father (from Halliday, 1975). (5) Nigel: Try eat lid Adult: What tried to eat the lid? Nigel: Try eat lid. Goat...man say no...goat try eat lid. . .man say no. In this fragment, the child referred out of the blue to an incident that had taken place in the recent past, on a visit to the zoo. Two other nice examples are noted in Weist (1989), the first from Piaget, the second from Bowerman. (6) Jacqueline at 1;7 Hopper, hopper jump boy Jacqueline was talking about the behavior of a grasshopper, and comparing it to something that had happened two days earlier. Piaget, who was the recipient of this utterance, knew of the previous event. The next example is similarly opaque, except to a person with knowledge of this prior situation: (7) Christy at 1;6 Man shout When Bowerman got Christy up in the morning, she made this utterance. Christy was referring to the voices of men taking away the garbage, which had been audible an hour earlier. There are very few examples of talk about the future in the literature on very early child speech. One of them is given by Sachs (1977), as the first time her daughter Naomi talked about a future event. Naomi’s MLU was three, at 27 months. Riding home from the babysitter with her mother, Naomi said: (8) see Daddy
230
C.S. Smith
In contrast, Naomi talked about the past much earlier, when her MLU was just over one. Individual children differ considerably. For instance, Nelson (1984) presents quite a different picture of the development of another child, Emily. Emily talked about present and future before she talked about the past, at least in solitary crib talk. It’s worth noting that talk about the future is probably under-represented in reports of early speech, because it is often conflated with wishes and imperatives. Evidently, very young children talk about the past and future before they have the linguistic forms for doing so. This shows that they have the conceptual structures required. I would like to mention some supporting evidence from another domain about the event representations of very young children. The evidence comes from experimental work on children’s memory. In several studies, Jean Mandler and her colleagues have shown that children as young as 1;4 are able to recall causal events in the sequence in which they occurred (Bauer & Mandler, 1989; Mandler, 1990). They interpret this result as showing that the children’s representations of events included temporal sequence, following Mandler (1986). There is good reason, then, to think that children have the conceptual structures associated with temporal notions – including the crucial notion of scope – well before they begin to use the linguistic category of tense.
2.2 Tense is the first temporal expression to appear in children’s speech. The past tense usually first appears in the second half of the second year.6 There is evidence that tense appears at about this time in the spontaneous speech of children learning English, Greek, Italian, Japanese, Polish, Russian and Spanish; see the references in Slobin (1985). The acquisition of tense is gradual, occupying the time between 1;6 and 2;6. This is well after children begin to talk about situations in the past and future, as in examples above.7 Early on, children tend to use tense occasionally: it is not acquired all at once. The number and percentage of tenses increases gradually in spontaneous speech. One may not want to say that tense – or any other form – has been acquired when it first appears. One clear criterion is that a form appears systematically in the speech of a child. Roger Brown (1973) took this tack in his landmark study of language acquisition. By his rather stringent criterion – 90% appearance in obligatory contexts – some children learning English actually acquire tense a year after they began
6 Comprehension studies are difficult to do with children under the age of about 2;6. Most information on this stage comes from diaries and studies of spontaneous speech. 7 It has been claimed that early on children use tense morphemes to code aspectual distinctions (Bronckart & Sinclair, 1973; Antinucci & Miller, 1976; Bloom, Lifter & Hafitz, 1980). I do not believe that this claim is correct: in fact, there is a good deal of evidence to the contrary, cf. DiPaolo and Smith (1978), Smith (1980) and Weist (1989). I assume that tense is used to express the notions of temporal location.
The Acquisition of Tense: Bootstrapping into Syntax
231
to use it (de Villiers & de Villiers, 1985). Yet by following this criterion we beg the question of what happens in acquisition; I return to this point below. Future tense or auxiliaries tend to appear later than past and present tenses (Slobin, 1985; Weist, Wysocka, & Lyytinen, 1991). The future is often more complex syntactically than the past, e.g. in English and other Germanic languages. Moreover, the future is inextricably tied to modal notions such as need, wishes, and predictions. This may mean that the acquisition of the notion of future is underestimated by looking only at uses of clear futurity and/or future. It seems clear, however, that linguistic expressions of futurity (distinct from wishes etc) appear relatively late in the acquisition of many languages. I have claimed that children have scope in their conceptual representations at the stage when they talk about situations in the past and future, and when they begin to use tense. (I have ignored the question of whether very young children comprehend tense, because it is difficult to test.) Now let us consider an interesting implication of this claim: at the stage when children have the concept of scope, one would expect it to be used for other notions besides time. And other uses of scope would support the claims for its acquisition. There is in fact at least one other case, brought to my attention by Susan Carey: that of pretending, which emerges during the second year for most children. According to Josef Perner, pretending is one of the early notions which requires complex representations; see Perner (1991) about the development of representations. The minimum conceptual apparatus required for pretending, Perner suggests, is a way of keeping distinct the real and pretend situations – that is, representations of these situations. Thus the child’s representations must have distinguishing markers of some kind. Although the approach is different, Perner’s idea is very similar to the one developed here. In our terms, children must be able to use scopal predicates to make the needed distinction. Pretending, then, requires the notion of scope, and children begin to pretend at about the same time that they talk about times other than the present – which also requires the notion of scope. Summarizing, I have considered two early stages of the acquisition of temporal location. At the earlier stage, children have the basic concepts of temporal location but not the linguistic forms with which to express these concepts. At the next stage, they begin to use tense to convey temporal location. We have seen that, in conceptual representation, the concept expressed by tense has scope over the situation talked about; cf. (4b) above. I will now discuss the grammar of tense, and show that its syntax mirrors conceptual structure.
3 The Grammar of Tense Syntactic structure involves hierarchy and scope. One of the main expressions of hierarchy is the phrasal category. The phrasal category is an abstract, hierarchical unit that functions as a syntactic constituent. Phrasal categories are projections of their lexical heads: a noun, for instance, heads a noun phrase (NP); a verb heads a verb phrase (VP), etc. There is at least one intermediate level of structure. (8)
232
C.S. Smith
gives an example of a simple NP constructed along these lines, ignoring details; N represents the intermediate level. (8) Schematic structure of the noun phrase NP Det the
N' Adj
Noun
striped
ball
This labeled, hierarchical tree represents the structure of the NP in X-bar theory. The important point to notice is that the phrasal category – in this case, the NP – is a maximal projection of its lexical head. In addition to the intermediate N there is an optional specifier node, realized differently for different categories, in the NP structure the determiner functions as specifier.8 This scheme also holds for the category tense in the current Government-Binding theory (see Cowper, 1992). Tense heads a phrasal category, inflection phrase; like other categories it has an intermediate level. This category dominates the verb constellation of a sentence; the verb constellation, headed by the node VP, appears as a complement of the inflection phrase. (9) gives the relevant structure for the sentence Mary fed the cat; Spec is an optional specifier node of the maximal projection. Inflectional phrase is abbreviated as IP; the intermediate node is I . (9) Schematic syntactic structure associated with Mary fed the cat, focusing on the IP phrase IP (Spec)
I' Infl tense
VP NP
past Mary
V' V feed
NP the cat
8 In some current views of syntax, nounphrase-like structures like this are of the category determiner phrase, headed by the determiner (cf. Abney, 1987).
The Acquisition of Tense: Bootstrapping into Syntax
233
(9) is not a surface structure representation, of course. In surface structure the tense morpheme appears as a suffix to the verb feed, and the subject of the inner sentence Mary appears in the position indicated by Spec. I will assume that there are operations to effect these changes; they are irrelevant here. It is interesting to note that IP has a dual role in the structure of a sentence. As the projection of tense, IP represents the syntactic relation between tense and the rest of the sentence. This relation is the same as the conceptual structure: Infl has scope over the verb constellation. Yet the head of IP is the tense itself. Thus the abstract phrasal category resolves the disparity, noted above, between the syntactic and semantic structures associated with tense. The tense morpheme appears low in the syntactic tree, but the projection of this category, the IP, has scope over the entire inner sentence. The syntactic structure of tense, then, reflects its semanticconceptual structure. The IP plays more than one role in syntax. A functor category, it is the locus for the operation of several important grammatical principles, including Casemarking and the principles involving empty categories (cf. Chomsky, 1981). Moreover, members of the category have special positions in certain constructions, e.g. negation and questions in English. Negation follows the tense-carrying element (auxiliary or modal) of a sentence, as in Mary didn’t leave and John won’t win. Questions invert the tense-carrying element with the subject, which otherwise appears first, as in Did Mary leave? and Will John win? The linguistic expression of pretending and related notions is made by verbs such as pretend that and believe that. Syntactically, they have scope over an embedded sentence through complement structures. Complements are also functor categories in syntax. Generalizing, the foregoing suggests that development of the concept of scope may underlie the emergence of all the functor categories.
4 The Acquisition of Tense How is tense acquired? There are three possibilities: (a) It’s there all along – this is the Strong Continuity Hypothesis (Pinker, 1984; Crain, 1991); (b) It emerges at a stage later than the earliest stages of acquisition; (c) It is acquired gradually by a general hypothesis-testing, strongly inductive process. The argument that I have presented here, which relates the conceptual and syntactic structures of tense, is in the spirit of the second approach. I shall not discuss the others here. Syntactic structure involves both hierarchy and scope. I have shown that the syntactic structure associated with tense follows from the conceptual and semantic structures of temporal location. By assumption, the concept of hierarchy appears earlier, during the first stages of language acquisition. Thus when the structures for tense are in place, the child is in a position to acquire other functional structures of the grammar. It is in this way that the scopal structure of temporal ordering provides a way into the structures of full syntax. The account of language acquisition that I have sketched here lacks the essential element of a learning mechanism. Unfortunately, I do not have one to suggest.
234
C.S. Smith
However, a fuller understanding of what’s involved in language acquisition may enable a more informed search for the mechanisms of learning. I’d like now to relate this hypothesis about tense to general theories of acquisition of the second type mentioned above. In a number of recent proposals, syntax is acquired in at least two maturational stages. At the early stage, the child has lexical categories and structures; later on the functional categories of a language develop, with their associated structures. According to this proposal grammatical structures emerge maturationally, but grammatical principles are present from the earlier stages. Therefore, although the child’s grammar is distinct from the adult grammar, it “will never violate Universal Grammar” (Guilfoyle & Noonan, 1989; see also Lebeaux, 1988; Radford, 1990; and Grimshaw, 1992). There are several different versions of the Stages Hypothesis approach, but I cannot go into them here. The Stages Hypothesis makes clear predictions about language acquisition. First, it predicts that lexical categories are acquired before the functor categories. Second, it predicts that those grammatical elements, processes, and principles which depend on the functor categories should appear at about the same time – with or soon after the functor categories are acquired. The Stages Hypothesis can only be maintained if these predictions are correct. This is an empirical question, but not a simple one. Evidence pro and con has been adduced. Conflicting reports about the facts can presumably be resolved with more and better information. But answers to the question also depend on how the facts are interpreted. There are real difficulties of interpretation. I suggest that some of these difficulties relate to another question, namely, that of gradual acquisition. Relating the gradual appearance of form to its acquisition is an important challenge for the Stages Hypothesis. One might argue that there is an early, nonsystematic use of tense at which the child has learned individual lexical items, or positional marking, rather than the syntactic structure of tense. Given that there is often considerable delay between early use and full, systematic acquisition, this topic deserves attention. Perhaps we need to identify intermediate stages which are significant. There is now a good deal of evidence about children’s use of tense morphemes before they meet Brown’s criterion (1973). I mention some of it briefly here, to give an idea of the time ranges involved. Bloom et al. (1980) distinguish between “emergence” – the appearance of a form – and “acquisition.” They define the latter following Brown’s criterion: acquisition is 90% use in obligatorily contexts. The four children studied used tense with increasing frequency, but none of them had “acquired” it by the age of 2;6, when their MLU was slightly under three. Valian (1991), in a careful study, looked at past tenses in spontaneous speech (among other things).9 She found a stable percentage of verbs, 5–10%, had past tense
9 The main focus of Valian’s study was the appearance of full and empty subjects in the speech of learners of English and Italian; she also considered tense, modals, auxiliaries, case-marking and other topics.
The Acquisition of Tense: Bootstrapping into Syntax
235
(regular and irregular), for the children with MLUs up to 3.39. The absolute numbers are small, because the younger children produced few verbs. Similar patterns are reported in an examination of the transcripts of Adam and Eve and another child, Seth, by O’Grady, Peters and Materson (1989). The interpretation of evidence like this is unclear. It seems to me that the distinction between emergence and acquisition is a useful one, but it needs further elucidation. What can we say about the grammatical knowledge of children who use tense 5–10% of the time? What about children who use tense 50% or more of the time? If these stages differ from full acquisition, we need to say how. Perhaps we can identify an interval on the continuum of gradual emergence at which such knowledge is plausible. One avenue is to look at the use of related forms, as Bloom et al. (1980) suggest. For instance, when a child correctly uses several forms – e.g. present tense, regular and irregular past tense – we might want to say that the child has grammatical knowledge of tense. We also might consider ways of testing whether consistently comprehend these forms before they produce them. Another important area for empirical study is the acquisition of forms which are related to tense in the grammar. These include verbal to, auxiliaries and modals; the adult constructions of question and negation; case-marking; complements and other functor categories. The Stages Hypothesis predicts that these forms are acquired in the second stage of syntax, with or after tense. Consider the English auxiliary subsystem, which involves tense and auxiliaries and underlies the structure of questions and negation. It’s been known since the 60’s that there are relationships in acquisition, as in the target language, between the elements of this sub-system (Klima and Bellugi, 1966). Looking for relationships between the elements which are related in the adult language seems like a good avenue to pursue; but this approach has not produced consistent results. Here too the issues about partial acquisition arise. Still, if the child’s production demonstrated relationships that hold in the target grammar, before systematic use of a given form, we might take it as evidence of acquisition. Proponents of the Stages Hypothesis claim that the evidence supports its predictions. Their claims have been called into question, however, by data which show partial use of related forms. For instance, the children in Valian’s (1991) study used both modals and case-marking appears before MLU 2.0. This is earlier than would be expected on some interpretations of the Stages Hypothesis. Evidently, the interpretation of these data, like other data on partial use, depends on the criterion of acquisition. One of the most serious challenges now facing the Stages Hypothesis, in my opinion, is the fact that language is acquired gradually. I hope that, once we recognize the problem, we can begin to consider solutions to it. Criteria for stages which are intermediate between full acquisition and occasional use need to be developed. This talk has dealt with the acquisition of English. It is important to note that work on other languages presents a somewhat different picture, and one that is less promising for the Stages Hypothesis. For instance, Weissenborn (1990) claims that the hypothesis fails to account for the acquisition of German; Demuth (1992) argues that acquisition in Sesotho constitutes a clear counter-example. Here as in many other areas, we may expect cross-linguistic work to be invaluable.
236
C.S. Smith
The Stages Hypothesis is clearly a fruitful one, however things turn out. The hypothesis makes interesting predictions that can be tested, and suggests questions for research. In summary, I have argued that the acquisition of tense provides the child with the representational notion of scope, which is needed for full syntactic representation. The projection categories of X-bar syntax resolve the disparity between the semantic and syntactic structures of tense. The phrasal category has scope over the inner sentence, while its lexical expression appears low in the syntactic tree. Acknowledgements I would like to thank the members of the Symposium for their questions and comments.
Appendix I suggest that there are five stages of increasing complexity in the linguistic expression of temporal location. The stages are based on the analysis of the target system sketched above, which uses the concepts of sequence and orientation, and of temporal perspective. I. Speech Time only. John is here II. Speech Time Orientation & Ordering: Past & Present. a. Tense and optional adverbs We went to the zoo yesterday. We’re going to the zoo tomorrow. b. adverbs only: I see zoo yesterday. III. Re-orientation, e.g. orientation to a past or future time. We went to the zoo. Then we had ice-cream. IV. Different perspectives. Mary arrived. Mary has (already) arrived. V. Multiple orientations. John said yesterday that Mary was leaving in 3 days. These are not necessarily stages of acquisition, for several reasons. Full development will vary with the syntactic complexity of a given language. In English, for instance, the perfect construction (Mary has arrived, Mary had arrived) is quite complex syntactically and emerges relatively late. (The adverbial already, however, turns up a lot earlier). But in languages with simpler perfects, the construction is not acquired so late. Similar schemes are suggested in Smith (1980) and Weist (1989).
The Acquisition of Tense: Bootstrapping into Syntax
237
References Abney, S. (1987). Functional elements and licensing. Ph.D dissertation, MIT. Antinucci, F., & Miller, R. (1976). How children talk about what happened. Journal of Child Language, 3, 167–89. Bauer, P., & Mandler, J. (1989). One thing follows another: Effects of temporal structure on 1- to 2-years-olds’ recall of events. Developmental Psychology, 25, 197-206. Bloom, L., Lifter, K., & Hafitz, J. (1980). Semantics of verbs and development of verb inflection in child language. Language, 56, 386–412. Bonevac, D., & Smith, C. (1993). Discourse Representation Theory: An introduction and handbook. University of Texas at Austin, ms. Bronckart, J. P., & Sinclair, M. (1973). Time, tense and aspect. Cognition, 2, 197–230. Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris. Comrie, B. (1985). Tense. Cambridge: Cambridge University Press. Cowper, E. (1992). A concise introduction to syntactic theory: The government-binding approach. Chicago: The University of Chicago Press. Crain, S. (1991). Language acquisition in the absence of experience. Behavioral and Brain Sciences, 14, 597–650. deVilliers, J., & de Villiers, P. (1985). The acquisition of English. In D. Slobin (ed.), The crosslinguistic study of language acquisition. Volume I: The data (pp. 27–139). Hillsdale, NJ: Erlbaum. Dermuth, K. (1992). Phrase structure and the acquisition of functional categories. Paper presented at the Symposium on Syntactic Theory and First Language Acquisition: Cross-linguistic Perspectives. Ithaca, NY: Cornell University. DiPaolo, M., & Smith, C. S. (1978). Cognitive and linguistic factors in the acquisition of temporal and aspectual expressions. In P. French (Ed.), The development of meaning (pp. 338–351). Tokyo: Bunka Hyoron. Friedman, W. (1990). About time. Cambridge, MA: MIT Press. Fukui, N., & Speas, M. (1986). Specifiers and projection. MIT Working Papers in Linguistics, 8. Grimshaw, J. (1992). Commentary. Paper presented at the Symposium on Syntactic Theory and First Language Acquisition: Cross-linguistic perspectives. Ithaca, NY: Cornell University. Guilfoyle, E., & Noonan, M. (1989). Functional categories and language acquisition. McGill University ms. Halliday, M. A. K. (1975). Learning how to mean: Explorations in the development of language. London: Edward Arnold. Klima, E. S., & Bellugi, U. (1966). Syntactic regularities in the speech of children. In J. Lyons & R. Wales (Eds.), Psycholinguistics papers. Edinburgh: Edinburgh University Press. Jakobson, R. (1932). Zur Struktur des russischen verbums. Charisteria Guilelmo Mathesio oblata, 74–84. [Selected Writings of Roman Jakobson, vol.1 (Word and Langauge), 3–15. The Hague: Mouton, 1971.] Lebeaux, D. (1988). Language acquisition and the form of the grammar. Ph.D. Dissertation, University of Massachusetts at Amherst. Mandler, J. (1986). The development of event memory. In F. Klix & H. Hagendorf (Eds.), Human memory and cognitive capabilities: Mechanisms and performance. North-Holland: Elsevier. Mandler, J. (1990). Recall of events by preverbal children. In A. Diamond (ed.), The development and neural bases of higher cognitive functions. Annals of the New York Academy of Sciences (vol. 608). New York: New York Academy of Sciences. Nelson, K. (1984). The transition from infant to child memory. In M. Moscowitz (Ed.), Infant memory: Its relation to normal and pathological memory in humans and other animals. New York: Plenum Press. O’Grady, W., Peters, A., & Materson, D. (1989). The transition from optional to required subjects. Journal of Child Language, 16, 513–529. Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press.
238
C.S. Smith
Pinker, S. (1984). Language learnability and language development. Cambridge, MA: Harvard University Press. Radford, A. (1990). Syntactic theory and the acquisition of syntax. Oxford: Blackwell. Reinchenbach, H. (1947). Elements of symbolic logic. London: MacMillan. Sachs, J. (1977). Talking about the there and the that. Papers on research in child language development. Stanford, CA: Stanford University. Slobin, D. (Ed.) (1985). The crosslinguistic study of language acquisition. Volume I: The data. Hillsdale, NJ: Erlbaum. Smith, C. S. (1980). The acquisition of time talk. Journal of Child Language, 7, 263–278. Smith, C. S. (1991). The parameter of aspect. Dordrecht, Holland: Kluwer. Smith, C. S., & Weist, R. (1987). On the temporal contour of child language. Journal of Child Language, 14, 387–392. Speas, M. (1991). Functional heads and the mirror principle. Lingua, 84, 181–214. Valian, V. (1991). Syntactic subjects in the early speech of American and Italian children. Cognition, 40, 21–81. Weissenborn, J. (1990). Functional categoroies and verb movement: The acquisition of German syntax reconsidered. In M. Rothweiler (Ed.), Spracherwerb und Grammatik. Linguistische Untersuchungen zum Erwerb von Syntax und Morphologie. Linguistische Berichte, 190–224. Weist, R. (1986). Tense and aspect: temporal systems in child language. In P. Fletcher & M. Garman (Eds.), Language acquisition. Cambridge: Cambridge University Press. Weist, R. (1989). Time concepts in language and thought. In I. Levin & D. Zakay (Eds.), Time and human cognition. North-Holland: Elsevier. Weist, R., Wysocka, H. & Lyytinen, P. (1991). A cross-linguistic perspective on the development of temporal systems. Journal of Child Language, 18, 67–92.
Sentences in Discourse: an Analysis of a Discourse by Bertrand Russell Carlota S. Smith
In this paper I present a stylistic analysis of a short discourse by Bertrand Russell. My purpose is twofold: first, to suggest an approach to syntactically based stylistic analysis that goes beyond mere frequency counts, and, second, to draw out some linguistic ramifications of the approach. The aim of the stylistic analysis is to find the structures that are characteristic of Russell’s discourse, and to give an explanation of why these structures were chosen. This is, of course, a traditional goal of stylistic analysis. Frequently such analysis is based mainly on syntactic information. I suggest that one can arrive at interesting results by considering, in addition, certain aspects of sentences that are not strictly syntactic: the structures that occur in key locations, and their importance in the discourse. In analysing the sentences of Russell’s discourse, I shall look for the structure and location of the most important information unit, in sentences where such a unit can be identified. I shall also be interested in the structures that occur at the beginnings and ends of long sentences; these locations are, I argue, naturally prominent. Finally, a syntactic analysis of each sentence will be used. I will show that Russell presents important material in a distinct and patterned way; the force of this pattern explains, to a certain extent, the occurrence of particular constructions in the text. The surface structure of sentences plays a crucial rôle in the pattern of presentation and emphasis, as I show below. The rôle of surface structure is particularly striking when one examines a succession of sentences, that is, a discourse. At the end of this paper I turn to a linguistic discussion of subject and topic. The notion of subject is usually taken to pertain to both deep and surface structures, while topic is a surface structure notion. I shall try to clarify differences between the two, drawing on syntactic evidence as well as the analysis of this paper. In the first section, I discuss the types of information on which the stylistic analysis is based; Section 2 presents the analysis itself; Section 3 is devoted to the linguistic discussion mentioned above.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_10,
249
250
C.S. Smith
1 Analytic Units In this section I attempt to justify my approach: I discuss the notion of locations of natural prominence, important information unit in a sentence, and the type of syntactic analysis used in the study that follows.
1.1 Locations of Natural Prominence One feature of Russell’s sentences that immediately strikes the reader is that many of them are quite long. In long sentences there are certain locations, I think, that readers tend to notice more than others: I call these locations of natural prominence. The way these locations are used can be an important feature of style. I will show that Russell uses these locations consistently and to advantage. People tend to notice the beginning and end of anything that is too long for them to attend to or remember en toto. This is not a particularly new or startling observation, of course. In speeches, poems, plays, movies, books, processions, etc., the first and last moments tend to be of particular significance. One reason for this is simply that the initial and final moments are most likely to receive the attention of an audience. Apparently the human faculties of memory and attention are such that, in a work of some length, the middle tends to recede. Experimental evidence from the field of psychology can be presented in support of the comments above. When they hear a series of digits, syllables, words, a long sentence or a series of sentences, subjects tend to remember the material that begins and ends the series (Neisser, 1967: 222). In a sentence or discourse there are many ways, of course, to emphasize material, wherever it occurs. Special type faces, spacing, violations of standard punctuation, and many other things can be used to draw the reader’s attention. Note that they involve the use of some device that is special and unusual, in the context of conventions, for the written language. The initial and final positions of long sentences, however, tend to be noticed without the use of special devices. Thus, a natural way of emphasizing one part of a long sentence is to locate it at the beginning or the end of the sentence. Grammatical transformations allow one to accomplish this in various ways: the topicalization transformation in English, for instance, moves material to the beginning of a sentence; extraposition moves material to the end. Location is a feature of the surface of a sentence, but perhaps not precisely of surface structure. Location does not involve hierarchical relationships, but only successiveness.1 Presumably when one understands a sentence there is some interaction between location and other features. To appreciate the force, for instance, of a final deeply embedded relative clause, or a preposed adverbial, one must know its grammatical relation to the rest of the sentence as well as its location in the sentence.
1 For a stimulating discussion of successiveness and simultaneity in language, see Jakobson (1960).
Sentences in Discourse
251
The initial or final unit of a sentence need not be related to the surface structure in a particular way: e.g. a final verb phrase adjunction may be relatively high on the surface structure tree, while a final relative clause may be relatively low on the surface structure tree. Thus the part of a sentence that tends to receive attention is not necessarily dominant in surface (or in deep) structure. If a series of sentences comprise a discourse, the naturally prominent locations can be used to establish patterns of emphasis and expectation. For instance, if time or place adverbials consistently begin the sentences of a particular discourse, time or place will probably be important in some way in that particular discourse, and will be noticed as such by the reader. If the name of a particular person occurs frequently in initial location, emphasis will gradually accrue to that name, partly because it is placed consistently and partly because the reader will come to expect and recognize it. I shall examine the way Russell uses the naturally emphatic locations in long sentences. To find whether there is a consistent pattern of use, it is necessary to find whether the same structures occur frequently in these locations. Therefore I shall count the frequency with which different structures occur in initial and final locations in long sentences.
1.2 Important Information Unit In many sentences one part is taken to be the most important.2 The most important part may contain the answer to a question, introduce new material in a text, carry an argument one step farther, make a dramatic point of comment, etc. (Chomsky (1969) discusses the FOCUS of a sentence; the notion appears to be similar to that of important information unit, except that Chomsky deals only with short units.) What constitutes the most important unit of a sentence depends on the linguistic and extralinguistic context: it is not intrinsic to the semantic or syntactic structure. In speaking, one indicates the most important unit, or focus, of a sentence by suprasegmental cues. In the written language a variety of attention-getting devices are available, as noted above. One way to indicate importance is to exploit the naturally emphatic locations: to place the important unit at the beginning or the end of a sentence. The study of an author’s placements of the most important units in sentences can lead to interesting discoveries. For instance, Henry James tends to use the naturally emphatic locations by ignoring them; he often places important information units in the middle of long sentences, where their natural emphasis is minimal. (This has the effect of indicating something about the judgement of the narrator of the story.) A discourse may be patterned by the consistent occurrence of important units in particular locations. For Russell’s discourse, I shall identify and locate the most
2 Except various technical sentences, as of mathematics or chemistry, for instance; or on the other hand, sentences of poetry.
252
C.S. Smith
important units, in sentences where one unit can be said to be most important, and look for consistency of placement. What one takes to be a unit in a long sentence cannot be neatly defined: it may be a simple noun phrase, a relative clause, and adverbial, etc. Probably there is an upper bound on how long one unit can be, but such a bound is difficult to state.3 A sentence may be optionally organized into units of different length, depending on the intention of the speaker or writer.4 In talking, a speaker indicates with phonological cues how his utterance is organized. In writing, punctuation can often give this information, although there are many cases where it does not do so. I shall not discuss here the cues that enable one to arrive at units where there is not punctuation or other visual cue. To decide on the units of Russell’s sentences, I used his punctuation whenever possible (Russell uses punctuation frequently). In cases where the punctuation was not sufficiently informative, I take one unit to be no greater, and sometimes less, than an embedded full sentence.
1.3 Syntactic Considerations The stylistic analysis of Russell’s discourse is based on syntactic information. For each sentence a derivation was constructed, roughly according to the model of generative grammar in Aspects of the theory of syntax (Chomsky, 1965). I consulted the derivations to find structures that occurred in particular locations in sentences, and structures that were particularly frequent in the discourse as a whole. Structures were noted for the analysis by the transformations that form them. The derivations do not follow Aspects very closely: many transformations that are required to generate a sentence are not included. The omissions are due to the fact that there is a problematic relation between a linguistic grammar and the way people understand sentences. It is clear that the relationship is not of a simple one-toone nature, and that not all the transformations involved in a linguistic derivation are important in the understanding of a sentence (Fodor and Garrett, 1966; Smith, 1970; Watt, 1970). Style has to do with the deployment of language with a certain force or effect, and a psycholinguistic rather than a strictly linguistic approach might be preferable for an analysis of style. Yet little is known about how people understand sentences: there is no grammar that is psycholinguistically sophisticated. I have chosen, therefore, to use a linguistic grammar, but to limit the derivations to certain transformations. The transformations were selected according to this principle: they were included if they directly affect the surface presentation of underlying grammatical relations. Thus for instance subject-raising and pronominalization were excluded, but transformations that affect the form in which a sentence is
3 See Ross (1967) for a discussion of the problem of pinning down the characteristics of linguistic units of a certain length. 4 An interesting discussion of how such units are signalled is given in Halliday (1967a).
Sentences in Discourse
253
embedded, such as relative clause or factive nominal, were included.5 My intention was to restrict the transformations considered to those likely to be important in a linguistic grammar and also in a grammar that directly reflects the way people understand sentences. Probably these are the transformations that affect style. It is generally agreed that to understand a sentence one must be aware of the underlying grammatical relations among its parts. It seems likely, then, that at least those transformations that affect the presentation of grammatical relations play some role in the understanding of a sentence. I thus suppose the minimal awareness of linguistic derivations in understanding: knowledge of surface form, and of underlying grammatical relations.
2 Stylistic Analysis 2.1 Preview A stylistic analysis attempts to characterize a given style or discourse so that the properties that make it unique are apparent.6 The analyst looks for the constructions that are most frequent, and, therefore, characteristic (although of course it may be characteristic of a discourse that certain structures are employed rarely but tellingly). One may pose and attempt to answer the question, why were these constructions used? Some interesting interpretive work has focussed on a writer’s stylistic choices as keys to his literary work and personality (Milic, 1967; Ohmann, 1964). Here too I shall ask about the choice of particular constructions. I discuss their selection in terms of the way they function in the sentences of Russell’s essay, ‘The Elements of Ethics, Part I’ (Russell, 1967: 13–59). [The relevant excerpt is provided as Section 4 of this article.] There are no one or two constructions that contribute to the unique style of the discourse: no construction is most frequent. However, one can identify a general principle that explains the choice of constructions in many of Russell’s sentences. I state this principle now; the detailed analysis that follows explicates it more fully. With great regularity, Russell constructs sentences so that the material that is new and/or emphatic – the most important information unit – occurs in final location. The underlying relations among the units of his rather complex sentences do not follow a regular pattern. Thus, in different cases different combinations of transformations are required to locate the desired unit at the end of the sentence, and no transformation is most frequent. 5 Different types of nominals are listed separately: there are listings for factive, genitive, action, question, infinitive, and N of N nominals. Whether these nominals should be considered separately is debatable. I consider them separately here since the underlying elements are presented in different surface forms, depending on the type of nominal. However, this difference may turn out to be unimportant in understanding. 6 A discourse is unique with respect to a particular period and a particular genre, among other things; Thomas Pynchon would be more unique had he written during the eighteenth century than he is today, and Samuel Johnson would be more unique today than in his own time.
254
C.S. Smith
The function of the transformations that Russell uses is to organize his sentences so that the important material comes last. The result is a strongly yet subtly patterned discourse.
2.2 Procedure The data for this conclusion come from study of the key aspects of Russell’s sentences discussed in Section 1. For each sentence, I noted what structures occurred in the naturally prominent locations, and where the most important information unit (if any) occurred. I also noted the location of sentence connexion, if any, the type of main surface verb, and the transformations that appeared in its derivation. The next step was to look for patterns of usage among the sentences. From the material on individual sentences I compiled lists that showed the frequency with which certain structures occurred in certain locations. (I did not deal with structures that might be called characteristic, although infrequent.) Interpreting the frequency counts was the final step. In the presentation below I show that Russell’s patterns of usage are complementary, all supporting the rhetorical principle that important materials be located at the end of long sentences.
2.3 Analysis First the results of the counts are given, with discussion of each; following the analysis of four representative sentences is presented, so that the reader can have a clear idea of how the analysis was conducted; finally, some details of the syntactic analysis are mentioned. Russell’s discourse is reprinted at the end of the paper, with numbered sentences for easy reference. The counts are as follows: (1) Transformations most frequently used, (2) Structures occurring in sentence-initial location, (3) Structures occurring in sentencefinal location, (4) Location of sentence-connecting material, (5) Type of main verb in surface structure, (6) Location of most important information unit, (7) Sentence length. I have chosen a short discourse so that it can be discussed in some detail and reprinted along with the discussion. Inevitably, the numbers that result from counting frequencies are not very impressive. However, I am interested in demonstrating a general tendency or pattern only, and the small numbers are sufficient, I think, for that purpose. According to Russell’s punctuation there are 31 sentences in the discourse. Seven of these are counted as 2 sentences (they contain separate sentences, requiring separate derivations), so that the total number is 38.
Sentences in Discourse
255
(1) Transformations frequently used By simply counting the occurrence of transformations in the derivation of each sentence, and listing those that occur nine times or more, I obtain the following list (in order of frequency): SubS7 Prenom Adj. Preposing Zeroing Passive Rel. Embedding
30 22 21 19 18 17
Conjunction Factive nom. Genitive nom. Question nom. Rel. reduction
17 16 16 14 10
No particular pattern is discernible here: in fact, except for the largest category, SubS (Subordinate sentence), most of the numbers are quite close together. One can arrive at slightly more illuminating figures by grouping certain transformations together. I group below transformations that produce similar effects in surface structure. By this criterion I form three groups: transformations that embed or adjoin full sentences, with verbals; transformations that result in complex NPs without verbals; transformations that change the order of elements. The frequencies for the groups of transformations are these (certain transformations appear only in the second list because they occur less than nine times): Ts that embed or adjoin full sentences Adjunction Relative Factive Conjunction Question nom. Action nom. Comparative Apposition As-embedding8
30 17 16 17 14 8 6 4 2 121
7 For an explanation of what is covered by this category, and by others, in the lists, see the section on details of the syntactic analysis, 2.4 8 I refer here to constructions like that in the relative clause of sentence (3). Such constructions are to be distinguished from as that can be replaced by about (as in the main clause of the same sentence) and from cases with a PP with as as an obligatory verb complement, e.g. with the words speak of, conceive, regard, in Russell’s discourse.
256
C.S. Smith
Ts that form complex NPs without verbals Prenominal adj. Genitive nom. Rel. reduction N of N Conjunction of simple NPs
22 16 10 6 3 57
Ts that change order Passive Preposing Extraposition There-insertion
18 21 3 1 44
By far the most frequent type of transformation is that which embeds a sentence in more or less full form. The fact that such embeddings occur often in the essay is perhaps one formal correlate for the impression of energy and activity that Russell’s style conveys.
(2) Structures in sentence-initial position It might seem appropriate to compare the beginnings and endings of sentences, following the procedure used to make the counts above. Thus one could compare the frequency with which transformations underlie the structures that occur in initial and final position. However, such a proposal is not well conceived, for not all constructions can begin and end a sentence. For instance, relative clauses and adjective phrases cannot begin a sentence, nor can the reduced part of a conjunction; dummy subjects cannot end a sentence. There are some constructions that can occur in both initial and final position, e.g. sentence nominals, but they usually play different grammatical rôles. Moreover, the grammatical rôle of a structure is signalled differently to the reader in different positions. If a sentence adjoined with because, for instance, is preposed, then its subordinate or adverbial function is immediately evident be cause of the constant. Even if the adjunction is itself a complex sentence, the comma (in most written discourse) between the adjoined and main sentence keeps the relation between them clear. However, if a complex adjunction follows the main sentence, the relation of
Sentences in Discourse
257
e.g. a deeply embedded relative clause to the material that precedes it may be less clear to the reader: he has more to keep track of. Three kinds of information about initial and final structures may be noted (1) type of structure (relative, SubS, etc.); (2) what structure dominates it (main S, SubS, relative, etc.); (3) its internal structure (simple NP, nominal, etc.) In the case of initial structures, preposed SubS can be dominated only by the main sentence. It is sometimes ungrammatical to prepose material that is dominated by a lower sentence; if the result is not ungrammatical, the preposed material is taken to pertain to the main sentence. These examples illustrate: I asked Mary to come late because she arrived early last week. ? Last week I asked Mary to come late because she arrived early. I asked Mary to come home after John called me because he has worried. ? Because he was worried I asked Mary to come home after John called me.
The second kind of information is important only for final structures, then. The lists below are limited to information of the first kind, type of structure. Since the positions of initial and final are not comparable, I have looked at each separately. The degree of overlap is obvious on inspection. The frequency of structures in initial position are these: Preposed adverbial (no S) Preposed SubS Reordered main S Subject NP, main S Nominal subj, main S
10 8 7 7 6
The most frequent structures are preposed adverbials (in ordinary life, in the first place) and preposed SubS; these two, for syntactic and semantic reasons, should perhaps be in one category. Sentences also begin with derived subjects (the result of reorderings), nominalized sentences as subjects, or simple subjects. All but one of the reordering transformations have the effect of placing important material at the end of the sentence: three are Extrapostion and There- insertion, and three are passives in which the derived subject is essentially introductory (sentences 1, 2, and 11). Contrast these with sentence 29, where preposing directs one’s attention to the beginning of the sentence. In roughly two-thirds of the sentences, then, Russell has taken one of several options of rearrangement for a beginning. These different transformations have the same function, in the sentences under examination: to position important material elsewhere. In six of the seven sentences with simple NP subjects, the same arrangement of information occurs. A definite pattern is thus established, a pattern in which the sentence tends to begin with introductory or subordinate material, and the reader finds the most significant point in the middle or at the end.
258
C.S. Smith
(3) Structures in sentence-final position These structures occur in sentence-final position: Relative SubS Compar or As-embedding Factive, main S obj including conjoined factive Simple main S predicate Conjoined nominal other than factive Apposition
9 7 7 5 7 2 1
Of the relative clauses, four occur on NPs in the predicate of a main sentence, and the others in more embedded positions (on other relative clauses, on a comparative in a SubS, ect.). Two points are striking about this list: first, most of the structures are embedded full sentences, and second, there is great variety among them. Only with the other information assembled here do these facts have any significance for an analysis of Russell’s essay. Sheer frequency or the lack of it, in vacuo, can tell us little. But information about structure and information about position and importance together indicate the organizing principle of Russell’s essay. The varied structures that occur at the end of Russell’s sentences have different grammatical function, but the same function of presenting important material in final position. The variety keeps the pattern from being too obtrusive, and at the same time allows Russell great flexibility. (4) Location of sentence connexions I now look briefly at the explicit connexions between sentences in Russell’s essay. The connexions are made in a consistent manner, one that reinforces the pattern of information placement noted above. There are many ways in which sentence connexion can be indicated in English. One of the most common is simple anaphora, e.g., I saw a dog, The dog was spotted. Anaphora can relate nouns even if they do not immediately follow each other, as the direct object and subject of the preceding simple sentence do, thus, I saw a dog, I called the dog. These well-known facts are repeated here simply to emphasize the point that anaphora can be used in different grammatical and surface positions. (There are several recent discussions of sentence connexion: Halliday, 1967b; Hiz, 1968; Keenan, 1969.) Parallelism and repetition can also used to indicate connexions between sentences. These devices have long recognized: see Jakobson’s insight ful comments (Jakobson, 1966; Levin, 1962). They are, of course, enormously flexible. Russell uses all three of these types of sentence connexion, and they usually occur at the beginning of a sentence. Rather than linking sentences consecutively, so that a word or phrase at the end of one sentence is picked up at the beginning of the next,
Sentences in Discourse
259
Russell places the linking material at the beginning of each sentence. This pattern of connexion is another aspect of the fact that the essay is organized so as to present new and important material towards the ends of sentences. I give several examples of Russell’s sentence beginnings: in the first paragraph of the essay Russell use these devices, viz. the first four sentence beginnings: 1. 2. 3. 4.
The study of Ethics is perhaps most commonly conceived as . . . It is conceived, that is to say, as . . . Owing to this view of the province of ethics . . . This view, however . . .
The ends of the sentences do not have other explicit connectives. Consider also the beginnings of sentences 10–14: 10. 11. 12. 13. 14.
When we are told . . . We shall be told . . . If we ask why . . . If we still ask why . . . feel irritation . . . His irritation . . .
Sentence 14 contrasts with the others, since it picks up the predicate of the previous sentence. Connexions such as this are unusual in Russell’s essay. In some cases Russell uses inversion to place the explicit connectives in initial position, as in sentences, 15 and 19: 15. In the second of these feelings . . .; in the first . . . 19. But in this he is mistaken . . . Not all of Russell’s sentences have explicit linkings with each other. In 28 of the 38 sentences, the link is made from the beginning of one sentence to the beginning of the next. (5) Type of main verb in surface structure The main surface verb has little semantic import in most of the sentences of the essay. Sixteen sentences have is as the main verb, and fifteen have a verb that is almost empty semantically (such as concern). Such main verbs allow great variety in the placement of sentence embeddings. (6) Location of most important information unit The judgements as to the position of the most important unit are my own, as are the other judgements given here. With this particular category I was especially careful: several counts were made on different occasions and the results compared, in order to check the stability of the judgements. They appeared to be quite stable.
260
C.S. Smith
In categorizing the position of the main information unit I used three categories: Final, Non-Final, and ?. The third category was for sentences in which no single information unit seemed to be most important. The numbers of sentences in each category were: Final 22
Non-final 5
? 11
In 60% of the sentences, the most important unit appears in final position. Even more striking, of the sentences where I could identify one unit as most important, that unit occurred finally 86% of the time. The sentences in the third category are those in which Russell develops the philosophical foundations of his argument. One expects in close reasoning and scientific discourse to find that the steps are as important as the conclusion. (7) Sentence length Number of words up to 10 10–19 20–29 30–39 40–49 over 50
Number of sentences 4 9 13 7 5 1
Over two-thirds of the sentences have 20 words or more. Because of this the short sentences are highly foregrounded.9
2.4 Analyses of Representative Sentences In order to make clear the type of analysis on which the counts are based, I present here the details of the treatment of four sentences. After each sentence, the transformations that occur in its derivation are listed, with whatever comments seem appropriate. Similar analyses were made for each sentence of the essay.
9 In the sense of the Russian formalists, cf. Mukarovsky (1967). The approach of the formalists is introduced with notable clarity in the first chapter of Gopnik (1970).
Sentences in Discourse
261
(23) Thus in the case of ethics, we must ask why such and such actions ought to be performed, and continue our backward inquiry until we reach the kind of proposition of which proof is impossible, because it is so simple or so obvious that nothing more fundamental can be found from which to deduce it.
Transformations: N of N (Adv) Qnom Prenom Adj
Passive (Qnom) (Compar (Sub1 Sub2 )) Preposed Adv Zeroing
Genitive
SubS (mainS) (SubS) Conjunction Rel (SubS) (Compar) Compar (SubS)
In the overall count, all of these transformations are listed for sentence (23). The notations in parentheses indicate what dominates the material introduced by the transformation, e.g. the relative clauses occur on a NP in SubS and in a comparative. The structures in initial and final position were also noted: Initial structure: Preposed Adv N of N
Final structure: Compar (SubS(Rel(SubS)))
(10) When we are told that actions of certain kinds ought to be performed or avoided, as, for example, that we ought not to steal, or that we ought to speak the truth, we may always legitimately ask for a reason.
Transformations: N of N (SubS) Factive (Ap) 2 (SubS) 2
Passive (SubS) Preposed SubS
Apposition (Fac) SubS
Conjunction (Fac) (Ap) Initial structure: Preposed SubS with factive conjunct
Final structure: Object main S
This sentence is typical of the essay. The adverbial adjunction is preposed, placing the main information unit in final position (the material preposed is repetitive
262
C.S. Smith
of previous sentences). Note that the passive in the preposed sentence positions the words ought to be performed or avoided at the end of the sentence, giving them more emphasis than the specific actions of certain kinds. (3) Owing to this view of the province of ethics, it is sometimes regarded as the practical study, to which all others may be opposed as theoretical.
Transformations: Prenom Adj
Passive (main S) (Rel) Preposed SubS
Genitive (SubS) Zeroing
Relative SubS Equative as
Here a SubS is preposed and a relative clause takes prominence as the final and most important unit of the sentence. Note that the relative clause is several transformational steps away from its deep structure. The deep structure of the relative may be roughly represented in this way: NP N Δ
S
NP
Δ
Pred PP VP
NP S
may oppose
all other studies
Studies theoretical NP
to study S
Study practical
In surface structure the relative is almost parallel to the main sentence, although the first as has a different derivation than the second.
Sentences in Discourse
263
(11) We shall be told that truth-speaking generates mutual confidence, cements friendship, facilitates the dispatch of business, and hence increases the wealth of the society which practises it, and so on.
Transformations: Genitive (Fac) 2 Prenom Adj (Fac) Factive 4 Initial structure: Main S passive
Passive Zeroing 4
Relative (Fac) Conjunction 4
Final structure: Object 4th factive conjunct
In this sentence the final unit is simply a pro-form that refers back to the four other conjuncts. I suggest that here Russell exploits the fact the important units usually end his sentences. Since we expect something significant, the dismissive and so on in final position has the effect of making the catalogue that precedes it less impressive.
2.5 Details of the Syntactic Analysis I mention briefly the types of constructions that are not covered. One group consists of forms that result from housekeeping transformations, as mentioned above. Another group results from optional expansion rules, in the framework of Aspects: quantifiers, adverbs, determiners and predeterminers, tense, auxiliary. Finally, I have excluded interpolations and sentence-initial words such as but, now, hence. Some transformations were grouped together into portmanteau categories, to make the lists as perspicuous as possible. The largest portmanteau category is SubS. This category includes a number of transformations that adjoin a sentence, in full or nominalized form, to the predicate of another sentence. The adjoined sentence is usually preceded by a constant that indicates the semantics of the adjunction – whether it relates to time, causation, etc., e.g. John left to meet Mary, John left because of Mary’s singing, John left because he had a dentist appointment. These constructions are usually taken to be adverbial, in a very general sense. Another portmanteau category is that of Comparative. This includes comparatives with er . . . than, as. . .as, superlatives, and sentences embedded with so. . .that. Conjunctions include sentences conjoined with and, or, but, and the forms (not) only . . . but (also). Conjunctions of simple noun phrases are kept separate from sentence conjunction, since the former may not be transformationally derived (Lakoff and Peters, 1969; Smith, 1969). Zeroing is a fourth portmanteau category, perhaps the least satisfactory since I am interested in identifying transformations that affect presentation. This category
264
C.S. Smith
covers all cases where material has been optionally deleted, and gives equal value to the deletion of many words and the deletion of one or a few. Nominals are listed separately according to their surface form, as mentioned in footnote 5. Genitives are distinguished from nominals such as propositions about practice, the wealth of society; the latter are listed separately as N of N forms.
2.6 Summary The facts adduced here allow an explanation of Russell’s stylistic choices, and enable one to find a pattern in the very diversity of structures that he uses. The sentences tend to be long, with relatively many empty main surface verbs and many embedded sentences. The sentences tend to be explicitly connected at the beginnings, frequently the result of a transformation that reorders. The endings of Russell’s sentences vary in structure, although most of them have an embedded sentence of some kind. And finally, the most important material tends to occur at the end of the sentence. This pattern of presentation is taken to be the organizing principle which explains, to a certain extent, Russell’s choices of particular constructions. This study shows the importance of surface features of sentences, and of judgements of importance, in the analysis of a discourse. By combining answers to questions about these matters with a syntactic analysis, we have been able to arrive at an interesting account of the organization of Russell’s sentences.
3 Linguistic Comments and Ramifications I have said that the function of the transformations in Russell’s essay is to make possible the pattern of information position that characterizes the essay. In fact, this sort of explanation constitutes a general raison d’être for transformations. Transformations allow the presentation of linguistic material in a variety of positions and surface structure relations without affecting the underlying grammatical relations, that is, the way a sentence is understood. This formulation is slightly different from, but not in conflict with, some other recent explanations of transformations. It has been noted with good reason, I think, that underlying structures are frequently complex and involuted, and that transformations, by stringing out and rearranging the material, make it easier to grasp (Langendoen, 1970). This explanation is particularly attractive, of course, if one posits underlying structures that are more abstract than those of Aspects, e.g. those of Lakoff & Peters (1969) and Ross (1967). Transformations affect only surface phenomena. I have stressed in this paper that surface phenomena are important, and that their importance is more apparent in discourse than in single sentences. In discourse, as I have tried to show, significant patterns of emphasis and expectation can be established. Such patterns are due to similarities in the surfaces of successive sentences; and the flexibility of surface
Sentences in Discourse
265
structures, the possibility of consistency and variety, is due to the fact that transformations are available to the user of the language. If the function of transformations is to allow positioning of linguistic elements, the function of surface structure is twofold: to organize, and make retrievable the underlying grammatical relations; and to display emphasis, topic, and comment. It might be useful to designate these last as ‘discourse features.’ I should like here to make a few remarks about deep and surface structures, and the notion of topic and subject. These remarks will, hopefully, support the suggestion that discourse features be regarded as separate from, and simultaneous with, the underlying structure of a sentence. It seems to me that some of the confusion surrounding the notion of ‘subject’ can be dispelled if deep and surface subjects are taken to have separate functions, that coincide when the deep and surface subjects are the same. I shall argue that derived subjects do not behave as do underlying subjects with respect to several kinds of adverbial modification. Since they also differ from underlying subjects with respect to notional or basic grammatical relations, they are derived ‘subject’ in only a very weak sense. None of the attributes of underlying subjects devolve on them. In fact, it would be clarifying to reserve the term ‘subject’ for underlying structures, and to refer to surface subjects (derived or not) as ‘topics’.
3.1 Subjects and Adverbial Modification Consider first the behaviour of subjects with manner adverbials. Certain adverbials are taken to pertain to the underlying subject of a sentence (in these first examples, also the surface subject): John followed Bill reluctantly. John watched Mary delightedly. John fought Michael happily.
If the sentences are made passive, the relation of the adverb to the underlying subject is unchanged: Bill was followed by John reluctantly. Mary was watched by John delightedly. Michael was fought by John happily.
Other relevant data on transformations and underlying subjects emerge when one considers the interaction of transformations that affect or pertain to the underlying subject of a sentence. I deal here with reordering transformations that result in derived subjects, and a certain group of adjunctions. There are several transformations in English that adjoin the predicate of one sentence to another, in cases where the two sentences have identical subjects (henceforth =subj adjunctions). Examples of sentences with =subj adjunctions:
266
C.S. Smith
He won the race by tripping his opponent. He mumbled to infuriate the teacher. He opened the door with a smile.
How are the adjunctions taken when such sentences are made passive? Do they then pertain to underlying or surface subject? Consistently, the adjunction pertains to the underlying subject (with varying degrees of acceptability), as the following examples show: John teased Mary to infuriate the teacher. ? Mary was teased by John to infuriate the teacher. John chased the robber with a yell. ? The robber was chased by John with a yell. John terrified Bill by brandishing a knife. ∗ Bill was terrified by John by brandishing a knife.10
One can also ask, what happens when =subj adjunctions are made to sentences that have derived subjects? Consider first adjunctions with (in order) to: ∗ Mary
∗ Mary
was followed by John to fool her pursuers. was beaten by Bill at chess to cheer him up.
The semantics of the purported adjunctions are fairly clear; it is also clear that, in English, a different construction is required to express them. In both sentences above, a plausible adjunction involves the derived subject’s allowing something to take place, letting or arranging for something to happen. Thus, in the first sentence, Mary arranged for John to follow her; in the second, she allowed Bill to win the chess game. The fit between syntax and semantics is no better when an instrumental adjunction is involved: ∗ Mary ∗ Mary
was heard by John by forgetting to tiptoe. was caught by John by letting him find her.
In the other type of =subj adjunction presented here, the adjunction is simply taken to pertain to the underlying subject, even if semantically it is more plausible as pertaining to the derived subject, e.g. Mary was hit by John with a groan. Janet was seduced by Tim with a smile.11
Time adverbials seem to constitute a class of counter-examples to the claim that =subj adjunctions pertain only to underlying subjects. Time adverbials can occur, with deleted subject, pertaining to the surface subject of a sentence rather than the underlying subject (when the two do not coincide): 10 It
is not clear why such sentences are ungrammatical. Fodor (1970) suggests that a surface constraint blocks two by-phrases, but note that not all successions of by-phrases are ungrammatical: Mary was seen by John by the river bank. A possible explanation, offered only tentatively, is that in the offending sentence both by-phrases come from the same sentence, and that it is this that must be blocked. 11 This sentence was suggested by Senta Plotz.
Sentences in Discourse
267
Mary was scolded by John after spilling the milk. The criminal was seen by a passer-by while burying the body. John was questioned before leaving the country. While walking in the park, Mary was seen by Fred.
However, time adverbials need not pertain to the surface subject. With appropriate intonation and choice of words, they are taken as pertaining to the deep structure subject, or ambiguously to either deep or surface subject: The roses were pruned by Daniel after putting on gloves. Tony was invited by Jim before consulting the committee. Mary was seen by John while walking in the park. The thief was found by Gideon after leaving the station.
There is a good deal of variation among speakers on the acceptability of these sentences; for discussion see Fodor (1970) and Elliot et al. (1969). Time adverbials can occur, without deletion, pertaining either to the surface or underlying subject of a sentence (they can also occur pertaining to neither). In this full form they seem to be acceptable to speakers without variation: John surprised Susan when she was reading the comics. John startled Mary when he opened the door.
Thus time adverbials are different from the =subj adjunctions mentioned above: the latter cannot occurs as full sentences, and must pertain to the underlying subject. It seems that time adverbials are not a type of =subj adjunction, and do not relate in a particular way to the subject of a sentence.
3.2 Subjects and Topics Derived subjects, then, do not supersede underlying subjects. This is true with respect to transformations that pertain to the subject of a sentence, and of course to grammatical relations. In view of these facts, it is appropriate to ask, what is the function of the derived or surface subject? I suggest that one function of surface subjects is to signal what a sentence is ‘about’: in other words, surface subjects signal the ‘topic’ of a sentence. Fillmore points out (Fillmore, 1970) that this is the traditional meaning of ‘subject’; his own work (Fillmore, 1968) shows that there is no such unity to the notion of underlying subject. In many instances the underlying subject coincides with the surface subject: that is, the topic of the sentence is also its (underlying) subject. In a simplex sentence, the subject NP is taken to be the topic: Sam chased Willie is about what Sam did. In sentences with derived subjects, the derived subject is taken to be the topic: Willie was chased by Sam is about what happened to Willie. Depending on the context, it might be appropriate to begin a sentence with a certain topic, irrespective of its underlying grammatical rôle in the sentence. (By context I mean, for example, what question preceded, what is the topic of the conversation or discourse, what presuppositions are involved.) Note that topic is quite different from
268
C.S. Smith
emphatic. The passive transformation is sometimes said to emphasize the underlying object of a sentence, presumably because it is placed in initial position by the transformation. However one can easily imagine a context in which just the opposite were the case: the agent phrase might receive greater stress and attention. The initial NP functions as topic, then, in simplex sentences and passives. Consider now sentences with derived dummy subjects, and sentences with preposed material: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Bill is (the one) who(m) John watched. It was Bill that John watched. There is a unicorn in the garden. It surprised me that the bell rang. In the evening John played tennis. To please his father, John mowed the lawn. In his profession Robert was unfortunate. Chicken salad Tom likes. John Mary was beaten by. Lucille Harry reminded of a gorilla (pace Postal).
I suggest here only a general statement about distinguishing the topic NP which seems to cover the facts of sentences in isolation. In sentences 1–4, all of which have derived subjects, the first NP in the sentence is the topic; when the derived subject is also a dummy subject, the NP that followed is the topic. Indirectly the dummy subject emphasizes the topic NP, by postponing yet anticipating it. In the sentences with preposing, the main sentence subject – whether the same or different from the underlying subject – is the topic. Examples 8–10 show that preposed material is emphasized but does not become the topic. Tentatively, then, I will say that the topic of a sentence is the first NP of the main sentence, provided it has not been preposed. There is thus an important difference between reordering transformations that replace underlying subjects with derived subjects, and reorderings that do not. A derived subject is the topic of a sentence; a dummy subject focuses on the next NP as topic; an underlying subject, if not replaced, functions as the topic of a sentence. I assume then that a NP can simultaneously play two or three different types of rôles in a sentence, e.g. underlying object, topic, and perhaps emphatic. The preceding discussion has been limited to sentences in isolation. One might say in summary that the natural position for a topic is at the beginning of a sentence, just as, earlier in this paper, the beginning and ending of a sentence were said to be naturally emphatic positions. In an actual discourse or context, however, the topic of a sentence might occur in some position other than the first NP. For instance, if a story or conversation about Harry, the sentence Jane likes Harry occurs, Harry will probably be taken as the topic although it is the object NP, and at the end of the sentence. Similarly, in written discourse if a series of six sentences deal with Harry, or the notion, these NPs will probably be taken as the topic of the seventh sentence, wherever it occurs in that sentence.
Sentences in Discourse
269
But, if in discourse the position of the topic is so free, it seems that the function of derived subjects is extremely limited. However, this is not the case: transformations resulting in derived subjects allow the presentation of the units of a sentence in various positions. I hope that the foregoing analysis of Russell’s essay demonstrates the great importance of this function in discourse.
4 ‘The Elements of Ethics’, by Bertrand Russell 4.1 The Subject-Matter of Ethics (1) The study of Ethics is perhaps most commonly conceived as being concerned with the questions, ‘What sort of actions ought men to perform?’ and ‘What sort of actions ought men to avoid?’ (2) It is conceived, that is to say, as dealing with human conduct, and as deciding what is virtuous and what vicious among the kinds of conduct between which, in practice, people are called upon to choose. (3) Owing to this view of the province of ethics, it is sometimes regarded as the practical study to which all others may be opposed as theoretical; the good and the true are sometimes spoken of as independent kingdoms, the former belonging to ethics, while the latter belongs to the sciences. (4) This view, however, is doubly defective. (5) In the first place, it overlooks the fact that the object of ethics, by its own account, is to discover true propositions about virtuous and vicious conduct, and that these are just as much a part of truth as true propositions about oxygen or the multiplication table. (6) The aim is, not practice, but propositions about practices; and propositions about practice are not themselves practical, any more than propositions about gas are gaseous. (7) One might as well maintain that botany is vegetable or zoology animal. (8) Thus the study of ethics is not something outside science and coordinate with it: it is merely one among sciences. (9) In the second place, the view in question unduly limits the province of ethics. (10) When we are told that actions of certain kinds ought to be performed or avoided, as, for example, that we ought to speak the truth, or that we ought not to steal, we may always legitimately ask for a reason, and this reason will always be concerned, not only with the actions themselves, but also with the goodness or badness of the consequences likely to follow from such actions. (11) We shall be told that truth-speaking generates mutual confidence, cements friendship, facilitates the dispatch of business, and hence increases the wealth of the society which practices it, and so on. (12) If we ask why we should aim at increasing mutual confidence, or cementing friendship, we may be told that obviously these things are good, or that they lead to happiness, and happiness is good. (13) If we still ask why, the plain man will probably feel irritation, and will reply that he does not know. (14) His irritation is due to the conflict of two feelings–the one, that whatever is true must have a reason; the other, that the reason he has already given is so obvious that it is
270
C.S. Smith
merely contentious to demand a reason for the reason. (15) In the second of these feelings he may be right; in the first, he is certainly wrong. (16) In ordinary life, people only ask why when they are unconvinced. (17) If a reason is given which they do not doubt, they are satisfied. (18) Hence, when they do ask why, they usually have a logical right to expect an answer, and they come to think that a belief for which no reason can be given is an unreasonable belief. (19) But in this they are mistaken, as they would soon discover if their habit of asking why were more persistent. (20) It is the business of the philosopher to ask for reasons as long as reasons can legitimately be demanded, and to register the propositions which give the most ultimate reasons that are obtainable. (21) Since a proposition can only be proved by means of other propositions, it is obvious that not all proposition can be proved, for proofs can only begin by assuming something. (22) And since the consequences have no more certainty than their premises, the things that are proved are no more certain than the things that are accepted merely because they are obvious, and are then made the basis of our proofs. (23) Thus in the case of ethics, we must ask why such and such actions ought to be performed, and continue our backward inquiry for reasons until we reach the kind of proposition of which proof is impossible, because it is so simple or so obvious that nothing more fundamental can be found from which to deduce it. (24) Now when we ask for the reasons in favour of the actions which moralists recommend, these reasons are, usually, that the consequences of the actions are likely to be good, or, if not wholly good, at least the best possible under the circumstances. (25) Hence all questions of conduct presuppose the decision as to what things other than conduct are good and what bad. (26) What is called good conduct is conduct which is a means to other things which are good on their own account; and hence the study of what is good on its own account is necessary before we can decide on rules of conduct. (27) And the study of what is good or bad on its own account must be included in ethics, which thus ceases to be concerned only with human conduct. (28) The first step in ethics, therefore, is to be quite clear as to what we mean by good and bad. (29) Only then can we return to conduct, and ask how right conduct is related to the production of goods and the avoidance of evils. (30) In this, as in all philosophical inquiries, after a preliminary analysis of complex data we proceed again to build up complex things from their simpler constituents, starting from ideas which we understand though we cannot define them, and from premisses which we know though we cannot prove them. (31) The appearance of dogmatism in this procedure is deceptive, for the premises are such as ordinary reasoning unconsciously assumes, and there is less real dogmatism in believing them after a critical scrutiny than in employing them implicitly without examination.
Sentences in Discourse
271
References Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, Mass: MIT Press. Chomsky, N. (1969). Deep structure, surface structure, and semantic interpretation. In R. Jakobson and S. Kawamoto (eds.), Studies in General and Oriental Linguistics (pp. 154–196). TEC Co. Tokyo. Elliot, D., Legum, S. and Thompson, S. A. (1969). Syntactic variation as linguistic data. Papers from the 5th Regional Meeting of the Chicago Linguistic Society (pp. 52–59). Chicago: University of Chicago Press. Fillmore, C. J. (1968). The case for case. In E. Bach and R. Harms (eds.), Universals in linguistic theory. New York: Holt, Rinehart & Winston, 1–88. Fillmore, C. J. (1970). Subjects, speakers, and roles, Working papers in linguistics No. 4. Columbus: The Ohio State University, 31–63. Fodor, J. (1970). Three reasons for not deriving ‘kill’ from ‘cause to die’. Linguistic Inquiry 1, 429–438. Fodor, J. and Garrett, M. (1966) Some reflections on competence and performance. In J. Lyons and R. Wales (eds.), Psycholinguistics papers (pp. 135–179). Edinburgh: Edinburgh University Press. Gopnik, I. (1970). The theory of style and Richardson’s Clarissa. The Hague: Mouton. Halliday, M. A. K. (1967). Notes on transitivity and theme in English, Part 2, Journal of Linguistics 3, 199–244. Halliday, M. A. K. (1967). The linguistic study of literary texts. In S. Chatman and S. R. Levin (eds.), Essays on the language of literature (pp. 302–307). Boston: Houghton Mifflin. Hiz, H. (1968). Referentials. Transformations and discourse analysis papers, 76. Philadelphia: The University of Pennsylvania. Jakobson, R. (1960). The language of poetry. In T. R. Sebeok (ed.), Style and language. Cambridge, Mass: MIT Press. Jakobson, R. (1966). Grammatical parallelism and its Russian facet. Language 42, 399–429. Keenan, E. (1969). A theory of extended discourse. Unpublished paper, University of Pennsylvania. Lakoff, G. and Peters, S. (1969). Phrasal conjunction and symmetric predicates. In D. Reibel and S. Schane (eds.), Modern studies in English (pp. 113–142). Englewood: Prentice Hall. Langedoen, D. (1970). The accessibility of deep structures. In R. A. Jacobs and P. S. Rosenbaum (eds.), Readings in English transformational grammar (pp. 99–104). Waltham: Ginn & Co. Levin, S. (1962). Linguistic structures in poetry. The Hague: Mouton. Milic, L. (1967). A quantitative approach to the style of Jonathan Swift. The Hague: Mouton. Mukarovsky, J. (1967). Standard language and poetic language. In S. Chatman and S. R. Levin (eds.), Essays in the language of literature (pp. 241–249). Boston: Houghton Mifflin. Neisser, U. (1967). Cognitive psychology. New York: Appleton Century Crofts. Ohmann, R. (1964). Generative grammars and the concept of literary style. Word 20, 423–39. Ross, J. (1967). Constraints on variables in syntax. Unpublished MIT dissertation. Russell, B. (1967). Philosophical essays. NewYork: Simon & Schuster. Smith, C. S. (1969). Ambiguous sentences with And. In D. Reibel & S. Schane (eds.), Modern studies in English (pp. 75–79). Englewood Cliffs: Prentice Hall. Smith, C. S. (1970). An experimental approach to children’s linguistic competence. In J. R. Hayes (ed.), Cognition and the development of language (pp. 109–133). New York: John Wiley Press. Watt, W. (1970). On two hypotheses concerning psycholinguistics. In J. R. Hayes (ed.), Cognition and the development of language (pp. 137–220). New York: John Wiley Press.
The Vagueness of Sentences in Isolation Carlota S. Smith
Many fully grammatical sentences are vague, in that they give too little information for complete semantic interpretation. However, speakers appear to have no difficulty interpreting such sentences. When they appear in a context, information from neighboring sentences allows complete interpretation. When vague sentences are presented in isolation they are interpreted with great consistency, a striking fact since no completing information is available. This consistency is due, I shall argue, to a general strategy for maximizing available information rather than to the linguistic properties of the sentences in question.
1 I will begin by trying to convince you that some grammatical sentences are incomplete semantically, from the point of view of temporal specification. By “incomplete” I mean that a sentence does not give enough information to allow a temporal interpretation. The first kind of evidence for this incompleteness comes from linguistic intuition. Consider intuitions about the temporal interpretation of 1 and 2. 1 Seth planted roses1 2 Reuben is driving the truck into Howling Sentence 1 indicates, on one reading, a time prior to the moment of speech, but does not specify this time: we do not know when the event occurred. Yet we cannot assume an arbitrary time in the Past:2 rather, the existence of a particular time is indicated but the time itself is not given.3 On the habitual reading, in which 1 indicates recurrent events rather than a particular event, we do not know how often the plantings occurred, nor within what time span. On both the specific and habitual readings, then, sentence 1 does not give enough information for temporal interpretation. 1 The
examples in this paper were inspired by Stella Gibbons’ novel, Cold Comfort Farm.
2 Capital letters are used to refer to times – Past, Present, Future; lower case letters refer to past and
present tense. 3 A similar point was made by Barbara Hall Partee (1973), in a paper arguing that tenses and pronouns are similar in function.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_11,
273
274
C.S. Smith
Sentence 2 is also incomplete: we do not know whether it indicates a time in the Present or the Future; if the Future, we do not know when. There is another type of evidence for the vagueness, or incompleteness, of 1 and 2. This evidence comes from sequences of sentences. In a sequence, the interpretation of 1 and 2 depends on information in neighboring sentences. A neighboring sentence can give a specific or habitual reading to 1, and a Present or Future reading to 2. The sequences 3–6 illustrate this. 3 4 5 6
Something unusual happened on Tuesday before noon. Seth planted roses. It happened every spring. Seth planted roses. No, Reuben can’t help you now: he is driving the truck into Howling. No, Reuben can’t help you tomorrow afternoon: he is driving the truck into Howling.
The examples show that 1 and 2 can be parts of different interpretations. In the sequences, information given in one sentence is used in the interpretation of another sentence: the information is shared by the two sentences. Not all sequences allow this sharing, which occurs also between matrix and complement sentences. For discussion and rules, see my recent papers (given in the bibliography).
2 Now I would like to introduce briefly an account of temporal specification that makes precise just what is missing in 1 and 2; the account is that of Hans Reichenbach. According to Reichenbach, three notions of time are involved in temporal specification, Speech Time, Reference Time, and Event Time. Speech Time (ST) is the moment at which a sentence is uttered; Reference Time (RT) is the time indicated in a sentence; Event Time (ET) is the time at which the relevant event or state occurs. In a given sentence the three times may be simultaneous, or they may be related sequentially, with one following the other. For instance, in 7 ET and RT are the same, and precede ST: 7 Mrs. Starkadder saw something nasty in the woodshed last year In 8 all three times are different: 8 Jeb had already fallen down the well last week Sentences like 8 show the need for three times in the temporal system.4 The temporal interpretation of a sentence can be given in an abstract time representation (ATR), which indicates RT and ET and the relations between them. ATRs for 7 and 8 are presented in 9; here = indicates simultaneity, ← indicates anteriority, and → indicates posteriority. 4 George Miller argues that all three times are needed in a paper “On Pastness,” given in June 1976 at the Lenneberg Memorial Symposium, Ithaca, New York.
The Vagueness of Sentences in Isolation
9 ATR for examples 7 and 8 7: [RT←ST: last year 8: [RT←ST: last week
275
ET=RT] ET← RT]
To understand a sentence temporally one must have information about the three times and their relations, just as to understand a sentence one must know the underlying grammatical relations of its parts. Speech Time is the keystone of the system, in that Reference Time is oriented to it. When ST and RT are simultaneous, RT indicates the Present; when RT precedes ST, RT indicates Past time; when it follows ST, RT indicates Future time. Deictic time adverbials are also oriented to ST: in 8, for instance, last week indicates Past time, and is oriented to the moment of speech. An analysis of temporal specification for a given language should show how the different times and their relations are indicated in that language. In English, combinations of tense and time adverbial give the information necessary for temporal interpretation. Tense and the time adverb together indicate RT, and the preposition introducing the adverb gives the relation between ET and RT. ST is not specified on the surface; ET is actually specified, if it differs from RT, only under certain conditions which I will not discuss here. To see how some tense-adverb combinations work, consider 10–13 and their interpretations. The ATRs are given to the right of each sentence. 10 Aunt Ada Doom flew to Paris before midnight [RT←ST: midnight 11 Flora burned the sukebind yesterday [RT←ST: yesterday
ET←RT] ET=RT]
In 10, tense and the adverb midnight establish Past RT; the preposition before indicates that ET precedes RT. 11 has the same pattern, except that no preposition introduces the adverb. In such cases, ET and RT are taken to be simultaneous. 12 Rennet is working now [RT=ST: now 13 Amos is preaching before next month [RT→ST: next month
ET=RT] ET←RT]
I cannot take time here to detail or to justify my analysis, which is discussed at length in Smith (1978b). In the analysis, the temporal expressions of English are given consistent relational values. The values are those of the temporal system, that is, simultaneity, anteriority, and posteriority. Combinations of temporal expressions are related to ATRs by surface structure interpretation rules. The relational values of English temporal expressions, and the interpretations of certain combinations, are summarized at the end of this paper. There is an important difference between the syntax and semantics of temporal specification. Tense is the only temporal expression that is necessary in a sentence that is syntactically well-formed; but for a sentence to be semantically well-formed, from the temporal point of view, it must have both tense and a time adverbial.
276
C.S. Smith
We can now see why the sentences discussed at the beginning of this paper, 1 and 2, are incomplete: they lack time adverbials, and therefore do not specify RT, or the relation between ET and RT. One might raise the question of whether such sentences are really vague, rather than ambiguous. I think that they are vague, because they do not give enough information to be associated with any particular interpretation. Sentences with tense alone are vague as to ET and RT, just as sentences without intentional adverbs are vague as to intentionality: the information is simply not present. I take it as established, then, that sentences with tense alone are vague, or incomplete, as to temporal specification.
3 We can now ask, how do people interpret such sentences? They are rarely felt to be incomplete and, as noted, they are grammatically impeccable. Usually sentences like 1 and 2 occur in a context, and the context provides enough linguistic or extralinguistic information to make a full interpretation possible. But temporally vague sentences can and do occur in isolation. For instance, sentences like 1 and 2 might be presented by a linguist to informants. Since they are vague, one might expect that such sentences would be interpreted in a number of ways. But this expectation is not borne out: speakers deal with incomplete sentences in a strikingly consistent manner. I will now present and discuss interpretations that are consistently given to some vague sentences, and then consider how people arrive at them. Consider, for example, 14–16; the interpretation of each in isolation is given to the right. 14 Reuben worked in the garden 15 Amos is preaching 16 Cousin Judith fed the cat
[RT ST: – ET=RT] [RT=ST: now ET=RT] [RT ST: – ET=RT]
Sentences like 14 and 16 specify neither RT nor the relation between ET and RT; yet they are usually taken to lack one time only, so that ET must be simultaneous with RT. Sentences like 15 are usually taken as Present, although they do not give RT. Sentences like 14 and 16 may also be habitual rather than specific. As habituals they are incomplete, lacking a frequency adverbial and specification of the RT interval. Although both readings are possible for such sentences – and can be given to them in a sequence – a specific reading is usually chosen if it is plausible. There is an interesting pattern to these interpretations: at least one relation among the times ST, RT, and ET is taken to be simultaneity. For 14 and 16, ET is simultaneous with RT; for 15, RT is also taken as simultaneous with ST. These interpretations maximize the temporal information given in the sentences. If a relation other than simultaneity between times were chosen, more information would be required. To see this, consider how the information is organized in the interpretations just given. In 14 and 16, ET = RT, so that the only information missing is the specification of RT. If RT and ET were different, we would need to know at least whether ET preceded or followed RT. In 15, the interpretation is complete on a Present reading:
The Vagueness of Sentences in Isolation
277
ET and RT are simultaneous with ST, and ST is of course the moment of speech. If RT were posterior to ST, we would have no specification of RT. The relations between temporally vague sentences and the interpretations people give them can be summarized in the following principle: Associate with a sentence the temporal interpretation that requires the least additional information. This principle constitutes a very general strategy for dealing with sentences, and with other things as well.
4 A comment must be made about the occasion for these interpretations. They are the common interpretations for sentences in isolation, in the absence of particular information in the sentence that makes a certain reading plausible, or of particular experience that someone interpreting the sentence might have. In other words, I am not saying that people will always give these interpretations, but rather that they will give them in the absence of other information. This is, I think, the situation that linguists attempt to set up when they present speakers with sentences in isolation. So far we have seen that incomplete sentences are interpreted as completely as possible, by a strategy that maximizes available information. Before further discussing the strategy, I would like to consider some apparent counter-examples to the specific claims that I have just made. I have claimed that sentences without frequency adverbials are taken as specific rather than habitual, and that sentences with present tense are taken as Present rather than Future. There are a number of sentences which are not interpreted in this way. I will show that these sentences contain information that makes the expected reading implausible. They are not counter-examples because of this additional information. Consider first the choice between a specific and a habitual interpretation. There are at least three types of sentences, that lack frequency adverbials, for which the habitual interpretation is preferred. Two types are exemplified in 17 and 18: 17 Seth built houses 18 Mrs. Starkadder ate breakfast last month In 17 the indefinite object, and knowledge of the world, make a specific reading implausible: people can’t build more than one house at a time, except under special circumstances. In 18 the RT is a longish interval, a month, and this cannot plausibly be matched with a single instance of the event in question, eating breakfast; therefore a habitual reading is preferred. Both of these examples contain information which makes the specific reading implausible. The third type of sentence has present tense and non-progressive aspect, as in 19. 19 Seth sleeps in the woodshed Sentences with this combination of tense and aspect may be interpreted either as habituals, or with the dramatic, telescoped reading usually associated with scripts or
278
C.S. Smith
eyewitness reports.5 People tend to interpret sentences like 19 as habitual, probably because the dramatic interpretation is implausible without special context. However, certain sentences, such as 20 and 21, might be given a dramatic reading even in isolation. 20 John rides off into the sunset 21 The pitcher throws to third Such sentences occur frequently in familiar dramatic contexts. Perhaps this frequency can be said to constitute particular information that makes a dramatic reading plausible. I turn now to sentences with present tense and a modal. Modals, although not really part of this analysis, are included because they affect the temporal interpretation of sentences with present tense. The examples 22–25, for instance, are usually taken in isolation to have Future RT, whereas sentences without modals are usually taken to have Present RT. 22 23 24 25
Reuben will fly the airplane Flora will be leaving Rennet may be practicing Amos may be preaching
Besides the modals, variation in aspect and type of verb make a difference. 22 would have no Present specific reading, because it has non-progressive aspect, so that the specific Future reading is preferred. The main verb of 23 is instantaneous rather than durative, making a Future reading much more plausible than a Present progressive reading. 24 and 25 are equally plausible on a Present or Future reading, and can occur naturally with explicitly Present or explicitly Future adverbials. But in isolation they are more often taken as Future than as Present. I have no explanation for this within the system of temporal specification. Quite outside the system, there seems to be a difference in frequency that might be relevant. Future predictions are much more common, I think, than predictions about the Present or the Past. This seems to be true although 26 and 27 show that such predictions can also be made in English. 26 The store will have Flora’s book now 27 Jeb will have left last April If predictions about the Future are much more frequent than other predictions, people may find Future interpretations natural on the basis of frequency alone. The interpretations that tend to be given to incomplete sentences in isolation must be separated from those given to sentences with particular information or frequency. I return now to the former, which are the main topic of this paper. What unifies the different interpretations is that people tend to take sentences as complete, or nearly 5 See
my paper “Present Curiosities” in CLS 12 for discussion.
The Vagueness of Sentences in Isolation
279
complete, whenever they can. The preference for complete interpretation can also be seen in the way sentences like 28 are usually taken in isolation. 28 Flora had left Cold Comfort Farm in November This sentence has two readings: on the “already” reading, Flora left sometime before November, and November is RT; on the “remote” reading, she left in November, and RT is not given. The first reading is complete, the second reading is incomplete (for full interpretation, sentences like 28 are dependent on other sentences). Sentences like 28 are almost always given the independent reading in isolation, showing again the preference of speakers for complete interpretation. A related phenomenon is discussed by Susan F. Schmerling (1978) in a recent paper: Schmerling notes that there are two main interpretations of the verb allow, one fuller than the other, and that people tend to make the fuller interpretation when it is plausible to do so.6
5 I would like now to consider more closely the strategy and interpretations that people generally make for incomplete sentences in isolation. I have said that people tend to interpret sentences as complete, when possible, even though they contain too little information to warrant such an interpretation. The strategy they use is to maximize available information. For temporal interpretations, this means that the relation of simultaneity is exploited whenever possible. One interesting characteristic of the resulting interpretations is their simplicity. They are simple in two quite different senses: they use relatively little information, and they are relatively complete. I think it can be agreed on without difficulty that complete structures are simpler than incomplete structures (from the perceiver’s point of view, at least), because when something is complete its parts are best organized relative to each other. The inverse of this is that incomplete structures are relatively complicated because certain aspects of them are unknown. For instance, if one is interpreting a sentence and does not know the relation between ET and RT, one must keep open three possibilities. Similar strategies and interpretations have been observed in other domains besides the linguistic. In the visual domain, it is known that people tend to perceive complete contours even where there is no objective stimulus for them. For instance, consider the three dots in 29: 29
∴
One tends to perceive them as the points of a triangle, that is, the visual system spontaneously organizes them. The psychologist Gaetano Kanisza argues that incompleteness is the main factor that produces such subjective phenomena. When the elements of a visual array are complete, subjective contours do not arise. Consider for example the arrays presented in 30, which are taken from an article by Kanisza (1976). 6 Schmerling gives a slightly different interpretation, but one not incompatible with the notion of completeness developed here.
280
C.S. Smith
30
One perceives the first two figures as crosses, one with a square and the other with a circle in the middle; they could, however, be seen as several lines, each incomplete in some way. The crosses, circle, and square, are what Kanisza calls “subjective contours.” The third array could be seen as three black sectors and three angles, each incomplete in some way. But most observers report a white triangle covering three black disks and another triangle with a black border. The second interpretation is apparently favored because it involves stable, complete figures. The subjective contour phenomenon is like the interpretation of temporally incomplete sentences discussed in this paper. When presented with something that is incomplete – a sentence, a visual array – people tend to understand that thing as complete, or nearly complete, if they can. It is not clear just why people behave this way. At least one psychologist, Julian Hochberg, suggests that the behavior is based on simplicity. He argues that people tend to perceive best things that can be
The Vagueness of Sentences in Isolation
281
completely organized with the least amount of information. For instance, he notes that a closed circle can be more quickly defined than an open one for which the location and extent of the gap must be specified.7 I cite these examples from the domain of visual perception in psychology to support my claim that the strategy of maximizing information, observed with temporally incomplete sentences, is a very general one. I suspect that examples could be found in other domains as well. The point about all this for linguistics is that people’s intuitions about sentences in isolation are not entirely due to the linguistic properties of the sentences. I have shown that grammatically impeccable sentences may be incomplete semantically, and that their interpretation in isolation is quite predictable. I have argued that the interpretations are based on some very general facts about how people deal with information. People tend to interpret things as complete, or nearly complete, when they can. This means that linguists may be misled when they interpret people’s intuitions about sentences in isolation. Rather than finding out the linguistic properties of the sentences in question, they may be collecting data about strategies for interpreting incomplete sentences. There are two methodological morals, I think, that can be drawn from the foregoing discussion. First, it is necessary to have a good idea of what constitutes completeness in a particular domain in order to recognize and study incompleteness. Second, those aspects of sentences that are tied most closely to context should not be investigated, at least primarily, by means of sentences presented in isolation. To close on a positive note, it seems to me that investigations of the kind I report here are helpful in disentangling the linguistic from other factors that affect the interpretation of sentences.
Summary of Analysis of Temporal Specification of English Classification of time adverbials Explicitly Past: yesterday, – ago, last – Explicitly Present: now, this moment, today8 Explicitly Future: tomorrow, in – , next – Unanchored: on Tuesday, in June, before midnight, etc.
7 I am indebted to Donald Foss for acquainting me with the work of Hochberg and McAlister (1953). 8 The adverbial today can also be used to indicate Future, as in the sentence Jeb will be leaving today. The situations under which today and this+present tense indicate Present, and Future, need further study.
282
C.S. Smith
Relational values of temporal expressions in English Anterior past tense Past adverbs Unanchored adverbs Past prepositions Past RT
Simultaneous present tense Present adverbs Present prepositions Present RT
Posterior Future adverbs Unanchored adverbs Future prepositions Future RT
Combinations of tense and adverbial that establish RT Tense present present present past past
Adverb Present Future Unanchored Past Unanchored
RT Present Future Future Past Past
Cousin Judith is sleeping now Mrs. Nettle is coming tomorrow Mrs. Nettle is coming Tuesday Seth worked in the garden last week Seth worked in the yard on Tuesday
Combinations of tense and adverbial that do not establish RT Tense past past have past have past present
Adverb Future Unanchored Past Present Past
RT – – – – Past
Elfine was leaving in 3 days Flora had read the letter at noon Flora had read the letter last week Aunt Ada Doom was annoyed now Last night, Jeb lets the bull out9
Acknowledgments I would like to thank Andy Rogers, Robert Wall, and especially Susan F. Schmerling for their helpful comments and suggestions. Responsibility for the analysis and errors is mine. [Note from the editors: We thank Barbara Partee for identifying two minor, but substantive, errors in the original CLS publication; these have been corrected, given that Carlota Smith’s intent was absolutely clear. We have also updated the references and corrected other typographical and formatting errors.]
References Hochberg, J., & McAlister, E. (1953). A quantitative approach to figural ‘goodness.’ Journal of Experimental Psychology, 46, 361–364. Kanisza, G. (1976). Subjective contours. Scientific American, 234, 48–52. Partee, B. (1973). Some structural analogies between tenses and pronouns in English. Journal of Philosophy, 70, 601–609. 9 The
remote reading is intended here.
The Vagueness of Sentences in Isolation
283
Reichenbach, H. (1947). Elements of symbolic logic. New York: Macmillan. Schmerling, S. (1978). Synonymy judgments as syntactic evidence. In P. Cole (Ed.), Syntax and semantics, v. 9. Pragmatics. New York: Academic Press. Smith, C. S. (1976a). A theory of auxiliary have in English. Bloomington: Indiana University Linguistics Club. Smith, C. S. (1976b). Present curiosities. Proceedings of the Regional Meeting, Chicago Linguistic Society, 12, 568–581. Smith, C. S. (1976c). The temporal interpretation of complements in English. Texas Linguistic Forum, 6, 150–168. Smith, C. S. (1978a). How context contributes to the interpretation of temporal expressions. Southwest Regional Laboratory Professional Papers, 40, 1–11. Smith, C. S. (1978b). The syntax and interpretation of temporal expressions in English. Linguistics and Philosophy, 2, 43–100.
Temporal Structures in Discourse Carlota S. Smith
In conversation, narrative, and other discourse, temporal reference often extends over several sentences. I will be interested in this paper in how extended temporal reference is established and maintained. Most of the discussion focuses on English, but I develop a general account. The study of extended temporal reference is essentially that of syntax beyond the sentence. The linguistic expressions of temporal reference are syntactically based: tense, verbal auxiliaries, adverbials. Yet they participate in semantic structures that have domains beyond the sentence, as well as within sentences simple and complex. Examining temporal references within and across sentences leads to some new insights about the system of English, and about temporal systems in general. In the first section of this paper I will discuss temporal reference across sentences, and will formulate principles to account for it that are based on the temporal expressions that occur in sentences. I will then make some remarks about the temporal system of English and other languages, and will give two arguments in favor of a Reichenbach-type approach to temporal reference. Finally, I will raise some semantic questions about the principles of section 1, and claim that syntactically-based principles cannot fully account for semantic structures beyond the sentence. 1 The temporal structures that I am interested in extend over syntactically independent sentences. In such structures, temporal information from one sentence is part of the temporal interpretation of another. The sentences are dependent
Editors’ Note: We thank Barbara Partee for pointing out several typographical errors in the original publication in Rohrer (1980). They have been corrected here. We have also lightly copyedited the original publication, especially for consistency in citation and reference style. In the original, all except the last four references were omitted. To the best of our ability we have tracked down the omitted references and have updated references to then-unpublished sources. Smith was not fully consistent in her use of capitalization in the names she used for different classes of adverbials; we have not made changes to this.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_12,
285
286
C.S. Smith
semantically, much as pronominal anaphors depend on antecedents.1 I will refer to such temporally dependent combinations as extended temporal structures. Most of the sequences presented here have two or three sentences, but in principle they are of indefinite length. In this section I will present sequences of sentences with various temporal expressions and ask whether and when they can be interpreted as single temporal structures. I will then formulate principles that account for the interpretations. The account that I give is based on my analysis of temporal reference (Smith, 1978). In this analysis surface structures are related to semantic representations by interpretive rules. The semantic representations and the general approach are deeply indebted to the work of Hans Reichenbach. 1.1 Before discussing the way temporal interpretation can be extended beyond the sentence, I will outline some basic assumptions of my analysis for sentences. The assumptions are explained and justified in the paper cited above. I assume that sentences have one tense and one time adverbial each, the adverbial of unlimited complexity. I also assume that complete temporal reference in English is made by a combination of tense and a time adverbial.2 The semantic unit for temporal reference, then, has a temporal expression that is optional syntactically. Sentences without a time adverbial are vague, or incomplete, because they do not have enough information to establish temporal reference. In practice they are usually interpreted with information from context, linguistic or extra-linguistic.3 Sentences are also incomplete temporally if they have certain combinations of tense and time adverbial that fail to establish temporal reference on their own (see the table in 2 below). The temporal expressions of the language are assigned consistent relational values: simultaneity (=), posteriority (→), and anteriority (←). Their interpretations vary, depending on syntactic and semantic factors, an essential feature of a temporal system. In English, past tense is assigned its traditional value of anteriority, and present tense the value of simultaneity.4 They are interpreted as anterior to, or simultaneous with, the time of speech or another orientation point. Time adverbials are also assigned relational values, according to their normal interpretation. A representative sample of temporal order adverbials is given in 1; adverbials of frequency or duration are not included.
1 I discuss the relation between temporal and pronominal anaphora in Smith (1981). Recent studies
of pronominal anaphor across sentences include Stenning (1978) and Webber (1978). 2 Aspect will not be considered in this paper. 3 For discussion, see Smith (1977) and Partee (1973). 4 In this analysis Future is indicated by the combination of present tense and future adverbial; there is no future tense.
Temporal Structures in Discourse
287
1. Relational value of English time ordering adverbials (not an exhaustive list) Anterior (←) Past
Simultaneous (=) Present
Posterior (→) Future
a.
last week 3 days ago yesterday
now this minute
next week in 3 days tomorrow
b.
previously earlier
at the same time
afterwards later
c.
on Tuesday5
on Tuesday
d.
before John left6
before John leaves
There are six possible combinations of tense and time adverbial, according to relational values. All six are grammatical in English. Four combinations establish a time independently, and two do not; sentences with these two need further information to be interpretable temporally. The combinations are given in 2, with examples of each.7 2. Combinations of tense and time adverbials Tense
Adverb
Time
a. b. c.
present present present
present future past
Present8 Future Past
d. e. f.
past past past
present future past
– – Past
a. b. c.
Jane is swimming now Jane is swimming tomorrow Last week, Jane is swimming all alone when. . .
d. e. f.
Molly understood the situation now Molly was arriving in 5 days Molly played soccer yesterday
(dependent) (dependent)
5 This example indicates a class of adverbials in which prepositions (at, on, before, etc.) introduce NPs that refer to clock or calendar times (hours, minutes, months, etc.) or to regularly scheduled events that can be taken as indicating a time (breakfast, school, etc.). These adverbials have the relational value of posteriority or anteriority (not both at once.) 6 This example indicates a class of adverbials in which prepositions introduce sentences. The sentences may be reduced (John left before Mary) if they contain material repeated in the main sentence. 7 I ignore here the conditional and irrealis interpretations of past tense.
288
C.S. Smith
These combinations are crucial for establishing temporal reference, within and across sentences. 1.2 I can now introduce extended temporal structures. The examples of 3–5 are sequences in which the (b) sentences do not establish temporal reference on their own, but are fully interpretable temporally with information from the (a) sentences; they are extended temporal structures. 3. (a) I talked to Mary last night. 4. (a) I talked to Mary on Friday. 5. (a) John arrived at noon.
(b) She was happy. (b) She was leaving in 3 days. (b) Mary came later.
3b has no time adverbial, and cannot be fully interpreted temporally. The other (b) sentences have time adverbials but cannot be fully interpreted either. The times established in the (a) sentences allow full interpretation of the (b) sentences. I am assuming a sort of Gricean co-operative approach in these interpretations, according to which one looks for a related interpretation of sentences presented together.9 I will ask whether sentences can function together as a temporal unit, and will not consider other interpretations that the sentences might have. The relationship between 3 and 5, considering temporal interpretation only, can be described with a military metaphor. The (a) sentences can be said to have “captured” the (b) sentences, since the latter are dependent on the former for interpretation. Continuing the military metaphor, I will focus in this section on sentences that may become captives, that is, on dependent structures. There are three possibilities: a sentence may demand capture, it may be available for capture, it may be protected from capture. Consider first sentences that demand capture. Such sentences do not establish temporal reference on their own; one class that demand capture is that of sentences with no time adverbials, as shown in 6–8. The (b) sentences are captured by the (a) sentences, the latter having time adverbials. 6. (a) Scott isn’t at the class picnic. (b) He is working. 7. (a) Scott can’t come to the picnic tomorrow. (b) He is working. 8. (a) John had a good time last night. (b) He went to the movies. 6 and 7 show that the present tense indicates Present or Future time, depending on the adverbial that completes the interpretation. Other sentences that demand capture have adverbials but do not establish temporal reference. Certain adverbials such as previously and afterwards never establish a time; they are intrinsically dependent on times given elsewhere. 9b exemplifies this
8 I capitalize temporal interpretation – Past, Future – and use lower-case letters for temporal expressions – e.g. past tense. 9 See Hobbs (1976) and Stenning (1978) for discussion.
Temporal Structures in Discourse
289
type. There are also combinations of tense and adverbial that do not establish a time, as in 2d and 2e above, and 10b below. 9. (a) Isobel went riding at noon. 10. (a) I talked to Michael at noon.
(b) Sylvia rode later. (b) He was going in an hour.
In these sequences the (a) sentences are the captors of the (b) sentences, giving the information that allows full interpretation. Thus 9b tells us that Sylvia rode later than some time, but doesn’t fix this time; 10b tells us only that Michael planned to leave after some time. Thus the (a) sentences, giving times, provide what I will call the “anchors” for the temporal expressions of the (b) sentences. In sum, sentences that demand capture do not establish temporal reference on their own; they depend on information given elsewhere. (Such information may be part of extra-linguistic context as well as in a neighboring sentence.) In contrast, sentences that are available for capture do establish time reference in certain contexts, and in isolation. They are interpretable as anchored to the time of speech, ST. But they can also be captured to form part of an extended temporal structure. As captives, they anchor to a time established in another sentence. 11 and 12 exemplify this situation; the (b) sentences are available for capture, and are captured by the (a) sentences. 11. (a) (b) 12. (a) (b)
Every Saturday afternoon Paul took the bus to town. He (Paul) went to the movies at 5 pm. Next Sunday we’re flying to Amsterdam. We visit friends on Monday.
11b alone indicates that Paul did something at a certain time – in the simplest interpretation, at 5 o’clock prior to the time of speech (ST). This is the simplest interpretation because it requires no information other than that given in the sentence.10 In sequence with 11a, however, 11b has quite a different interpretation: it specifies an habitual activity of Paul’s on Sunday afternoons. The frequency adverbial is captor, on the extended temporal interpretation of 11. Similarly, the interpretation of 12b is different in sequence with 12a from its interpretation alone. On the simplest interpretation, 12b alone anchors to ST, indicating the Monday following ST. In sequence 12 it indicates the Monday after next Sunday: the time of 12a acts as anchor. The sentences that are available for capture have time adverbials that may be taken as anchored to ST; they can, however, anchor to a time established in another sentence. To interpret a time adverbial it is essential to know how it anchors, and this knowledge will also allow prediction of the capturing possibilities of sentences in which the adverbial appears.
10 See
Smith (1977) for discussion of the notion of simplicity in interpretation.
290
C.S. Smith
There are some sentences that cannot be captured: they are, in the terminology suggested above, protected from capture. Consider for instance sequences 13 and 14. 13. (a) Robert talked to Bonnie at noon. (b) He (Robert) arrived 3 hours ago. 14. (a) Every summer Robert travels. (b) He (Robert) is leaving for Greece tomorrow. Although these sequences are not incoherent, they cannot be taken as extended temporal structures. In isolation 13b indicates a time 3 hours before ST, and it has the same interpretation in sequence with 13a. The situation is the same with sequence 14, in that the interpretation of 14b is unchanged by the preceding sentence. Both sentences of 13 and 14 are anchored to ST. The same type of adverbial appears in the (b) sentences: tomorrow and 3 hours ago are anchored to the moment of speech, making deictic reference. Deictics protect a sentence from capture because they anchor to ST and not to times established in other sentences (actually the situation is a little more complicated than this, as I will show directly). Temporal adverbials can be classified according to how they anchor, and whether they can re-anchor. These are essential factors for interpretation. There are three main classes: Deictics, that anchor to ST only; Dependents, that anchor to a time given explicitly (the time much be established in another sentence, since sentences have only one time adverbial each); Flexible Anchoring adverbials, that anchor either to ST or to a time given explicitly. The chart in 15 shows this classification of temporal adverbials. Habitual adverbials have quite different properties, and are not included in the analysis given here; for some discussion see Smith (1978). 15. Types of adverbials (not an exhaustive list) Deictics ← last week, a week ago, yesterday = now, this moment → in three days, tomorrow Dependents ← previously, earlier = the same time → afterwards, later Flexible Anchoring ← → on Tuesday, before John left Adverbials with all relational values occur as Deictics and Dependents; Flexible Anchoring adverbials have the values of anteriority and posteriority.
Temporal Structures in Discourse
291
These categories serve to organize the facts about participation in extended temporal structures. Sentences with no adverbials, or Dependents, demand capture; sentences with Flexible Anchoring adverbials are available for capture; sentences with Deictic adverbials are protected from capture. Only adjacent sentences have been considered, so far, as possible captives. Sentences can also be captured from afar, although deciding what sentences go together may pose difficult problems. (See Section 3 for discussion, especially sentences 40– 42.) Consider for example sequence 16. 16. (a) We are deeply concerned with last Sunday’s robbery of an important public figure. [. . .intervening sentences. . .] (b) Last night I talked to a reporter who had new information. (c) The maid left before noon. I am interested in the interpretation of (c). In sequence with (b), it may be interpreted as anchored to the time established in (b). If (a) is in the sequence, there is also a natural interpretation of (c) as anchored to (a); the shift of attention from reporter to robbery is made without an explicit transition, elliptically. If the robbery is in the forefront of consciousness no explicit reference to (a) is necessary. There is also a vague interpretation of (c), in sequence with (b) only. On this interpretation (c) is taken to be anchored to a time not given explicitly, but anterior to that time – which may be known extralinguistically, or may be not known precisely. Vagueness like this is not at all unusual, especially in conversation. I shall return to this vague reading in Section 2. 1.3 Captors: Sentences that act as captors in extended temporal structures must establish a time, either independently or through sharing with another sentence. Deictic, Flexible Anchoring, and Dependent adverbials can all appear in a capturing sentence. Sentences with Deictics can capture other sentences, with the general limitations established above. Sentences with Flexible Anchoring adverbials can capture Dependents and sentences without adverbials. If two Flexible adverbials appear in temporally related sentences, they always have an orderly interpretation11 because of the fact that clock and calendar time use recurring sequences. Strictly speaking, such an orderly relation is not capturing, since neither of the adverbials is anchored to the other. Sentences with Dependent adverbials can also capture other sentences. When this happens the time of the captor sentence is part of the interpretation of the captive, even though the captor does not establish a time independently. As the examples show, there are lexical items that indicate chains of dependency with Dependent adverbials. The sequences of 17, 18, and 19 show capturing by sentences with Deictic, Flexible Anchoring, and Dependent adverbials respectively. Habituals, which can capture other sentences, are not included here. Deictics do not appear in captive sentences, since they protect a sentence from capture.
11 Assuming
that they refer to identical or compatible units – see Section 3 for discussion.
292
C.S. Smith
17. (a) John arrived last Tuesday.
(b) Mary was away in the country. (c) Mary came on Sunday. (d) Mary was leaving in 3 days.
18. (a) John arrived on Sunday.
(b) Mary was away in the country. (c) Mary came on Tuesday. (d) Mary was leaving in 3 days.
19. (a) . . .Sunday. Mary was leaving in 3 days. John was leaving later. (b) . . .noon. John called afterwards. He said he was coming in 5 minutes. (c) . . .10 o’clock. Oona bought her ticket earlier. John got his even earlier.
1.4 General Principles: I can now state a generalization about extended temporal structures: A sentence can be captured to form part of such a structure, so long as it does not have a Deictic adverbial. A principle of this type might be used in interpreting texts; extended temporal structures would be sought according to the temporal expressions that occurred in sentences of the text. A second principle is also necessary. This principle applies to tense, the other element needed to establish temporal reference. Principle 2: A sentence can be captured only if it has the same tense as its captor. Sequence 20 shows the need for this second principle. The two sentences of 20 have different tenses, and cannot be interpreted as a temporal unit. 20. (a) Last summer Isobel went to Greece. (b) She will visit friends. The principle involving Deictic adverbials has a certain logic, and one would expect it to hold generally among languages. If a language has adverbials that anchor only to the time of speech, those adverbials would not be available for participation in extended temporal structures. The second principle, that participants in such structures have the same tense, depends on the way tenses function in a particular language. In fact both principles must be modified, as I now show. 1.4.1 I consider now some examples which appear to cause difficulty for the principles given above. There are sentences that call into question the principle barring deictic adverbials from participation in extended temporal structures. Consider first sentences whose main verbs involve communication or consciousness, such as say, announce, realize. In sentences with verbs such as these, a deictic adverbial can have an anchor other than ST; in sequence with such sentences, a deictic can form an extended temporal structure. The examples illustrate. 21. (a) The office announced the news last night. (b) The ambassador will resign in a month. 22. (a) John said something surprising at noon. (b) Mary had resigned an hour ago.
Temporal Structures in Discourse
293
Sequence 21 exemplifies direct quotation, and 22 indirect. In both, the adverbials of the (b) sentences are taken as anchored to a time established in the (a) sentences. It should be noted that deictic adverbials vary as to whether they allow this type of re-anchoring. Some deictics are flexible, others are rigidly anchored to ST. The deictics in the examples (in _ and _ ago) are flexible for most speakers; tomorrow and yesterday, in contrast, are usually taken as anchored to ST and rarely appear in sentences like 21b or 22b. Verbs of communication and consciousness (CC verbs) allow the re-anchoring of deictics. One might attempt to save Principle 1 by treating these verbs as expressions governing the exceptions. Sentences with re-anchored deictics could be identified as exceptional because of their relationship to CC verbs. However, this approach is inadequate. It cannot deal with ellipsis, as sequences 23 and 24 show. 23. (a) (b) 24. (a) (b)
We invited Tom to join us but he couldn’t come. He had eaten dinner an hour ago. Della left the party early. She was flying to Istanbul in a few hours.
In both sequences, the (b) sentences can be taken as reported speech, or narrated monologue.12 23b presents the reason Tom gave for not coming to dinner; 24, on this interpretation, gives Della’s explanation for her early departure. Although the (b) sentences present material that is communicated or in consciousness, they are not governed by CC verbs or explicitly related to such verbs. The mode of reported speech exemplified in the (b) sentences above has a characteristic pattern of deixis which sets it off from direct and indirect speech. Reported speech presents material unrefracted through the consciousness of a speaker or narrator. It has neither the directness of quotation nor the ambiguity of indirect speech.13 Characteristically, in reported speech pronouns are 3rd person, place and time adverbials are deictic (unshifted), and tense is past. The sentences of 25 illustrate. 25.
Direct: Indirect: Reported:
I am ready now. She is ready now. She was ready now.
I will leave next week. She will leave next week. She would leave next week.
The mode of reported speech has been discussed mainly in connection with literary texts.14 However, reported speech also appears in other types of texts, 12 This
mode is also known as “erlebte Rede”, and the “style indirecte libre”. essential ambiguity of indirect speech is emphasized by Lips (1926). 14 For instance by Banfield (1973), Cohn (1947), Friedman (1955), Hamburger (1957), Weinrich (1964). [Editors’ note: The reference for Cohn (1947) was omitted from the original publication of this article; we have been unable to identify it.] 13 The
294
C.S. Smith
especially historical narratives. I cite two examples from texts that are primarily historical rather than fictional. 26. An education at Oxford appealed to a new class of rich and well-to-do men who wished to use it to improve the prospects of their sons. The Colleges were now therefore able to charge fees proportionate to the social advantages likely to accrue. (Darlington, Encounter, 1967) 27. Having put his religious house in order, Mohammed now began to enjoy his power as the undisputed ruler of a large number of tribes. (van Loon, The Story of Mankind, 1921) From the point of view of this paper, reported speech must be distinguished not only because of its pattern of deixis, but also because its temporal forms cannot be fully interpreted alone. The sentences with now in the examples above, for instance, do not establish temporal reference. They lack a reference point that would give value to now. In the terms of this analysis, the sentences of reported speech are incomplete temporally and demand capture; to be fully interpreted, they require additional information. Sentences with re-anchored deictics are in the mode of reported speech, and are dependent temporally as noted above. Their temporal dependency can be provided for in a revision of Principle 1, that accounts for the counter-examples and preserves the main point of the original principle. The revised principle: A sentence can be part of an extended temporal structure so long as it does not have an independent Deictic adverbial. The phenomenon of reported speech shows that CC verbs need not appear in a text for consciousness to be conveyed. Reported speech involves ellipsis, which is common in discourse of all kinds. Therefore an approach that depends on governing CC verbs is doomed to failure. Yet the connection between CC verbs and reported speech is undeniable. One interesting piece of evidence comes from the interpretation of complex sentences that have re-anchored deictics. Compare, for instance, the following sentences. 28. On Sunday Mary told me she had lost the watch a week ago. 29. On Sunday Mary lost the watch she had bought a week ago. The interpretation of 28 is straightforward, with the deictic anchored to Mary’s telling of Sunday. That of 29 is less so: the deictic must be anchored to Sunday because of the anterior have,15 but there is no explicit consciousness to which the deictic can be attributed. I suggest that we understand sentences like 29 in a conventionalized way, so that even without an explicit shift the narrative is taken to have assumed (if only momentarily) the point of view of Mary. Since language is often elliptical, semantic 15 Auxiliary
have indicates anteriority; see Smith (1976).
Temporal Structures in Discourse
295
dependencies cannot always be associated with explicit governing forms.16 One can explain certain consistent interpretations by recognizing the possibility of ellipsis. Thus I shall say that re-anchored deictics signal a consciousness at the time of reanchoring, whether or not such a time is associated on the surface with a CC verb. Benveniste, in his important paper “Les relations de temps dans le verbe français”, proposes a fundamental distinction between texts, that of histoire and discours. Histoire is objective (science, history and the like) and discours subjective. In histoire no voice, no consciousness appears: “. . .A vrai dire, il n’y a même plus alors de narrateur. Les événements sont posés comme ils se sont produits à mesure qu’ils apparaissent a l’horizon de l’histoire. Personne ne parle ici: les événements semblent se raconter eux-mêmes.” (Benveniste, 1966, p. 241) Benveniste bases his distinction on the different tenses that appear in histoire and discours (he recognizes that the two often are mixed, as in a history with commentary). He notes that pronouns do not appear in histoire, but makes no other reference to deictics and none to temporal adverbials. Yet it seems evident that deictics of most if not all kinds must not occur in histoire, since they are evidence of consciousness. In an analysis where both tense and adverbial combine to make temporal references Benveniste’s main insight about types of discourse can be easily stated. The essence of the distinction is that temporal reference involving subjectivity or consciousness appears in discours but not in histoire. I suggest that the distinction be recast in terms of deictic temporal expressions rather than with tense: deictics would occur only in discours. Languages differ in tense systems; the double system of French that Benveniste cites appears quite rarely (this is true whether or not one agrees with his particular claims about French). Therefore in generalizing these notions about discourse to other languages it would be necessary to refer to other indications of subjectivity and consciousness besides tense. The suggested recasting faces no such problems, since Deictics appear quite generally in language. 1.4.2 The second principle for extended temporal structures also requires modification. According to this principle, sentences that form extended structures must have the same tense. But there are sequences that violate this principle, sequences with the Historical Present. Historical Present is the traditional name for the combination of tense and adverb exemplified in 2c above, present tense and past adverb, indicating Past time. The combination occurs frequently in informal discourse, such as joke or story-telling; it can also be found in more formal and sustained narratives. Sentences with the Historical Present often occur in sequence with past tense sentences.17 30 and 31 illustrate. 30. Something funny happened last night. I’m sitting in my rocking chair as usual, when all of a sudden the door opens, apparently without human agency.
16 Smith
and Whitaker (1979) investigate some different types of ellipsis. (1979) discusses the narrative function of switching from past to present tense in accounts of past events.
17 Wolfson
296
C.S. Smith
31. . . .this American couple the other day. They walk in, total strangers, ask what kind of local ale is good, I put in my bit of advice, and straightaway, we’re carrying on talking and laughing like we’d all been chums for years. Then I taught him how to play darts. He won. (An advertisement in The New Yorker, 1978) The sentences of these examples have different tenses, yet they all indicate Past time. Thus the temporal interpretation is consistent although the tense forms are not. These facts suggest a modification of Principle 2 that is more general than its first formulation above. The modified form is semantic rather than syntactic, referring to interpretation rather than particular tense forms. Revised Principle 2: The temporal forms in sentences of an extended temporal structure must establish the same time. The formulation allows for variation among languages, an important advantage over the earlier one. It is a peculiarity of English (although not exclusive to English) that present tense can indicate times other than Present. This peculiarity means that extended temporal structures indicating Past may have past or present tense. In languages where tenses function differently, the application of Principle 2 would not allow the same sequences. I have formulated two principles that account for important aspects of extended temporal structures, in English and other languages (but see Section 3 for some caveats). These principles allow the recognition or formation of temporal structures, according to the forms that occur in individual sentences. The classification of adverbials according to their anchoring properties proved to be essential for understanding of temporal structures within and across sentences. 2 I turn now to some different questions about temporal reference. The discussion above has prepared the ground, I hope, for a consideration of the general properties of systems that make temporal reference. 2.1 The distinction between temporal expressions that anchor freely and those that anchor only to Speech Time (ST) has been emphasized above. This distinction is traditionally made, in slightly different form, for tenses. Tense is said to be “absolute” or “relative”, depending on whether it anchors to ST or to other temporal expressions as well. The fact that the distinction applies to both tenses and adverbials is expected on the analysis given here, where temporal reference is based on both tense and adverbial. In languages such as Japanese and Hindi tense is relative. This means that tense is interpreted either as anchored to ST or to the time established by another sentence: in embedded contexts, tense anchors to the matrix sentence and not to ST. For Japanese and Hindi the relational value of present tense is simultaneity, that of past tense is anteriority. These values are consistent although the anchors for the tenses vary. The way that tense in embedded sentences is interpreted, for these languages, is indicated in 32 and 33.18 18 I
would like to thank Hermann can Olphen for information about Hindi. The Japanese information is given in Ota (1972).
Temporal Structures in Discourse
297
32. John [say+past] that Mary [is+present] sick Interpretation: S1 :Past S2 : simultaneous with S1 33. John [say+past] that Mary [is+past]sick Interpretation: S1 :Past S2 : anterior to S1 It is interesting to compare the interpretations with those of the corresponding embedded sentences in English. 34. John said that Mary is sick. Interpretation: S1 : Past S2 : simultaneous with ST 35. John said that Mary was sick. Interpretation: S1 : Past S2 : simultaneous with S1 36. John said that Mary had been sick. Interpretation: S1 :Past S2 : anterior to S1 35 corresponds to 32 in Japanese or Hindi. The examples show that past tense is absolute in English: it is taken as anterior to ST in both embedded and nonembedded contexts. To indicate anteriority the anterior form have is used; thus 36 corresponds to 33 in Japanese or Hindi. There are dialects of English that do not use have as in 36; for such dialects, past tense is relative in embedded contexts. No counterpart to 34 exists in Japanese or Hindi. In 34, the embedded sentence is taken as simultaneous to ST rather than to the time of the matrix sentence; 34 would be appropriate only if Mary were sick at the moment of speech, with the same sickness that afflicted her when John spoke of it. This interpretation of present tense as limited to speaker responsibility occurs only when it is embedded under a Past time. It is an absolute interpretation of present tense, with the relational value of simultaneity and the constant anchor ST. This exceptional fact corresponds to another: the Historical Present cannot be embedded. The speaker responsibility interpretation takes precedence, within sentences, in the system of English. Hindi does not have a use of present tense like the Historical Present, perhaps because tense is relative in that language to ST or to other sentences, rather than to adverbials. The notion of relative and absolute anchors is illuminating for understanding the English tense system. The English system is well-known for its asymmetry and irregularity, but just where the asymmetry lies is less well-known. Most analyses content themselves with a simple list of interpretations of the tenses, as in 2 above. The facts of interpretation are these: past tense indicates a time anterior to ST except when embedded as a complement under a sentence establishing Future time. In this exceptional case, it indicates anteriority to the matrix sentence.19 Present tense
19 An example of this use of past tense: “The minister will announce at midnight that he burned the
documents an hour ago.” On one reading the documents were burned an hour before ST, but on another they will be burned an hour before midnight, and the past tense indicates a time anterior to the RT of the main sentence. Discussion of the limited syntactic contexts for this interpretation is presented in Smith (1981).
298
C.S. Smith
indicates various times – see 2 above – except when embedded under a sentence establishing Past time. In this exceptional case, it indicates a time simultaneous to ST. Embedded contexts are exceptional for both tenses. It is clear that past and present tense have different anchoring properties in English, exceptions aside, and therein lies its irregularity. Past tense anchors consistently to ST: it is an absolute tense.20 Present tense anchors to times established by adverbials or times in other sentences: it is a relative tense. The system has both relative and absolute tense. This is true for both tense and adverbials, of course. Deictic adverbials are absolute, Dependent adverbials are relative, and Flexible Anchoring adverbials may be either absolute or relative. 2.2 Temporal reference in language involves two kinds of anchoring, as we have seen. Absolute tense and adverbials are anchored to the time of speech; relative tense and adverbials are anchored to a time given in text or context. These facts lead to a new argument for how systems of temporal reference should be represented: I will argue that they support the approach of Hans Reichenbach to this question. According to Reichenbach, temporal reference involves three times, Speech Time (ST), Reference Time (RT), and Event Time (ET). ST is the keystone of the system: other times are related to it and to each other by the relations of simultaneity or sequence. Examples that show the need for three times include (for English) the perfect, and embedded sentences that do not establish a time independently. 37. Michelle had already called last week. 38. They said on Sunday that they were leaving in 3 days. In both of these sentences, three times must be distinguished for temporal interpretation: ST, a time anterior to ST (last week, on Sunday), and a third time, anterior or posterior (37 and 38, respectively) to the time established in the matrix sentence. One might claim that a system of tense (or time) embedding would be adequate to account for sentences such as 35 and 36. In this view, there is no need for a temporal system such a Reichenbach’s. The English perfect, for instance, could be analyzed as a past embedded under a past, without any special machinery. The analyses of tense logic exemplify this view. An embedding analysis can indeed account for sentences with three times, but whether it gives the best analysis of temporal reference is dubious. It would need special, ad hoc provisions to account for vague and discontinuous reference, and would need to be severely restrained in power since it predicts multiple embeddings of times that do not occur in natural language.21 A system of three times with ST at its center is predicted implicitly by the way times anchor in language. Absolute reference has ST (or another orientation time) as its anchor, and a related time, RT. Relative reference has a time given on the
20 In
the conditional and irrealis uses of past tense, it does not have the absolute interpretation. present a detailed discussion of this point in Smith (1975) in which I argue against a proposal of McCawley’s (1971) that English tenses be treated as higher verbs, and embedded as part of the general embedding system of English.
21 I
Temporal Structures in Discourse
299
surface as its anchor, and a related time. It is this third time that corresponds to ET in Reichenbach’s system. The close fit between the anchoring facts and a system of temporal reference such as Reichenbach’s constitutes strong evidence in favor of that system.22 2.3 A second argument for Reichenbach’s system also emerges from the analysis of this paper. For this argument I return to example 16, repeated here. 16. (a) We are deeply concerned with last Sunday’s robbery of an important public figure . . . (b) Last night I talked to a reporter who had new information. (c) The maid left before noon. I noted in Section 1.3 above that sentence (c) anchors to (a); such a sentence might be uttered in sequence with (b), that is, without an explicit anchor. Its interpretation would then be vague, as noted in the discussion of this example. I have suggested23 an analysis of this type of sentence, in which it specifies ET and shares RT specified in another sentence. Cases like this are not unexpected in a system with three times. A sentence has only one time adverbial, specifying RT or ET. A related sentence may specify ET different from RT, or an RT that anchors ET. I see no natural way to account for discontinuous and vague interpretations in an embedding system. For discontinuous readings, a missing anchor or related time would have to be located and correctly interpolated. For vague readings the mediating anchor time would have to be specified as missing. But why, in such a system, should there be a missing time? And, if in principle one could allow for a missing time, why should not there be infinitely many such times? These and other questions seem to me extremely serious. No doubt an embedding system could handle time, but this is an indication of the general power of such systems and not of their appropriateness for dealing with temporal reference. In this section I have developed some ramifications of the analysis of extended temporal structures. I have shown that tenses as well as time adverbials may anchor either to ST or a time given on the surface, and have explained the well-known asymmetry of English tenses in terms of different types of anchoring. The tense system of English is mixed, with past tense of the absolute type and present tense of the relative type; there is one exception for each tense, that of an embedded complement context. I have also argued that a Reichenbach approach accords well with this analysis, which I take to have general scope well beyond the system of English.
22 See
McGilvray (1978) for an argument that Reichenbach’s approach to temporal reference captures important points about consciousness and temporal referrer. [Editors’ note: McGilvray (1978) was apparently an unpublished manuscript that is not now available and that is unknown to its author. For a more recent treatment, see McGilvray (1991). Our thanks to James McGilvray for his assistance on this matter.] 23 See Smith (1978).
300
C.S. Smith
An embedding system, on the other hand, is ill-suited to predicting or explaining the way temporal reference works. I return in Section 3 to extended temporal reference. 3 Temporal reference is established by combinations of tense and time adverbials, and under certain conditions can have a domain beyond that of a single sentence. In Section 1, two principles were presented that account for the way temporal expressions form extended temporal structures. However, the principles do not, and cannot, tell us when such structures actually occur. There are two reasons for this limitation. Semantic connections are essential to linguistic structures that have a domain larger than a sentence. But such connections are not within the scope of syntactically-based principles that deal with temporal expressions. Elliptical phenomena are also beyond the scope of such principles, which rely on explicit forms. I discuss briefly some points of semantic interpretation that are not temporal, strictly speaking, but that affect extended temporal structures. My purpose here is to indicate the wide range of connections that are relevant rather than to give a complete account. To begin, note that temporal expressions themselves cannot be related without consideration of the type of unit to which they refer. For instance, in 39, (b) cannot be plausibly capturing according to the criteria developed in Section 1. 39. (a) Every weekday morning Jenkins did the same thing. (b) Last year he (Jenkins) went to Louisiana. The units morning and year are not commensurate, and capture must therefore be blocked. Order is crucial in these sequences: if year preceded, a plausible extended interpretation could be made. Semantic facts about temporal expressions, then, are relevant to capturing. A closer look at possible interpretations reveals a Pandora’s box of difficulties. Temporal structures can be established only when certain connections obtain between sentences. These connections may have to do with subjects, objects, or other complements in the sentences; or with the relation between verbs in one sentence and nouns in another; etc. Thus in the pair of sequences below, 40 can be interpreted as an extended temporal structure and 41 cannot, yet the difference is minimal and not related to the temporal expressions. 40. (a) (b) 41. (a) (b)
Last week Nick went to Oxford. He saw Danforth on Tuesday. Last week Nick went to Oxford. I saw Danforth on Tuesday.
Taking 40 as one temporal structure, Tuesday refers to a day of last week; and Danforth is probably someone whom Nick saw at Oxford. 41 has no such interpretation. Rather, Tuesday is deictic; Danforth is probably someone who has news of Nick’s activities. As a final example consider 42, followed by either 43 or 44:
Temporal Structures in Discourse
301
42. (a) The Taylors went on a trip. (b) They sold their porcelain dog collection to raise the money. 43. 44.
The sale was last month. The trip was last month.
For semantic reasons, 43 forms a temporal unit with 42b, and 44 forms a unit with 42a. Sell and sale connect the first pair, the two trips connect the second. The immediate conclusion to be drawn from these examples is that information other than the temporal is needed to establish extended temporal structures. But there is a more general conclusion, which bears on the larger question of linguistic structures beyond the sentence. The conditions for temporal structures given here are sentence-based, depending on expressions that are associated with individual sentences. But it is clear that conditions of this type cannot, in principle, suffice for recognizing or forming structures beyond the sentence. Semantic connections are necessary, and semantic connections are not limited as to type of expression or syntactic context. Although this paper discusses only temporal structures, I suggest that the conclusion extends to other areas as well. I would like, finally, to consider briefly the relation of extended temporal structures to structures in discourse. What in fact is the relation of syntactico-semantic units such as extended temporal units to discourse structure? Is there a correspondence between the two? Is it consistent, so that in locating the first, one has located the second? These are difficult questions. However, I would like to suggest a partial answer; somewhat negative, possibly useful for inquiry in these interesting and little-understood areas. There is no general relation, I would argue, between discourse structure and extended temporal units such as those I have investigated in this paper. In some discourse, temporal structures play a role that corresponds to important structures in discourse. In other discourse it does not. Only when dealing with a limited genre, where it is known that temporal units have an organizing role, can a correlation between temporal units and discourse structure be assumed in advance. Even in such cases there are always exceptions. In certain types of discourse, temporal and other deictic structures often have an organizing role: extended temporal units can be taken as signals of units of discourse. Narratives, especially informal and personal narratives, are of this type.24 But in other discourse (interviews, didactic writing, thematically organized texts, for example) units of time or place may play a subsidiary role in the structure of the whole. To locate such units, then, does not necessarily mean that one has located the structures important to the general organization of a given text.25 To decide
24 Labov
and Waletzky (1967) make a persuasive case that temporal and spatial structures are important in personal narratives. They are less persuasive in their argument that such structures are of general importance in discourse. 25 See Levinson (1979) for an interesting discussion of differences in the structure of different types of linguistic activity.
302
C.S. Smith
on the organizing structures of a text, the nature of the particular language activity must be considered, and no correlation can be assumed in advance.
References Banfield, A. (1973). Narrative style and the grammar of direct and indirect speech. Foundations of Language, 10, 1–39. Benveniste, E. (1966). Problèmes de linguistique générale. Paris: Gallimard. Friedman, N. (1955). Point of view in fiction: The development of a critical concept. PMLA, 70, 1160–1184. Hamburger, K. (1957). Die Logik der Dichtung. Stuttgart: E. Klett. Hobbs, J. R. (1976). A computational approach to discourse analysis. Research report 76-2, Department of Computer Science, City University of New York. Labov, W., & Waletzky, J. (1967). Narrative analysis: oral versions of personal experience. In J. Helm (ed.), Essays on the verbal and visual arts (pp. 12–44). Seattle: University of Washington Press. Levinson, S. C. (1979). Activity types and language. Linguistics, 17, 365–399. Lips, M. (1926). Le style indirect libre. Paris: Payot. McCawley, J. (1971). Tense and time reference in English. In C. J. Fillmore & D. T. Langendoen (eds.), Studies in linguistic semantics (pp. 97–113). New York: Holt, Rinehart, & Winston. McGilvray, J. (1991). Tense, reference, and worldmaking. Montreal: McGill-Queen s. Ota, A. (1972). Comparison of English and Japanese, with special reference to tense and aspect. Studies in English Linguistics. Tokyo: Asahi Press. Partee, B. H. (1973). Some structural analogies between tenses and pronouns in English. The Journal of Philosophy, 70, 601–609. Reichenbach, H. (1947). Elements of symbolic logic. New York: Macmillan. Smith, C. S. (1975). The analysis of tense in English. Texas Linguistic Forum, 1, 71–89. Smith, C. S. (1976). A theory of auxiliary have in English. Bloomington: Indiana Linguistics Club. Smith, C. S. (1977). The vagueness of sentences in isolation. Proceedings of the Chicago Linguistic Society. Department of Linguistics, University of Chicago. Smith, C. S. (1978). The syntax and interpretation of temporal expressions in English. Linguistics and Philosopy, 2, 43–100. Smith, C. S. (1981). Semantic and syntactic constraints on temporal expressions. In P. Tedeschi. & A. Zaenen (eds.) Syntax and semantics, vol. 14: Tense and aspect (pp. 213–239). New York: Academic Press. Smith, C. S., & Whitaker, J. T. (1979). Interpreting ellipses in a text. Pragmatics Microfiche. Stenning, K. (1978). Anaphora as an approach to pragmatics. In M. Halle, J. Bresnan, & G. Miller (eds.), Linguistic theory and psychological reality (pp. 162–200). Cambridge, MA: MIT Press. Webber, B. (1978). A formal approach to discourse anaphora. Technical Report 3761. Cambridge, MA: Bolt, Beranek and Newman. Weinrich, H. (1964). Tempus. Stuttgart: Kohlhammer. Wolfson, Nessa. (1979). The conversational historical present alternation. Language, 55, 168–182.
Temporal Interpretation in Mandarin Chinese Carlota S. Smith and Mary S. Erbaugh
Abstract This article presents an account of temporal understanding in Mandarin Chinese. Aspectual, lexical, and adverbial information and pragmatic principles all contribute to the interpretation of temporal location. Aspectual viewpoint and situation type give information in the absence of explicit temporal forms. The main, default pattern of interpretation is deictic. The pragmatic principles are the bounded event constraint, the simplicity principle of interpretation, and the temporal schema principle. Lexical and adverbial information can lead to non-default interpretations. Two other temporal patterns — narrative dynamism and anaphora — appear in text passages that realize the “discourse modes” of narrative and description. We state the semantic meaning of grammatical forms and explain the deictic pattern. Three times are needed to explain temporal interpretation, following Reichenbach (1947). Mandarin forms code the relation between a designated perspective time, or reference time, and situation time. These are typically marked redundantly in written texts. Relation to speech time is not coded linguistically, but conveyed by context.
1 Introduction In all languages, sentences convey information that allows people to locate situations in time. We discuss here how temporal information is conveyed in Mandarin.1 The traditional account has it that Mandarin relies heavily on adverbs and pragmatics. The tradition is partly correct, but we show that aspectual information plays a key role as well. We state the different kinds of information, and the principles, that underlie temporal interpretation. The principles are pragmatic in nature, since they rely on inference and context, as well as the information conveyed by linguistic forms.
1 The ideas that we develop in this article were first presented in Smith 2000a, and Smith and Erbaugh 2001.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_13,
303
304
C.S. Smith and M.S. Erbaugh
As a starting point we ask what semantic information is needed for temporal location, and then consider the language resources that convey it. Our discussion is relevant to both tensed and tenseless languages, although we focus on Mandarin. As the resources of languages differ, so do the ways the principles are implemented in a given case. When no direct temporal information appears, aspectual information, such as whether a situation is ongoing or closed, gives pragmatic cues to the temporal location of the situation expressed. Time is a single unbounded dimension. Like space, time requires an orientation point for location. The speaker is the canonical center of linguistic communication, and the canonical temporal orientation point is speech time, now. Thus, the basic pattern of temporal interpretation is deictic: the situations expressed in sentences are located in relation to speech time.2 This is the default interpretation of tense, and the default interpretation of tenseless languages. The default arises in a manner that we spell out below. The linguistic forms that play a role in temporal location for Mandarin are adverbial, lexical, and aspectual. Temporal adverbs give direct information about location in time. Lexical forms such as modals and future-oriented verbs suggest temporal location. Aspect introduces information about situation type—events (run a mile, build a house) and states (believe, be happy)—and viewpoint (perfective, imperfective, neutral). Situation type and viewpoint express situations as discrete and bounded, ongoing and unbounded, or indeterminate. We propose three pragmatic principles that explain the deictic pattern: (1) the bounded event constraint; (2) the simplicity principle of interpretation; and (3) the temporal schema principle. Adverbs and other information may supplement or override aspectual information. In this article, we state the principles and discuss the relevant linguistic forms; we give natural examples from written texts in Mandarin. Not all temporal interpretation is deictic. Two important patterns are the temporal advancement of narrative, and the static interpretation that one finds in description and certain other contexts (Smith 2000b, 2003); perhaps there are others. Later in this article, we show that all three patterns of temporal interpretation are found in written texts of Mandarin. Section 2 discusses the deictic pattern and introduces the basic notions of aspect and temporal location. Section 3 presents an account of the deictic pattern, with examples of how it is realized in Mandarin. Section 4 introduces narrative and static temporal interpretation, with examples from Mandarin texts. Section 5 concludes.
2 The Deictic Pattern: Aspectual Information and Pragmatic Constraints The deictic pattern—in which speech time is central—is a linguistic universal, as far as we know. The deictic pattern locates situations either at speech time, or gives 2 Deictic terms such as ‘later’ or ‘here’ can only be interpreted with reference to a specified time or place.
Temporal Interpretation in Mandarin Chinese
305
temporal locations in terms of their relation to speech time. The present is located at speech time; the past precedes it, and the future follows. In deixis, the default interpretation locates ongoing events and states in the present (running, believe), and bounded events in the past (run, break). The other possibilities are constrained and require additional explicit information. The key factor underlying the deictic pattern is whether a situation is bounded or unbounded. “Boundedness” is an aspectual notion: it refers to a property of the situations expressed in sentences. Bounded situations are temporally closed, by implicit or explicit bounds (ran, broke); unbounded situations are ongoing, temporally open (running, breaking). Boundedness depends on both aspectual viewpoint and situation type. We will say more about this below. The deictic pattern is stated in (1), and more compactly in (1 ):
(1)
Default deictic pattern of temporal form and interpretation a. b. c. d.
Ongoing events are in the present: located at speech time States (unbounded) are in the present: located at speech time Bounded events are in the past: located before speech time Explicit temporal information may override (a)–(c)
(1 ) a. Unbounded situations are located in the present b. Bounded events are located in the past
The temporal possibilities are not limited to the default statements in (1a)–(1c) and (1 ). Bounded events may be located in the future, and states or ongoing events may be in the past or future. These departures from the pattern are expressed with additional information, as (1d) indicates. However, no bounded events are located in the present. This is an important, nonaccidental, gap in the paradigm. It is due to a general constraint that events in the present cannot be bounded. The realization of the deictic pattern depends on pragmatic constraints that involve both aspectual and temporal information. The restriction on events in the present arises from a pragmatic convention of communication that has semantic consequences. In taking the temporal perspective of the present, speakers follow a tacit convention that communication is instantaneous. The perspective of the present time is incompatible with a bounded event, because the bounds would go beyond that perspective. As Kamp and Reyle (1993) put it: A present tense sentence describes an eventuality (situation) as occurring at the time at which the sentence is uttered, and thus at a time at which the thought is being entertained which the sentence expresses. So the thought must conceive the eventuality as it appears from the perspective of the time at which it is going on. A sentence which describes something as going on at a time—in the sense of not having come to an end when that time is up—cannot represent something as an event. For the event would have to be entirely included within the location time and thus would not extend beyond it. (Kamp and Reyle 1993: 536–537)
306
C.S. Smith and M.S. Erbaugh
Due to this constraint, all simple present tense sentences express unbounded situations. We call it the bounded event constraint (c.f. Smith 2003). The constraint is stated in (2): (2) The bounded event constraint: bounded events are not located in the present. The bounded event constraint is an essential factor in the deictic pattern, since it partially explains clause (1c), the default for bounded events. Almost the same notion is called the “punctuality constraint” in Giorgi and Pianesi (1997: 163); for an earlier formulation, see Lyons (1977). The constraint explains the limited interpretations available for situations located in the present in English. They may be ongoing events (John is talking); specific states (Ella is sick); or generalizing statives and generics (Tom feeds the cat, elephants eat grass), but not bounded events. There are some exceptions, which have well-known special interpretations.3 Generalizing statives have the simple verb form. With present tense, they express a pattern of situations rather than a specific event or state, and are thus semantically stative. Languages realize the bounded event constraint in different ways. In Russian, sentences with the perfective viewpoint and present tense convey future time; the Romance languages do not have a perfective present tense form. The deictic pattern provides that the default for bounded events is past rather than future. We must also explain this. To do so, we invoke a general simplicity principle of information processing. When faced with information that does not fully determine an interpretation, people choose the simplest interpretation to resolve it. The point has been made convincingly for perception. If people view a partial or indeterminate picture or shape, the visual system constructs a simple, complete percept (Kanizsa 1976). People construct a simple complete form rather than a complex or incomplete one. We adapt this finding as a simplicity criterion for the interpretation of temporal location. The past is simpler than the future, we claim. The two are symmetrically related to speech time—the past precedes, the future follows. But the future has the additional factor of uncertainty, a modal meaning (e.g. will, may) (Lyons 1977; Kamp and Reyle 1993). Due to the modal factor, we take it that the future is more complex 3 The simple verb form conveys the perfective viewpoint in English. For statives, the perfective viewpoint focuses the entire temporal schema of the state, which does not include endpoints; see Smith (1997 [1991]) for discussion. Thus the interpretation of sentences like Tom feeds the cat is consistent with other statives in English. The shift to a stative situation type arises by coercion, triggered by the bounded event constraint. There are exceptions, for example, sports-announcer reports and stage directions (Now Jones throws to third base), literary commentary (Here the author creates an interesting metaphor). In these cases, time is telescoped. Note that performatives (I pronounce you husband and wife) are not exceptions: a performative is just that, a performance rather than a report (Austin 1961). Present tense narratives have another interpretation: they express events that are related to each other, rather than to the moment of speech. See Section 4 for discussion of the temporal pattern of narrative.
Temporal Interpretation in Mandarin Chinese
307
than the past. The prediction then is that people take sentences expressing bounded events to be in the past unless there is explicit indication of the future. And indeed, this is the case. We state the finding as a general simplicity principle that constrains interpretation. (3) Simplicity principle of interpretation: choose the interpretation that requires the least additional information. From the Gricean point of view (Grice 1975), this principle constrains the inferences that people make when they infer information that is not explicitly asserted. Given a sentence that says the minimum necessary, perhaps according to Grice’s second maxim of quantity, one “fills out” the interpretation (c.f. the informativeness principle of Levinson [1983], or the R-principle of Horn [1984]). Similar principles are used to constrain computational reasoning procedures. We suggest that these two pragmatic principles underlie the deictic pattern of temporal interpretation. In the discussion below, we state a third principle—the temporal schema principle—that applies to clauses that have neither explicit temporal or aspectual viewpoint information. We refer to clauses without an overt viewpoint morpheme as “zero-marked,” following Klein et al. (2000). Information relevant to temporal interpretation is coded primarily by adverbials, tense, and aspect, depending on the language. While this article focuses on Mandarin, we will occasionally compare it to English. The deictic pattern is expressed according to the resources of each language. We assume that Mandarin is tenseless, while English is a tensed language. In Mandarin, aspectual factors give key information, in a manner explained below. In English, tense and aspect interact: the present tense conveys unbounded situations in the present and the past tense conveys situations in the past.4 Since we are talking about interpretation, we implicitly take the viewpoint of the receiver rather than that of the speaker. However, we assume a Gricean (1975) approach in which the calculations of speaker and hearer are essentially the same. In this article, we try to state what and how information contributes to interpretation. We do not intend a psycholinguistic account, which would deal with processing: we make no claims about the actual order in which processing occurs.
2.1 The Keys to the Deictic Pattern: Aspect and Temporal Location Aspect concerns the internal temporal structure of the situations introduced into a discourse. We assume that aspectual systems have two components—situation
4 The present tense doesn’t always express a time in the present, nor does the past tense always express past time. This well-known fact is not relevant to the discussion here.
308
C.S. Smith and M.S. Erbaugh
type and viewpoint—modeled within discourse representation theory (Smith 1997 [1991]; Kamp and Reyle 1993). Situation type indirectly classifies a sentence as expressing a state or an event of a certain type. The categories of event and state are idealizations, types of situations with distinctive temporal properties. The situation types are realized at clause level by the verb and its arguments, the “verb constellation.” Verb constellations associated with a given situation type have unique distributional and semantic properties, and are thus recognizable as covert linguistic categories in the sense of Whorf (1956). Three two-valued temporal features differentiate the main classes of situations: dynamic/static, telic/atelic, durative/instantaneous. The distinction between events and states, and between telic and atelic events, can affect temporal interpretation. Events are dynamic, occurring at successive stages. States are static, holding consistently throughout an interval without endpoints—the changes into and out of a state are events in themselves. Telic events have natural endpoints; durative atelic events, or activities, have arbitrary, potential endpoints.5 For situation type, a temporal schema models its defining features; the temporal schema is associated by rule with the appropriate verb constellations. The features have linguistic correlates, as discussed by Vendler (1957) and others. The features of heterogeneity and homogeneity, based on part-whole relations, also form classes among the situation types. Telic and bounded events are heterogeneous because the part is different in kind from the whole. Only a complete telic event has the properties of the event itself, for example, a part of “walking to school” doesn’t count as “walking to school.” In contrast, states and atelic events sometimes pattern together linguistically. Aspectual viewpoint makes available for semantic interpretation all or part of a situation; more precisely, the temporal schema associated with a situation type. The aspectual value of a clause is a composite of information from both situation type and viewpoint. The semantic meaning of each viewpoint must be stated for each language, but some generalizations can be made. The same situation may be described from different aspectual viewpoints as a language allows. Most viewpoints are usually grammatical at the clause level except for a few fixed contexts (though they may not be felicitous in a given discourse). There are three classes of viewpoint: perfectives (open the door) make events visible as including endpoints, bounded. Imperfectives (opening the door) make situations visible without information as to endpoints, unbounded. Neutral viewpoints are flexible, giving enough information to allow a bounded or an unbounded interpretation. English does not have a neutral viewpoint; see below for Mandarin examples. We relate the neutral viewpoint
5 We assume five main situation types (Smith 1997 [1991]). The classification is based on Vendler (1957) with one addition, the semelfactive. States: know the answer, love Jean, be expensive; events: activity—atelic, durative: sing, walk; semelfactive—atelie, instantaneous: cough, flap a wing; accomplishment—telic, durative: build a house, walk to school; achievement—telic, instantaneous: win the race, reach the top.
Temporal Interpretation in Mandarin Chinese
309
to zero-marked clauses, those without an overt viewpoint morpheme.6 Viewpoint morphemes are optional in most clauses of Mandarin. Boundedness and unboundedness may be expressed by situation type, viewpoint, and otherwise, according to the resources of a language. Bounds are usually conveyed by the perfective viewpoint; they may also be stated independently with adverbial or other information (Depraetere 1995). In the two-component theory, situation type and viewpoint are distinct. Imperfectives (Leigh is cooking) and states (Leigh is happy) are similar — for instance, both have the subinterval property: when they hold for an interval I, they hold equally for any smaller interval of I. But they also differ. One important difference is that states can be taken as inchoative in certain contexts, whereas imperfectives cannot. To see this, compare Leigh was smiling when John broke the glass and Leigh was angry when John broke the glass; only in the latter can the main clause have an inchoative interpretation (Smith 1997 [1991]). Temporal location requires an orientation point, as noted above, which in the default case is “speech time,” the moment of speech.7 The moment or interval at which a situation takes place, “situation time,” is understood in relation to speech time. Past situations precede speech time, future situations follow it. The expression of temporal location in language also involves another time: “reference time.” Our understanding of temporal interpretation is informed by the views of Hans Reichenbach (1947), extended and modified by many, including Partee (1984), Hinrichs (1986), and Smith (1997 [1991]). In his discussion of tensed languages Reichenbach showed that the temporal location of situations involves a designated reference time, in addition to speech time and situation time. The designated time conveys a temporal perspective from which “the speaker invites his audience to consider the event” (Taylor 1977: 203). We will show that the notion of reference time is relevant to Mandarin as well. Situation time may be a moment or interval, depending on contextual information. Situations are wholly or partially located at SitT according to their aspectual value. Bounded situations (perfective, cooked), are totally included in situation time (e ⊆ SitT); unbounded events (imperfective, is cooking) and states (believe) overlap or surround it (e 0 SitT). This insight goes back at least to Jespersen (1931: 79). Another account of aspect and temporal location is given by Klein (1994) and Klein 6 The neutral viewpoint is a default with a positive semantic value. It arises in aspectual systems which allow sentences without an overt viewpoint morpheme. Empirically, the interpretation of zero-marked, aspectually vague sentences can be shown to be neither perfective nor imperfective. Such sentences are more flexible than sentences of either viewpoint, in that they allow both open and closed readings. See Smith (1997 [1991]) for discussion. The range of meanings that is found for zero-marked sentences, and the theory-internal requirement of visibility, suggest that they have a viewpoint that is open but not unlimited. 7 In Reichenbach’s (1947) system, tense conveys information about three times, speech time (SpT), reference time (RT), and event time (our “situation time”); and the relations between them. The semantic meaning of a tense gives the relations between the times. The approach has been modified and formalized for English in the framework of discourse representation theory, or DR theory (Smith 1997 [1991], 2003; Kamp and Reyle 1993).
310
C.S. Smith and M.S. Erbaugh
et al. (2000). In this account aspect codes the relation between the run-time of a situation and topic time, a designated reference time.8 For perfective events, the event is totally included within topic time; for imperfectives the situation overlaps topic time. Reichenbach (1947) gave several kinds of evidence for reference time; he focused on the English tense system. One piece of evidence is the contrast between sentences in the past tense and the present perfect.9 Another is that reference time clarifies the temporal relations between the situations expressed in language. Adverbial clauses and other contexts give information that locates situations relative to one another. Using the notion of reference time, Reichenbach shows that overlapping situations share reference time and those in sequence do not. The phenomenon of shifted deixis also supports the notion of reference time. Deictic adverbials such as now, in three days, which normally anchor to the moment of speech, can anchor to a past (or future) time, as in Leigh sat down at the desk. Now he was ready to start work. In such contexts, the shifted now suggests Leigh’s perspective. Reference time is the anchor for this perspective. We take the view, along with Y. R. Chao, that “Chinese verbs have no tense” (Chao 1948: 54). In this we follow the traditional idea that the term “tense” refers to a morpheme, either an inflection or auxiliary, that appears in the main verb phrase of a sentence and has a temporal meaning. Huang (1984) and others have claimed that Mandarin has a finite-nonfinite distinction, which might suggest that it has abstract tense. Hu et al. (2001) present strong arguments to the contrary; Lin (2003), discussing the semantics of temporality in Mandarin, agrees. We concur with the latter authors. We do not think that there is evidence for syntactic tense in the Mandarin language, nor for a systematic finite-nonfinite distinction between clauses. Tense as a linguistic category is discussed further in Smith et al. (2003). The notion of a designated reference time is not dependent on tense, but is basic to temporal location in language. Indeed, it has explanatory value for Mandarin. Reference time partly accounts for the contrast between the perfective morphemes le and -guo, which have a temporal component. Adverbials such as yijing (‘already’) code a relation to reference time; and for temporal relations between situations and shifted deixis, reference time is useful as well. Aspectual morphemes may code the relation between reference time and situation time. They do not code a relation between reference time and speech 8 For Klein and his colleagues (1994, 2000), every sentence introduces a topic time, the time about which the assertion of a sentence is made. The notions of topic time and reference time as adduced here, and in Smith (1997 [1991]) are quite similar. The theory of Kamp and Reyle (1993), although based on Reichenbach (1947), takes a slightly different approach. 9 The two have the same truth conditions, yet they differ in conceptual meaning. The notion of reference time explains the difference. Sentences with the past tense are set squarely in the past, for example, Leigh arrived. Reference time and situation time are both past. But present perfect sentences take the perspective of the present: for example, in Leigh has arrived, reference time is present and situation time is past. In some sentences three times are needed to state truthconditional meaning, for example, the past perfect Leigh had already arrived. The perfect construction has both temporal and aspectual meanings; aspectually perfects are stative.
Temporal Interpretation in Mandarin Chinese
311
time, however. In Mandarin, the latter relation is inferred from information in the text and/or context. In tensed languages such as English, tense codes the relation between reference time and speech time. English forces a choice between present and past present tense for states (The Great Wall is / was impressive). Mandarin does not. What differentiates Mandarin from a tensed language, then, is that the grammatical forms relevant to temporal location do not relate reference time to speech time.
3 Temporal Information in Mandarin In this section we discuss the temporal information conveyed by aspect, adverbs, and certain specific lexical forms in Mandarin. We then show how the deictic pattern is realized. Although our presentation is relatively informal, we are interested in an account that can be developed formally within discourse representation theory, which constructs a dynamic representation of the information conveyed in discourse. Syntactic surface structure is the input to the construction rules of the theory. The rules for temporal information enter situation entities, times, and their relations into semantic representation.
3.1 Reference Time in Mandarin The aspectual viewpoint morphemes of Mandarin give both aspectual and temporal information. Temporally, le and -guo indicate different relations between reference time (RT) and situation time (SitT). Le conveys that the SitT interval is simultaneous with RT; -guo conveys that SitT precedes RT, at a different interval. For instance, consider the constructed examples below; the translations are approximate: (4) a. Ta da le majiang. She play LE mahjong. ‘She played/has played mahjong.’ b. Ta daguo majiang. She play GUO mahjong. ‘She has played mahjong (at least once, at some past time).’ c. Ta jie le hun. She contract LE marriage. ‘She got married (and probably still is married).’ d. Ta jieguo hun. She contract GUO marriage. ‘She has been married (at least once).’ The difference between these clauses is only partly captured by the translations using the English past tense and present perfect. More precisely, le conveys in (4a) that the event is bounded and that SitT is simultaneous with RT. Recall that bounded
312
C.S. Smith and M.S. Erbaugh
events are included in SitT. By the bounded event constraint, the default interpretation is past: SitT and RT precede the time of speech. In (4b), -guo conveys that there is at least one time included in SitT, that the event takes place at that time; and that SitT occurs before RT. Again the situation is past by default.10 Adverbs such as yijing (‘already’) and cai (‘have just’) may convey that an event or state precedes reference time, in past and future contexts, or with perfective -LE: (5) a. Ta yijing zou le. He already left LE. ‘He has already left.’ b. Wo cai dao. I only just arrive. ‘I have just arrived.’ The temporal meaning of the adverbs in such cases is to locate a situation at a SitT before RT. One reason that Reichenbach (1947) posits three times for tense is that past perfect sentences require three times for temporal interpretation, as noted above. Mandarin, too, has sentences that require three times. We give as an example a sentence with a past adverbial and the aspectual morpheme -guo, as in (6); we owe the examples to Haihua Pan. (6) San nian qian ta (jiu) quguo Meiguo. Three year ago he (just then) go GUO America. ‘He had been to America 3 years ago./He (just then) went to America 3 years ago.’ For some speakers the sentence has two readings, like the English translations. One reading requires three distinct times: reference time is ‘3 years ago,’ prior to speech time; this time is indicated by the adverb. The time of leaving, situation time, is prior to RT, indicated by -guo. In the other reading, SitT and RT are the same, and prior
10 For
simplicity, we present constructed examples without temporal adverbs. Under certain circumstances, sentences with -guo may have temporal adverbs, as noted in Smith (1997 [1991]). In the sentence below, mentioned by a reviewer, the past adverb zuotian (‘yesterday’) specifies SitT; here prior to RT = SpT. (i) Zhangsan zuotian daguo majiang. Zhangsan yesterday da GUO mahjong. The interpretation is similar to that of a present perfect sentence in English, although English does not allow a past adverb in such contexts.
Temporal Interpretation in Mandarin Chinese
313
to SpT.11 To take a somewhat clearer case, an explicitly past sentence with yijing (‘already’) indicates a prior SitT, as in (7): (7) San nian qian ta yijing quguo Meiguo. Three year ago he already go GUO America. ‘He had already gone to America 3 years ago.’ These examples show that three times are needed for the temporal interpretation of some Mandarin sentences. Reference time is also needed to model the relations between situations expressed in Mandarin. The context of a clause gives information that locates situations relative to one another, as in (8), from Mangione and Li (1993): (8) Ta chi le fan cai zou de. She eat LE rice only then go DE. ‘Only after eating did she go.’ The structurally determined RT for ‘going’ is the time of ‘eating’ (Mangione and Li 1993: 67). The two events occur in sequence, or overlap. Overlapping situations share reference time; those in sequence do not. Thus, RT provides a locus for relating situations in a principled manner, explicated for English in Hinrichs (1986). The final argument we make for RT comes from clauses with shifted deictic forms. Deictics normally anchor to speech time; but they can also anchor to a different time. Shifted deictics convey the speaker’s perspective at a designated reference time. We present as an example a fragment from a novel. The speaker contrasts her current good fortune with her earlier, rough life in Shanghai; the shifted deictics are underlined. This fragment comes from one of the naturally-occurring, written texts that we examined for this article. The translations are our own, some adapted from published bilingual editions, but all checked with native speakers. The end of this article lists details in the sources of text examples (see Appendix). (9) a. . . . xiang dao gei nage sha qian dao keren da bachang de . . . think to give that kill 1,000 knife guest big slap DE shi, incident b. zhongzhong bu xing ren de qingjing manner unfortunate person DE situation c. he xianzai shenghuo yi bijiao with current life one compare d. zhen shi bu shen huishou. really be boundless comparison 11 We
have checked these examples with several other native speakers and find considerable variation. Some speakers accept the three-time reading, especially in the presence of jiu (‘just then, already’); others do not.
314
C.S. Smith and M.S. Erbaugh
e. Wo xianzai suoxing nenggou jia gei Bi Xiansheng I now simply able marry to Bi Mr. [‘. . . thinking back to the time when I slapped that violent killer guest, of being a person unlucky in every imaginable way, and comparing it to my current life, it was really a boundless difference. Now I could simply marry Mr. Bi . . .’] Here, the writer uses xianzai (‘current, now’) to talk about her state of mind at a time in the past. We will assume that reference time plays a role in the temporal interpretation of Mandarin sentences. In the default case, reference time is the same as situation time, the moment or interval at which a situation occurs. Reference time may be given linguistically or contextually. Note that reference time alone does not specify temporal location because it is not anchored to speech time or another orientation time. In Mandarin, the relation of reference time to speech time—temporal location—is specified with explicit temporal information or determined pragmatically. We also assume that the default reference time is now.
3.2 The Linguistic Resources of Mandarin We discuss aspectual, adverbial, and lexical information, all of which contribute to the determination of temporal location. The locus of the discussion is the clause. We treat each clause separately, noting complement clauses as dependent on a main verb. In some cases, the nature of the main verb determines the temporal interpretation of the complement clause. We discuss complement clauses only briefly; see Lin (2003) for a detailed account of the temporal semantics of complements and other dependent clauses. Here and below we give examples from written texts. The examples come from an informal study in which we intensively analyzed six complete discursive articles from current newspapers and magazines. The articles exemplify standard written Chinese style. A dozen other texts were also analyzed, though less intensively. The texts used in the intensive study were published in Hong Kong, though some are by mainland authors. Three are editorials from the Ming Bao, a staid newspaper which serves as stylistic model for Chinese literature students. Three longer articles come from the Ming Bao Yuekan (Ming Pao Monthly), Yazhou Zhoukan (Asiaweek), and Qianshao Yuekan (Frontline Monthly). We deliberately avoided articles in the slangy, Cantonese-influenced style popular in Hong Kong tabloids and comic books. Instead, we chose articles in the standard written Chinese, which is very close to spoken Mandarin. Chinese writers, regardless of dialect, are taught to write in this prestigious style.12 12 The
editorials we used average about 700 characters each, the articles about 2,000 characters each. The six articles have a total of 7,640 characters, 176 sentences, in 699 clauses.
Temporal Interpretation in Mandarin Chinese
315
3.2.1 Aspect Aspectual viewpoints are conveyed in Mandarin by verbal le, the experiential toneless -guo; the aspectual morpheme zai, the toneless suffix -zhe; and by zero-marked clauses.13 For each one, we state semantic meanings that are compatible with the range of pragmatic interpretations that arise. We assume that the semantic meaning is associated with each morpheme. We give summary, uncontroversial characterizations here; these morphemes have been discussed at length by many authors and our main interest is their contribution to temporal interpretation.14 Recall that in our approach aspectual viewpoint makes semantically visible all or part of a situation, at situation time. Situation time is located at reference time unless the two are explicitly differentiated. Sentences with perfective morphemes express bounded situations. The verb suffixes le and -guo are perfective. Perfectivity is also conveyed by resultative verb complements such as the suffix -wan (‘finish’) (and by reduplication in a construction known as the “tentative,” which is not considered here). Le conveys that an event is bounded, although the boundary need not coincide with the completion of a telic event; (see examples in Smith 1994; Wu 2003). The event is contained in SitT, SitT = RT. Le appears with event verb constellations, which we take to include inchoatives.15 Clauses with -guo convey that SitT precedes RT. For reversible situations, -guo codes a discontinuity such that the resultant state no longer obtains (Chao 1968). We coded every clause for situation type (event, state, generalizing-habitual), explicit aspectual viewpoint or RVC, temporal adverbial or other temporal form, modal, and special lexical verb factors, whether within the scope of a preceding temporal adverbial. We also looked at other sources less intensively. For narrative, we examined five short stories and one novel. For nonfiction, we examined two additional editorials from the Ming Bao Daily News, museum catalogs and travel guides, a book on Hakka culture, and a book of essays on Shanghai life. See sources of text examples above for more precise information. 13 Standard orthography for Hanyu Pinyin romanization prints le as a suffix immediately following the verb, without a hyphen. The exception is when le appears sentence-final, where it is printed as a separate morpheme, after a space (as the homophonous le mood particle also is). We do not consider prepositional uses of -zhe to be aspect markers (e.g. yanzhe ‘following along,’ bianzhe ‘edging’). 14 There are many studies of verbal le and -guo, too many to cite here. In addition to the traditional accounts, recent work includes Huang (1987), Shi (1990), Mangione and Li (1993), Yeh (1996), Klein et al. (2000). Lin (2003) gives a more complete list, though still partial. Wu (2003) discusses many traditional and current accounts of the aspectual morphemes. 15 Perfective le appears with state verb constellations in ingressives and inchoative clauses. Ingressives focus on the initial endpoint of an event; inchoatives focus on the coming about of a state. They are derived events, arising through a regular process of coercion. There are examples with le in which a situation can be taken as continuing, as Lin (2003) points out. The following is one of Lin’s examples: (i) Ta yang le yi zhi tuji. He raise LE one CL wild chicken. He is raising a wild chicken.
316
C.S. Smith and M.S. Erbaugh
Often indefinite, -guo conveys that a situation must have obtained at least once. It appears with event verb constellations, and, much more rarely, with state verb constellations.16 Guo clauses have another aspectual component: they are stative, focusing on the state that arises after a situation has taken place; (has had the experience of being married/playing mahjong/eating snake). Like statives generally, such clauses do not advance narrative time. The verb constellation of a -guo clause is often dynamic at the basic level of interpretation, so that such clauses are derived statives. (Basic level and derived situation types are discussed in Smith 1997 [1991].) Following Mangione and Li (1993), and Klein et al. (2000), we model the temporal meaning of -guo and le with two times, the time at which a situation occurs and a designated reference time. Guo locates situation time before reference time, whereas with le, and the other aspectual morphemes, situation time and reference time are simultaneous. Sentences with imperfective morphemes express unbounded situations. Zai is a progressive imperfective, appearing with event clauses in preverbal position. In contrast -zhe is a more restricted imperfective, used often for resultatives of situations, end locations of motions, or for backgrounding; it is more common in subordinate clauses than in main clauses. Clauses with zai have an ongoing meaning, due to their dynamic nature. Zhe clauses are static. SitT is simultaneous with RT for both imperfectives. The zero or Ø morpheme conveys a neutral aspectual viewpoint: it gives enough information to allow either an open or a closed interpretation. The notion of a neutral viewpoint for Mandarin is discussed in Smith (1997 [1991], 1994). The information focused by the neutral viewpoint includes part of a situation; this correctly provides that the situation may or may not extend beyond the situation time interval. We summarize informally the aspectual and temporal information conveyed by the four overt viewpoint morphemes, and the zero aspectual morpheme; the term “situation” includes events and states.
We analyze the continuing reading as the result of the ingressive: bounded le focuses the beginning of the activity event. After the beginning, the event continues indefinitely. Note that our account differs from Lin (2003), in that we allow le clauses to express telic events that are not necessarily complete. That is, the bounds need not coincide with the natural final endpoint. Our account is similar to the characterization of Klein et al. (2000), and of Mangione and Li (1993): “. . . le marks a specific event time, which is ordered before and closely related to the sentence’s reference time, while -guo provides an existential quantification over times earlier than the clause’s reference time” (Mangione and Li 1993; 68). Thus, events occur just before the topic time indicated by le. This is directly relevant to narrative: as narrative time advances, RT advances to “just after” a bounded event (see Section 4). Lin (2003), considering le clauses in both dependent and independent contexts, proposes a relative anteriority parameter for le. In addition, le may express assertion, as Klein et al. (2000) argue. 16 This is only a partial characterization of -guo: there is a great deal more to say. For instance, Klein et al. (2000) argue that the time distinguished by -guo is a time at which the “result” of a situation is already past, whereas the English perfect distinguishes a time at which the result may still be obtained.
Temporal Interpretation in Mandarin Chinese
(10)
317
The aspectual morphemes of Mandarin: le: bounded event, if telic not necessarily completed, included within SitT; at RT; -guo: bounded prior situation; included within SitT; before RT; zai: unbounded event in progress; surrounding SitT; at RT; -zhe: unbounded situation; surrounding SitT; at RT; Ø zero: situation wholly or partially included within SitT; at RT.
These summaries focus on boundedness and relation to temporal location only. We now give a more formal statement of the semantic meaning of clauses with aspectual viewpoint morphemes.17 We need semantic meanings that are flexible enough to underlie pragmatic interpretation. The formal meanings stated below deal only with the temporal and aspectual; they do not cover the distributional range of the morphemes or the subtle additional meanings that arise in certain contexts. We assume a DR theory approach in which construction rules interpret syntactic structures. Syntactic structures license the entry of entities and other information into the developing representation of a discourse, the discourse representation structure (DRS). Let us say that a clause consists of a syntactic structure φ that expresses a situation e, and an aspectual viewpoint morpheme. A structure φ licenses the entry of a situation entity e into the DRS, with the properties focused by the aspectual morpheme that appears. Recall that the aspectual morphemes make all or part of a situation visible for semantic interpretation. Aspectual morphemes also indicate a relation between situation and reference time; we symbolize situation time as t1 , reference time as t2 . There are three parts to the formal statements below. We state first the relation between t1 and t2 that is conveyed by the aspectual morpheme; for all but -guo, t1 = t2 . The situation entity e expressed by a clause φ is associated with the appropriate situation type, by providing that a given e belongs to the class of situations with the properties of the appropriate event or state situation type (E or S). In some cases, e is related to the complete notion of the event, e’. Additional information about e varies with the viewpoint. Perfective viewpoints focus situations with visible endpoints, so that all of e is included in the SitT (t1 ); symbolized as e ⊆ t1 . Imperfectives focus situations without endpoints, so that e overlaps SitT; symbolized as eot1 . The neutral viewpoint focuses the initial endpoint of a situation and, for durative situations, at least one internal stage. The part focused is included in situation time (t1 ⊆ e); other parts, if any, are not included. This partial information allows for the indeterminacy of the neutral viewpoint; I(e) indicates the initial endpoint of an event. (11) Semantic meaning of the Mandarin aspectual viewpoint morphemes a. φ le is true iff t1 = t2 & ∃e ∃ e [e ∈ E & e ≤ e & e ⊆ t1 ] 17 We
thank Bernhard Schwarz for useful discussion of the ideas formalized here and for his help with the actual formalization.
318
C.S. Smith and M.S. Erbaugh
b. c. d. e.
φ -guo is true iff t1 < t2 & ∃ e [e ∈ E/S & e ⊆ t1 ] φ zai is true iff t1 = t2 & ∃ e [e ∈ E & e o t1 ] φ -zhe is true iff t1 = t2 & ∃ e [e ∈ E/S & e o t1 ] φ Ø is true iff t1 = t2 & ∃e [e ∈ E/S & ε| e ⊆ t1 v e o t1 ]
We unpack these statements in words. Condition (a) says that φ le conveys the following information. (1) SitT = RT; (2) the entity e expressed by the clause φ le belongs to the class of event entities E, and e consists of all or part of the complete situation expressed as e ; this allows for bounded but incomplete telic situations; (3) e is included in the moment or interval SitT (t1 ). Condition (b) says: φ -guo conveys that SitT precedes RT; that the entity e expressed by the clause φ le belongs to the class of event or state entities E/S, and that e is entirely included in the moment or interval t1 . Conditions (c) and (d) differ only in that zai is available for events, -zhe for events and states; in both e overlaps t1 . Condition (e), for the zero morpheme, says: SitT = RT; the entity e belongs to the class of state or event entities E or S, and SitT (t1 ) includes all or part of e. This statement is very weak: it accounts for the indeterminacy of the neutral viewpoint, since e may or may not extend beyond SitT. It is supplemented by the default temporal schema principle in (12) below. Situation type is realized by the verb constellation (see Smith 1997 [1991], 1994 for discussion of situation types in Mandarin). The main situation type categories are events and states. Telic events have natural endpoints, atelic events have arbitrary potential final endpoints. States and atelic events differ, though neither is telic (c.f. Smith 1999). Atelic events such as walking and chuckling are only partly homogenous—“only down to a certain size” as Taylor (1977) comments. Instantaneous events consist of a single point; states have no endpoints. The category of state includes generalizing/habitual and generic sentences. Such sentences may have the basic verb constellation of an event or a state. They are semantically stative, expressing a pattern rather than a particular situation. As noted by Krifka et al. (1995), generalizing stative clauses have habitual adverbs or allow them to be added without changing the sense of the clause. The adverbs include zong (‘always’), mei tian (‘every day’), and changchang (‘often’). Generalizing clauses can also be rephrased as mei ci X, Y (‘whenever X, Y’). Event verb constellations may appear with all four of the aspectual morphemes, though they frequently appear without any of them. State verb constellations may have -guo or -zhe only rarely and under very restricted conditions (Chao 1968: 664– 668). Zero-marked clauses have a consistent default interpretation. The default is this: verb constellations that express telic and/or instantaneous events are taken as bounded; state verb constellations are taken as unbounded; activity events (run, cook) are unbounded. Strikingly, these interpretations accord with the temporal schema of each situation type. They are the basic-level interpretations of events and states, and thus the ones interpreted as default. Explicit temporal information can override the default. The default interpretation of zero-marked clauses appears as a pragmatic principle in our interpretive account.
Temporal Interpretation in Mandarin Chinese
319
(12) Temporal schema principle: In a zero-marked clause, interpret a verb constellation according to the temporal schema of its situation type, unless there is explicit or contextual information to the contrary. Note that these interpretations are in accord with the bounded event constraint. Zeromarked clauses are very common; a study of Mandarin narrative finds that telic clauses with neutral viewpoint are the norm (Yang 2002: 83, 117). An earlier version of this account is given by Smith and Erbaugh (2001). In a related approach, Bohnemeyer and Swift (2004) posit a notion “event realization” which triggers a “default aspect” viewpoint: perfective for telic events and imperfective for atelic events and states; Lin (2003) adopts the idea. There are some differences between this view and our own. “Default aspect” interprets zero-marked clauses in terms of particular aspectual viewpoints, whereas our approach is more flexible. The temporal schema principle gives the default interpretation of zeromarked clauses, using the properties of situation type schemas. But we also provide that additional information in a clause can override the default. In certain contexts, zero-marked clauses may be interpreted as bounded (perfective) or unbounded (imperfective).18 Context may give information which pragmatically suggests a perfective or imperfective interpretation. For instance, if a telic event is expressed in a context suggesting the present, it may be taken as future. Or, in answer to a question that explicitly expresses an ongoing event, or in the scope of a present adverbial, a zero marked telic event clause might have an ongoing present interpretation. This is especially relevant since Mandarin prefers to answer a question with a verb, rather than with an equivalent to ‘yes’ or ‘no’: Women (jintian) qu bu qu? Qu. ‘We (today) go not go? Go’; in more colloquial English: ‘Are we going (today)? [We’re] going.’ The default principle must be related to the semantic meaning of a zero-marked clause. The temporal schema principle is a default. It allows that zero-marked clauses have a flexible interpretation: depending on context, zero-marked event clauses may be taken as open or closed. We assume that the information conveyed by a clause is interpreted by construction rules from syntactic surface structure, and introduced into a semantic represen18
There are several contexts in which le is obligatory if one wishes to convey the meaning of boundedness. For instance, a single syllable verb without a resultative complement conveys an ongoing situation without le. (i) a. Wo chi
vs. wo chi le. I eat vs. I eat LE. I’m eating vs. I’ve eaten. b. Wo pao vs. Wo pao le. I run vs. I run LE. I’m running vs. I’ve run.
320
C.S. Smith and M.S. Erbaugh
tation such as that of discourse representation theory. The situation type of a clause is composed from the verb constellation and associated with a temporal schema. Rules license the entry of situation entities in the representation, together with conditions stating their temporal properties. Aspectual morphemes, including the zero morpheme, appear under the aspect node in the syntactic tree. They license the introduction of times and conditions that specify their interpretation into an ongoing discourse representation structure (Smith 1997 [1991]; Kamp and Reyle 1993). The interpretation is determined by the semantic meanings of the relevant morphemes and verb constellations, and by the pragmatic principles stated above. 3.2.2 Lexical Information: Verbs and Modal Verbs Lexical information relevant to temporal location is conveyed by resultative verb complements, by past- and future-oriented verbs, and by modal verbs. We list here some of the main factors that make a difference in interpretation. Resultative verb complements (RVCs) appear with many event verb constellations. RVCs are verb suffixes that add lexical and aspectual meaning to the verb constellation. RVCs such as -wan (‘finish’) convey that the final endpoint of a telic event has been reached. Examples, with RVC underlined, include Ta zuowanle gongke (‘She has finished her homework,’ literally ‘do-finish’). There are also “result” RVCs. They specify the result state of an event and convey that the state was reached, as in dapole huaping (‘break the vase,’ literally ‘hit-break’), lachangle shengzi (‘stretch the rope out long,’ literally ‘pull-long’), Zhangsan shasi ta le (‘Zhangsan killed him,’ literally ‘kill-dead’). Other RVCs are directional, specifying the direction of action laxia lai (‘pull down,’ literally ‘pull down come’). RVCs of completion or result give explicit information of boundedness. Both Yang (2002) and Christensen (1994) find that most main clause narrative events occur with resultative verb complements.19 Temporal information is also conveyed by future-oriented verbs such as jihua (‘plan to’), zhunbei (‘prepare for, plan to’). These verbs are modal in nature. Their complements express attitudes and plans, projected and unrealized situations; the meanings are only indirectly temporal. However, the temporal strand affects the interpretation of complement clauses. The default interpretation of a future-oriented verb complement is unrealized or future. It overrides the pragmatic principles stated above. (13) illustrates; the main verb zhunbei (‘prepare for, plan to’) expresses a present state. Superscripts indicate the type of situation expressed by a clause, event (E) or state (S); complements are given in square brackets. 19 Yang
(2002) finds that a resultative complement strongly favors inclusion of le in foregrounded clauses, while a directional complement strongly favors omitting it. Single syllable verbs and main verbs without a verb complement are much more likely to include a le. In sequences of events, le is often omitted in the first clause, but included in the last. Telic verbs strongly favor inclusion, atelic verbs typically omit le. Non-native speakers, in contrast, are often insensitive to the importance of RVCs and overly attentive to adverbs and le (Yang 2002: 247).
Temporal Interpretation in Mandarin Chinese
321
(13) Wo zhunbei tian hei yihou wo he Zhangsan yiqi lai.S[E] I plan sky dark after I and Zhangsan together come. ‘I plan to come with Zhangsan after it gets dark.’ The verb zhunbei is future-oriented: it locates a clausal complement at a future time as unrealized. Pragmatic inference suggests that the event was realized, since there is no information to the contrary. The example is from Hu et al. (2001). Future-oriented verbs appear in various contexts. In the scope of a past adverb they convey a future time from a past RT, a future-in-past. The fragment in (14), taken from a magazine article published in 2000, illustrates. We are interested in the clause of line (14d). The verb jueding (‘decide’) is a future-oriented verb with an unrealized complement; it is in the scope of the adverbial phrase in (14a) that specifies a past time. In this context, jueding has the interpretation of future-in-past. We note that this and other textual examples below do not offer the clarity and simplicity of constructed examples. They are often quite complex and redundant, as is true of many actual texts. However, we believe that this makes them more interesting and relevant than constructed examples, not less so. Temporal location is very rarely ambiguous in written Mandarin. (14) a. Jiujiu nian jiu yue er ri, 1999 September 2 day, b. Lianxiang zhaokai xinwen fabu hui xuanbuE cong jiu yue Lianxiang hold news briefing announce from September yi ri qi 1 day begin c. zhengshi jiepin yuan zong gongcheng shi Ni Guangnan official dismiss original chief engineer Ni Guangnan academy yuanshi,E scholar d. bing jueding jiaogei Zhongguo Kexue Yuan wu bai wan yuanE and decide transfer to Chinese Science Academy 5,000,000 RMB. [‘On September 2, 1999, the Lianxiang Company held a press conference and announced the dismissal of its original chief engineer, Ni Guangnan, a member of the Chinese Academy of Sciences, effective September 1. It was also decided to transfer 5,000,000 RMB to the Chinese Academy of Sciences.’]
In this past context, it is natural to infer that the decision was carried out and the transfer of funds made, that is, that the event was realized. Classes of verbs with a temporal component include verbs of future having, for example, yuzhi (‘advance’), baozheng (‘guarantee’); verbs of future situation, for example, tiyi (‘propose’), jihua (‘plan to’), zhunbei (‘prepare to’); verbs of wish and desire, for example, yao (‘want’), wangxiang (‘covet’), xuyao (‘need’), xiangwang (‘yearn to’); verbs of future events, for example, yugao (‘predict’), xiang (‘expect’), yuyan (‘foretell’); verbs of future prevention, for example, paichu (‘preclude’), zuzhi (‘prevent’); c.f. Levin (1993).
322
C.S. Smith and M.S. Erbaugh
Modal verbs also indicate futurity and affect temporal interpretation.20 The modals hui (‘can’), yao (‘want, should, need, will’), and jiang (‘will’) have a clear temporal component: they convey a future time for the complement, as default. They differ somewhat: yao is primarily subjective, volitional; hui is predictive; jiang is used for scheduled, planned situations. The other epistemic and deontic modals tend to have a component of futurity, including ying (gai) (‘should, must’), keneng (‘possibly, may’), keyi (‘may’). In (15i), (15ii), and (15iii), the modals yao, jiang, and ying appear; yao and ying have complex complement clauses. The modals are underlined. (15i) . . . zhengfu he minjian yao you zugou de kongjian he shijian shenru . . . government and people should have enough space and time taolun . . .S[S[E]] to deep discuss. . . . ‘. . . the government and the people should have enough space and time to thoroughly discuss . . .’ (15ii) a. Guotai sihu yijing renshi dao zhege yinsu de Cathay Pacific appear already realize this factor DE zhongao xing,S[E] importance, b. jinnian jiang zhaomu ke zengjia zhi sishi ba renS[E] this year will recruit number additional 48 people [‘Cathay Pacific appears to have realized how important this is. This year it will recruit 48 additional people.’]
(15iii) a. . . .ciwai, zhengfu geng ying kaolyu touzi . . .in addition, government even more should consider invest b. chuangban geng duo de zai zhiyuan gong zixiu daxue, S[S[E]] establish even more DE at worker work self improvement university, c. jin keneng ling geng duo ren jieshou gaodeng jiaoyu. S[E[E]] maximum possible enable even more people receive higher education. [‘. . .in addition, the government should consider investing even more in self-study universities for workers, and to make maximum possible efforts to enable even more people to receive higher education.’]
The modals have a clear temporal effect, locating the complement clause in the future. (In this context, keneng is not a modal: it means ‘to the maximum extent.’) The futurity of these modals overrides the temporal schema principle,which would
20 In
Mandarin, futurity is typically modality, Hu et al. (2001) comment. This is true of many languages, c.f. Stassen (1997).
Temporal Interpretation in Mandarin Chinese
323
apply since the clauses are zero-marked for aspectual viewpoint. We ignore dynamic modals such as neng (‘can, able to’), which are internal to a proposition (Lyons 1977) and not future-oriented. Explicit conditionals, for example, ruguo (‘if’), jiaru (‘if’), jishi (‘even if’), also indicate a future, unrealized interpretation. There is a great deal more to be said about the contribution of modals to texts; but this brief overview suffices for our purposes. Future-oriented modals affect the temporal location of a complement clause.
3.2.3 Adverbs Mandarin has several types of temporal adverbial. “Locating” or “frame adverbs” (‘Tuesday’) form a large class of adverbs that temporally locate situations. These adverbs specify a time or an interval more or less precisely, and may locate situations at times other than the default. Their scope often extends beyond the clause in which they appear. In the fragment below, an adverbial clause specifies a past event, which serves as temporal anchor locating the events and states expressed in the clauses that follow. (16) a. . . .zai yijian BEI foujue hou,E Ni Guangnan bian kaishi xiang . . .at idea BEI veto after, Ni Guangnan then begin toward b. Zhongguo Kexue Yuan lingdao konggao zhuyao fuze ren,E(E) China Science Academy leader accuse important responsible people, c. neirong cong geren gongzuo zuofeng dao yanzhong de jingji wenti.S content from personal work style to serious DE economic problems. [‘. . .after his idea was vetoed, Ni Guangnan began taking accusations against major, responsible people to the leaders of the Chinese Academy of Sciences. The content (of his accusations) ranged from personal work style to major financial problems.’]
The adverbial clause in (16a) sets up a past context, supported in the actual text by explicitly past adverbs in earlier clauses. The stative in (16c) is anchored to the time of the adverbial clause. There are also adverbs that appear as connectives between clauses, including yihou (‘after, then’), jiu (‘then, at once’), cai (1. ‘just now,’ 2. ‘only then’). Adverbs such as gang (‘just now’), yijing (‘already’) also contribute temporal information, locating a situation before reference time, as we note above. Frequency adverbs such as changchang (‘frequently’) are relevant to aspectual situation type and thus affect temporal location. When they appear in a clause, the clause is a generalizing stative rather than an event and taken as located in the present unless there is information to the contrary. Temporal adjectives such as xianshi (‘current’) and congqian (‘previous’) are also relevant to temporal interpretation.
324
C.S. Smith and M.S. Erbaugh
3.2.4 The Deictic pattern for Mandarin We state the forms that realize the deictic pattern in Mandarin, and then give examples of the main cases. The possibilities are arranged in terms of their relations to speech time. (17) The deictic pattern in Mandarin a. Present situations are unbounded, located at speech time; expressed by: i) Clauses with aspectual zai or -zhe; ii) Cases with state or activity verb constellations, zero-marked; iii) Generalizing/habitual clauses, zero-marked. b. Past situations are bounded, located before speech time; expressed by: i) Clauses with aspectual le or -guo and/or RVC; ii) Clauses with telic event verb constellations, zero-marked. c. Other cases: i) Unbounded situations in the past or future, ii) Bounded situations in the future. Expressed by: aa) Clauses with future modals; bb) Future-oriented verbs and expressions; cc) Clauses with, or in the scope of time adverbials or temporal connectives. The pattern follows the pragmatic constraints stated above. The bounded event constraint locates bounded events in the past. The temporal schema principle locates zero-marked states and activities in the present, zeromarked telic events and instantaneous events at a nonpresent time. Instantaneous events are past; they are intrinsically bounded since they consist of a single point. The simplicity principle of interpretation locates nonpresent situations in the past rather than the future. The statement of the deictic pattern constitutes a set of predictions about temporal interpretation in Mandarin. We now show that the predictions are borne out, presenting examples of each case. We begin with the temporal location of unbounded situations, located at speech time. There are four cases. The first is that of a situation with one of the imperfective morphemes, zai and -zhe. The examples of (18a) and (18b) illustrate: (18) Imperfectively marked situations: located in the present a. zai i. Xianggang de jiaoyu yijing guodu Hong Kong DE education already exceed-limit zhidu hua,S systematized
Temporal Interpretation in Mandarin Chinese
325
ii. yiqie dou zai zhaoban everything all zai indiscriminately-imitate heli hua de moshi.E “rationalization” DE model [‘Education in Hong Kong is already too systematized, everything is all about indiscriminately imitating some model of “rationalization.”’]
b. -zhe i. Tianjing zhong zhi you yi yan shuijingS Open courtyard middle only have one CL well ii. bie wu qita sheshi.S other not have additional install iii. Yanzhe bian fangzhe ji kuai xiyi ban,S edgeing side set zhe several CL wash board iv. bu da pingzheng de luanshi dimian shanyaozheS not very level DE cobblestone floor surface shine zhe v. yi dai you yi dai one generation another one generation zhumin suo tachu de guangze.S resident which tread DE glow [‘The courtyard has only one well in the center, no other one has been drilled. Several washboards are set down along its sides, where the rough cobblestone surface shines with a gloss trodden smooth by generation after generation of residents.’]
The second kind of unbounded event is a zero-marked, atelic, durative event, of the activity situation type. The examples illustrate. In (19i), provided by J.-S. Wu from a corpus, the events with xing (‘go’) and zou (‘walk’) in (b) and (d) are of this type, in small capital letters. In (19i) the events in (g) and (h) are zero-marked, and unbounded. In (19ii), from an essay on disco dancing in Shanghai, a zero-marked activity event appears in (h) with yaobai (‘sway’). (19i) Zero-marked activity a. Na ren dangran daxi ruokuangS yiwei That person certainly very happy mistakenly think zhaodao le daoluE find LE road b. genzhe zuji er xing dan zou dao houlaiE following footprints go but walk until later c. zhe zuji yuanlai shi ziji liuxia deS his footprints originally is self leave DE d. ta zoulai zouqu zhishi zai rao quanzi.E he walked back and forth only ZAI go in circle
326
C.S. Smith and M.S. Erbaugh [‘That person was certainly very happy, (and he) mistakenly thought he found a road and went following some footprints. He walked until it was late. (It turned out that) the footprints were originally his own. He walked back and forth, he was only going in circles.’]
(19ii) a. Xiong de you tui zhizhu diban,S Bear DE right leg set floor b. zuo tui xiangqian kuachu yi bu.S left leg toward front stride one step c. Xiong de zuo shou pingshen xiang ce qianfang,S Bear DE left hand flat extended toward side forward d. shi zhi zuo gou zhuang, tiaodou zhuang;S index finger left hook position e. you bi quqi,S tease position; right arm crook up f. wu zhi zuo kai qiang zhuangS five finger make fire gun position g. zai disike jiezuo zhongS at disco rhythm inside h. Xiong kuangfang di qianhou yaobai,E Bear wild style DI front back sway i. Xiong chenjin zai ziji chuangzao de wudao zhong.S Bear deeply immerse in own create DE dance in [‘Bear’s right leg is set on the floor, his left leg striding one step forward. Bear’s left hand is flat out to the side, index finger hooked in a teasing pose. His right arm is twisted up, the five fingers imitating firing a gun. Inside the disco rhythms, Bear sways wildly back and forth, deeply immersed in a dance of his own creation.’]
Examples like this are rare in the discursive texts that we studied. However, clauses with zero-marked atelic events are common in conversation, where people are likely to talk about ongoing events. A printed text automatically imposes knowledge that the author wrote at some earlier time, making a past interpretation often the preferred one. In our texts, event verb constellations tend to appear with information that indicates a past time, for example, an RVC, a past adverb, le, or -guo. The next case is that of zero-marked states. We give an example from a newspaper editorial that appeared after the Seattle riots against the World Trade Organization. The clauses expressing states are zero-marked and located in the present, as predicted. (20) Clauses with state verb constellations, zero-marked; located in the present: a. Buguo, Xianggang mei you biguan zishou de tiaojian.S However, Hong Kong not have close self-self DE situation
Temporal Interpretation in Mandarin Chinese
327
b. Xianggang shi shijie shang ziyou du zui gao de Hong Kong is world on freedom degree most high DE jingji tixi.S economic system c. Miandui fada guojia gao keji chanpin, Confront developed nation high technology product d. yiji linjin diqu dicheng ben laodong long with neighboring region low-cost labor e. miji xing chanpin de liang mian jiaji,S intense type product DE two side pincer attack f. Xianggang de chulu zhi you yi tiao,S Hong Kong DE out road only have one route g. jiu shi nyuli xiang fada guojia kanqi,E that is make effort toward developed nation emulate h. jiasu fazhan gao keji yu zhishi xing speed up develop high technology and knowledge shaped jingji.E economy [‘However, Hong Kong does not have the option of closing its doors. Hong Kong has the freest economic system in the world. With high-technology products from developed countries and low-cost, labor-intensive products from its neighbors in a pincer attack, Hong Kong has only one way out: to make an effort to emulate developed countries and to speed up its high tech development and knowledgebased economy.’]
The last two clauses express unrealized events set in the future. The linguistic cue is nyuli xiang (‘make an effort toward’), a future-oriented expression. Zero-marked, generalizing statives are the third case. They are, as predicted, temporally located in the present—like other statives. We present three examples. (21i) contains an explicit frequency adverb; the clauses of (21ii) and (21iii) make generalizations about classes of individuals and entities—women and polystyrene containers. (21i) Generalizing statives, zero-marked, located in the present a. Fu tachu Luohu Guanka Just step out Lohu Immigration b. zai Shenzhen Huoche Zhan chukou zong hui kanjian rentou cuandong,GenS at Shenzhen railway station exit always can see people head assemble move c. qizhong bufa shenqing jiaoji de meiyan nyuzi qiaoshou shouhouGenS
328
C.S. Smith and M.S. Erbaugh
among them not lack look anxious DE beautiful gaudy woman raise head wait d. zi Shenzhen He bian qianlai from Shenzhen River other shore forward come de ailang.GenS DE lover [‘Just stepping out from Lohu Customs at the Shenzhen Railway Station exit, (you) can always see a moving crowd not lacking in anxious-looking, beautiful, gaudy women, raising their heads and waiting for their lovers to come across from the other shore of the Shenzhen River.’]
(21ii) a. . . . cun zhong zhuyao jiedao ji shiji kanlai hen pingjing,S . . . village within main street and market appear very peaceful b. yixie nyuzi xietong xiao tong zai lu pang zouguoGenS some woman in company child at roadside walk past c. huo zuo zai gongyuan liaotian,GenS or sit at park chat d. xiao hai ze zai pilin de qiu chang wanshua.GenS child then at adjoining DE ballfield play [‘. . . the village main streets and market look very peaceful, some women with their children walk past along the roadside or sit in the park chatting, their children playing in an adjoining ballfield.’]
(21iii) a. Fapao jiao canju fangbian,GenS Polystyrene food container convenient b. chengben diGenS price low c. dan weihai ji da,GenS but harm extremely great d. shou re chao guo liushi du jiu huiGenS receive heat exceed RVC 60 degrees then can e. shifang chu you hai zhi ai dusu,GenS release out have harm cause cancer poisonous substances f. you hui zaocheng wenshi xiaoying,GenS also can create greenhouse effect g. pohuai diqiu shengtai.GenS damage earth ecology [‘Polystyrene containers are convenient and cheap, but extremely harmful. Polystyrene releases carcinogens when heated above 60 degrees Celsius; it can also create a greenhouse effect and damage the earth’s ecology.’]
Temporal Interpretation in Mandarin Chinese
329
Although they have event verb constellations, the clauses are semantically stative because they involve a pattern of events rather than a particular event. No temporal location information is given. The interpretation is present, in accord with our predictions. We now turn to the expression of bounded events. The deictic pattern predicts that such events are located in the past. In the first case, the clauses are explicitly marked as perfective. Since the examples are from actual texts, they contain several kinds of information. (22i) has an RVC as well as a le; (22ii) has a past adverb, a date, a RVC, and a le; (22iii) has an adverbial hou (‘after’), as well as an RVC. (22i)
Clauses with RVC, perfective morpheme le Gong’an Fa shiqu le ke xinfeng de Public Order Ordinance lose RVC le can believe in DE qiyue jingshen.E contract spirit. ‘The Public Order Ordinance lost all spirit of a believable binding contract.’
. . . zhengfu zao zai 77 nian tongguo le . . . government early in 1977 year approve RVC le guwen gongsi de huanjing baogao.E consult company DE environmental report ‘. . . as early as 1977 the government had already approved a consultant’s report on the environment.’ (22iii) a. . . . zhe shi wo he duo wei nianqing xuezhe jiaotan this be I and many CL young scholar exchange-talk houE after (22ii) a.
b. suo dedao de jielun.E which reach RVC DE conclusion ‘. . . this is the conclusion which many young scholars and I reached after exchanging views.’ In our texts, many event clauses are zero-marked: only a few have aspectual morphemes. We cannot discuss the reasons for such variation here, but we note that the presence of an aspectual viewpoint morpheme highlights a clause. Yang finds that redundant temporal marking is the norm: in his study, 67% of foregrounded clauses in narrative are explicitly marked with an adverb (51%) or both an adverb and le (16%). Only 9% of the clauses are marked with le alone (Yang 2002: 162). The next case is that of zero-marked clauses with telic event verb constellations. There are many examples of such clauses in our texts, usually with a past adverb. We first illustrate zero-marked event clauses without an adverb, in (23i)(e) and (23ii)(b).
330
C.S. Smith and M.S. Erbaugh
(23i) Zero-marked telic clause a. Bieyou dong tian de Fengxian Si Dalu Shena xiangkan shiS Unique cave (open to) sky DE Fengxian Temple Locana Buddha niche be b. Chang’an Shiji Si Shandao chanshi, Fahai Si zhu, Chang’an Shiji Temple Shandao monk, Fahai Temple Abbot, Hujian fashi, Huijian daoist monk, he Si Nong, Wei Ji, deng ren and Minister of Agriculture, Wei Ji, and others c. fengming send down orders d. zhuchi take charge of e. kaizhi de lutian moyan xiangkan.E decorate DE open-air carved cliff image. [‘The Locana Buddha Niche at the unique temple of the Fengxian Si is an openair niche, an image (of Buddha) which was carved in an open air cliff under the orders and direction of the Monk Shandao of the Shiji Temple of Chang’an, of Daoist Monk Huijian, the Abbot of the Fahai Temple; and Weiji, the Minister of Agriculture, as well as others.’]
(23ii) a. . . . wo fei dan xin li zanchengS I not only heart inside approve b. erqie ye ceng weiwenE but also before write article c. xiang huanqi Xianggang ge jie renshiS[E] want rise Hong Kong every circle people d. dui Xianggang jingshen wenming de zhuyi. toward Hong Kong intellectual cultivation DE attention. [‘. . . I not only approve, but I, too, in past wrote articles calling on Hong Kong people from every walk of life to pay attention to intellectual cultivation.’]
The adverb ceng (‘before, once, in past’) in (23ii) reinforces the past interpretation. The fragment in (24), which has a past adverbial with scope over several clauses, is more typical of what we found in our texts.
Temporal Interpretation in Mandarin Chinese
331
(24) Zero-marked telic clause with past adverb a. . . . zai huihuang niandai, Wang Jizhi in glorious era, Wang Jizhi faming de Zhongwen da zijiE invent DE Chinese word processor b. rang Sitong feng guang wu xian. . . . allow Sitong prospects without limit [‘. . . in that glorious era when Wang Jizhi invented the Chinese word processor, (this) let the Sitong Company (gaze toward) limitless prospects . . .’]
Example (23i) is from an illustrated guidebook to the Buddhist grottos of Luoyang; in such a context, locating an event in the past is sufficient for understanding. (24) appears in a news magazine article, where more information about temporal location is appropriate; context locates “that glorious era.” Finally, there are cases of temporal interpretation that run counter to the predictions of the deictic pattern—they consist of unbounded situations in the past or future, and bounded situations in the future. The interpretations are based on explicit information in the immediate context: modals, future-oriented expressions, and adverbials. We present here one example of each type, organized according to the type of situation that is temporally located; see also examples (13), (14), (15) and (16) above. The fragment in (25) illustrates states located in the recent past within the scope of the time adverbial jinnian (‘in recent years’). (26) expresses an unbounded event in the past. The fragment is from a novel; very few instances of unbounded events appeared in our discursive texts. (25) States in the past a. Jinnian, Shenzhen jingji jisu fazhan,E Recent years, Shenzhen economy rapidly develop, b. Xianggang he Shenzhen liang di jiaotong fangbian,S Hong Kong and Shenzhen two place transportation convenient, c. renliu riyi pinfan,S people exchange increasingly frequent, d. jia shang xiaofei yu shenghuo zhishu shuiping add on consumption and living index level chaju . . .S difference . . . [‘In recent years, as the Shenzhen economy developed rapidly, transportation between Hong Kong and Shengzhen became convenient, the flow of people increased, and adding on the gap in standards of consumption and quality of life. . .’]
332
C.S. Smith and M.S. Erbaugh
(26) Unbounded event in the past a. Wo yi tou gongzuo,E I one head work, b. yi tou wenzhe yi zhen yi zhen guo li de yu one head smell zhe one burst pan in DE fish xiang, S fragrance, c. huang yu zai women Anhui shi mei you de . . .S yellow fish in our Anhui be not exist DE . . . . [‘One part of me was working, one part was smelling wave after wave of the fragrant smell of the fish in the pan. Yellow croaker fish was something we didn’t have back home in Anhui . . .’]
The zero-marked form gongzuo (‘work’) and imperfective wenzhe (‘was smelling’) express an ongoing activity event. These clauses could also be translated in the present, but as they occur in a reminiscence-based novel, they are conventionally taken as past. The next examples illustrate states and events in the future; both have an explicit future form. The future modals are stative. Square brackets indicate their complements. (27) Bounded events in the future a. Neidi jiangyu ben zhou liu quanmian jin yongS[E] Mainland about to this Saturday completely forbid use b. jiyong jiqi de fabao canju disposable DE polystyrene food-containers . . .’ ‘This coming Saturday the mainland will completely ban use of disposable polystyrene food containers . . .’ (28) State in the future a. . . . buguo zai siren qiye qiu cun de yali xiaS however in private business survive DE pressure under b. zhe jiang shi bu ke kangju de da qushiS[S[E]] this will be not possible resist DE great trend
... [‘. . . however, as private businesses are under pressure to survive, this great trend will be impossible to resist . . .’] ...
Examples like these show that Mandarin has linguistic resources for the full range of temporal locations. We comment briefly on the incidence of aspectual morphemes in the six texts that we studied most intensively. There were few: only 49 of the total of 7,640 characters were aspect morphemes. One editorial contained only one le, and no other aspectual morphemes; another editorial contained 20, or 40%, of all aspect morphemes. Together, the six articles contained 39 le but just 2 progressive zai, 4 -zhe, and 5
Temporal Interpretation in Mandarin Chinese
333
experiential -guo. This is an average of one aspectual morpheme every 14 clauses, or 7% of clauses. These texts had fewer aspectual markers than appear in narratives, certain other genres, and probably in ordinary conversation.21 We emphasize that the default interpretations are just that: defaults, which may be overridden by contextual information. For instance, consider the following constructed examples which do not fit. Assume that both (29a) and (29b) are uttered in a present context; note the definite zhege (‘this’) in (29a) and mei ge (‘every single’) in (29b). (29) a. Wo chi zhege pingguo. I eat this apple. ‘I’ll eat this apple.’ b. Mei ge Zhongguo ren dou dai Mao Zhuxi Every CL Chinese all wear Mao Chairman xiangzhang. badge. ‘Every single Chinese is/was wearing a Chairman Mao badge.’ The event in (29a) is telic; we may interpret the sentence as a future situation, in which the particular apple will be eaten. The situation in (29b) is a generalizing stative; one takes it as referring to a period in the past, assuming that the proposition the sentence expresses is true. It is knowledge of Chinese history that determines the interpretation, as Professor L. J. Xu, to whom we owe these examples, points out. We suggest that sentences like this do not appear out of the blue, but typically have contextual support, including temporal information. Therefore, they do not undermine our claims of default interpretation. We have shown how deictic temporal interpretation is realized in Mandarin. We now widen our discussion to include other cases.
4 Two Other Patterns of Temporal Interpretation Although deixis is basic in language, the full range of temporal interpretation includes patterns other than the deictic. Two others (at least), “continuity” and “anaphora,” are required. They are found in text passages of particular “discourse modes.” The discourse mode, proposed in Smith (2000a) and developed in Smith 21 There
are various studies of the incidence of aspect morphemes in texts. In a recent study of the corpus of Mandarin texts of the Academia Sinica, Wu (2003) found more aspect morphemes in report and fiction texts than in texts of commentary and other genres. A rhetorical comparison of Chinese and English newspapers in Beijing and Hong Kong also finds extensive variation (Scollon et al. 1994). Christensen (1994) found aspect morphemes in 36% of the intonation units, which generally correspond to clauses, in written Chinese narrations of Chafe’s ‘Pear Stories’ experimental film. Erbaugh (1990) examined spoken Mandarin narratives based on the ‘Pear Stories’; in the most representative spoken narrative, 26% of intonation units contain aspect morphemes.
334
C.S. Smith and M.S. Erbaugh
(2003), characterizes passages of written texts according to linguistically-based features. Five modes are proposed: narrative, report, description, information, argument.22 Each mode introduces certain types of entities into the universe of discourse, and follows one of three principles of temporal progression. The main types of entities are eventualities (states and events), generalizing statives, and abstract entities. The analysis of discourse modes is developed for English: we will show that the same temporal principles apply to Mandarin. The similarity suggests that the discourse modes are realized in Mandarin. We cannot go into the matter more fully (see Smith 2003 for a detailed account of discourse modes). The temporal modes, narrative, description and report, progress by different temporal principles, discussed immediately below. Narrative primarily introduces event and state entities into a discourse. The situations are temporally related to each other according to their aspectual properties, and information specified by adverbs. In narrative, tense conveys continuity. Description primarily introduces states and ongoing events into the universe of discourse and temporally locates them at a single established time. Tense in description conveys a time anaphoric to a previously established time. Report primarily introduces events and states, and general statives; they are temporally related to speech time. In report, tense is deictic; this is the default, found in the atemporal modes as well. The deictic pattern in Mandarin is discussed above; we focus here on other patterns of temporal interpretation. They too are found in passages of Mandarin texts. In our informal study, we looked at novels and short stories, as well as informative texts such as travel and museum guides. We begin with narrative. Narrative consists of consequentially related events, recounted in the sequence in which they occur—the main story line (Labov and Waletzky 1967; Moens 1987). The essence of a narrative is dynamism: as the events of the narrative unfold in sequence, we understand that narrative time advances. The dynamism that advances narrative time is due to aspectual information, and to explicit time adverbials and inference. A narrative advances with bounded, perfective events and with explicit, temporal adverbials. It fails to advance with states and ongoing events unless additional information warrants an inference of advancement. In narrative, situations are related to each other; the information that tense conveys after the initial sentence of a narrative is continuity. Narrative dynamism is modeled with a mechanism based on reference time. Time in narrative advances when warranted by aspectual or adverbial information. A narrative begins with a given RT. With a subsequent clause that expresses a bounded event, RT advances to just after the event. This principle delivers a result similar to the analysis of Klein et al. (2000), in which bounded events immediately precede RT. With states or ongoing event clauses, RT does not change: the situations are located at the current RT. For a clause with an adverbial, RT advances to the time specified by the adverbial. Additional information can license an inference
22 The discourse modes are proposed for written texts. They do not include conversation, procedural
texts, and some other types.
Temporal Interpretation in Mandarin Chinese
335
of advancement. Rules for narrative interpretation in the framework of discourse representation theory are provided in Kamp and Reyle (1993) and Smith (2003); the two differ slightly. The same interpretation of narrative dynamism is found in Mandarin, with adjustments for the resources of the language. This is not surprising. Narrative is surely universal, and narrative advancement depends on aspect and temporal adverbs. Information of both kinds is available in Mandarin, of course. Bounded events are the key to narrative advancement. But in Mandarin, unlike English, event clauses need not be explicitly perfective. Narrative time is advanced by clause with perfective le and/or RVCs, and by zero-marked event clauses. No new mechanisms are needed to provide for this. By the temporal schema principle, zeromarked event clauses are taken as bounded. With the principle that bounded events advance narrative time, this predicts the dynamic interpretation of narrative. Clauses with -guo do not advance time: they are stative. Clauses with the imperfectives zai, and -zhe, and zero-marked clauses with state verb constellations, do not advance narrative time. We state the interpretation as a principle for narrative dynamism in (30). (30) Narrative dynamism principle: a. Explicitly bounded event clauses, and zero-marked event clauses, advance narrative time; State and unbounded event clauses do not advance narrative time. b. If e is a bounded event, RTx immediately preceding: RTx < RTy;e at RTy This statement predicts the dynamics of narrative (Kamp and Rohrer 1983; Partee 1984)—an additional statement is needed for adverbial advancement. A more formal version of this principle mentions the types of entities typical of narrative passages (Smith 2003). Like other pragmatic principles, this one can be overridden by explicit information.23 We give an example from a fictional narrative and a short narrative that appears in a discursive text. (31) illustrates the stark narrative pattern, without supporting adverbials. The time is conventionally past. The order of presentation and aspectual information convey the information necessary to interpret the relations between the situations. All the clauses are zero-marked, invoking the temporal schema principle: telic event clauses express bounded events, atelic and state clauses express unbounded situations, unless there is information to the contrary. The events are numbered as they advance narrative time.
23 In
addition to bounded events in sequence—the default interpretation for narrative—there are other possibilities that are not considered here. Events may be overlapping or simultaneous, there may be a change in level of detail, situations may be notionally rather than temporally related. Recognition of these different temporal and atemporal relations is pragmatically based.
336
C.S. Smith and M.S. Erbaugh
(31) Fictional narrative a. Ta hao rongyi ququ zhezhe de huichu shou lai,E1 He with difficulty turn turn twist twist DE pull out hand come b. shou li jiu you yige xiaoxiao de changfang bao, kuilyu se de,S hand in just have a small small rectangular package sunflower green DE c. yi jing di gei Si Taitai.E2 one movement give to Si Mrs. [‘With difficult twists and turns he pulled his hand out of his clothing. It held a small rectangular package of sunflower green. In one movement, he handed it to his wife.’]
The first event—“pulling” the hand—is located at a contextually licensed reference time, RTn. In context, this RT follows RTn–1, the RT licensed by the previous event. The state—there is a package—is also located at RTn. The second event, handing over the package, follows at RTn+1. Narrative passages often appear in nonfictional, discursive texts. The following example is from an essay on the Shanghai Hotel. (32) Nonfictional narrative a. Yige shou nanzi laidao “Hua Ting.”E1 A thin male come to “China Room” b. Ta duoshao you dian huangkong bu an di zoujin haohua de boli men,E2 He more or less have a bit terror uneasy DI go in luxurious DE glass door c. chuanguo shui sheng zongzong de da tang,E3 pass RVC water sound gurgling DE great hall, d. zai P.R. shiying sheng, zong tai fuwu yuan de aimei de zhushi xia,E4 at P.R. worker, main desk service worker DE muddled DE attention under, e. tajin er lou de chenghu xing de zoulang.E5 step into second floor DE arc shape DE corridor. f. Ta kanjian le zoulang shang de yi pai zhaopianE6 . . . He see RVC LE corridor above DE a row photographs [‘A young man came to the “China Room.” Terrified and uneasy, he walked through the luxurious glass door, passed through the main lobby, which was filled with the sounds of gurgling water, and, under the muddled attentions of the young P.R. workers and reception desk personnel, stepped into the arc-shaped corridor of the second floor. He saw a row of photographs hung along the corridor . . .’]
The events are related to each other sequentially in these examples. Note that only the last event of (32) has an aspectual morpheme, perfective le with kanjian (‘seeRVC,’ ‘perceive’). We now turn to a temporal pattern which is neither dynamic nor deictic. In this case, interpretation is static: situations are temporally located, but the time does not change. The static pattern is typical of the discourse mode of description. In
Temporal Interpretation in Mandarin Chinese
337
passages of description, time is stable or suspended, without dynamism. Situations are located at a time already established in a text so that all the clauses of a given passage have the same reference time. Description passages appear in fiction, travel writing, and accounts of states of affairs. In the description mode, tense is anaphoric to a time already established in the discourse. The entities in description are unbounded events and states. Often a locative adverb or other anchoring information appears at the beginning of a description. We also assume a tacit durative time adverb that has scope over the passage. Event verb constellations in passages of description are atelic, sometimes by coercion due to the tacit time adverb. If motion is involved, it is without significant changes of state and there is no sense that time advances.24 The pattern is stated in (33). (33) Static interpretation principle a. States and unbounded events are located at an established time b. If e is a state or unbounded event, RTx immediately preceding: RTy = RTx We illustrate with two examples. The fragment in (34) begins with an event that establishes the prior RT; this is the time of the description that follows. Recall that states surround their situation time, so that the interval during which a state holds is not necessarily limited to a given SitT. Pragmatically, we infer that the states in (34) continue before and after the RT established by the event. (34) Description a. . . .zuo zhongren de Wei Lao Pozi dai ta jinlai le,E go-between DE Wei Old Wife bring her enter come LE b. tou shang zazhe bai tousheng, wu qun, lan jia ao,S head on braid ZHE white cord, black skirt, blue padded jacket c. nianji dayue ershi liuqi,S lian se qinghuang,S age about 20 six seven, face color pale d. dan liang jia que hai shi hong de.S but two cheeks still be red DE [‘. . .old Mrs Wei, who acted as go-between, brought her in. Her hair was tied with white bands, she wore a black skirt, blue jacket and pale green bodice, and was about twenty-six years old, with a pale face but rosy cheeks.’]
The description in (35) is from an account of the great carved Buddha at Luoyang. The RT is present, simultaneous with speech time. 24 Occasionally, telic verb constellations appear in a descriptive passage. These have an atelic value,
due to a regular process of coercion. Durative adverbials trigger a shift to an atelic situation, for example, She wrote a letter for an hour. The durative adverbial may be tacit. See Smith (2003) for discussion.
338
C.S. Smith and M.S. Erbaugh
(35) a. Dalu Shena Fo luoxing fa ji,S Locana Buddha snail-like hair coils b. shen pi tong jian shi jiasha, body draped whole shoulder katsaya patchwork vestmentS c. ta zuiba shao qiao, cheng weixiao zhuang.S his mouth slightly pulled upward, assume form of smile shape [‘The great Locana Buddha is robed in a kasaya vestment, his hair is coiled in spirals, and the corners of his mouth are tilted slightly upward in a smiling expression.’]
The continuing RT may be past or present—or, in principle, future—according to the context of the passage. In a sense this pattern complements narrative dynamism: it applies to unbounded situations and locates them at a single time. Three patterns of temporal interpretation have been demonstrated: the deictic, dynamic, and static. The deictic pattern is found in the discourse modes of report, information, and argument. Report passages give an account of situations and their significance from the temporal standpoint of the reporter. Passages in the other two modes are essentially atemporal. They, too, follow the deictic pattern. See the discourse representation theory account developed in Smith (2003).
5 Conclusion Temporal understanding in Mandarin, as we have shown, is based on information of two kinds: the semantic meaning conveyed by aspectual, lexical, and adverbial forms; and pragmatic principles of interpretation. The main pattern of temporal interpretation is deictic. The situations expressed linguistically are located relative to speech time: bounded events precede speech time, while unbounded events and states are located at speech time. The pragmatic principles that guide interpretation are the bounded event constraint, the simplicity principle of interpretation, and the temporal schema principle. Lexical and adverbial information can determine a temporal interpretation that departs from the default. Aspectual information of viewpoint and situation type is the key to the interpretation of temporal location. Aspectual viewpoints explicitly code whether a situation is bounded or unbounded. Zero-marked clauses, which are common, convey situation type and are interpreted by the temporal schema principle together with the other principles. Temporal interpretation is not limited to the default patterns, of course. Information in the context may override the default, providing the flexibility required. Three times are needed to account for temporal interpretation in Mandarin. Deixis has speech time as the main orientation point: the time of the situation expressed—situation time—is simultaneous with, before, or after speech time. There are grammatical forms that code the relation between a designated perspective time, or reference time, and situation time. These forms include the aspectual
Temporal Interpretation in Mandarin Chinese
339
viewpoint morphemes le and -guo, and adverbials such as yijing (‘already’) and cai (‘only then’). Relation to speech time is not coded linguistically, but is conveyed by context. We thus show that the Reichenbach approach to temporal interpretation applies to Mandarin. Two other patterns of temporal interpretation are also found: narrative dynamism and static description. They appear in text passages that realize the discourse modes of narrative and description. In narrative, time is dynamic and situations are related to each other rather than to speech time. Bounded events advance narrative time. In description, time is static and situations are related to a time already established. The deictic temporal pattern holds for the other discourse modes—report, information, and argument. Text passages realize different discourse modes according to the type of aspectual entities that they introduce, and their principle of text progression. The resources of tensed and tenseless languages differ; but this inquiry suggests a basic similarity in the interpretations and the principles that underlie them.
Appendix: Sources of Text Examples Examples published in bilingual English-Chinese editions list the English title after a slash instead of in parentheses. The Ming Bao publishes an English translation of each day’s Chinese editorial on the following day.
(9), (26) From a novel: Zhou Tianlai (1997). Tingzi Jian Saosao (Sister Next Door). Hefei: Anhui wenyi chuban she. 50, 13. (14), (16), (24) Magazine article: Chen Lirong (2000). “Feng bao chuixi Zhongguan Cun (Stormy personnel conflicts rock Beijing’s Silicon Valley)”. Qianshao Yuekan (Frontline Monthly). February. 48–49. (15i), (22i) Editorial: Anonymous (2000). “Gang fu minjian yao zai ding gong’an qiyue / The Hong Kong government must renegotiate its regulations for public order”. Ming Bao (Ming Pao Daily News). 9, 10 October. (15ii) Editorial: Anonymous (2001). “Guotai jishi weixie bagong bu de ren xin / Cathay Pacific pilots have alienated the public”. Ming Bao (Ming Pao Daily News). 29, 30 June. (15iii), (20), (28) Editorial: Anonymous (1999a). “Gang fu xuanjin qu Xiyatu baoluan jiaoxun / Government should learn lesson from Seattle”. Ming Bao (Ming Pao Daily News). 12, 13 December. (18i), (22iii), (23ii) Magazine article: Li Oufan [Leo Lee] (2000). “Xianggang weihe zai chu bu liao da xuewen zhe? (Why can’t Hong Kong produce great scholars again?)”. Ming Bao Yuekan (Ming Pao Monthly). August. 21–22. (18ii) Book: Lin Jiashu (1995). Tulou yu Zhongguo chuantong wenhua (Hakka houses and Traditional Chinese Culture). Shanghai: Shanghai renmin chuban she. 92.
340
C.S. Smith and M.S. Erbaugh
(19ii) Essay: Wang Weiming (1996). 1990 zoujin xin kongjian (Entering a new space in 1990). In Yuwang de Chengshi (City of Longing). Shanghai: Wenhui chuban she. 33. (21i), (21ii), (25) Magazine article: Wang Ruizhi (2000). “Shenzhen er nai cun qing se youhuo (Love and sex temptations of Shenzhen’s ‘Village of Concubines’)”. Yazhou Zhoukan (Asiaweek). 28 August–3 September. 24–25. (21iii), (22ii), (27) Editorial: Anonymous (1999b). “Xianggang huan bao zheng ce ying da che da wu / Hong Kong must revise its environmental policy”. Ming Bao (Ming Pao Daily News). 28 December. (31) Short story: Lu Xun (1990 [1924]). Feizao (Soap). [www.yifan.com.] English version in Lu Xun (1990). Soap. In Diary of a Madman and Other Stories, Translated by William A. Lyell, 264. Honolulu: University of Hawaii Press. (32) Essay: Wang Weiming (1996). Shanghai Da Fandian (The Shanghai Hotel). In Yuwang de Chengshi (City of Longing). Shanghai: Wenhui chuban she. 60–61. (33), (35) Travel guide: Gong Dazhong (1986). Longmen Shiku / The Longmen Grottos, translated by Zhang Linjia. 42. Beijing: Zhongguo lyuyou chuban she. (34) Story: Lu Xun (1973 [1926]). Zhu Fu / The New Year’s Sacrifice. Hong Kong: Chao Yang Publishing Company. 20–23. Acknowledgements We thank Dr. Qing Wu and Ms. Hsi-Yao Su for assistance in preparing the texts that were used in this study. We also wish to thank the audiences who have heard one or both of us talk about the material, including the annual meeting of the Society for Text and Discourse, Lyons, France, in July 2000; the conference Linguistics in the Next Decade, Academia Sinica, Taipei, Taiwan, in August 2000; the Third Texas Workshop on Text Structure in October 2000; and the First International Conference on Modern Chinese Grammar for the New Millenium, Hong Kong, in March 2001. We also thank two anonymous reviewers for their comments. [Editors’ note: We thank Mary Erbaugh for proofreading this chapter. She has made a few corrections to the glosses and references that appeared in the original publication. Lastly, we thank Barbara Partee for noticing an error in the formalism in example (11e) and for suggesting a correction. That correction has been made.]
References Austin, John (1961). Philosophical Papers. Oxford: Oxford University Press. Bach, Emmon (1981). On time, tense, and aspect: an essay in English metaphysics. In Radical Pragmatics, Peter Cole (ed.), 66–81. New York: Academic Press. Bohnemeyer, Juergen, and Swift, Mary (2004). Event realization and default aspect. Linguistics and Philosophy 27(3), 263–296. Chao, Yuen Ren (1948). Mandarin Primer: An Intensive Course in Spoken Chinese. Cambridge, MA: Harvard University Press. Chao, Yuen Ren (1968). A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of California Press. Christensen, Matthew B. (1994). Variation in spoken and written Mandarin narrative discourse. Unpublished doctoral dissertation, Ohio State University.
Temporal Interpretation in Mandarin Chinese
341
Depraetere, Ilse (1995). On the necessity of distinguishing between (un)boundedness and (a)telicity. Linguistics and Philosophy 18, 18–19. Erbaugh, Mary S. (1990). Mandarin oral narratives compared with English: the Pear/Guava Stories. Journal of the Chinese Language Teachers Association 25(2), 21–42. Giorgi, Alessandra and Pianesi, Fabio (1997). Tense and Aspect. Oxford: Oxford University Press. Grice, H. P. (1975). Logic and conversation. In Speech Acts, Peter Cole and Jerry L. Morgan (eds.), 41–58. New York: Academic Press. Hinrichs, Erhard (1986). Temporal anaphora in discourses of English. Linguistics and Philosophy 9, 9–82. Horn, Larry (1984). Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In Meaning, Form and Use in Context: Linguistic Applications, Deborah Schiffrin (ed.), 11–42. Washington DC: Georgetown University Press. Hu, Jianhua, Pan, Haihua, and Xu, Liejiong (2001). Is there a finite-nonfinite distinction in Chinese? Linguistics 39(6), 1117–1148. Huang, C. T. James (1984). On the distribution and reference of empty pronouns. Linguistic Inquiry 15, 15–574. Huang, L. Meei-jin (1987). Aspect: A general system and its manifestation in Mandarin Chinese. Unpublished doctoral dissertation, Rice University. Jespersen, Otto (1931). A Modern English Grammar on Historical Principles. Book 4: Syntax. London: George Allen and Unwin. Kamp, Hans and Reyle, Uwe (1993). From Discourse to Logic. Dordrecht: Kluwer Academic Publishers. Kamp, Hans, and Rohrer, Christian (1983). Tense in texts. In Meaning, Use and Interpretation of Language, R. Bauerle, R. Schwarze, and A. von Stechow (eds.), 250–269. Berlin and New York: Mouton de Gruyter. Kanizsa, Gaetano (1976). Subjective contours. Scientific American 234(4), 48–52. Klein, Wolfgang (1994). Time in Language. London: Routledge. Klein, Wolfgang, Li, Ping, and Hendriks, Henriette (2000). Aspect and assertion in Mandarin Chinese. Natural Language and Linguistic Theory 18, 773–770. Krifka, Manfred, Pelletier, Francis J., Carlson, Gregory, Meulen, Alice ter, Chierchia, Gennaro, and Link, Godehard (1995). Genericity: an introduction. In The Generic Book, G. Carlson and F. J. Pelletier (eds.), 1–124. Chicago: University of Chicago Press. Labov, William, and Waletzky, Joshua (1967). Narrative analysis: oral versions of personal experience. In Essays on the Verbal and Visual Arts: Proceedings of the 1996 Annual Meeting, American Ethnological Society, Jane Helm (ed.), 12–43. Seattle: University of Washington Press. Levin, Beth (1993). English Verb Classes and Alternations. Chicago: University of Chicago Press. Levinson, Stephen (1983). Pragmatics. Cambridge: Cambridge University Press. Lin, Jo-Wang (2003). Temporal reference in Mandarin Chinese. Journal of East Asian Linguistics 9, 259–311. Lyons, John (1977). Semantics. Cambridge: Cambridge University Press. Mangione, L., and Li, Dingxuan (1993). A compositional analysis of -guo and -le. Journal of Chinese Linguistics 21, 21–122. Moens, Mark (1987). Tense, aspect, and temporal reference, Unpublished doctoral dissertation, Centre for Cognitive Science, University of Edinburgh. Partee, Barbara (1984). Nominal and temporal anaphora. Linguistics and Philosophy 7, 243–286. Reichenbach, Hans (1947). Elements of Symbolic Logic. London: Macmillan. Scollon, Ron, Scollon, Suzanne Wong, and Kirkpatrick, Andy (1994). Contrastive Discourse in Chinese and English (Han Ying Bian Zhang Duibi Yanjiu). Beijing: Foreign Language Teaching and Research Press. Shi, Zhiqiang (1990). Decomposition of perfectivity and inchoativity and the meaning of the particle le in Mandarin Chinese. Journal of Chinese Linguistics 18, 18–124. Smith, Carlota S. (1997 [1991]). The Parameter of Aspect. Dordrecht: Kluwer Academic Publishers.
342
C.S. Smith and M.S. Erbaugh
Smith, Carlota S. (1994). Aspectual viewpoint and situation type in Mandarin Chinese. Journal of East Asian Linguistics 3, 3–146. Smith, Carlota S. (2000a). Temporal information in sentences. Paper presented for Linguistics in the Next Decade. Academia Sinica, Taipei, Taiwan, August 2000. Smith, Carlota S. (2000b). Discourse mode: a linguistically interesting level of text structure. In Proceedings of the Third Workshop on Spoken and Written Texts, 12–34. Austin: Department of Linguistics, University of Texas. Smith, Carlota S. (2003). Modes of Discourse. Cambridge: Cambridge University Press. Smith, Carlota S., and Erbaugh, Mary S. (2001). Temporal information in sentences of Mandarin. In Hanyu Yufa Yanjiu de Xin Tuzhan—21 Shiji Shaojie Xiandai Hanyu Yufa Guoji Yantao Hui [New Views in Chinese Syntactic Research—International Symposium on Chinese Grammar for the New Millenium], K. K. Luke, Shao Jingmin, Shan Zhourao, and Xu Liejiong (eds.), 514–542. Hangzhou: Zhejiang Jiaoyu Chuban she. Smith, Carlota S.; Perkins, Ellavina, and Fernald, Theodore (2003). Temporal interpretation in Navajo. In Proceedings of the SULA Conference, Amherst Working Papers in Linguistics, Jan Anderssen, Paula Menéndez-Benito, and Adam Werle (eds.), 175–192. Amherst, MA: GLSA. Stassen, Leon (1997). Intransitive Predication. Oxford: Oxford University Press. Taylor, Barry (1977). Tense and continuity. Linguistics and Philosophy 1, 199–220. Vendler, Zeno (1957). Verbs and times. The Philosophical Review 66, 143–160. Reprinted in Vendler, Zeno (1967). Linguistics in Philosophy, Ithaca, NY: Cornell University Press. Whorf, Benjamin L. (1956). Grammatical categories. In Language, Thought and Reality: Selected Writings of Benjamin Lee Whorf, John B. Carroll (ed.), 87–111. New York: Wiley and Son. Wu, Jiun-Shiung (2003). Modeling temporal progression in Mandarin: aspect markers and temporal relations. Unpublished doctoral dissertation, University of Texas. Yang, Jun (2002). The acquisition of temporality by adult second language learners of Chinese. Unpublished doctoral dissertation, University of Arizona. Yeh, Meng (1996). Analysis of experiential GUO in Mandarin Chinese: a temporal quantifier. Journal of East Asian Linguistics 5, 5–182.
Some Significant Omissions: Ellipses in Flaubert’s Un Coeur Simple Jeanne T. Whitaker and Carlota S. Smith
WHAT IS LEFT OUT OF A TEXT can be as important as what is explicitly present. Readers interpret omissions according to principles of inference that are generally assumed by participants in communication. This essay discusses omissions, or ellipses, and their literary effects in Flaubert’s Un Coeur simple. When we read:
elle pleura en écoutant la Passion. Pourquoi l’avaient-ils crucifié, lui qui chérissait les enfants, nourrissait les foules ...
We know that the question refers to Christ, even though his name is not mentioned in the text. Flaubert’s allusions to the Bible rely on the assumption that his readers have the background knowledge required for their interpretation. All texts require some interpretation, even at the relatively straightforward level of determining what reference is being made. People are able to determine references partly because of a body of shared assumptions and expectations. Equally important, however, are principles of inference that are used almost automatically by participants in communication. This point is made in a compelling way by the recent work of H. P. Grice.1 Grice formulates a set of conversational principles that underlie much of communication. Through these principles he elucidates the patterns of reasoning used by speakers and hearers. At the basis of Grice’s analysis is what he calls the cooperative principle, a general agreement of cooperation between the speaker or writer and the hearer or reader: the latter may assume that the former are being informative, truthful, relevant, clear, and organized in their effort to communicate. When the principle seems to be violated, it will be for good reason. Grice gives as an example the following letter of recommendation for an applicant seeking a lectureship in philosophy: “Dear Sir, Jones’s command of English is excellent, and his attendance at tutorials has been regular. Yours faithfully,” which the recipient will surely construe to imply that Jones has no other qualifications for the post. The process by which this conclusion is reached involves understanding of conver-
1 Grice’s lectures are now available: Hubert P. Grice, “Logic and Conversation,” in Syntax and Semantics, 3: Speech Acts, ed. P. Cole and J. Morgan (New York: Academic Press, 1975), pp. 41–58. They have produced much interesting discussion in the linguistic and philosophical literature.
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_14,
349
350
J.T. Whitaker and C.S. Smith
sational implicature. Metaphors and irony, and other complex uses of language, may depend partially upon implicature. Our discussion of Flaubert’s story is based on a theory of how pronouns and other inexplicit references are interpreted, a theory in which principles of inference are essential. Following Grice, we assume that the principles are cooperative and are flexible enough to allow for the indirect and the unexpected. Flaubert often violates the expectations of the reader. We will argue that violating expectations has rhetorical value, and that such value plays an important role in Un Coeur simple. We use the term “ellipsis” very generally to indicate a word or phrase that requires inference, including the construction of additional material, to be fully interpreted. Different types of omission share an important feature: they require inferential work on the part of the hearer or reader. We will show how ellipses are interpreted by making use of different types of information, both in the text and beyond it. We discuss three different types of ellipses and their effects in this essay: pronouns without a clear antecedent, definite noun phrases that are not explicitly identified in the text, and phrases that require the inference of additional material through implicature, as proposed by Grice. Our theory of interpretation of the first two types is outlined below; we follow the Gricean analysis of implicature. Although pronouns do not have a unique referent (unlike names, for instance), we usually understand the intended referent of pronouns in a given context without difficulty. A natural explanation is that general principles govern the interpretation of pronouns, and these principles are assumed by speaker and hearer. For single sentences, linguists have identified rules relating pronouns and possible antecedents; for example. compare (i) and (ii): (i) John saw Bill and then he saw Dick. (ii) He said that John was working hard. In (i), we might assume that he refers to John (unless the pronoun has contrastive stress), but in (ii) he could not be interpreted as coreferential with John. In a more extended discourse or written text, we believe that the main factors that determine the interpretation of a pronoun as having a particular antecedent are contiguity and discourse focus. In the first case, a pronoun is near or contiguous to a given noun phrase, and by a principle of local cohesiveness is interpreted as referring to the same entity as that noun phrase. Alternatively, a pronoun can have as antecedent something referred to earlier and of general importance in a text; such an entity is in the forefront of consciousness2 and is a natural referent in that text. In this case we shall say a pronoun is related to its antecedent by a principle of discourse focus.3 2 Chafe uses this phrase in a discussion of the notion “topic.” Wallace Chafe, “Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View,” in Subject and Topic, ed. C.N. Ki (New York: Academic Press, 1976), pp. 25–55. 3 There are, of course, other properties of discourse organization that affect the interpretations of pronouns.
Ellipses in Flaubert’s Un Coeur Simple
351
Normally, discourse focus takes precedence over contiguity, as we demonstrate in discussing examples from this story. The interpretation of a given pronoun is generally supported, or corrected, by information in the text4 : particular lexical items and comments about the thoughts or activities of a character, for instance, give essential information. It is important to notice that contiguity and discourse focus also constitute information that is essential to the interpretation of a text. Definite noun phrases—that is, noun phrases referring to definite entities—are sometimes introduced in a text without explicit antecedents.5 To make sense of the text, the reader must infer the referent. Some definite noun phrases do not need explicit antecedents because they refer to entities that are unique and well-known in a culture, such as the sun or the sea. Unique references also may be tied to specific situations. Knowing the type of situation often enables the hearer to interpret ellipses, and the speaker counts on this ability. This type of reference involves “a collection of knowledge about a stereotyped situation,” and has been widely discussed under the general heading of “framing.”6 For example, after Flaubert has introduced the topic of the first communion, he does not need to identify the book or the rosary. There are also definite noun phrases that can be interpreted with information from the text, such information often being supplemented by general knowledge. In all these cases the reader infers and/or constructs a referent; as with pronominal reference, the connections are normally made almost automatically. We will show that several types of information are involved in constructing a referent, and through detailed discussion of examples we will make explicit the chains of connections. The third type of ellipsis we will consider is implicature. In conversational implicature information is conveyed by word meaning together with assumptions about relevance and completeness. For example, we interpret the sentence, “Elle résolut de le porter elle-même à Honfleur” as meaning she actually did take something to Honfleur, since there is no information to the contrary.7 If there were such information in the text the implicature would be cancelled, in Grice’s terms. Implicatures have been discussed extensively by Grice, who notes that they are common in texts of all kinds.
4 We
do not include here pronouns that refer without an antecedent, although such pronouns occur often, especially in conversation (one may designate a person in full view of the hearer, for instance, simply by a pronoun). 5 In this essay “definite noun phrase” is a noun phrase that refers to a definite entity. It has a semantic rather than a strictly syntactic definition, since so-called indefinite forms can refer definitely. See, for instance, the use of the indefinite article in reference to moths cited later in this essay. 6 For interesting discussions of this notion, see Eugene Charniak, “With Spoon in Hand This Must Be the Eating Frame,” in Theoretical Issues in Natural Language Processing (TINLAP), 2, Association for Computational Linguistics, 1978, pp. 187–193; Marvin Minsky “A Framework for Representing Knowledge,” in The Psychology of Computer Vision, ed. P.H. Winston (New York: McGraw-Hill 1975), pp. 211–277. 7 This sentence comes from Un Coeur simple, paragraph 176.
352
J.T. Whitaker and C.S. Smith
In the appendix we discuss in detail the interpretation of a number of elliptical constructions. In the body of the essay we wish to demonstrate that ellipsis is important as a literary device in Un Coeur simple. Not only does it focus and advance the narrative in a particular way, but it also contributes to the tone of the work and to the reader’s apprehension of a particular point of view.
1 Point of View in Un Coeur simple When Félicité, the heroine of Un Coeur simple, dies, she is blessed with a marvellous apparition, at once sublime and ridiculous: elle crut voir, dans les cieux entr’ouverts, un perroquet gigantesque, planant au-dessus de sa tête.
The cautious “crut voir” stresses the subjectivity of the vision, but does not make it less convincing, for the whole narrative has progressively prepared us to accept it. Un Coeur simple is the tale of a long, relatively uneventful life, told in very few pages. Flaubert communicates simultaneously the routine of Félicité’s daily existence and the depth of experience underlying it, the successive losses of love she suffers and the growth of love in her heart, her limitations and her greatness. From the largest to the smallest structures of the story, Flaubert shows the paradox of Félicité’s life. In the construction of the narrative there is little chronological suspense. We know from the very first sentence that Félicité spent most of her life as a servant in one household.8 Similarly, the first sentence of the second part, “Elle avait eu, comme une autre, son histoire d’amour,” tells us at once, though indirectly, through the verb tense and the dismissive comparison, about the outcome of this love affair. This sacrifice of the unfolding of adventure allows Flaubert to substitute the unfolding in depth of Félicité’s experience, even as its scope grows narrower. In the first section, we see her in her socially defined role, as a model servant; in the second, we are told of her attachment to her only suitor, Théodore, then, after she is hired by Mme. Aubain, to the latter’s daughter, Virginie, and later to Victor, son of a long-lost sister. The third section begins with Virginie’s first communion; Félicité’s religious feeling intensifies as she identifies with the child. Victor and Virginie leave and soon they both die. While Mme. Aubain falls into depression, Félicité’s active generosity multiplies. In the fourth section she receives Loulou, the parrot, to whom she transfers the best of her love, and whom she identifies with the Holy Ghost, particularly when, after his death, she stuffs him and keeps him in her room. Thus she progresses, at the same time, from a solid, practical, efficient life to a destitute and grotesque death, and from the material world to the spiritual.
8 This point is discussed by Raymonde Debray-Genette, “Les figures du récit dans ‘Un Coeur simple,’ ” Poétique, 3 (1970), 349.
Ellipses in Flaubert’s Un Coeur Simple
353
On the smallest scale, indeed in his very grammatical constructions, Flaubert elicits our sympathy for Félicité and makes us feel what she feels, without explicitly suggesting that we should. Moreover, he does so without sentimentality or condescension. Among the various means he uses to engage readers’ interest and assent, we focus on the elliptical constructions listed in our introduction. Many paragraphs have more than one type of ellipsis, so it is difficult to study these constructions separately and yet retain the flavor of the text. We will discuss first a few passages in which these grammatical structures reveal Félicité’s imagination and emotions in a particularly interesting manner. Félicité’s identification of the Holy Ghost with her parrot is described in some detail. She notices that the picture of the Holy Ghost she sees in church avait quelque chose du perroquet. Sa ressemblance lui parut encore plus manifeste sur une image d’Epinal, représentant le baptême de Notre-Seigneur. Avec ses ailes de pourpre et son corps d’émeraude, c’était vraiment le portrait de Loulou.
In these last two sentences, only the possessives (“sa ressemblance,” “ses ailes de pourpre,” “son corps d’émeraude”) evoke the Holy Ghost. To grasp the import of these sentences, knowledge of the biblical description of Christ’s baptism, in which “the Holy Ghost descended in a bodily shape like a dove upon him,” is required.9 The association is intensified in the following sentence: L’ayant acheté, elle le suspendit à la place du comte d’Artois,—de sorte que, du même coup d’oeil, elle les voyait ensemble.
The use of the masculine pronoun, “le suspendit,” indicates she has bought “le portrait de Loulou” rather than “une image d’Epinal.” The reader is given her vision of the picture. To understand the rest of the sentence fully, one has to remember three or four paragraphs back the description of Félicité’s room when, after a long agonizing period of waiting while the dead parrot is off in Le Havre being stuffed, he is finally returned to her: il y avait des fleurs artificielles au bord de la commode, et le portrait du comte d’Artois dans l’enfoncement de la lucarne. Au moyen d’une planchette, Loulou fut établi sur un corps de cheminée. . . .
When “le portrait de Loulou” replaces “le comte d’Artois,” it is next to the stuffed Loulou, thus “elle les voyait ensemble.” The plural pronoun fuses them for the reader, as they are fused in Félicité’s mind. This fusion draws us into Félicité’s mind. This fusion draws us into Félicité’s obsession, indirectly but powerfully, before the author explains it, as he does in the following sentence: Ils s’associèrent dans sa pensée, le perroquet se trouvant sanctifié par ce rapport avec le Saint-Esprit, qui devenait plus vivant à ses yeux et intelligible.
Throughout the third section, where Félicité’s faith is described, Flaubert evokes deep feelings indirectly through the grammatical form of his references; he rings subtle changes on the use of pronouns and possessives. Félicité accompanies 9 Luke
4:22.
354
J.T. Whitaker and C.S. Smith
Virginie to cathechism every day. She feels at home in church and her imagination is easily stimulated. Flaubert tells us so directly: “Elle croyait voir le paradis, le déluge. . . .” Biblical images sanctify familiar aspects of life for her: “elle aima plus tendrement les agneaux par amour de l’Agneau, les colombes à cause du Saint-Esprit.” Félicité’s feeling of familiarity is also communicated indirectly by the pronouns, as in the example cited at the start of this essay: Puis elle pleura en écoutant la Passion. Pourquoi l’avaient-ils crucifié, lui qui chérissait les enfants, nourrissait les foules. . . .
The fact that la Passion is a feminine noun makes the lack of antecedent in the pronoun reference to Christ more noticeable. Such reference assumes the reader’s recognition of Jesus’ life story, and also indicates in what familiar terms Félicité thinks of him. The quality of Félicité’s imagination and feeling is further manifested as she prepares for first communion with Virginie. Initially Flaubert is explicit: dès lors elle imita toutes les pratiques de Virginie, jeûnait comme elle, se confessait avec elle. A la Fête-Dieu, elles firent ensemble un reposoir [emphasis added].
The next paragraph expresses identification more indirectly: La première communion la tourmentait d’avance. Elle s’agita pour les souliers, pour le chapelet, pour le livre, pour les gants. Avec quel tremblement elle aida sa mère à l’habiller.
We know from the larger context that it is Virginie’s first communion that is worrying Félicité, that the shoes, gloves, rosary, and prayer book are for Virginie. The use of the definite article in such cases implies there is only one (or one set) of each object and that everyone knows which one. It also confers a particular importance to those objects. With the definite articles we adopt Félicité’s point of view. At the end of the following paragraph, this same technique is even more striking. Again Flaubert prepares us clearly: avec l’imagination que donnent les vraies tendresses, il lui sembla qu’elle était elle-même cette enfant; sa figure devenait la sienne, sa robe l’habillait, son coeur lui battait dans la poitrine.
The subject nouns and possessive adjectives refer to Virginie, the pronouns, personal and possessive, to Félicité. “La poitrine” is also evidently Félicité’s, given the indirect object pronoun that accompanies it. What follows is surprising: au moment d’ouvrir la bouche, en fermant les paupières, elle manqua s’évanouir.
Elle must still be, as at the beginning of the sentence, Félicité. However, it is Virginie who receives the wafer from the priest. We can surmise that Félicité so identifies with the child that she opens her own mouth and closes her own eyes. Normal French grammatical usage reinforces this reading: here the definite article refers to the eyes and mouth of Félicité. Yet the previous paragraph indicates Félicité’s concentration on the mouth and eyelids that alone count for her at the moment: Virginie’s. Finally, we may take it that both characters are doing the same thing at the same time: the definite articles serve to fuse them, to communicate Félicité’s complete identification with Virginie.
Ellipses in Flaubert’s Un Coeur Simple
355
After the description of Virginie’s death and burial, Félicité and Mme. Aubain are shown walking together in the garden speaking of Virginie. We are thus led to a description of the child’s room: Toutes ses petites affaires occupaient un placard dans la chambre à deux lits. Mme. Aubain les inspectait le moins souvent possible. Un jour d’été, elle se résigna; et des papillons s’envolèrent de l’armoire. Ses robes étaient en ligne sous une planche où il y avait trois poupées, des cerceaux, un ménage, la cuvette qui lui servait. Elles retirèrent également les jupons, les bas. . . .
We understand that Mme. Aubain was very reluctant to inspect her dead daughter’s possessions when we learn that she did so as seldom as possible, and that she had to resign herself. Flaubert does not tell us to what she resigned herself, nor does he mention what she in fact did: open the cupboard. But the two parts of the sentence are linked, by being together in one sentence, and by et. We infer that the moths flying out of the cupboard are the direct consequence of Mme. Aubain’s resignation. The suppression, in the narrative, of her action reflects her inhibitions. The flight of moths following her psychological surrender conveys visually and symbolically the passage of time, the neglect, the deterioration, the feeling of desolation that kept Mme. Aubain from opening the cupboard for so long. Mme. Aubain being the subject of the last two sentences, we are surprised by the next paragraph’s first words: “Ses robes étaient en ligne sous une planche. . . .” Ses robes are Virginie’s, as the context makes abundantly clear. The shift of subject reflects the mother’s obsession, as the reader is jolted into recognizing it. Flaubert notes the presence of Virginie’s toys on the shelf and of her dresses below, then adds: “Elles retirèrent également les jupons, les bas. . . .” Only également indicates that other things have already been taken out. In this phrase the plural subject elles reappears, taking us back to “elles se promenaient ensemble” two paragraphs above. Though Félicité has not been mentioned in the interval, we know she is in Virginie’s old room with her mistress, and is helping her clean out the cupboard. This fusion of the two women, unexpected at that moment, serves as a delicate preparation for the end of the paragraph: after finding a little moth-eaten hat of the child’s, which Félicité asks to keep, they dissolve into tears and fall into each other’s arms, feeling for the first time equal and as one.
2 Three Types of Ellipsis There are in Un Coeur simple many such passages in which material essential to the understanding of the text is omitted or presented indirectly. As stated in the introduction, the term ellipsis is used to describe three types of constructions: definite noun phrases without explicit antecedents (for example, “La première communion la tourmentait d’avance. Elle s’agita pour les souliers, pour le livre, pour les gants . . .”); pronouns whose interpretation is less than obvious (for example, “au moment d’ouvrir la bouche, en fermant les paupières, elle manqua s’évanouir”); and material introduced indirectly in the text that must be taken as asserted (for example, “Elles retirèrent également les jupons, les bas . . .”). In this section we show how
356
J.T. Whitaker and C.S. Smith
these three types of ellipses, and related stylistic features, contribute to the reader’s experience of the story. Detailed discussion of the inferences involved in particular examples will be found in the appendix.
2.1 Definite Reference Many definite references in Un Coeur simple are anchored by frames, particularly the frames of Catholic practice and bourgeois life in a small French town. When we read that, upon Mme. Aubain’s return to Pont-l’Evêque from the seashore, M. Bourais l’éclaira sur le choix d’un collège. Celui de Caen passait pour le meilleur. Paul y fut envoyé. . . .
we share the unquestioned bourgeois assumption that a young son must be sent away to a collège if there is none in the family’s hometown. Similarly, we are drawn into Félicité’s frame of reference when, after the scene where she stops by a calvary to pray, having just seen Victor off on a long voyage, we read: Le parloir n’ouvrirait pas avant le jour. Un retard, bien sûr, contrarierait Madame; et, malgré son désir d’embrasser l’autre enfant, elle s’en retourna.
At that point in the story, Flaubert has not yet stated that Félicité associates Virginie with Victor in her heart. But we know she loves the little girl and has missed her since her departure for the convent. We also know that a parloir is the place in an institution where boarders can receive visitors. Le parloir must be the reception room in which she could see Virginie. There are also many references that cannot be fully understood by someone unfamiliar with the town and terrain in which the story is set. In the first section of Un Coeur simple, Flaubert places us in Pont-l’Evêque without comment on its location, size, demography, economy, or any other generalization. He mentions “la ferme de Toucques” and “la ferme de Gefosses,” “sa maison de Sainte-Melaine,” “l’assemblée de Colleville,” “la route de Beaumont,” “la ferme des Ecots” without any elucidation for readers, who are forced to conclude that their existence in the real world is the basis of such precise, yet rapid reference to them. Such references connote familiarity as well as reality, and make those places appear important. Just as it seems unnecessary to say “All roads lead to Rome, Italy,” in this story it would be superfluous to explain that Pont-l’Evêque is a small town in the department of Calvados. Pont-l’Evêque is the center of this world and as such needs no explanation. Readers unfamiliar with local geography have to make some effort as they follow the text to distinguish between “la ferme de Toucques,” which the Liébards farm for Mme. Aubain, the village of Toucques, where Théodore lives after his marriage, and the river Toucques, in which Félicité washes her laundry, without leaving Pont-l’Evêque. Local people of course would have no trouble making these distinctions; in real places such multiple uses of the
Ellipses in Flaubert’s Un Coeur Simple
357
same name are common. The effect is to communicate a local point of view. Soon enough the reader who does not know Normandy realizes that the sense of locality created by this presentation matters a great deal, while the geographical details do not. If we were part of Flaubert’s fictional universe we would recognize all these landmarks; the presentation of them without explanation makes us part of that universe. Flaubert treats history as he does geography. Very precise dates are given, but they seem incidental. We do not know the dates of Félicité’s birth and death, but we do learn that Mme. Aubain’s husband died at the beginning of 1809, that she herself died in the month of March 1853. The July Revolution (of 1830) is announced one night, its only mentioned effect being the appearance in town of a new sous-préfet who brings Loulou among his retinue. In one sense this story is as specifically set in the first part of the nineteenth century as in Pont-l’Evêque. The economic and social conditions, the child-rearing and religious customs of a bourgeois household headed by a widow in a small northern French town in that period are taken for granted. When Mme. Aubain engages Félicité to work for her, they both assume it is to be a live-in and lifelong arrangement. The servant keeps constantly active, and the mistress spends her time knitting, playing boston once a week with regular visitors, and worrying about her children and her properties. The contrast between the two women contains an implicit critique of society at that time and place. As Félicité is a figure of primitive strength who transcends specific circumstances, the precise dates, rather than adding to our specific understanding, enhance our sense of the story’s historical reality.10
2.2 Pronominal Reference The understanding of pronouns often requires that the reader go beyond explicitly presented information. Some pronouns have no antecedents, others involve a shift of focus, others violate the expectations of a discourse. For example, in Félicité’s question as she thinks about the Passion, “Pourquoi l’avaient-ils crucifié, lui qui chérissait les enfants . . .” neither ils nor the group l’, lui, qui has an antecedent in the text. To interpret the sentence the reader must know the story of the crucifixion. Other cases are more subtle: what is involved is a choice between possible antecedents. As explained previously, in cases where there are two possible antecedents, one contiguous and the other not contiguous but the focus of the preceding paragraph, we believe that the normal pattern and expectation is for discourse focus to have precedence. Flaubert sometimes violates this expectation, producing the effect of a relatively unorganized chain of references. We find such an example at the beginning when the focus shifts from Félicité to her mistress.
10 A very interesting discussion of how Flaubert creates a sense of reality in this story was published
by Roland Barthes in Communications, 11 (1968), pp. 84–89 under the title, “L’Effet de Réel.”
358
J.T. Whitaker and C.S. Smith
The story opens: Pendant un demi-siècle, les bourgeoises de Pont-l’Evêque envièrent à Mme. Aubain sa servante Félicité. . . . Pour cent francs par an, elle faisait la cuisine. . . et resta fidèle à sa maîtresse,—qui cependant n’était pas une personne agréable.
This is followed by a paragraph beginning, Elle avait épousé un beau garçon sans fortune, mort au commencement de 1809, en lui laissant deux enfants très jeunes avec une quantité de dettes. Alors elle vendit ses immeubles. . . .
Since mistresses, not servants, sell property and have expectations of marrying into money, we realize that elle no longer refers to Félicité but to Mme. Aubain. This change of referent throws us off balance, forces us to pay careful attention, and coming as it does at the start of the story, puts us on notice that the author does not always fulfill our most habitual grammatical expectations. Félicité is brought back into focus in an equally indirect manner. Two long paragraphs describing Mme. Aubain’s house lead to Félicité’s room, then back into the routine of her daily existence: Une lucarne au second étage éclairait la chambre de Félicité, ayant vue sur les prairies. Elle se levait dès l’aube. . . .11
The expectation Flaubert uses most of the time, however, is the more common one, that of discourse focus. Once a character is established as the focus of the story, it seems normal and easy for readers to refer pronouns to that character over many paragraphs. In sections one and two of Un Coeur simple, from paragraphs 6–18, elle and other pronouns and possessives are the only references to Félicité. This passage begins toward the end of section one, “Elle se levait dès l’aube. . . et travaillait jusqu’au soir sans interruption” and continues with a description of her. The chain of pronouns is unbroken as section two begins: “Elle avait eu, comme une autre son histoire d’amour,” and proceeds with a one paragraph summary of her life history up to the age of eighteen, followed by a description of her courtship of Théodore. Throughout, she is clearly the one possible referent for the feminine pronouns. This use, to the exclusion of her name, reinforces our focus on her, but also makes her seem less individual, and confirms the view of her as someone who is defined by her social role, and whose love story is utterly commonplace. A further effect of these long sequences of pronouns used rather than Félicité’s name, or any other designation for her, is to make her seem more familiar: we take her for granted.12 She is central to the story, but not foregrounded. The third person 11 These
examples are discussed by Svend Johansen, “Ecritures d’Un Coeur simple,” Revue Romane, 2, No. 1 (1967), pp. 108–120. 12 The name of Madame Aubain appears 44 times in the text; she is called “Madame,” as Félicité would, 15 times; 15 times she is referred to by some other noun: “la maîtresse,” “sa mère,” “une bourgeoise,” or some collective noun. She is referred to 53 times by a third person singular pronoun and 10 times by a third person plural pronoun. Virginie is called by that name 22 times, and 27 times by other nouns, singular or plural, such as “fille,” “enfant,” “petite,” or “mademoiselle,” these
Ellipses in Flaubert’s Un Coeur Simple
359
pronouns serve to maintain a delicate balance between attachment and detachment in our attitude toward the protagonist. Félicité, obviously, would not be who she is if she talked about herself. To describe so inarticulate a person in words at all is a paradoxical enterprise. To reveal the strengths embedded in her limitations is to risk exaggeration or falseness.13 Any attempt to show her inner life could easily be over dramatized or sentimentalized, and thereby become much less convincing. Often Flaubert uses pronouns to refer to Félicité when he wants to present her thoughts indirectly; in other passages he uses them to describe her from the outside. In the paragraph following Virginie’s death that begins “La bonté de son coeur se développa. . .,” we find a rapid evocation of her generous actions toward many: soldiers, the sick, Poles, and finally the complete derelict, “le père Colmiche.” We are not told what Félicité feels or thinks, we simply see what she does. Here the use of the feminine pronoun rather than a name may at once reflect the individual woman’s self-forgetfulness, and make of her a more universal feminine figure, an archetypal female image.
2.3 Material Indirectly Introduced Some linguistic forms have a dual role in the presentation of information: they are part of one direct assertion and also function to make a second, indirect assertion on which the first is dependent. One understands that both assertions are added to the developing of the story or conversation.14 These forms occur frequently in discourse of all kinds, and we consider them part of the elliptical, indirect tools Flaubert uses in this narrative. Many adverbs and verbs that embed subject or object complements are of this type. A striking example occurs in one of the passages we discussed: “elles retirèrent également.” We know, because of the adverb également, that Mme. last two in indirect quotes. She is referred to by 37 singular and 13 plural pronouns. As for Félicité, her name is used 62 times, and other nouns only 6 times: “bonne,” “servante” (3 times), “jeune fille,” and “femme.” But we find 315 third person singular personal pronouns referring to her and 20 plural. Whereas all in all Félicité is mentioned four times as often as Virginie and almost three times as often as Mme. Aubain, the latter is called by a name or a noun more often than Félicité (74 times as opposed to 68), and the former considerably more than half as often (49 as opposed to 68). Comparing the number of third person pronouns, one finds that they are used for Félicité over 5 or 6 times more often than for Mme. Aubain or Virginie. 13 In an article entitled “Schizophrenia and the Mad Psychotherapist,” reprinted in his Ways of the Will (New York: Harper & Row, 1968), Leslie Farber discusses the tendency to overdramatic behavior on the part of some therapists working with nonverbal patients. An unsuspecting visitor to a sanatorium for schizophrenics “would have to conclude that . . . these therapists had apparently abandoned what must once have been their more usual habits of expression in favour of some more florid and declamatory style.. . .” Farber attributes this change to a desire to celebrate the nonverbal at the expense of the nuances and complexities of language. Writers can fall into the same trap. Flaubert obviously does not. 14 For use of presupposition in literary analysis, see Gerald Prince, “On Presupposition and Narrative Strategy,” Centrum, 1, No. 1 (1973), pp. 23–31; and Milton C. Butler, “Factive Predicates and Narrative Point of View,” Texas Linguistic Forum, 13 (1979), pp. 34–39.
360
J.T. Whitaker and C.S. Smith
Aubain and Félicité took other things out of the wardrobe besides those explicitly mentioned in the sentence. Another example occurs later when Félicité is knocked down on the road: Son premier geste, quand elle reprit connaissance, fut d’ouvrir son panier.
Here the main sentence is asserted and the adverbial sentence presupposed. By saying Félicité opened her basket when she regained consciousness, Flaubert tells us she lost consciousness and regained it. Detailed discussion of this type of example, and others, can be found in the literature on linguistics, particularly in the area of pragmatics and speech acts.15 Here we note simply that certain forms carry presuppositions, and that these forms introduce information indirectly. Knowledge of these presuppositions is part of speakers’ basic knowledge of their language. In Un Coeur simple such ellipses generally serve to quicken the pace of the narrative: the author focuses on part of the story, while making clear we are seeing only part. This is the main effect of the ellipses marking repetition, such as, “Il l’embrassa encore une fois.” “Mme. Aubain lui défendit de les baiser à chaque minute” is an interesting example because “lui défendit” alone does not necessarily suggest Félicité did in fact kiss anyone. But the frequency of her embraces would not have been mentioned if there was no basis for Mme. Aubain’s warning: “à chaque minute” may be an exaggeration, but it indicates Félicité’s physical affection for the children, and that Mme. Aubain was talking about an actuality.
3 Ellipsis and Narrative Effects The narrative of Un Coeur simple is often choppy: the author presents only a series of sharply focused details, leaving the reader to compose the total picture. At the center of the paragraph, which describes the most dramatic episode of the story, the scene of the encounter with a bull, we find a series of bold shifts of focus that involve sudden cuts in the narrative flow and unexpected grammatical sequences: un beuglement formidable s’éleva. C’était un taureau, que cachait le brouillard. Il avança vers les deux femmes. Mme. Aubain allait courir. —“Non! non! moins vite!” Elles pressaient le pas cependant, et entendaient parderrière un souffle sonore qui se rapprochait. Ses sabots, comme des marteaux, battaient l’herbe de la prairie; voilà qu’il galopait maintenant! Félicité se retourna. . . .
First we hear the terrible bellowing. Then we see the bull advance and Mme. Aubain’s flight reaction, interrupted by the sudden cry, “Non! non! moins vite!” unattributed in the text, but which must come from Félicité. We see the two women 15 For
a summary and discussion of presupposition, see Lauri Karttunen and Stanley Peters, “Requiem for Presuppositions,” Proceedings of the Berkeley Linguistic Society, Univ. of California, Berkeley, (1977), 360–371.
Ellipses in Flaubert’s Un Coeur Simple
361
together hurrying, and with them we hear heavy breathing coming closer. A sudden shift to another frightening noise is effected by a shift to another view of the bull, and from the indefinite un souffle to the possessive ses, as we hear “Ses sabots, comme des marteaux battaient l’herbe.” The menace of the images in this scene is strongly enhanced by the suddenness, and grammatical oddity, of their juxtaposition. As we have pointed out, the antecedent of a possessive need not immediately precede it. Following the principle of discourse focus, we understand that ses sabots are the bull’s and the focus of our attention on the bull is intensified. But we are then struck by the reference to un souffle sonore, rather than son souffle. The indefinite article dissociates the heavy breath from the animal. There are many other less dramatic but similarly cut scenes, which to a contemporary reader seem cinematic. One instance is the description of the trip to Trouville: les chevaux trottaient, l’âne galopait; tous enfilèrent un sentier, une barrière tourna, deux garçons parurent, et l’on descendit devant le purin, sur le seuil même de la porte.
The rhythm of the sentence reflects the rhythm of the action. In such passages, Flaubert’s style is mimetic. The drama of this rapid arrival at the farm is enhanced by the fact that nothing in the text prepares the reader to expect it. We have been told the family is going to Trouville, but there has been no hint that a halt was planned. We do not know where they are until la mère Liébard appears in the next sentence, though devant le purin localizes la porte on a farm. As we have seen, ellipses accelerate the pace of the narrative. When Félicité, after Loulou’s death, decides to send the parrot by boat to Le Havre to be stuffed, elle résolut de le porter elle-même à Honfleur. Les pommiers sans feuilles se succédaient au bord de la route.
Flaubert tells of her decision to go to Honfleur with the implicature of résolut. The definite noun phrase, la route, provides confirmation, and acceptance of this allows us to identify la route as leading to Honfleur. This is one more example of how a meshing of elliptical constructions contributes to the narrative pace of the story. Another salient effect of the ellipses is to make us identify with a character’s point of view. In the most striking examples we have studied—the description of the church, the first communion, the bringing together of Loulou and the Holy Ghost— unclear pronominal references and unidentified definite references express Félicité’s vision, and might even be considered iconic. In the scene where she goes through her dead daughter’s belongings, the ellipses reflect Mme. Aubain’s feelings. In the following example an ellipsis reflects the point of view of incidental characters: Et jamais elle [Félicité] ne parlait de ses inquiétudes. Mme. Aubain en avait d’autres sur sa fille. Les bonnes soeurs trouvaient qu’elle était affectueuse, mais délicate. La moindre émotion I’énervait. Il fallut abandonner le piano. Sa mère exigeait du couvent une correspondence réglée.
362
J.T. Whitaker and C.S. Smith
The point of view shifts between sa fille and sa mère. This shift is intelligible to the reader because of the intervening paragraph, which introduces the nuns’ point of view. It has been noted that such shifts are unacceptable in a single sentence16 ; they are difficult to follow between paragraphs. The difficulty one experiences with such passages is itself difficult to characterize, but clearly a conscious effort, a deliberate attention to understanding, is involved. This is unusual: normally the understanding of such things as who is being talked about is effortless and automatic. Certain ellipses require inferential work and conscious attention. The requirement of readers’ attention is also an effect of ellipsis in Un Coeur simple, and an important one. The most striking examples of ellipses occur at crucial points in the story: at the beginning, where the tone is set, and in the central emotional scenes. Ellipses are, in fact, a stylistic device of some importance. We believe that the difficulty and subtlety they contribute to the text are essential to readers’ appreciation, as essential as the mimetic effects of Flaubert’s elliptical style in passages such as the encounter with the bull or Félicité’s identification with Virginie during her first communion. In those scenes the ellipses reflect important aspects of the characters’ experience and draw us into it. In general, the interpretation of ellipses requires the cooperation of readers, and readers are drawn into the text by their cooperation. It is a flattering requirement: it engages readers’ attention, their sympathy, and ultimately their assent. If Félicité’s story seems neither hagiographic nor sentimental, if we come to believe in her, it may be because we have been required as readers to assemble the different elements of the story ourselves and to draw our own conclusions. The elliptical quality of Flaubert’s style has often been discussed, most notably by Proust, who declared that: A mon avis la chose la plus belle de L’Education sentimentale n’est pas une phrase mais un blanc.17
The blanc Proust admired is a large gap in narrative time “sans l’ombre d’une transition.” There are many strikingly cut scenes in Flaubert’s novels that have fascinated critics; the most famous is perhaps Emma and Léon’s carriage ride through Rouen in Madame Bovary.18 The ellipses we have considered here are neither so striking nor so mysterious. At this level of the text, the level of grammatical structure, one usually infers almost automatically what is not made explicit. The ellipses we discuss are common in conversation. Thibaudet makes a similar point in response to complaints that the author used incorrect French:
16 See Susumo Kuno, “Subject, Theme, and the Speaker’s Empathy,” in Subject and Topic, ed. C.N. Li (New York: Academic Press, 1976), pp. 417–444. 17 Marcel Proust, “A propos du ‘style’ de Flaubert,” Chroniques (Paris: Gallimard, 1927), p. 205. 18 Discussed by Jonathan Culler, Flaubert (Ithaca: Cornell Univ. Press, 1974), pp. 121–22.
Ellipses in Flaubert’s Un Coeur Simple
363
A l’origine du style de Flaubert, on voit une oreille extraordinairement ouverte aux nuances de la langue parlée.19
In a conversation there is usually a great deal of shared extralinguistic information that the participants can count on. In a written text references have to be explicitly anchored or available by inference. Readers are sometimes aware of making these inferences and sometimes not. Flaubert’s elliptical stylistic devices in Un Coeur simple make Félicité’s world and adventures at once very ordinary and quite surprising. Flaubert communicates this paradoxical vision in many ways. The plot, though static, represents a spiritual odyssey. Descriptions of the heroine vary in technique and tone. The last two paragraphs of the first section present the physical portrait of a stiff person who “semblait une femme de bois, fonctionnant d’une manière automatique.” Yet when Théodore abandons her, “Ce fut un chagrin désordonné”; when she finds herself in a household with two young children “elle les portait sur son dos comme un cheval” and kisses them constantly. Later Flaubert tells us that “La bonté de son coeur se développa.” Nevertheless, when one of the Poles she has befriended wants to marry her, she throws him out one morning because “elle le trouva dans sa cuisine, où il s’était introduit, et accommodé une vinaigrette qu’il mangeait tranquillement.” We receive a multiplicity of impressions and evaluations of Félicité. Flaubert’s masterly use of the style indirect libre,20 and his subtle choice of vocabulary, which indicates shifting voices and tones, give the story a delightful humor.21 As Félicité in church contemplates the Holy Ghost and observes “qu’il avait quelque chose du perroquet,” “c’était vraiment le portrait de Loulou,” we smile at her association, with perhaps a touch of superiority. The substitution of pronouns that we found in the following sentence has the effect of drawing us in to Félicité’s vision, of making us feel what she feels. Flaubert’s elliptical means of pulling us closer to his heroine interact with the indirect humor that maintains her at a distance, creating the tone of amused affection and ironic admiration that makes this story so distinctive. Acknowledgement We would like to thank Janet Letts and Richard Pearce for invaluable discussions; Isabel Kenrick, Gerald Prince, Michael Smith, and Bonnie Webber also have made helpful comments.
19 Albert
Thibaudet, Gustave Flaubert (Paris: Plon, 1922), p. 311. realize that the aspects of Flaubert’s style we are discussing often involve the style indirect libre. The large number of third person pronouns and possessives in the text, as well as their occasional uncertain references, surely result from this narrative strategy, as do the unanchored definite references. The interpretation of conversational implicature is assisted by our recognition of indirectly reported speech in a sentence such as “Mme. Aubain lui défendit de les baiser à chaque minute.” While acknowledging the links between our analysis and the studies of Flaubert’s style indirect libre, we do not feel we can, in the scope of this essay, treat all the relationships satisfactorily. 21 Culler discusses Flaubert’s verbal irony with admirable clarity in his Flaubert, pp. 185–207. 20 We
364
J.T. Whitaker and C.S. Smith
Appendix: Examples and Analyses We present here a set of exemplary cases from the text that involve the construction of a noun phrase or recognition of an unexpected referent (implicatures are omitted since Grice and others have dealt with them extensively). For each example we give the various types of information and inference that are relevant to its interpretation. This is a level of detail that is crucial to the understanding of the examples but could not have been given in the main body of the essay without marring the exposition. The first set of examples involve identification of the referent of an NP. In some cases there is one clear route to the correct interpretation; in others, more than one route is possible because more than one type of information can be drawn on. The interpretations are presented in terms of identification of the NP; we do not attempt to fill in the information that is missing from the text. To state what is missing would be misleading, if not impossible. There might be more than one way of arriving at an identification, and different routes might involve different omissions— stating “the” route would fix things too solidly. More serious, there is no unique set of omissions for a given route: events can be chunked together in different ways and can be broken down into many smaller events. We distinguish between actual words in the text that lead to identification, and general information from the text that does so. The two are not mutually exclusive: a lexical item may appear in a sentence that gives relevant information, and itself key identification. The NPs of these examples do not have anchors in the text, nor are they anchored by framing or unique reference. We took as anchors appropriate previous references; grammatically related possessives, genitives, and locatives; and certain transitive verbs of which the NP in question is an object. The examples below are organized according to the notional relation between the referent of the NP and the referent of related material in the text. In these examples we focus on the way material in the text leads to identification of the NPs. Part to whole: the definite NP may indicate a part, which must be identified as belonging to some whole. The text may give both but without making explicit the relation between them, as in example 2; or the text may give only clues as to the whole, as in examples 3 and 4. The first sample of text has two NPs needing identification; the text samples are numbered according to paragraph. (a) 130 Le couvent se trouvait au fond d’une ruelle escarpée. Vers le milieu, elle entendit des sons étranges, un glas de mort. “C’est pour d’autres,” pensa-t-elle; et Félicité tira violemment le marteau. 1. the knocker of the door of the convent (a) Text: the convent; (b) Information in text: Félicité’s destination is the convent; (c) General knowledge: convents are buildings with doors, and doors have knockers, and one uses a knocker to gain entrance to a building; (d) Inference: Félicité is now at the convent door.
Ellipses in Flaubert’s Un Coeur Simple
365
2. the middle of the alley where the convent is Text: the bottom of a steeply sloping alley; toward; (b) Information in text: Félicité’s destination is the convent; (c) General knowledge: rules of language (the same, roughly, for French and English) that relate the NP in question to the preceding NP; (d) Inference: Félicité is now in the middle of the alley.
Both these NPs have the same narrative function. They move the narrative forward by requiring the inference that a character, Félicité, has moved to the place of the definite NP. (b) 63–64 à partir de Noël, elle mena tous les jours la petite fille au catéchisme. Quand elle avait fait à la porte une génuflexion, elle s’avançait sous la haute nef entre la double ligne des chaises, ouvrait le banc de Mme. Aubain, s’asseyait, et promenait ses yeux autour d’elle. 3. the door of the church (a) Text: catechism, genuflect, nave, pew; (b) Information in text: Félicité went to the place where catechism is held; (c) General knowledge: catechism is often held in church, and nave, etc., are uniquely associated with church; the actions are appropriate to entering church.
This example, by assuming an identification that would be automatic for inhabitants of the town, presents their point of view to the reader. (c) 49–50 . . .les chevaux trottaient, l’âne galopait; tous enfilèrent un sentier, une barrière tourna, deux garçons parurent, et l’on descendit devant le purin, sur le seuil même de la porte. La mère Liébard, en apercevant sa maîtresse, prodiga les démonstrations de joie. 4. the door of the Liébard’s farmhouse (a) Text: manure pile, Mother Liébard; (b) Information in text: Liébard is a farmer and tenant of Mme. Aubain; (c) General information: farms have manure piles, and farmhouses have doors; (d) Inference: Mother Liébard is near the door.
Note that one could not be sure from the text just where Mother Liébard is: she might have been beside the door, waiting to welcome the travelers, or she may be inside the door. This example requires much of the reader because there is no preparation for it in the text. In 3, the notion of catechism is introduced just before the NP in question. But here the travelers have a final destination (known to the reader) that is not the Liébard farm, and there is no mention of an intermediate stop. Both 3 and 4 have clues in the text; the difference between them is a matter of how much prior information has been given. Another general relation that obtains between definite NPs and material in the text is that of a conventional connection: event B is generally known to follow event A. Such a connection can serve to identify B, if A is established clearly. There are several examples of this type of connection, all involving Catholic life, which was a frame taken for granted at the time.
366
J.T. Whitaker and C.S. Smith
(d) 69–71 Ce fut de cette manière, à force de l’entendre, qu’elle apprit le catéchisme, son éducation religieuse ayant été négligée dans sa jeunesse; et dès lors elle imita toutes les pratiques de Virginie, jeûnait comme elle, se confessait avec elle. A la Fête-Dieu, elles firent ensemble un reposoir. La premiere communion la tourmentait d’avance. Elle s’agita pour les souliers, pour le livre, pour les gants. Avec quel tremblement elle aida sa mère à l’habiller! Pendant toute la messe, elle éprouva une angoisse. . .. 5. the thought of the first communion Virginie will make (a) Text: catechism, in advance, the shoes, etc.; (b) Information in text: Virginie, who attends catechism, is the young daughter of the household where Félicité is a servant; Félicité is totally devoted to the child with no expectations for herself; (c) General knowledge: catechism leads to first communion, which is a solemn ceremony with much pomp; (d) Inference: Félicité is worrying about Virginie, not her own communion.
This fragment is less ambiguous in context than it seems here. The focus of the story on Félicité is well established, and her character is clear when this fragment occurs. 6. the shoes and gloves, etc., Virginie will wear at communion (a) Text: first communion; (b) Information in text: the family is small-town bourgeoisie; (c) General knowledge: one wears special shoes, etc., at a first communion, especially bourgeois little girls; (d) Inference: Virginie’s first communion is at hand.
This use of the definite article here may be compared to situations common in American life. As Thanksgiving nears, one might worry about “the turkey,” and no identification would be needed; before a wedding, one might talk about “the dress.” 7. the mass being said at Virginie’s first communion (a) Text: the mass; (b) Information in text: Virginie is dressed for her first communion; (c) General knowledge: Mass is part of the first communion ceremony; (d) Inference: the scene has moved to the first communion.
The following examples are varied. The first two are anaphoric elements without explicit antecedents, and the third differs from all preceding examples in introducing entirely new material. (e) 46–49 . . .le docteur conseilla les bains de mer de Trouville. Dans ce temps-là, ils n’étaient pas fréquentés. Mme. Aubain prit des renseignements, consulta Bourais, fit des préparatifs comme pour un long voyage. Ses colis partirent la veille, dans la charetle de Liébard. Le lendemain, il amena deux chevaux . . .et Paul enfourcha l’âne de M. Lechaptois. . . . La route était si mauvaise que ses huit kilomètres exigèrent deux heures. Les chevaux enfonçaient. . .. 8. the night before the trip to Trouville (a) Text: trunks, as if for a long trip; (b) Information in text: Mme. has prepared for a trip, and been advised to go to Trouville; (c) General information: one takes trunks on a trip, and they are sent just before the trip; (d) Inference: Mme. is going to take a trip to Trouville, and the trip is about to begin.
Ellipses in Flaubert’s Un Coeur Simple
367
It is difficult to assess the contribution of the second phrase noted in (a). This phrase suggests an antecedent for the definite NP, but cannot actually serve as one because of “as if.” 9. the next morning, following the night when the trunks left (a) Text: the next day; (b) Information in text: the trunks left at night; (d) Inference: the reference point for “next” is the night the trunks left.
The juxtaposition of NPs 8 and 9 is odd, because neither has an explicit antecedent and there is a shift of focus between them. “The night before” precedes the trip; “the next morning” does not follow the trip, but rather the night before. The function in the narrative will be discussed in the section on pronouns, where there are other examples of shifts of focus. 10. the road to Trouville (a) Text: sea baths at Trouville; (b) Information in text: a trip to Trouville is planned, the horses are mounted, trunks have been sent; (c) General knowledge: traveling on horseback involves a road, one mounts a horse before setting out; (d) Inference: the travelers were ready to leave, are now on the road. (f) 61–62 comme Virginie toussait et que la saison n’était plus bonne, elle revint à Pontl’Evêque. M. Bourais l’éclaira sur le choix d’un collège. Celui de Caen passait pour le meilleur. Paul y fut envoyé.. . . 11. the choice that Mme. Aubain will make of a school to which Paul will be sent (a) Text: Paul; (b) Information in text: Paul is 7 years old, and was sent to a school in Caen; (c) General knowledge: 7 is school age, children are sometimes sent away to school and if so, a school must be chosen; (d) Inference: Mme. has decided to send Paul away to school, and, following advice, chose Caen.
Note that the choice itself is passed over in the text. We surmise that the school at Caen was chosen because Paul was sent there. The narrative is advanced by this treatment of the choice, which is abrupt and unprepared-for. Most of these ellipses function to move the narrative forward. Identification of the NP means recognizing a new time and/or place, and inferring that the scene has shifted to that time or place. Most such shifts are relatively small and not unexpected; the most dramatic shift occurs in 11, where the unanchored NP introduces a new topic.
Pronominal References Since such reference is discussed extensively in the body of this essay, we present here only a few examples. In the first two, an antecedent for the pronoun must be constructed; the third example has an unexpected focus shift.
368
J.T. Whitaker and C.S. Smith
(a) 66 Le prêtre fit d’abord un abrégé de l’ Histoire Sainte. Elle croyait voir le paradis, le déluge, la tour de Babel, des villes tout en flammes.. . . Puis, elle pleura en écoutant la Passion. Pourquoi l’ avaient-ils crucifié, lui qui chérissait les enfants.. . . 1. him: Christ (a) Text: the Passion; (c) General knowledge: the Passion is the story of Christ.
Here knowledge of matters Catholic not only identifies a reference, but also allows the construction of a referent. (b) 191–92 A l’ église, elle contemplait toujours le Saint-Esprit, et observa qu’il avait quelque chose du perroquet. Sa ressemblance lui parut encore plus manifeste sur une image d’Epinal, représentant le baptême de Notre-Seigneur. Avec ses ailes de pourpre et son corps d’émeraude, c’était vraiment le portait de Loulou.. . . L’ayant acheté, elle le suspendit.. . . 2. its: dove in the baptism of Christ of the Epinal print (a) Text: baptism of Christ, Epinal print, purple wings; (c) General knowledge: doves appear in pictures of the baptism of Christ.
This is the only example that may seem obscure to modern readers. A fairly detailed knowledge of Christian iconography, and familiarity with a certain type of reproduction, are assumed. 3. it: an exemplar of the Epinal print (a) Text: Epinal print, truly a portrait of Loulou; (b) Information in text: Félicité is poor, she bought a picture; (c) General knowledge: Epinal prints are cheap, brightly colored reproductions (made in the town of Epinal) available in quantity; one usually buys exemplars of prints; “truly” is not to be taken literally in this context; (d) Inference: the portrait bought was an Epinal print.
The delicacy of the distinction between the Epinal print and the portrait of Loulou is clearer in French than in English. “A print” has feminine gender and “a portrait” has masculine gender. The pronoun in the text is masculine and must refer to the portrait; what Félicité actually bought was the print. Examples of this type are fairly frequent in Flaubert’s story but do not survive translation into English.22 (See the discussion in the body of the essay of the first paragraphs of the story.) The reason is that French pronouns have gender whether they refer to animate or inanimate objects; most of the cases where contiguity prevails over discourse organization (usually to one’s surprise) involve a pronoun whose immediate antecedent is inanimate and whose more remote possible antecedent is animate. 22 In
several translations we have consulted, the ellipses are either evaded or made easier to interpret. Evasions occur when a proper name is given, while the original has a pronoun. Interpretation is made easier when more information is given in the translations than in the original text. These changes make the English more fluent, but fail to convey the spirit of the French.
Ellipses in Flaubert’s Un Coeur Simple
369
There are two discourse functions served by the examples of this section. They underline or actually make shifts of focus, and they force readers to follow the text quite closely, because normal expectations are not always borne out. We have suggested that both of these have an important literary effect. They indicate to readers that attention is necessary, that the most plausible and normal expectations may not occur; readers must sort things out in the material presented to them.
Accounting for Subjectivity (Point of View) Carlota S. Smith
I discuss in this article some of the forms and interpretations usually covered by the phrase ‘point of view’. The term is used by linguists for expression of speech and thought, perspective, evidentiality, and other indications of an authorial voice. ‘Point of view’ is often used almost interchangeably with ‘viewpoint,’ ‘perspective,’ and ‘subjectivity’. This has led to considerable confusion. In what follows I will be concerned mainly with the notion as discussed by linguists, while recognizing that there is a strong literary tradition.1 I take it that all expressions of point of view are subjective, since they involve mind; I will use ‘subjective,’ and ‘subjectivity’ as general terms rather than ‘point of view.’ Subjectivity is expressed by grammatical forms at the sentence level (verbs and their complements, tense, aspectual viewpoint, anaphors, etc.), yet subjectivity arises primarily in discourse contexts. Anaphors and many other forms can only be interpreted with information from outside the sentence. More generally, discourse sets up expectations for structure and interpretation. These expectations depend largely on genre (e.g., narrative, newspaper editorials). The dynamic established by relations between several sentences can set up a pattern that guides interpretation. One of my goals is to sort out the main types of linguistically conveyed subjectivity as a basis for the systematic interpretation of sentences and discourse. I will distinguish two general classes: sentences that express point of view, and perspectival sentences that present a situation from a particular standpoint. Both involve the mind and are therefore subjective, but in clearly different ways. The discussion is mainly about English, though I also comment on other languages. I also consider the question of how grammatical forms give rise to interpretations of subjectivity when they occur in sentences. I suggest a ‘composite’ account: subjectivity is conveyed by composites of syntactic and semantic factors and interpreted by rules which look at several grammatical forms together. Typically the forms convey more than one kind of information. The composite approach is well-suited to point of view because it deals naturally with such variety. I will suggest principles for interpreting subjectivity in the framework of Discourse Representation Theory.
1 As a literary term ‘point of view’ refers to presentation of the speech and thought of a fictional character or, more generally, as “the perceptual or conceptual position in terms of which narrated situations and events are presented” (Prince 1987: 73).
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0_15,
371
372
C.S. Smith
Section 1 sets the stage with introductory examples and commentary; Section 2 organizes the phenomena and discusses the basic distinctions; Section 3 sketches the composite analysis; Section 4 concludes.
1 Introduction The examples below illustrate different types of subjectivity. The first three examples are in the first person. Fragments (1a–b) are from novels; (1c) is from a newspaper article of opinion.2 (1) a. . . . 1 “My God,” Alec said. “What is he doing?” 2 “Who?” 3 “Your boss,” Alec said. “Standing in the fountain.” 4 I crossed to the window and stared downwards: down two floors to the ornamental fountain in the forecourt of the Paul Ekaterin merchant bank. 5 Down to where three entwining plumes of water rose gracefully into the air and fell in a glittering circular curtain. 6 To where, in the bowl, calf-deep, stood Gordon in his navy pin-striped suit . . . (Dick Francis) b. . . . 1 I sipped my drink and nodded. 2 The pulse in his lean grey throat throbbed visibly and yet so slowly that it was hardly a pulse at all. 3 An old man two-thirds dead and still determined to believe he could take it. (Raymond Chandler) c. . . . I feel reasonably certain of the final verdict on the current impeachment affair because I think history will see it as the climax of a six-year period marred by a troubling and deepening failure of the Republican party to play within the established constitutional rules. (Peter Ehrenhalt) I note briefly the forms that indicate subjectivity in these fragments. (1a) begins with direct quotations, introduced by the verb said and quotation marks. The perception verb stared appears in sentence 4; we understand what follows as expressing what the subject perceives—including the material in sentences 5 and 6, within the scope of stared.3 Perception is involved in (1b), but less directly: sentences 2 and 3 suggest it although no perception verb appears. The fragment in (1c), with the verbs feel and think, expresses the speaker’s feelings and thoughts. In all three we would ascribe a point of view to the referent of the first person pronoun—the speaker-reporter. The next example is from a detective novel in the third person. Consider the interpretation of sentence 4, which is not preceded by a verb of thought or perception.
2 Text
sources are listed at the beginning of the references, below. Harris, one might account for this inclusion by the presence of zero allomorphs of stared etc. in each sentence. 3 Following
Accounting for Subjectivity (Point of View)
373
(2) 1 The workshop was dusty and messy, crusted with bits of old clay, and it suited a part of Mara’s personality. 2 Mostly she preferred cleanliness and tidiness, but there was something special, she found, about creating beautiful objects in a chaotic environment. 3 She put on her apron, took a lump of clay from the bin and weighed off enough for a small vase. 4 The clay was wet. (Peter Robinson) Sentence (4) may express a perception or awareness of Mara’s, the person working with the clay; or, it might be due to the reporter. This kind of uncertainty as to who is responsible for a phrase or clause is not uncommon. The next example illustrates subjectivity of a different kind: an authorial voice that evaluates and suggests a temporal standpoint. Example (3) is from an article of popular science: (3) Cell Communication: The inside story 1 It might seem surprising that mere molecules inside our cells constantly enact their own version of telephone without distorting the relayed information in the least. 2 Actually, no one could survive without such precise signaling in cells. 3 The body functions properly only because cells communicate with one another constantly. 4 Cells of the nervous system rapidly fire messages to and from the brain. 5 But how do circuits within cells achieve this high-fidelity transmission? 6 For a long time, biologists had only rudimentary explanations. 7 In the past 15 years, though, they have made great progress in unlocking the code that cells use for their internal communications. (Scientific American) The sense of voice can be traced to a number of sources: the modal might, the verb seem, the predicate surprising, all in sentence 1; the adverb actually in sentence 2; the direct question which comprises sentence 5; the time adverb in the past 15 years and the concessive adverbial though of sentence 7. The well-known examples in (4) below have a subjective element, suggesting a particular standpoint from which the situation is presented: (4) a. Physicists like yourself are a godsend. b. John pulled the blanket over himself. (4a) takes the perspective of a speaker toward an addressee (from Ross 1970); (4b), from Kuno (1987), demonstrates what he calls an “empathy perspective”, in which the reflexive indicates the perspective of John, a participant in the situation. These examples show that a variety of linguistic forms convey subjectivity. This diversity is essential to the composite analysis sketched in section (3). When a sentence expresses a point of view, or takes a particular perspective, we need to identify the person/mind responsible. The question is, to which mind do we ascribe the material expressed or implied? Ascribing responsibility is part of the interpretation
374
C.S. Smith
of a sentence that conveys subjectivity.4 In the sentence Mary believed that John was sick, for instance, we ascribe the belief expressed in the complement clause to Mary and not to the reporter, or to John (assuming that we trust the reporter). In the sentence Mary unfortunately may win the race we ascribe responsibility for the modal and the adverbial commentary to the reporter. We have also seen cases where either the reporter or the participant in the situation may be responsible. For ‘neutral’ sentences—by convention objective—the question of responsibility does not arise. We may ask whether a neutral sentence is true but not, usually, who is responsible for it.
2 Subjectivity: Point of View and Perspective In this section I discuss different types of subjectivity and the key linguistic factors that convey them. I suggest two major categories: sentences expressing a point of view, and sentences that take a particular perspective on a situation. Among sentences that express point of view, I distinguish three subclasses. They have been developed on the basis of intuitions about meanings and the linguistic forms that appear in them. The subclasses of point of view are communication, contents of mind, and evaluation. Perspectival sentences fall into two subclasses: sentences that involve perception and those that do not. The focus of this discussion is grammatical: I do not consider lexical choice, although that is clearly an important factor. The particular words that appear in a sentence may strongly suggest choices that a particular participant would make, and thus provide access to the mind of that participant.
2.1 Expressions of Point of View In this section I consider the expression of point of view. Since the forms are sententially based, the discussion is in terms of sentences; however, factors in the linguistic context are often relevant. This will become clear directly. I begin with point of view as expressed in sentences involving communication, and then move on to sentences expressing content of mind and evaluation. 2.1.1 Communication This category of expression deals with speech and other communicative events, which are by definition external. Linguistic presentations may consist of quoted speech, represented speech, or indirect speech. The first two directly present what was said, often introduced by a verb of communication. Quoted speech is often referred to as ‘direct’ speech. I use the term ‘quoted’ here because I want to 4 In narrative fiction, it is customary to distinguish sentences that contain a ‘self’ at a moment corresponding to an act of consciousness (Banfield 1982:158). Sentences without a ‘self’ report events “objectively, independent of an explicitly-perceived, narrating self” (Fleischman 1991:31).
Accounting for Subjectivity (Point of View)
375
recognize quoted and represented speech as directly presented communication. In contrast, indirect speech may involve the reporter’s recoding of the communication. The features of communicative sentences have been analyzed extensively (Partee 1973; Banfield 1982). Verbs of communication are a distinct syntactic class. They allow a direct object complement which expresses the actual communication and an indirect object referring to the addressee (X said Y to Z). The class includes the verbs say, ask, request, command, declare, confess, advise, insist, claim, shout, read, sing, remark, observe, note, yell, swear, promise, announce, pray. Some verbs of thought have the same syntactic characteristics. I begin with examples of quoted and indirect speech. The former reproduces what was said, while the latter is indirect, a report. In discussing these and all later examples I will refer to the source of a sentence or main clause—the writer or speaker, the person responsible for the quotation or report—as the reporter. (5) Quoted speech and thought a. “I am getting ready for the party this afternoon.” b. Maryi said “Ii am excited.” c. Maryi told mej yesterday at the station, “Ii will meet youj here.” d. Maryi asked “Do Ii have to go?” (6) Indirect speech and thought a. Maryi told mej yesterday at the station that shei would meet mej there. b. Maryi asked whether shej had to go. Quoted speech is just that: as in (5), it typically appears with a verb of communication and a direct representation of what was said. Person, tense, and other deictics orient to the first-person speaker, as in the original utterance. In contrast, indirect speech does not present exactly what was said: pronouns, tense, and deictics shift in accord with the report of the communication. Compare for instance (5c) and (6a), from Banfield (1982). The complement of (5c) reproduces Mary’s utterance: it has the first person, present tense, and a proximal deictic oriented to the speaker. In (6a) these forms are shifted: the tense is past in accord with the tense of the main clause verb, the pronoun is third person, the deictic pronoun is non-proximal (distal). Syntax also shifts in indirect speech: constructions that are limited to main clauses, such as questions and exclamations, have other forms (or do not appear at all). Compare for instance examples (5d and 6b): the question is direct in (5d) and indirect in its counterpart (6b). The tenses of main and complement verbs are in concord, as the examples of (6) show; this kind of concord is also known as sequence of tense (Comrie 1986). It is required in English, but not in all languages. Japanese and Navajo, for instance, do not have shifted deictics nor tense concord in the complements of communication verbs. (Hirose 2000, Speas 1999); Russian does not have sequence of tense, Amharic does not have sequence of person (Schlenker 1999). Indirect speech is explicitly introduced by a reporter, and in sentences of indirect speech there is a systematic ambiguity. The report may be a recoding by the reporter
376
C.S. Smith
of all or part of what was said, or it may be precisely what was said. In the former case some responsibility for formulation and truth is due to the reporter rather than the person whose communication is reported. (7) gives a well-known example. (7) Oedipus said that his mother was beautiful. This sentence could be used to report an utterance of Oedipus in which the speaker identifies the person that Oedipus talked about as Oedipus’s mother (the de re reading). It could also be used to report exactly an utterance of Oedipus, “My mother is beautiful” (the de dicto reading). The truth of the sentence depends both on the interpretation in question and what was said: (7) might be true on the de re but not on the de dicto reading. When deictic features that are oriented to the reporter appear in the complement the reporter clearly has some responsibility, as in (8): (8) a. John said that Mary was leaving tomorrow. b. John said that Louise is pregnant. In these sentences an important component of the complement clause meaning is ascribed to the reporter. In (8a), the complement has the deictic adverb tomorrow, oriented to the time of speech, while the verb of communication is past. This conveys that the reporter has recoded all or part of what was said: the day of leaving is located with respect to the time of speech rather than to the past time of John’s utterance. In (8b) the complement has present tense while the verb of communication is past. The sentence conveys that John’s utterance about Louise pregnancy was made in the past; and that the reporter relates the pregnancy to the time of speech, since (8b) would be false if Louise’s were not pregnant at the time of the report. Examples like this are known as ‘double-access’ sentences; they are discussed by Ogihara (1996), Abusch (1997), Giorgi and Pianesi (2000). Given the close relation between them, we might attempt to derive indirect speech from direct speech. However it has been shown that this approach cannot work. The argument turns on two points: first, not all indirect speech reports have plausible counterparts in expressions of direct speech; and secondly, the ambiguity between de re and de dicto readings arises only for indirect speech (Partee 1973; Banfield 1982). The third member of the category of communication is ‘represented speech and thought’. Represented speech has some features of direct speech and some of indirect. It presents the syntax of actual speech, but tense and pronouns are shifted as in indirect speech; (9) illustrates: (9) Represented speech and thought 1 Mrs Dallowayi said shei would buy the flowers herselfi . 2 For Lucy had her work cut out for her. 3 The doors would be taken off their hinges; Rumpelmayer’s men were coming. 4 And then, thought Clarissa Dalloway, what a morning—fresh as if issued to children on a beach. (Virginia Woolf)
Accounting for Subjectivity (Point of View)
377
The first sentence of fragment (9) has shifted tense and person: what Mrs Dalloway actually said, presumably, would have been I will buy the flowers myself. The subsequent sentences represent Mrs Dalloway’s thoughts. The syntax preserves the locutions of speech and thought. Represented speech and thought is not syntactically embedded: for instance, (9) ends with an exclamation, a form which can appear only in main clauses (topicalized sentences, elliptical fragments and a few other constructions are similarly limited). Represented speech may have a ‘discourse’ parenthetical, as in He would be late, John said (Reinhart 1975). The tense-aspect forms that appear in represented speech are limited in some languages. In French, for instance, the imparfait—a past tense with the imperfective viewpoint—is usually found in represented speech (Banfield 1982: 158). Imperfectives and statives often occur in expressions of subjectivity; see section 2.2.3 below for discussion. Represented speech usually appears in fiction, where it can play an important role in conveying the character of a protagonist. Banfield suggests that it represents the consciousness of the person whose thought is presented (1982: 10). Represented speech is also known as the style indirect libre (free indirect style), narrated monologue, and erlebte rede. Literary studies include Cohn (1978), Chatman (1978), Genette (1980), and others; Jespersen (1924) and Banfield (1982) have a more linguistic orientation. 2.1.2 Contents of Mind This category comprises expressions of mental states (such as thoughts, beliefs, and attitudes) and other propositional expressions. Standard examples of sentences expressing mental state have verbs like think and believe; their complements express the object of thought or belief, the content of mind. The complements are similar in form to those of indirect speech; rather than external events of communication, they report mental states as in (10). (10) a. We thought that Bella was in New York. b. Maryi believes that shei won the race. (10a) is subjective, expressing thought. There is tense concord between the main and complement clauses. In (10b), which expresses a belief, the main and complement clauses have coreferential subjects. This coreference relation between referents in clauses involving thought or communication is sometimes called logophoric. Logophoric pronouns convey coreference between the subject of main and complement clauses in the context of verbs of thought, belief, or communication. Some languages have particular logophoric pronoun forms; one of these would appear in the complement of (10b). The term ‘logophoric’ is due originally to Claude Hagège, who studied African languages such as Ewe, Mundang, Tuburi, and Ubangi languages.5 Logophoric pronouns have
5 Hagège (1974) coined the term ‘logophoricity’ in connection with the African languages Mundang, Tuburi, and Ewe; he defined it as “to designate a category of anaphoric pronouns, personal
378
C.S. Smith
also been identified in Igbo and Gokana (Hyman and Comrie 1981), and Mapun, a Chadic language (Frajzyngier 1985), among others. In some languages, it is only verbs of psychological state and verbs of perception that introduce logophoric pronouns; in others complementizers and the subjunctive may also introduce them (Stirling 1993: 259 et seq, Giorgi and Pianesi 2000). The existence of logophoric pronouns (and related forms, discussed in Section 2.2 below) shows that linguistic coding of access to mind is an important feature of language. Reports of belief and thought involve the mind and consciousness of the person holding the belief. In Japanese there are particular features of vocabulary and syntax which are used to convey access to mind (Kuroda 1973). For instance, the pronoun zibun in Japanese conveys access to consciousness in communicative contexts: the referent is aware of the propositional content of the clause containing the pronoun (Hirose 2000: 1646). There are also perspectival uses of zibun, see Section 2.2.2 below. Expressions of thought and belief are reported, like indirect speech; their formulation can be due to the reporter in the same way. (11), for instance, is ambiguous: (11) Mary believes that that fool Gwendolyn wants to take over the committee. On one reading, the epithet that fool is due to Mary: it is part of Mary’s belief. There is also a reading, perhaps more salient in an isolated sentence, in which the reporter designates Gwendolyn as a fool. The two interpretations are true under different circumstances so that one cannot be substituted for the other. This property is known as ‘referential opacity’; it is typical of complements referring to propositions. Thoughts and beliefs are due to the person who holds the proposition. Vendler notes that propositions belong to particular individuals and are thus limited by subjective factors; in particular, referential opacity reveals the subjectivity of propositions (1972: 73, 81). (12) illustrates; b and c are due to Peterson (1997) and Asher (1993) respectively. (12) a. It seems that Mary will win the race. b. Mary’s having refused the offer was unlikely. c. Everything that John believes is true. We ascribe responsibility for the propositions to the reporter in these sentences. Predicates which take propositions as complements include seem, appear, believe, fear, hope, want, think, affirm, deny, unlikely, impossible, inconsistent, sure, true, be certain, propose, and hypothesize. Sentences with predicates which refer to and possessive, which refer to the author of a discourse or to a participant whose thoughts are reported” (Stirling’s translation 1993: 253). The term logophoric is now widely used, often in reference to discourse-oriented anaphors, or anaphors that appear in point of view contexts (Reinhart & Reuland 1993: 671). The term is also used for all kinds of indications of the mind or perspective of a participant; see the references cited in the text.
Accounting for Subjectivity (Point of View)
379
propositions can be identified linguistically; sentences which directly express propositions cannot be. Questions, imperatives, and sentences with modal auxiliaries or adverbials are another source of subjectivity in sentences. These forms express Projective Propositions, a class noted by Asher (1993). Projective Propositions are unrealized, as in (13): (13) a. b. c. d.
Clean up your room! Will he leave the room? He may/might leave the room. John will probably win the race.
Projective Propositions also appear as the complement of verbs expressing unrealized notions (e.g. wonder, want, desire, guess, command, plead, entreat, allow, permit). Projective Propositions can be distinguished from other propositions by their behavior under quantification.6 Modals, questions, and other projective propositions are ascribed to the reporter when they appear in main clauses, as in (13). When they appear in the complement clause of a mental or projective propositional verb, they are ascribed to the human subject of the verb, as in (14). (14) a. John thinks that Mary will probably win the race. b. They asked/ordered Mary to clean up her room. In (14a) the modal is due to John, not the reporter; in (14b) they gave the order. 2.1.3 Evaluation This class of expressions comprises evaluation, emotional reactions, commentary, and evidentials. All are subjective since all involve mind. They are conveyed by 6 In the quantificational test, one constructs a sentence with predicates referring to abstract objects (facts, propositions, projective propositions). If there is no quantifier that can appear in such sentences with a truth value, then the predicates take arguments of distinct, incompatible types. This test distinguishes Propositions from Projective Propositions: # indicates lack of truth value.
(i) a. b. c. d.
#John desires everything that Mary believes. # Everything that Mary asks for is true. John asks for something that Mary wants. Everything that John believes is true.
In (ia) and (ib) the clauses refer to different types of entities: in the first clause, the complements refer to projective propositions while the complements of the second clauses refer to propositions. They are semantically anomalous. In contrast, (ic) and (id) both have complements referring to propositions and the sentences are semantically well-formed (Asher 1993: 33–34).
380
C.S. Smith
predicates and adverbials. The person responsible or affected is not always made explicit, as (15) illustrates. (15) a. b. c. d. e. f.
That Mary won the race was surprising to her/everyone. That Mary won the race was surprising. Surprisingly, Mary won the race. That Mary won the race was lucky for her. That Mary won the race was lucky. Luckily, Mary won the race.
The person affected is identified in the prepositional phrases of (15a, d). Without such phrases (15b, e) we ascribe the effect to the reporter or to another person (the latter interpretation is plausible in contexts where other participants are mentioned or assumed). Adverbials of this class are due to the reporter (15c, f) when they appear in a main clause. Many evaluative expressions have related ‘psychological’ verbs, e.g. surprise, frighten, annoy, etc. I do not think that these verbs are subjective in the sense being developed here: they do not require that responsibility be ascribed to a mind. For instance, we can account for a sentence like That John was early surprised Mary adequately with the notion of a thematic role of experiencer. Commentary and evidentials are also inherently subjective. They are expressed by adverbials or predicates, and imply a responsible mind: (16) a. b. c. d. e.
Frankly, Mary won the race. Clearly, Mary won the race Allegedly, Mary won the race. It was alleged that Mary won the race. It was alleged by many observers that Mary won the race.
The reporter is responsible for adverbials of commentary and for some of evidentiality (16a-b). The situation is different for allegedly which means that an allegation was made by a particular person or source other than the reporter (16c). The adverbials in these examples correspond to the ‘highest’ classes of adverbs distinguished in Cinque (1999): Speech Act adverbials (frankly, honestly), Evaluative adverbials (unfortunately), Evidential adverbials (clearly, allegedly). Cinque gives a syntactically-based account in which a syntactic projection is posited for each class. Evidentials may take other forms as well. Cinque points out that, in addition to predicates and adverbials, many languages have particles and idiomatic expressions of surprise, approval, etc. (1999). Cinque gives as an example the expression after all as in the sentence So he is coming after all! (despite our expectation to the contrary). Like other lexical expressions, cases like this are beyond the scope of this discussion.
Accounting for Subjectivity (Point of View)
381
2.2 Perspectival Sentences with a particular perspective depart from the standard understanding of perspecitive as neutral, objective. By convention we understand the perspective of a non-subjective sentence to be transparent—not filtered through a particular mind— unless the sentence contains information to the contrary. The central examples of this category are reports of perception. The reported perception may be direct or indirect. Perception necessarily involves the particular perspective of the perceiver: reports of perception are subjective since a situation is perceived from the perceiver’s standpoint. I include in this category sentences which convey or suggest a particular perspective. Perspective is conveyed by expressions which may be oriented either to the reporter or to a participant, primarily reflexives not syntactically conditioned, deictics, NPs and PPs that involve direction or location. 2.2.1 Perception Linguistic presentation of perception may be direct, indirect, or inferred. The most straightforward cases are reports of direct perception: a verb of seeing, hearing etc. introduces a complement which expresses the situation perceived, as in (17):
(17) a. John saw that the sun was shining. b. John saw Mary walk to school. c. John saw Mary walking to school.
The examples illustrate the three forms of perception verb complements in English: propositional, a ‘bare’ or ‘naked’ infinitive, or gerundive. First-person reports of perception are also subjective: they express the perspective of the reporter as a participant (including the ‘unreliable narrator’ of fiction). Less direct but very clear are cases where a sentence with a perception verb precedes another sentence; the second is taken to express the percept of the perception verb’s subject. Examples like this tend to occur in narrative contexts. (18) illustrates:
(18) a. John looked out the window. The children were building a sandcastle. b. Gabriel smiled at the three syllables she had given his surname and glanced at her. She was a slim, growing girl, pale in complexion and with haycolored hair. (James Joyce)
These and other examples are discussed extensively in Caenepeel (1989). The second sentence in each case conveys the percept implied in the first: they are ‘perspectivally situated’ in Caenepeel’s terms. Such sentences either have the imperfective
382
C.S. Smith
aspectual viewpoint (18a), or express a state (18b); both express unbounded situations, for slightly different reasons.7 These expressions of indirect perception are like represented speech in some ways: both offer access to the mind of a participant, often with shifted deictics. Stative sentences, and sentences with the imperfective viewpoint, are usually found in these contexts. The reason is partly semantic: both focus an internal interval of a situation and thus lend themselves to the perspective of a participant in the situation. In French, only the imparfait may appear in such contexts (Banfield 1982: 158).8 Perspectivally situated sentences must present unbounded situations, according to Caenepeel. As evidence she gives examples like (19); the second sentence expresses an event with the perfective viewpoint. (19) John looked out the window. Mary arrived. It is difficult to interpret the second sentence here as conveying John’s percept.9 The more natural interpretation is that the two events occurred in sequence; or perhaps at the same time. Caenepeel explains these facts by appealing to our concept of perception. Perception is instantaneous. Therefore one can perceive only a short segment of an unbounded situation, an instant; a bounded situation requires more than a single instant.10 This explanation suggests another set of cases: in the context of sentences that express continous perception, perspectivally situated sentences with the perfective
7 Aspectual
viewpoints focus all or part of a situation. Imperfectives focus part of a situation, excluding endpoints; perfective viewpoints focus a situation in its entirety, including endpoints or implicit bounds for events. In Russian there is a pragmatic convention in which the focus of the perfective is taken to be on the completion of an event; this convention is overridden under certain circumstances (Smith 1997). The imperfective viewpoint in English is also known as the ‘progressive’; it is conveyed by the auxiliary be+ing. The unbounded interpretation arises for different reasons in imperfective and stative sentences. In the former, the viewpoint does not include endpoints; in the latter, the temporal schema of a state has no endpoints (Smith 1991). In Caenepeel’s work the two are grouped together into a supercategory of ‘stative’; see also Herweg (1991). I argue against the supercategory approach in Smith (1996, 1999). 8 The French system is more strongly codified than the English. The French imparfait is a past imperfective tense which is commonly used to express perspective. The English stative and imperfective together correspond grammatically to the French imparfait. The imparfait can be used for statives and non-statives, wheres the English progressive is possible neutrally only with nonstatives. English stative sentences have the simple, perfective verb form. 9 Caenepeel claims that when a sentence presents a bounded situation the perspectival interpretation is unlikely at best. She says that bounded events—events presented perfectively—are impossible or awkward as perceptual reports unless there is a contingency relation between the event and the focalizing sentence. 10 This is essentially the same as Kamp’s account of why present perfective sentences cannot be used to express a bounded event. We conceive of communication as instantaneous; communication of a bounded event, even if instantaneous, requires at least one instant after the completion of the event.
Accounting for Subjectivity (Point of View)
383
viewpoint should be appropriate. I think it is possible to construct such examples, in which perception takes place over an interval. (20) illustrates; the perception verb is perfective in (20a), imperfective in (20b). (20) a. John looked out the window. Mary threw the ball to Sue and Bill played in the sandbox. The neighbor’s dog arrived and trotted back and forth. b. John was looking out the window. Mary threw the ball to Sue and Bill played in the sandbox. The neighbor’s dog arrived and trotted back and forth. Both examples convey ongoing perception, in which situations unfold as John looks from the window. They differ slightly. In (20a) the verb of perception has an inchoative interpretation, suggesting the beginning of John’s looking; in (20b) the looking is in progress. Context plays a role in the interpretation of examples involving perception. In narratives it is common for thoughts and perceptions to be represented, often with some ambiguity as to whether the perspective belongs to the reporter or a participant. Thus narrative contexts—the main topic of Caenepeel’s discussion— lend themselves to the perspectivally situated interpretation. There are also cases of inferred perception. Sometimes one infers that perception is involved from the developing situation, as (21) illustrates. As in the cases of indirect perception more than one sentence is required. The first example is (1b), repeated here as (21a); the second is from Dowty (1986). (21) a. I sipped my drink and nodded. The pulse in his lean grey throat throbbed visibly and yet so slowly that it was hardly a pulse at all. An old man twothirds dead and still determined to believe he could take it. b. John entered the president’s office. The clock ticked loudly. In both fragments the first sentence sets up a participant in a situation. The sentence(s) following are taken to express the percepts and/or thoughts of that person. Fragments that lead to the inference of perception tend to appear in narrative contexts, in which one expects to find expressions of participants’ perceptions. 2.2.2 Particular Standpoints I now turn to a different set of cases, where a particular perspective is suggested but perception is not involved. Rather, a situation is presented from the standpoint of a given participant, or by the reporter as participant. The notion of standpoint has a literal basis in the participant’s placement in the world. From a particular location, an observing individual sees things in a certain way: if I say that something is nearby or in the distance, it is because of my position in space. If I say that the bank is around the corner and you say it is across the street, we can both be right if we are standing in different places (Mitchell 1986: 1). The notion of standpoint can be extended to situations in which one talks as if one were in a location; and metaphorically to attitudes and views that are not grounded in space.
384
C.S. Smith
Perspectival examples have reflexive pronouns, deictics, the imperfective viewpoint, and other expressions that imply a particular standpoint rather than an objective stance. (22) illustrates with reflexives: (22) a. This paper was written by Ann and myself. b. Theyi heard the stories about themselvesi . c. Maryi put the blanket over herselfi . The particular perspective of the reflexives’ antecedent is suggested in these sentences. In (22a) the speaker must be the antecedent of the reflexive, and is plausible as participant (Ross 1970).11 In (22b–c) the sentence subjects they and Mary are the antecedents of the reflexives, and the reflexive suggests that the stories and blanket are located with respect to these people, that is, from their standpoint or perspective; (22b–c) are based on examples from Cantrall (1974) and Kuno (1987). When they convey a particular perspective, reflexive pronouns are optional rather than syntactically conditioned. Syntactically conditioned reflexives are obligatory in certain contexts; they are defined in Government Binding theory with the notions of c-command and locality. Very roughly, the Binding Theory requires that the antecedent c-commands the reflexive if it is within the domain of the relevant governing category (Chomsky 1981).12 Reflexives that violate these conditions represent a choice of the anaphor rather than a pronoun. To see the contribution of the reflexive, compare the sentences of (23). Both are well-formed. (23) a. Johni pulled the toy toward himi . b. Johni pulled the toy toward himselfi . The choice of the reflexive makes a difference in interpretation. (23b) suggests the perspective of John as he pulls the toy, (23a) has no such suggestion.13 I shall refer to these and similar cases as “perspectival”. 11 This
is one of a set of examples which led Ross (1970) to suggest that all sentences have a higher clause in underlying structure with a first-person pronoun and a verb of communication in the present tense. The overt reflexive pronoun would be coindexed with the covert first-person pronoun. Harris reached the same conclusion on a much broader basis (1982), as Bruce Nevin has pointed out to me. 12 The Binding Theory as stated in Chomsky 1981 has been the subject of much critical comment. Reinhart and Reuland 1993 offer an extensive revision in the same general framework; a different approach is taken in Pollard and Sag 1992. 13 Cantrall was perhaps the first to note the perspectival use of the reflexive. Cantrall presents many examples, among them the following sentences. Cantrall asks us to imagine that they describe a photograph which portrays a group of standing women who have their backs to the camera: a. The womeni were standing in the background, with the children behind themi . b. The womeni were standing in the background, with the children behind themselvesi . In (a) the children are located from the perspective of the speaker; in (b) they are located from the perspective of the women. As Zribi-Hertz notes, the sentences provide empirical evidence
Accounting for Subjectivity (Point of View)
385
According to Kuno (1987), these examples convey an “empathy perspective” such that the reporter takes the perspective of a particular participant. They need not involve access to the mind of the participant, though they may suggest it. The antecedent of a perspectival reflexive may appear in an independent sentence earlier in the discourse, as in (24), cited by Baker (1995):14 (24) She was not immediately able to say anything, and even when her spirits were recovered, she debated for a short time on the answer it would be most proper to give. The real state of things between Willoughby and her sister were so little known to herself, that in endeavoring to explain it, she might be as liable to say too much as too little. (Jane Austen) Reflexives that can contrast with pronouns are known as ‘Long Distance Bound’ (Zribi-Hertz 1989), ‘Long Distance Reflexives’ (Stirling 1993), ‘Locally Free Reflexives’ (Baker 1995). I will use the latter term, Locally Free Reflexives, LFR for short, since examples like (23) do not involve a long distance. Not all LFRs are perspectival: they may also be emphatic or intensive (Zribi-Hertz 1989; Reinhart and Reuland 1993; Baker 1995); such cases are beyond the scope of this discussion. Perspectival LFRs are coreferential in a way similar to the logophoric pronouns mentioned above, and indeed are called logophoric by many scholars. The antecedent to a perspectival LFR is a referent whose perspective is being represented. Such reflexives tend to occur in just those contexts in which logophoric pronouns may occur. However, they do not necessarily convey access to the consciousness of the participant.15 Perspectival LFRs have been identified in many
that the reflexive is correlated with an ‘internal’ point of view—that of a discourse protagonist as opposed to the speaker (1989: 704). 14 Alternatively, one might follow Harris in analyzing reflexives (1982) as reductions of metalanguage assertions of sameness. 15 Hirose gives example of the two uses of zibun. The logophoric involves access to consciousness, the perspectival does not; his term for the latter is ‘point of view.’ Hirose says that in the logophoric example (a) Kazuo is obviously aware that he is shy, because he says so. On other other hand in example (b), Kazuo does not have to be aware that the book he lost is the one he borrowed from his friend. This is shown [by the fact that] (c) is not contradictory (2000: 1646). (i) a. Kazuo wa zibun wa tereya da to itteiru K. TOP self TOP shy-person COP QUOT say-STAT Kazuoi says that hei is shy. b.
Kazuo wa zibun ga tomodati karita hon o nakusit K. TOP self NOM friend from borrowed book ACC lost Kazuoi lost a book that hei borrowed from a friend.
c.
Kazuo wa zibun ga tomodati karita hon o nakusita ga, sono K. TOP self NOM friend from borrowed book ACC lost but that hon ga tomodati kara karita mono da to wa kizuite-it-nai, book NOM friend from borrowed thing COP QUOT TOP realize-STAT-NBG Kazuoi lost a book that hei borrowed from a friend but he has not realized that the book is the one he borrowed from a friend.
386
C.S. Smith
other languages, among them Japanese (Kuno 1972), Scandinavian languages (Thráinsson 1976), and Italian (Giorgi 1984). Kuno (1987) offers a survey of logophoric phenomena across languages. Possessive pronouns may also suggest the perspective of a participant. They can do so, in English at least, because of the limited resources of the language. Possessives have no counterpart to the reflexive: there is only one form of possessive pronoun. (The possessive often appears with own in such cases, e.g. her own house.) Therefore possessive pronouns have the potential either for a reflexive or a perspectival reflexive reading. (25) gives examples in which possessive pronouns suggest the perspective of the antecedent: (25a) is from Kuno (1987), (25b) from Hirose (2000). (25) a. John criticized his brother. b. Kazuo lost a book that he borrowed from a friend of his. The perspectival reading of the possessives is only optional in these examples, I think. They can be read as simply giving information about the relationship of the participants. For the latter interpretation one might say that the possessive pronouns do not have a reflexive component.16 Kuno finds that the perspectival effect also occurs in other cases of directional relationships. For instance, the phrase John’s sister suggests John’s perspective whereas Mary’s brother suggests the perspective of Mary. If this is correct a sentence like (26) would offer the possibility of two perspectives. (26) John and his brother talked to Mary about her sister. If (26) suggests the perspectives of both John and Mary, it does not I think require the ascription of responsibility to either of them. The perspectival effects are relatively weak, then. Recognizing differences in strength among the relevant examples, Kuno (1987) posits a continuum of ‘degrees of empathy’. At the high end of the continuum the reporter totally identifies with a participant; in the middle the identification is partial; at the low end the reporter manifests a total lack of empathy with participants. On such a continuum the examples of (25) and (26) are toward the low end. Perspective may also be suggested by adjectives or epithets that would be expected from the participant, e.g. John talked to Mary about his beloved cat. Another source of participant perspective is the deictic adverbial. Deictics strongly suggest the perspective of a participant, especially when anchored to a time other than the time of speech, as in (27b–c):
According to Hirose, zibun in examples like (b) and (c) conveys ‘point of view’, whereas in (a) zibun is logophoric. 16 Kuno, among others, regards possessive pronouns as ambiguous between a [+reflexive] and [–reflexive] feature (1987: 81).
Accounting for Subjectivity (Point of View)
387
(27) a. Mary lost her watch 3 weeks ago. b. Mary had lost her watch 3 weeks ago. c. Mary packed her clothes. She would be leaving soon. In (27a) the deictic 3 weeks ago is anchored to the time of speech. The deictics in the other two examples, 3 weeks ago and soon, are anchored to past times not given explicitly; they take the perspective of Mary and therefore suggest her mind. The imperfective viewpoint is hospitable to particular perspectives, as we have seen in the focalizing sentences of perception in (18a). Since the imperfective focuses an internal interval of a situation its formal meaning is compatible with its use to suggest an experiencing mind and/or a particular perspective. Traditionally, imperfective viewpoints are said to involve an ‘internal perspective’, whereas the perfective is external (Comrie 1986). Shifted deictics are clear linguistic evidence for the internal interpretation. They are always good in imperfective sentences, but limited in perfectives. This is particularly clear in French: the imparfait past tense allows shifted deictics more freely than the perfective past tenses (Banfield 1982; Smith 1991).15 Oppositions such as perfective and imperfective often have the pragmatic function of marking what is traditionally referred to as ‘point of view’ in narrative (Fleischman 1991:26). I conclude this brief survey by observing that there is an additive effect in sentences with linguistic forms which suggest the perspective of a participant but do not require this interpretation. If there is one such form, the suggestion is weak: with two or more, the suggestion of a participant perspective becomes stronger, as (28) illustrates. In these examples, consider whether the question should be ascribed to the reporter or to Mary in a narrative context. Recall that questions suggest subjectivity, since they belong to the class of projective propositions (Section 2.3 above). (28) a. Mary played in the sandbox. Was it going to rain? b. Mary was playing in the sandbox. Was it going to rain? c. Mary was playing in the sandbox with her brother. Was it going to rain? In (28a) the direct question is the only subjective element; in (28b) the preceding sentence has the imperfective viewpoint, which as we have seen invites a subjective interpretation. The example with the strongest subjective interpretation is (28c), which also has a possessive phrase oriented to Mary.
2.3 Summary and Comment Sentences require interpretations of subjectivity when they express a point of view, or take a perspective on a situation. Among sentences that express a point of view, communication sentences involve public events, whereas contents of mind and evaluation express private events or mental states such as thoughts, beliefs, attitudes,
388
C.S. Smith
evidentials, comment. These sentences have recognizable classes of predicates and adverbials; they also have characteristic patterns of deictics and aspectual categories. The perspectival category includes perceptual reports and indications that a situation is viewed from a particular standpoint. Perspective is conveyed by verbs of perception, reflexives that are not syntactically conditioned, deictics, and other expressions involving directionality. In interpreting subjectivity we ascribe responsibility for a clause to the REPORTER or the SELF . There may be uncertainty and/or degrees of responsibility. Indirect speech complements, and certain perceptual reports, are ascribed to either the REPORTER or the SELF. Perspectival sentences suggest the perspective of a SELF in varying degrees; in the weaker cases, we may ascribe the perspective to the REPORTER.
3 The Composite Approach In most of the cases discussed above more than one linguistic form contributes to subjectivity either in the expression of point of view, or of a particular perspective. To make this concrete I list the forms most often found (29) Linguistic forms contributing to subjectivity Communication and consciousness verbs Main clause constructions mirroring idiomatic thought, speech Complementizers Deictic adverbials: place, time Direction and location PPs Epithets Evaluative verbs and adverbs, conjunctions (yet, anyway, still, after all, but) Evidential adverbials: evidently, possibly, frankly Imperfective aspectual viewpoint Lexical: verb direction (go vs come); psych verbs, dative verbs, etc. Projective propositions Pronouns, reflexives, possessives Propositional attitude complements Stative sentences Subjunctive The list is not complete but it gives a sense of the many and varied forms involved. The composite approach looks at several linguistic forms and constructs an interpretation. In previous work I have presented such an analysis for aspectual situation types, e.g. states and different types of events. The interpretation of a sentence as stative, telic, atelic, etc. arises from a composite of the verb, its arguments, and relevant adverbials (Smith 1991). The rules for aspectual situation types focus on three aspects of surface structure: syntax, e.g. verb complement relations; categorial
Accounting for Subjectivity (Point of View)
389
information (adverbial, NP, PP, etc.); and features which encode such information as verb class, definiteness, ±directional (for PPs), ±completive (for adverbials). The composite analysis looks at multiple information sources and constructs an interpretation. The composite rules are stated in the framework of Discourse Representation Theory. The interpretation of subjectivity can be made explicit with an extension of the Discourse Representation Theory (DR theory) framework. The theory develops a dynamic representation of the truth-conditional and conceptual meanings of sentences in discourse (Kamp 1981; Kamp and Reyle 1993). The individuals, situations, and times that a discourse introduces are represented as distinct entities in the Discourse Representation Structure. The roles of SELF and REPORTER must be added to the DR theory repertoire if we are to convey the meanings of subjectivity. A detailed proposal to extend DR theory in this way is made by Stirling (1993); Sells (1989) also gives a proposal.17 In this article I will confine myself to a sketch of compositional rules that can analyze subjectivity. DR theory posits construction rules which lead to a Discourse Representation Structure (DRS) for a discourse. The construction rules apply to sentences and deliver an interpretation to be encoded in the DRS. Construction rules can account for the interpretations of subjectivity discussed above. Input to the rules is a syntactically analyzed surface structure. The rules recognize the combinations of linguistic forms that trigger an interpretation of subjectivity. The output governs the interpretation: it will set up two roles, the REPORTER and the SELF, and ascribe responsibility for a clause to one or both of them. The role of SELF is identified with a participant in the situation; the REPORTER is always the writer/utterer of the clause. I will not attempt to deal with uncertainty or degrees of responsibility here. Relevant factors include the syntactic relations between clauses, categorial information, and such features as 1st, 2nd or 3rd person of pronouns; tense values; proximal vs. distal deictics. As illustration, I will sketch a tentative analysis of sentences of indirect speech. The composite of forms which convey indirect speech are a verb of communication; the complementizer that; in the complement, a non-first person pronoun, past tense in matrix and lower clause, and whether nonproximal deictics. Much of this information can be encoded with features, e.g. verb class V[com], tense [past], NP [–1st person], Adverb [-proximal]. I will state interpretation principles for two cases: the case where the REPORTER is clearly indicated; and the default case, where it is not. In the default case there is no formal indication that the reporter is responsible for the clause of indirect speech, though it is always a possibility. If tense, proximal deictics, or other forms directly indicate the reporter, then both REPORTER and SELF are responsible for the complement clause; I will refer to them as ‘reporter-based’.
17 Sells
(1987) proposes that three roles be recognized: SOURCE, the one who makes the report; the one whose mind is reported; PIVOT, the one from whose physical point of view a report is made. For arguments against this view, see Stirling 1993. SELF ,
390
C.S. Smith
(This statement is too simple since it doesn’t account for epithets or lexical cues, e.g. adjectives such as beloved.) The composite of linguistic forms which triggers the interpretation of indirect speech is sketched in (30). Syntactic surface structures are relatively simple in DR theory, with few functional categories. The NP subject must be [+human]; the verb must be a verb of communication [Vcom ]; the tenses of both main and subordinate clauses, past; the subject of the lower clause 3rd person; adverbs in the lower clause may be proximal or distal. (30) includes only these forms and features; the rule does not ‘see’ other forms in the sentence. (30) Indirect speech S[NPlsubj [+hum] Vcom [past] . . . [that S [NP2subj [3p] [past] Adv [±proximal]]]] The rule as stated is quite limited: it does not apply, for instance, to sentences that have first or second person pronouns (I/you said that I/you would be late). In (31) I state the two basic principles for interpreting indirect speech; the reverse arrow ascribes responsibility for a clause. Recall that the REPORTER is the speaker or writer; the SELF is the human referent of the main clause subject NP. The principles apply to a sentence if it fits the structure of (30). (31) Interpretation of indirect speech a. Default ← Main clause, possibly complement clause ← Complement clause b. Reporter-based forms in complement REPORTER ← Main clause, complement clause SELF ← Complement clause REPORTER
SELF
The default principle holds unless the complement clause has material explicitly related to the REPORTER ; in that case the second principle applies. If the main and complement clauses have coreferential subjects, the interpretation is logophoric; this would be stated as an additional principle for a language with logphoric pronoun forms. I now show how the rules apply to actual examples, two sentences presenting indirect speech. The first is (6a), repeated here as (32a); (32b) is closely related. (32) a. Mary told me yesterday at the station that she would meet me there. b. Mary told me yesterday at the station that she would meet me here. The sentences differ only in the adverb of the lower clause: it is distal in (32a) proximal in (32b). (33) shows informally how the construction and interpretation rules apply to (32a and b).
Accounting for Subjectivity (Point of View)
391
(33) a. S[ Mary Vcom [past] . . . [that [NP [she] [would:PAST] Adv [-proximal]]]] REPORTER ← Main clause SELF = Mary ← Complement clause b. S[Mary Vcom [past] . . . [that [NP [she] [would:PAST ] Adv [+proximal]]]] REPORTER ← Main clause; complement clause SELF = Mary ← Complement clause (33a) gives the default interpretation, (33b) the case where the reporter explicitly shares responsibility for the complement clause. Similar composite rules can be stated for the other cases discussed above. Since the rules apply at the level of syntactic realization, they must be stated separately for each case. For instance, there are different rules for propositional, modal, and evaluative predicates when they appear as verbs and adverbs. As a final example I show how when there is a propositional, modal, or evaluative adverb in the main clause of a sentence, responsibility is ascribed to the REPORTER. When such adverbials appear in an object complement that clause of a verb of thought, belief, or attitude, however, responsibility is ascribed to the SELF. The class of adverbs is notated as Advp . (34) Propositional, modal, evaluative adverbs a. S[. . . Advp . . .] REPORTER ← Adv b. S[. . . NP+humsubj . . . Vcom . . . S[that [. . . Advp . . .]]] SELF ← Adv
Rules for other cases of subjectivity will follow the same lines.
4 Conclusion In this discussion I have considered a variety of subjective phenomena. I reserve the term ‘point of view’ for sentences which express communication, content of mind, or evaluation. I use the term ‘perspectival’ for sentences which express perception or otherwise suggest a particular perspective on a situation. The domain for the linguistic expression of subjectivity is the sentence, although it is usually at the discourse level that it is recognized and interpreted. A composite analysis was sketched for sentences which express subjectivity, using compositional rules in the framework of Discourse Representation Theory. The rules recognize the key forms or combinations of forms that trigger interpretations of subjectivity. They ascribe responsibility through two roles, REPORTER and SELF , which are identified with participants in the Discourse Representation Structure. For a more complete treatment, see Smith (2003).
392
C.S. Smith
Acknowledgement I would like to thank the audience at the Conference on Information Structure in Oslo, Norway (December 2000) for discussion of an early version of this material; some of it appears in different form in the Working Papers for the conference.
References Sources for non-constructed examples are as follows, identified by example numbers: (1a) Dick Francis. Banker. New York: Fawcett. 1984, p. 3. (1b) Raymond Chandler. The Long Goodbye. Penguin Books. 1953, p. 64. (1c) Alan Ehrenhalt, “Hijacking the Rulebook”. The New York Times, Section 4, p.13, Dec 20, 1998. (2) John D. Scott and Tony Pawson. “Cell Communication”. Scientific American 72–79 June 2000. (3) Peter Robinson. A Necessary End. New York: Avon Books, 1989, p. 182. (9) Virginia Woolf. Mrs. Dalloway. New York: Harcourt, Brace, 1925 (1981), p. 3. (18b) James Joyce. “The Dead”. In Dubliners. Penguin Books. (pub 1916) 1958, p. 177. (24) Jane Austen. Sense and Sensibility. Oxford University Press. p. 173. Abusch, Dorit. 1997. “Sequence of tense and temporal de re.” Linguistics and Philosophy 20:1–50. Asher, Nicholas. 1993. Reference to Abstract Objects in Discourse. Dordrecht: Kluwer. Baker, C.L. 1995. “Contrast, discourse prominence, and intensification, with special reference to locally free reflexives in British English.” Language 71:63–101. Banfield, Ann. 1982. Unspeakable Sentences: Narrative and Representation in the Language of Fiction. Boston: Routledge & Kegan Paul. Caenepeel, Mimo. 1989. Aspect, Temporal Ordering, and Perspective in Narrative Fiction. Ph.D. Dissertation, University of Edinburgh. Cantrall, William. 1974. Viewpoint, Reflexives, and the Nature of Nounphrases. The Hague: Mouton. Chatman, Seymour. 1978. Story and Discourse: Narrative Structure in Fiction and Film. Ithaca: Cornell University Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Cohn, Dorritt. 1978. Transparent Minds: Narrative Modes for Presenting Consciousness in Fiction. Princeton: Princeton University Press. Comrie, Bernard. 1986. Tense. Cambridge, England: Cambridge University Press. Dowty, David. 1986. “The effects of aspectual class on the temporal structure of discourse: Semantics or pragmatics?” Linguistics and Philosophy 5:23–33. Fleischman, Suzanne. 1991. “Verb tense and point of view in narrative.” Discourse-Pragmatics and the Verb, ed. by S. Fleischman and L. Waugh, Routledge: London. Frazyngier, Z. 1985. “Logophoric systems in Chadic.” Journal of African Languages and Linguistics 7:23–27. Genette, Gérard. 1980. Narrative discourse: An Essay in Method. Translated by Jane E. Lewin. Ithaca: Cornell University Press. Giorgi, Alessandra 1984. “Toward a theory of long distance anaphors: a GB approach.” The Linguistic Review 3:307–361. Giorgi, Alessandra and Fabio Pianesi. 2000. “Sequence of tense phenomena in Italian.” Probus 12:1–32. Hagège, Claude. 1974. “Les pronoms logophoriques.” Bulletin de la Societé de Linguistique de Paris 69:287–310. Harris, Zellig S. 1982. A Grammar of English on Mathematical Principles. New York: John Wiley & Sons. Herweg, M. 1991. “Perfective and imperfective aspect and the theory of events and states.” Linguistics 29:969–1010.
Accounting for Subjectivity (Point of View)
393
Hirose, Yukio. 2000. “Public and private self as two aspects of the speaker: A contrastive study of Japanese and English.” Journal of Pragmatics 32:1623–1656. Hyman, Larry and Bernard Comrie. 1981. “Logophoric reference in Gokana.” Journal of African Languages and Linguistics 4:19–37. Jespersen, Otto. 1924. The Philosophy of Grammar. London: Allen & Unwin. Kamp, Hans. 1981. “A theory of truth and semantic interpretation.” Formal Methods in the Study of Language, ed. by J. Groenendijk, T.M.V. Jannssen and M. Stokhof, Mathematical Centre Tract 135, Amsterdam. Kamp, Hans and Christian Rohrer. 1983. “Tense in Texts.” Meaning, Use and Interpretation of Language. ed. by Bauerle, R., C. Schwarze, and A. von Stechow. Berlin: de Gruyter. Kamp, Hans & Uwe Reyle. 1993. From Discourse to Logic. Dordrecht: Kluwer. Kuno, Susumo. 1972. “Pronominalization, reflexivization, and direct discourse.” Linguistic Inquiry 3:161–195. Kuno, Susumo. 1987. Functional Syntax. Chicago: University of Chicago Press. Kuroda, Yuki. 1973. “Where epistemology, grammar and style meet: A case study from Japanese.” A Festchrift for Morris Halle, ed. by Anderson, S. and P. Kiparsky. New York: Holt, Rinehart & Winston. Mitchell, Jonathan E. 1986. The Formal Semantics of Point of View. Ph.D. dissertation, University of Massachusetts at Amherst. Ogihara, Toshi. 1996. Tense, Attitude, and Scope. Dordrecht: Kluwer. Partee, Barbara. 1973. “The syntax and semantics of quotation.” A Festchrift for Morris Halle, ed by Anderson, S. and P. Kiparsky. New York: Holt, Rinehart & Winston. Peterson, Philip. 1997. Fact, Proposition, Event. Dordrecht: Kluwer. Pollard, Carl and Ivan Sag. 1992. “Anaphors in English and the scope of the binding theory.” Linguistic Inquiry 12:2651–2305. Prince, Gerald. 1987. Dictionary of Narratology. Lincoln, Nebraska: University of Nebraska Press. Reinhart, Tanya. 1975. “Whose main clause? Point of view in sentences with parentheticals”. In S. Kuno (ed.), Harvard Studies in Syntax and Semantics, No. 1. Reinhart, Tanya and Eric Reuland. 1993. “Reflexivity.” Linguistic Inquiry 24:657–720. Ross, John Robert. 1970. “On declarative sentences.” Readings in English Transformational Grammar, ed. by Jacobs, R. and P. Rosenbaum. Waltham, Mass: Ginn & Company. Schlenker, Phillip. 1999. Propositional attitude and indexicality: A cross categorial approach. Ph.D. dissertation, MIT. Sells, Peter. 1987. “Aspects of logophoricity.” Linguistic Inquiry 18:445–479. Smith, Carlota S. 1991. The Parameter of Aspect. Dordrecht: Kluwer. Smith, Carlota S. 1996. The Relation between Aspectual Viewpoint and Situation Type. Linguistic Society of America address. Published electronically, Eric database. Smith, Carlota S. 1999. “Activities: States or Events?” Linguistics and Philosophy 22:479–508. Smith, Carlotta S. 2003. Discourse Modes. Cambridge: Cambridge University Press. Speas, Margaret A. 1999. “Person and point of view in Navajo.” Proceedings of the West Coast Conference on Formal Linguistics. CSLI Publications 18, Stanford, California. Stirling, Lesley. 1993. Switch-Reference and Discourse Representation. Cambridge, England: Cambridge University Press. Thráinsson, H. 1976. “Reflexives and subjunctive in Icelandic.” Proceedings of the New England Linguistics Society 6:25–39. Vendler, Zeno. 1972. Res Cogitans: An Essay in Rational Psychology. Ithaca: Cornell University Press. Zribi-Hertz, Anne. 1989. “Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse.” Language 65:695–727.
Author Index
A Abney, S., 232 Abusch, D., 376 Ahn, Y. O., ix Andersen, R. W., 206 Antinucci, F., 221, 230 Aristar, A., 20, 21 Aristar-Dry, H., v, xiii, 20, 21, 71, 74, 345, 346–348 Asher, N., 83, 166, 176, 378, 379 Austin, J., 306 Axelrod, M., 31, 41 B Bach, E., 65, 67 Baker, C. L., 385 Banfield, A., 293, 374, 375, 376, 377, 382, 387 Bartels, C., 244 Barthes, R., 357 Bauer, P., 230 Bellugi, U., 235 Bennett, M., 11, 66, 152 Benveniste, E., 122, 244, 245, 295 Bloom, L., 210, 219, 230, 234, 235 Boas, F., 21 Bohnemeyer, J., 188, 206, 319 Bonevac, D., 226, 227 Borker, R. A., lvii Braroe, E., 101, 107, 117 Bronckart, J. P., 206, 212, 213, 215, 218, 219, 220, 230 Brown, R., 79, 168, 210, 212, 215, 230, 234 Butler, M. C., 359 Bybee, J., 190 C Cabral, L., 217 Caenepeel, M., 170, 198, 381, 382, 383 Cairns, W. B., 171
Cantrall, W., 384 Carlson, G., 9 Chafe, W., 21, 333. 350 Chao, Y. R., 246, 310, 315, 318 Charniak, E., 351 Chatman, S., 377 Cho, E., 185 Chomsky, N., vii, viii, 125, 233, 251, 252, 384 Christensen, M. B., 320, 333 Cinque, G., 380 Clark, H., 167 Cochrane, S., 211 Cohn, D., 293, 377 Comrie, B., 8, 9, 30, 67, 69, 162, 183, 184, 191, 194, 211, 226, 375, 378, 387 Cowper, E., 232 Craig, C., 21 Crain, S., 233 Cromer, R., 209, 212 Crystal, D., 100 Culler, J., 362 D Dahl, O., 30, 184, 190, 191, 194 De Villiers, J., 210, 231 De Villiers, P., 210, 231 Debray-Genette, R., 352 Depraetere, I., 65, 66, 72, 309 DiPaolo, M., 218, 220, 230 Dowty, D., 3, 5, 7, 8, 10, 26, 63, 64, 65, 66, 70, 71, 72, 77, 79, 152, 153, 166, 214, 383 Dry, H., 20 See also Aristar-Dry, H. Ducrot, O., 16, 17 E Elliot, D., 267 Emonds, J., 104 Enc¸, M., 185, 191
R.P. Meier et al. (eds.), Text, Time, and Context, Studies in Linguistics and Philosophy 87, C Springer Science+Business Media B.V. 2009 DOI 10.1007/978-90-481-2617-0 BM2,
395
396 Erbaugh, M. S., ix, 181, 185, 245, 246, 303, 319, 333 F Farber, L., 359 Fernald, T., 5, 184, 193 Fillmore, C. J., 267 Fleischman, S., 186, 376, 387 Flier, M., 80 Fodor, J., 252, 266, 267 Forsyth, J., 28 Freed, A., 10, 28 Friedman, N., 293 Friedman, W., 83, 226 Fukui, N., 227 G Gallagher, M., 123 Galton, A., 80 Garcia, R., 206 Garrett, M., 252 Gelman, R., 81, 206 Genette, G., 377 Giorgi, A., 179, 306, 376, 378, 386 Givon, T., x, 21 Gleitman, L., viii, xxxiv, xli, 207, 241–247 Goldin-Meadow, S., 207 Gopnik, I., 260 Grevisse, M., 15, 18 Grice, H. P., 307, 349, 350, 351 Grimshaw, J., 234 Guilfoyle, E., 234 Gunlogson, C., 244 H Haas, M., 21 Hafitz, J., 230 Hag`ege, C., 377 Hale, K., 39, 47, 185 Halliday, M. A. K., 216, 217, 229, 252, 258 Hamburger, K., 122, 293 Hardy, F., 31, 56 Harms, R., xxxix, xliii Harner, L., 217 Harris, Z. S., xvi, 384, 385 Hatav, G., 72, 73 Heinamaki, O., 66, 69, 75 Herriot, P., 217 Herweg, M., 63, 65, 66, 67, 71, 76, 382 Hinrichs, E., 65, 66, 70, 71, 79, 92, 163, 168, 184, 309, 313 Hirose, Y., 375, 378, 385, 386 Hiz, H., 258 Hobbs, J. R., 288
Author Index Hochberg, J., 280, 281 Hockett, C., 210 Hofmann, T., 105, 127 Hoijer, H., 21, 34, 56 Hopper, P. J., x Horn, L., 307 Hornstein, N., 102, 162, 179, 191 Hu, J., 246, 310, 321, 322 Huang, C. T. J., 184, 246, 310 Huang, L. M., 315 Hyman, L., 378 I Iatridou, S., 93, 185, 186, 187, 188, 190 Inhelder, B., 205 J Jakobson, R., 250, 258 James, D., 185 Jenkins, L., 101 Johansen, S., 358 K Kalectaca, M., 21 Kamp, H., 4, 18, 63, 70, 82, 92, 94, 162, 168, 169, 174, 175, 177, 178, 179, 184, 190, 196, 305, 306, 308, 309, 310, 320, 335, 389 Kanisza, G., 279, 280 Kari, J., 29, 31, 42, 56 Keenan, E., 258 Kenny, A., 7, 62 Kiparsky, P., 117, 123 Klein, W., 184, 188, 307, 309, 310, 315, 316, 334 Klima, E. S., 235 Krauss, M., 41 Krifka, M., 62, 65, 166, 318 Kuczaj, S., 210, 216 Kuno, S., 362, 373, 384, 385, 386 Kuroda, Y., 378 L Labov, W., 69, 196, 301, 334 Lai, I., x Lakoff, G., 8, 101, 263, 264 Lakoff, R. T., xlvi Lawler, J., 120 Lebeaux, D., 234 Leech, G., 8, 100 Leslie, A. M., 81, 82 Levin, B., 321 Levin, S., 258 Levinson, S., 165, 187, 301, 307 Li, C. N., x
Author Index Li, D., 313, 315, 316 Lifter, K., 230 Lin, J-W., 246, 310, 314, 315, 316, 319 Lips, M., 293 Lyons, J., 10, 190, 306, 323, 340 Lyytinen, P., 231 M Malotki, E., 184 Maltz, D., lvii Mandler, J., 230 Mangione, L., 313, 315, 316 Mateo Toledo, E., x Materson, D., 235 McAlister, E., 281 McCawley, J., 123, 127, 298 McDonough, J., 29 McGilvray, J., 299 Merlin, F., 22 Midgette, S., 29, 30, 31, 36, 38, 45, 47, 48, 49, 55, 56 Milic, L., 253 Miller, G., 210, 274 Miller, R., 221, 230 Minsky, M., 351 Mitchell, J. E., 383 Mittwoch, A., 32, 33, 63 Moens, M., 5, 74, 168, 173, 196, 334 Morford, J., 207 Morgan, W., 21, 28, 29, 30, 31, 40, 41, 42, 45, 47, 53, 56, 349 Mourelatos, A., 65, 67 Mukarovsky, J., 260 N Neisser, U., 250 Nelson, K., 230 Newton, B., 211 Nichols, L., 186 Noonan, M., 234 O O’ Grady, W., 235 Ogihara, T., 376 Ohmann, R., 253 Ota, A., 296 P Palmer, F., 190 Pan, H., 246 Parsons, T., 65, 67 Partee, B., 11, 65, 66, 70, 71, 92, 94, 152, 161, 164, 168, 184, 194, 273, 286, 309, 335, 375, 376
397 Perkins, E., 5, 184, 193 Perner, J., 231 Peters, A., 235 Peters, S., 187, 263, 264, 360 Peterson, P., 166, 378 Piaget, J., 205, 206, 212, 229 Pianesi, F., 179, 306, 376, 378 Pinker, S., 233 Pinon, C., 64, 67, 80 Pollard, C., 384 Pollock, J-Y., 183, 186 Prince, G., 359, 363, 371 Proust, M., 362 Pustejovsky, J., 65 R Radford, A., 234 Rathmann, C., x Reichard, G., 21 Reichenbach, H., 70, 89, 95, 96, 119, 124, 148, 151, 162, 169, 179, 184, 185, 188, 196, 205, 209, 211, 227, 274, 285, 286, 298, 299, 303, 309, 310, 312, 339 Reinhart, T., 377, 378, 384, 385 Ren, F., x Reuland, E., 378, 384, 385 Reyle, U., 4, 63, 70, 82, 162, 169, 174, 175, 177, 178, 179, 184, 190, 305, 306, 308, 309, 310, 389 Rice, K., 41, 313 Riddle, E., 107, 127 Rohrer, C., 14, 70, 92, 94, 168, 171, 177, 196, 287, 335 Ross, J. R., 107, 244, 252, 264, 373, 384 Russell, B., xv, liv, 241, 253 Ryle, G., 26 S Sachs, J., 217, 229 Sadock, J., 118 Sag, I., 384 Sapir, E., 21, 56 Schilder, F., 76, 81 Schlenker, P., 375 Schmerling, S., 279 Scollon, R., 333 Sells, P., 389 Shatz, M., 206 Shi, Z., 315 Shipley, E., viii, xxxii, xxxiv, xli, 207, 241 Shipman, C., vi Shirai, Y., 206 Sinclair, H., 206, 212, 213, 215, 218, 219, 220, 230
398 Slobin, D., 230, 231 Smith, C. S., v, vi, vii, viii, ix, x, xxi–lx Speas, M., 29, 56, 227, 375 Stassen, L., 190, 322 Steedman, M., 5, 74, 159, 173 Steele, S., 185 Stenning, K., 286, 288 Stirling, L., 378, 385, 389 Stoel-Gammon, C., 217 Stowell, T., 179 Swift, M., 206, 31 T Taylor, B., 11, 63, 67, 309, 318 Thibaudet, A., 362, 363 Thompson, E., 179, 180 Thompson, S. A., x Thr´ainsson, H., 386 V Valian, V., 234, 235 van Kleeck, A., ix Velupillai, V., 184, 191 Vendler, Z., 3, 4, 5, 7, 10, 13, 14, 26, 61, 62, 64, 65, 152, 155, 166, 212, 214, 308, 378 Verkuyl, H., 65, 214 Virbel, M., 220 Visser, F., 20 Vlach, F., 6, 7, 11, 63, 66, 152, 153, 159
Author Index W Waletzky, J., 69, 196, 301, 334 Wall, R. E., 159 Watt, W., 252 Webber, B., 286 Weinrich, H., 220, 293 Weissenborn, J., 235 Weist, R. M., 206, 226, 229, 230, 231, 236 Whitaker, J. T., x, xi, 16, 22, 295, 345, 346, 347 Whorf, B.L., xxv, 26, 61, 166, 195, 308 Wolfson, N., 295 Woodbury, A., 21 Wu, J-S., 315, 321, 325 Wysocka, H., 231 X Xu, L., 246, 333 Y Yang, J., 319, 320, 329 Yeh, M., x, 315 Young, R., 21, 28, 29, 30, 31, 36, 40, 41, 42, 45, 47, 53, 54 Z Zagona, K., 179 Zribi-Hertz, A., 384, 385
Subject Index
A Acquisition future events, 229–231 past events, 216, 217, 229 past tense, 218, 226, 234, 235 imperfective, 215–217 perfective, 216 progressive -ing, 215 simple aspect, 218 temporal ordering, 212–213, 221, 233 Activities, 3, 4, 5, 6, 8, 11, 12, 13, 19, 26, 61–84, 152, 172, 212, 213, 214, 221, 300, 308, 324, 351 perfective, 69, 71, 72, 74, 76–81, 83, 84 type, 67, 164, 165 Advancement, xi, 69, 70, 71, 72, 73, 74, 77, 80, 81, 83, 84, 91, 165, 167–168, 169, 172, 174, 177, 178, 179, 181, 194, 195, 196, 197, 198, 199, 246, 247, 304, 334, 335, 345 Adverbials completive, 64, 91, 93, 155, 156, 189, 389 complex time adverbials, 100 durative (time), 92, 172, 173, 197, 337 habitual adverbs, see Habitual time adverbials, 95–100, 102–104, 106, 109, 121, 122, 125, 127, 129, 134, 140, 143, 144, 147, 148, 150, 173, 196 past/present/future, 97–101, 124–125, 129 unanchored, 98, 105, 106, 110, 124, 132, 282 Afar (Cushitic), 21 Aktionsarten, 5, 10
Anaphora, 173, 176, 177, 178, 179, 180, 181, 196, 243, 247, 258, 286, 333, 347 principle, 92, 177, 178, 179, 180 Anglo-Saxon, 20 Arabic, 21 Aspect acquisition of, 206–207 aspectual choices, non-standard, 3, 5, 8, 10, 14, 20, 27, 159 aspectual information, 55, 69, 71, 163, 164, 169, 171, 175, 177, 179, 180, 181, 196, 247, 303, 304–311, 334, 335, 338 aspectual viewpoint imperfective, 20, 22, 25, 30, 31, 32, 33, 35, 37, 40, 43, 44, 50, 57, 68, 81, 188, 382, 387, 388 neutral, 42, 68, 80, 189, 190, 193, 308, 309, 316, 317, 319 perfective, 6, 10, 11, 20, 30, 42, 48, 49, 50, 52, 54, 61, 68, 69, 80, 82, 163, 172, 188, 306, 309, 317, 382 progressive, 10, 11, 34, 44, 45, 53, 64, 66, 69 imperfective morpheme, 21, 22, 30, 40, 41, 42, 316, 324 perfective morpheme, 21, 22, 41, 42, 80, 310, 315, 329 progressive, 8, 10, 14, 210, 218, 219, 277, 278 sentential, 7, 10, 21 “simple aspect” (English), 4, 10, 11–13, 14, 15, 20, 214, 215, 218, 219 two component theory of, ix, 25
399
400 B Boundedness, 66, 77–79, 80, 84, 163, 246, 305, 309, 317, 319, 320 C ChiBemba (Bantu), 21 Chinese, see Mandarin Chinese Cohesiveness, 350 Complements, temporal interpretation of, 107–119, 136, 320 speaker-oriented, 117, 118 Continuity, 92, 169, 173, 176, 177, 178, 180, 181, 196, 197, 233, 315, 333, 334 principle, 92, 177 Cooperation, 346, 349, 362 Coptic, 21 Counterfactual interpretation, 187, 199 Covert linguistic categories, 25, 26, 61, 195, 308 D Deictic deictic pattern of temporal form and interpretation (Mandarin), 305 deictic principle of tense interpretation (English), 176, 178 Derived subjects vs. underlying subjects, 267, 268 as topics, 267–269 Discourse Discourse representation structure (DRS), 164, 174, 175, 176, 177, 178, 194, 317, 320, 389, 391 Discourse representation (DR) theory, x, 4, 70, 94, 161, 164, 174–178, 181, 189, 190, 199, 308, 309, 311, 317, 320, 335, 338, 347, 371, 389, 390, 391 dynamics, 69–71, 168 conventional time, 82–83 narrative time, 69, 82–83 reference time, see Time features as surface features, 264, 265 focus, 346, 350, 351, 357, 358, 361 modes argument, xi, 91, 165, 167, 194, 195 description, 176, 181, 337 information, 181 narrative, 176–179 report, 176, 181, 198
Subject Index E English adverbials, 147–152, 155–158 after-clauses, 215, 227 before-clauses, 215, 227 French, comparisons to, 20–22 futurate aspectual interpretations of, 151 progressive, 152, 153, 155, 156, 158 imperfective, 158 conjunction test for, 155–156 representation of, 151 simple, 157 gerundive, 76, 381 have, 89, 96, 98, 105–106, 110, 115, 127, 128, 131, 134, 175, 294, 297 imperfective, 214, 219 perfective, 20, 215, 219 progressive, 218, 219 simple aspect, 4, 11–13, 214, 215, 218, 219 will-future, 147, 149 Ellipsis constructions definite reference, 356–357 material indirectly introduced, 359–360 pronominal reference, 357–359 types definite nounphrase not explicitly identified, 345, 350 phrase triggering Gricean implicature, 350 pronouns without antecedents, 345, 350 Events bounded, 77, 83, 92, 163, 164, 169, 171, 173, 177, 180, 196, 197, 303, 304, 305, 306, 312, 316, 317, 319, 324, 334, 335, 338, 382 perfective, 68, 69, 84, 168, 214, 218 unbounded, 163, 164, 172, 173, 177, 180, 181, 197, 198, 309, 317, 325, 331, 332, 335, 337, 338, 348 F Foreground (Foregrounding), x, 260, 320, 329, 358 Frame adverbs, see Mandarin Chinese Framing (and Frames), 323, 351, 356, 364, 365 French conditional, 186, 187, 188, 189 English, comparisons to, 20–22 futur, 191, 192, 193, 199 historical present, 100, 125, 129, 134, 137, 138, 139, 142, 171, 220, 221, 244, 295, 297
Subject Index imparfait, ix, 4, 14, 15–17, 18, 20, 186, 187, 188, 189, 190, 192, 199, 219, 220, 377, 382, 387 pass´e compos´e, ix, 14, 18–20, 188, 189, 190, 192, 193, 196, 213 pass´e simple, ix, 14, 70, 188, 189, 190, 196, 199, 220 perfective viewpoint, 20, 22 plus-que-parfait, 186, 187, 188, 189, 190, 199 pr´esent, ix, 186, 187, 192, 213 si-clauses, 188, 189, 190 Future acquisition of future reference, 211, 212, 221, 222 English futurate construction, 147–159 English will-future, 147, 149 French futur, 191, 192, 193, 199 future-in-past, 99, 100, 321 Mandarin future-oriented modals, see Mandarin Chinese Mandarin future-oriented verbs, see Mandarin Chinese Navajo future mode, 29–31, 41, 42 prepositions, 282 reference time (Future RT), 98, 99–101, 105, 106, 111, 113, 114, 117, 119, 123, 124, 130, 135, 136, 140, 278, 282 temporal adverbials, 97–101, 129, 147, 290, 323
401 I Implicature, 12, 13, 345, 350, 351, 361, 363, 364 Inceptives, 10, 12, 19, 28, 33, 35, 38, 39, 40, 43, 44, 45, 47, 49, 53, 54, 56, 57 Inchoative interpretation, 74, 309, 383 Incomplete sentences, 103, 121, 128, 132, 133, 135, 136, 243, 276, 277, 278, 279, 280, 281 Inference, xxvi, 49, 65, 67, 71, 74, 78, 196, 210, 303, 307, 321, 334, 345, 347, 349, 350, 356, 363, 364, 365, 366, 367, 368, 383 Inflectional phrase (IP), 174, 175, 179, 180, 207, 232, 233 J Jacaltec, 21 L Lalana Chinancotecan, 21–22
G Generative grammar adverb movement, 124, 127, 132, 133, 143, 144 reduction transformation, 127, 131, 140–141, 143, 145 tense copying transformation, 143, 144 time adverbial, 127, 129, 134–136 German, 122, 235 Germanic languages, 20, 231 Gricean Maxim of Quantity, 187
M Mandarin Chinese aspectual morphemes, 246, 315, 316, 317, 318, 333, 336 bounded event constraint, 304, 306, 312, 319, 324, 338 deictic pattern, 304–311, 324, 329, 331, 334, 338 frame adverbs, 323 future-oriented modals, 323 future-oriented verbs, 304, 320, 321, 324 narrative dynamism principle, 335 past-oriented verbs, 320 perfective viewpoint, 306, 309, 317 resultative verb complement (RVC), 315, 320, 324, 326, 328, 329, 335, 336 static interpretation principle, 337 temporal schema principle, 303, 304, 307, 318, 319, 322, 324, 335, 338 zero-marked clauses, 309, 315, 318, 319, 330, 335, 338
H Habitual, 5, 9, 10, 12, 13, 14, 15, 18, 21, 27, 43, 63, 64, 71, 91, 97, 99, 119–122, 124, 125, 128, 130, 132, 133, 143, 144, 166, 211, 242, 273, 274, 276, 277, 289, 290, 291, 315, 318, 324, 358 See also English, simple aspect Hebrew, Biblical, 73, 185, 226 Hopi, 21, 22, 184
N Narrative dynamism principle French, 197 Mandarin, 335 Navajo aspectual viewpoint broad, 27, 28, 38, 39, 40, 45, 47, 48, 49, 54 narrow, 27, 28, 38, 44, 45, 46, 47, 50, 53, 54, 55
402 imperfective mode, 41 perfective mode, 41 progressive mode, 29, 30, 31, 32, 34, 42, 44, 53 verb structure, 28–30 verb lexeme categories (VLC), 30–32, 34–37, 50, 51–55 Non-standard aspectual choices, see Aspect O Orientation principle, 90, 91, 108, 112, 113, 114, 115, 116, 117, 118, 119 P Past tense, ix, 14, 20, 70, 81, 92, 93, 94, 98, 99, 100, 101, 105, 107, 112, 115–117, 119, 124, 125, 137, 140, 164, 169, 170, 171, 175, 183, 185, 186, 188, 189, 190, 191, 192, 193, 194, 196, 206, 210, 213, 215, 216, 217, 218, 226, 230, 234, 235, 242, 243, 244, 282, 286, 287, 288, 295, 296, 297, 298, 299, 307, 310, 311, 377, 387, 389 See also Acquisition; French, imparfait, pass´e compose, pass´e simple Perspective perception, 381–383 direct, 381 indirect, 381, 382, 383 inferred, 383 See also Subjectivity; Point of view Polish, 80, 230 Point of view communication, 374–377 contents of mind, 377–379 evaluation, 379–380 See also Perspective; Subjectivity Preposing of adverbials, 250, 257 for emphasis, 262 of subjects, 257, 267–269 Present tense historical present, 99–101 pr´esent, see French present progressive, ix, 171, 243, 278 See also Tense, present Process, 43, 49, 50, 89, 91, 152–154, 157–159 extended, 153, 154, 157 preliminary process, 154, 157, 158 Projective propositions, 379, 387, 388 Prominence emphasis, 249, 251
Subject Index focus, 251 phonological cues, 252 punctuation, 250, 252 syntactic cues, 252–253 Pronouns interpretation and contiguity, 350, 351 interpretation and discourse focus, 346, 351 reflexives, 18, 348, 373, 381, 384, 385, 386, 388 locally free, 348, 385 possessives, 386 Q Questions, viii, 94, 233, 235, 244, 375, 379 Quich´e, 21 R Relative clause, viii, xxxii, 93, 96, 191, 192, 193, 199, 250, 251, 252, 253, 255, 256, 257, 258, 261, 262 Responsibility, 297, 373, 374, 376, 378, 380, 386, 388, 389, 390, 391 Rhythm, 361 Romance languages, 20, 80, 306 Russian, ix, xxiv–xxv, liv, 3, 20, 27, 28, 32, 58, 80, 230, 260, 306, 375, 382 S Scope, 34, 41, 63, 91, 92, 147, 150, 154, 155, 159, 172, 173, 197, 225, 228, 229, 230, 231, 233, 236, 299, 300, 315, 319, 321, 323, 324, 330, 331, 337, 352, 363, 372, 380, 385 Semantic interpretation, x, 3, 67, 122, 123, 125, 126, 128, 132, 134, 136, 140, 141, 145, 147, 152, 161, 162, 183, 189, 300, 308, 317 interpretive rules, 134–143 Semelfactive, 5, 26, 31, 54, 55, 63, 308 Seneca, 21 Sentences connexion, 258, 259 embedding, 259 final position in (or end of sentence), 242, 250, 253, 257, 258, 261, 262, 263, 264, 268 habitual, 12, 95, 119–122, 130, 132, 133 incomplete, 103, 121, 128, 132, 133, 135, 136, 243, 276, 277, 278, 279, 280, 281 initial position in (or beginning of sentence), 242, 250, 256–257, 259, 261, 354
Subject Index as position of emphasis, 262 and sentence connexion, 254, 258, 259 and surface structure, 259, 262 and topicalization, 250 structures occurring in, 254 vs. final position, 250, 256, 261 as-embedding, 255, 258 relative clause embedding, 250, 257 non-habitual, 119, 121 vague, 246, 247, 273, 276, 277, 309 Sharing principle, 90, 91, 108, 110, 112, 113, 114, 115, 116, 117, 118, 119, 131 Situation aspect, 4, 6, 10, 11, 13, 14, 20, 21, 68 Situation entities abstract entities, 195 facts, 166 propostitions, 167 atemporal, 167 events, 20, 43–50, 317 eventualities, 166, 195, 196 general statives, 166, 167, 195, 196 states, 317 temporal, 167 Situations future, 154, 155, 156, 157, 158, 309, 321, 333 predictive, 157, 158, 159 present, 9, 19, 25, 26, 80, 154, 156, 157, 227, 324 statives, 8, 10, 11, 20, 30, 41–43, 195, 306 specifying approach, 158 structural approach, 157 Situation types accomplishments, 3, 4, 11, 12, 13, 15, 26, 61, 63, 64, 65, 67, 68, 70, 80, 152 achievements, 5, 7, 11, 12, 13, 15, 18, 19, 26, 61, 63, 64, 65, 67, 68, 70, 80, 152, 308 activities, 4, 5, 6, 8, 11, 12, 13, 19, 26, 61–86, 152, 172, 212, 213, 214, 308, 324 atelic, 26, 27, 33, 34, 40, 43, 48, 52, 54, 55, 57, 58, 62, 63, 64, 72, 81, 173, 197, 308, 318, 319, 325, 326, 335, 337 discrete, 64, 65–67 durative, 26, 28, 32, 33, 44–50, 52, 57, 62, 76, 317 dynamic, 26, 34, 62, 64, 67, 80, 81, 82, 83, 84 habitual, see Habitual
403 idealized situation type, 9, 10, 11, 13, 14, 16, 20, 22 endpoints, 10, 11, 12, 13, 27, 28, 44 instantaneous, 33, 43–44, 62 non-discrete, 65–67 static, 25, 26, 67 telic, 33, 40–41, 43, 48, 52, 54, 57, 58, 62, 63, 72, 308, 318, 319 temporal features of atelic, 26, 33, 34, 40, 41, 62, 81, 308 durative, 32, 35–40 dynamic, 32, 34 instantaneous, 35–40 static, 32, 34, 62 telic, 26, 32, 33, 40–41, 62, 308 Spanish, 20, 58, 186, 230 Speech direct, 293, 347, 376 indirect, 293, 347, 348, 374, 375, 376, 377, 378, 388, 389, 390 quoted, 374, 375 represented speech and thought, 376, 377 thoughts and beliefs, 378 Standpoint, 162, 169, 338, 347, 371, 373, 381, 383–387, 388 States, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 17, 18, 19, 20, 25, 26, 32, 33, 35, 41, 61–84, 91, 117, 119, 120, 121, 144, 157, 162, 163, 165, 166, 167, 168, 169, 172, 173, 174, 177, 192, 193, 194, 195, 197, 226, 274, 304, 305, 306, 308, 309, 312, 316, 317, 318, 319, 334, 335, 337, 387 Strong decentring hypothesis, 205, 206, 209, 213–219 Stylistic analysis, 249, 252, 253–264 Style indirect libre, 244, 363, 377 Subjectivity, x, xi, 183, 295, 345, 346, 347, 348, 352, 371–392 Surface structure importance of, 95, 123, 124, 127, 128, 174, 249 interpretation rules, 95, 275 function of, 241, 242, 249, 264, 265 T Tense acquisition of, ix, xv, 225–236 features/properties morpho-syntactic, 185 semantic, 185 future, ix, 101, 124, 169, 183, 185, 191, 193, 194, 231, 286
404 grammatical properties of, ix, 184, 190, 227, 231–233 interpretation in single sentences, 161, 164 in context/discourse, 164–166 past, ix, 14, 20, 70, 81, 92, 93, 94, 98, 99, 100, 101, 105, 107, 112, 115–117, 119, 124, 125, 137, 140, 164, 169, 170, 171, 175, 183, 185, 186, 188, 189, 190, 191, 192, 193, 194, 196, 206, 210, 213, 215, 216, 217, 218, 226, 230, 234, 235, 242, 243, 244, 282, 286, 287, 288, 295, 296, 297, 298, 299, 307, 310, 311, 377, 387, 389 perfect, 162, 176, 177 present, ix, 91, 98, 101, 106, 112, 115–117, 118, 123, 124, 136, 137, 147, 150, 151, 164, 171, 181, 184, 185, 186, 190, 191, 192, 194, 197, 206, 210, 218, 226, 227, 231, 235, 273, 277, 278, 281, 282, 286, 288, 295, 296, 297, 298, 299, 305, 306, 307, 311, 375, 376, 384 simple, 162, 164, 185, 214, 226 Tenseless languages, 161, 181, 304, 307, 339 Terminative interpretation, 74, 75, 76, 77 Time Event Time (ET), x, 3, 62, 89, 93, 95, 96, 97, 102–106, 110–111, 148, 149, 150, 151, 162, 184, 206, 211, 243, 274, 298, 309, 316 experiential, 226 linguistic, 226 logical, 226 Reference Time (RT), x, 3, 70, 71, 79, 83, 89, 90, 93, 95, 96, 97, 99, 101, 103, 108, 110, 115, 123, 148, 149, 150, 151, 152, 162, 184, 185, 188, 192, 196, 197, 206, 211, 227, 243, 245, 246, 274, 275, 298, 303, 309, 310, 311–314, 315, 316, 317, 323, 334, 336, 337, 338 Situation Time, 184, 185, 303, 309, 310, 311, 312, 314, 315, 316, 317, 337, 338
Subject Index Speech Time (ST), x, xi, 89, 90, 93, 95, 96, 97, 148, 150, 151, 162, 183, 184, 185, 188, 190, 192, 194, 195, 196, 198, 206, 211, 226, 227, 243, 244, 246, 274, 275, 296, 298, 303, 304, 305, 306, 309, 311, 312, 313, 314, 324, 334, 337, 338, 339 topic time, 184, 188, 310, 316 temporal relations/ordering anteriority, 94, 97, 98, 105, 106, 109, 110, 130, 131, 192, 210, 213, 226, 274, 275, 286, 287, 290, 294, 296, 297, 316 posteriority, 94, 97, 98, 106, 109, 110, 111, 130, 192, 206, 210, 226, 274, 286, 287, 290 simultaneity, 94, 96, 97, 98, 109, 117, 118, 130, 193, 211, 221, 226, 250, 274, 275, 276, 279, 286, 296, 298 Topicalization, 250 Transformations frequency (in Bertrand Russell’s writing), 264 function of, 265 types embedding, 255, 256, 258 resulting in complex NPs, 255, 256 reordering, 257, 265, 268 reader awareness of, 253 Tunica, 21
V Verb(s) constellation, 27, 28, 33, 34, 35, 38, 48, 56, 57, 58, 63, 64, 65, 68, 74, 80, 163, 164, 166, 172, 173, 191, 195, 196, 197, 228, 232, 233, 308, 315, 316, 318, 319, 320, 324, 326, 329, 335, 337 future-oriented, see Mandarin Chinese lexeme categories (VLC), see Navajo past-oriented, see Mandarin Chinese perception, 372, 378, 381, 383 Viewpoint, see Aspect