Perspectives on Grammar Writing
Benjamins Current Topics Special issues of established journals tend to circulate wit...
86 downloads
1403 Views
1MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Perspectives on Grammar Writing
Benjamins Current Topics Special issues of established journals tend to circulate within the orbit of the subscribers of those journals. For the Benjamins Current Topics series a number of special issues have been selected containing salient topics of research with the aim to widen the readership and to give this interesting material an additional lease of life in book format.
Volume 11 Perspectives on Grammar Writing Edited by Thomas E. Payne and David J. Weber These materials were previously published in Studies in Language 30:2 (2006)
Perspectives on Grammar Writing
Edited by
Thomas E. Payne University of Oregon & SIL International
David J. Weber Summer Institute of Linguistics
John Benjamins Publishing Company Amsterdam / Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Perspectives on grammar writing / edited by Thomas E. Payne, David J. Weber. p. cm. (Benjamins Current Topics, issn 1874-0081 ; v. 11) Includes bibliographical references and index. 1. Language and languages--Grammars--Authorship. I. Payne, Thomas Edward, 1951- II. Weber, David J. P207.P47 2007 808'.066418--dc22
2007024712
isbn 978 90 272 2241 1 (Hb; alk. paper)
© 2007 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
To William Bright (1928-2006)
Shortly after the original publication of this volume, Bill Bright left us, after a full life and illustrious career as an anthropological linguist, professor, editor and advocate for indigenous language communities. With sadness and joy we dedicate this republication to him, our teacher, mentor, role-model, colleague and friend.
Table of contents Introduction
1
Contextualizing a grammar William Bright
11
Writing grammars for the community James Lokuuda Kadanya
19
Collective field work: Advantages or disadvantages Aleksandr E. Kibrik
25
Grammars and the community Marianne Mithun
45
From parts of speech to the grammar Pamela Munro
71
Grammar writing for a grammar-reading audience Michael Noonan A grammar as a communicative act, or what does a grammatical description really describe? Thomas E. Payne
113
127
A typology of good grammars Keren Rice
143
Thoughts on growing a grammar David J. Weber
173
The linguistic example David J. Weber
199
Index
215
Introduction Thomas E. Payne University of Oregon and SIL International
A grammatical description is an essential component of any program of language documentation and preservation. With over half the languages in the world currently in danger of extinction within a century, the need for descriptive grammars is more critical than ever. Potential grammar writers often find themselves paralyzed by the massive prospect of describing a language. Additionally, grammar writing is a long term task, and more immediate concerns often preclude the expenditure of time and energy necessary to accomplish an adequate grammatical description. The papers in the present volume provide suggestions and encouragement from experienced grammar writers and users regarding concrete methods for approaching the complex task of writing a descriptive grammar of a language.
Introduction Current estimates are that about 3,000 of the 6,000 or so natural human languages now spoken will become extinct during the present century, unless some positive action is taken. When a language dies without documentation, all potential for enriching human experience embodied in the oral tradition and wisdom of that language is lost. Many have argued that the loss of diversity that language extinction represents is a scientific and human tragedy. Descriptive linguistics is an essential part of any program of language documentation and preservation. The mere existence of a good dictionary, collection of texts, and grammatical description confers a certain status on a language that may have previously been considered to be of little importance, by speakers and non-speakers alike. Good linguistic research communicates to minority language speakers and to surrounding groups that the minority language is worthy of respect. Furthermore, the products of descriptive linguistic research constitute part of the reference material necessary to develop indigenous educational materials and written literature.
Thomas E. Payne
Finally, from a scientific perspective, good linguistic descriptions provide the raw data for much research into the organization of the human mind. The tension between universality and diversity of language constitutes the subject matter for linguistics as a science. The central questions are: “How are all languages alike?” and “What are the limits to their variation?” Needless to say, from this perspective, a corpus of reliable and accessible data from as many languages as possible is essential. With every language that becomes extinct, the potential data source for this enterprise becomes narrower. Since at least Boas (1893, 1894 and 1911), the products of a research program of language documentation have consisted of three essential components: a dictionary, a collection of texts and at least some ethnographic and grammatical notes and observations. While the emphasis on one or the other of these components has varied over time, this collection of documents has circumscribed the work of descriptive linguistics of undocumented languages for more than a century. If none of these three components exist in one form or another, we may say that the language is “undescribed.” If some subset of them exist, or what does exist is inadequate for some reason, we may say the language is “underdescribed.” Current estimates are that approximately 5000 of the 6000 or so languages that are presently spoken fall into the categories of “undescribed” or “underdescribed.” Clearly there is much work yet to be done in the areas of language description and documentation. In the early years of the 20th century, texts were the foundation of descriptive linguistics. Typically, stories were recorded either by dictation, or on wax cylinders or wire recorders, and transcribed in a phonetic script. Free translations into a research language, such as English, were given in a separate column or page. Lexical, grammatical and ethnographic notes were provided to support the texts, but the texts were treated as the centerpiece of the program of language documentation (see, e.g., Boas 1894, Sapir 1921). As linguistics became an independent academic discipline, distinct from philology and anthropology, the notion of the written grammar began to undergo considerable elaboration, while texts and dictionaries tended to decline in importance for linguists. By the mid-twentieth century, complex linguistic grammars were being produced with minimal supporting lexical and textual data. Competing theories, or “models” of grammar were developed and became the focus of written grammatical descriptions, displacing the cultural patterns and knowledge implicit in the language of a speech community. Descriptive linguistics became less a matter of language documentation and more a proving ground for competing conceptualizations of linguistic theory. In recent years, perhaps with the publication of Krauss (1992), Hale (1992) and others in the same volume, the emphasis has shifted back to the centrality of
Introduction
data, principally naturally occurring texts and a detailed lexicon. Advancements in technology have made it much more feasible to provide close interlinear glosses, in addition to the traditional free translations of written texts. Furthermore, computational tools have made it possible to link the lexicon directly to the text corpus, thus promoting consistency and completeness. Technology for audio and video recording, archiving and accessing has also opened the way for richer, more reliable and faithful documentation of individual performances. These developments have had a synergistic relationship with a concurrent movement in linguistics toward becoming a more empirical, data-oriented discipline. This renewed emphasis on data has led a few to suggest that a linguistic grammar is no longer an essential part of a program of language documentation. Part of the argument is that a written grammar represents someone’s particular analysis, rather than documentation of objective fact. Written text and video/audio recordings are more likely to reflect the spontaneous reality of the subject matter, rather than an analyst’s interpretation. Furthermore, grammatical analysis is a very technical task, requiring years of study and experience to master. There are so many undocumented languages in the world, and so many of them are in imminent danger of becoming extinct, that the pressing need at this time is for pure documentation of data, with the technical aspects of grammatical analysis left for future generations. While there certainly is truth to the idea that one cannot meaningfully analyze a complex phenomenon such as a language without some theoretical “grid” through which to process one’s observations, nevertheless the pendulum swing away from the importance of the written grammar and toward a “just-the-facts” concept of language documentation can be interpreted as an (over) reaction to the excesses of earlier academic linguistics, in which formality, abstraction, and elegance took precedence over empirical considerations. We question whether it is possible to adequately prepare written texts for use by future generations without undertaking serious linguistic analysis. Certainly insightful and consistent glossing and even consistent transcription of data entail detailed grammatical knowledge. Therefore, a linguistically informed descriptive grammar will be a central component of any program of language documentation for many years to come.
The papers in this volume The papers presented in this volume represent products of a symposium on Grammatical Description of Underdocumented Languages, held in Dallas, Texas, at the SIL International Linguistics Center in October of 2003. The stated aims of the symposium were:
Thomas E. Payne
1. to raise the profile of methodology within the discipline of linguistics, 2. to elaborate the concept of the written grammar as a product of a program of linguistic analysis and documentation, 3. to encourage production of insightful and readable grammatical descriptions. The idea was to focus on the written grammar, in light of the kinds of changes the discipline has been undergoing in recent years. It is safe to say that all the contributors to this symposium affirm that an insightful grammatical description is important to a program of language documentation, and/or preservation and that it can and should address empirical and practical considerations, as well as being theoretically significant. In the course of the symposium, several recurring “themes” emerged. In the following paragraphs we will attempt to highlight some of those themes, and provide a glimpse into possible future directions for research and development in descriptive linguistics. The themes to be discussed include: 1. the necessity of community involvement in the conceptualization and elaboration of grammatical descriptions, 2. the link between a grammar and the other products of a program of language documentation (a dictionary and texts), 3. the complementary functions of artificially elicited vs. naturally occurring data, 4. grammatical description as “art” as well as “science.”
Community involvement While all of the contributors acknowledged that writing a linguistic grammar is of necessity a cooperative enterprise, two papers specifically discuss why and how a grammar might be produced that will serve the needs of the speech community as well as academic linguistics. James Kadanya provides a perspective of a native speaker currently involved in a collaborative grammar-writing project in an underdocumented language spoken in Sudan (Toposa, Nilo-Saharan). Marianne Mithun documents her work over several years developing a grammar of Mohawk in collaboration with the community, and with the community’s needs in view. Both of these papers emphasize the importance of balancing specificity, which is of more local and practical interest, with universality, which tends to be of interest to academic linguists, and others. Certainly, addressing both audiences in the same document is a major challenge.
Introduction
On the one hand, linguists depend on language communities for data. On the other hand, many communities whose languages are endangered are beginning to value linguistic training and methodologies in order to foster and maintain language use and awareness in the communities. Communities whose heritage language is no longer in use turn to linguists, either members of the community or outsiders, to help preserve what records there may be of the language, and maintain a sense of language identity in the face of social and cultural pressure from the majority language. In terms of future directions, several authors mention the need for more native speakers of underdocumented languages to obtain linguistic training. Noonan mentions that this would be facilitated if linguistics departments were more willing to accept descriptive grammars as MA Theses or PhD dissertations. As it is, many graduate students are discouraged from choosing straight descriptive topics in favor of more esoteric theoretical issues for their graduate work. Individuals whose priorities are the practicalities of language documentation and preservation are less likely to find theoretical topics satisfying. David Weber, in his paper on linguistic examples, also suggests that future grammars will be “web-delivered,” thus providing a measure of flexibility not available in print publications. For example, in an online grammar, users may be able to customize output by selectively viewing and hiding various kinds of information present in the document in order to meet specific needs.
The link between a grammar and a dictionary In her paper Pam Munro describes her experience of doing dictionary work in three languages, and points out various ways in which a grammar and a dictionary provide valuable support and cross-checking for one another. In particular, Munro reminds us that dictionary work underscores the need to classify and specify the part of speech of every word in a language, and gives concrete examples of how that works out in grammar writing. More generally, dictionary work can elucidate many features of grammatical analysis and reveal problems that must be dealt with that might not have been noticed otherwise. While grammar writers may feel free to ignore difficult or annoying words, a dictionary maker cannot do this (and be honest). In the absence of dictionary work a grammatical description is very likely to miss important patterns and features. Michael Noonan seconds the notion that a good dictionary is a powerful adjunct to a good grammar. Every grammar should, at bare minimum, contain a vocabulary consisting of all the lexical items employed in the grammar itself.
Thomas E. Payne
However, such vocabularies are no substitute for a full dictionary. A good dictionary contains lots of example sentences from which a great deal of useful grammatical information can be gleaned even about topics that the grammar writer/dictionary maker has no knowledge of or interest in. Bill Bright notes that one very valuable way to situate a written grammar of a language within the social and cultural context of its speakers is to link it closely to a dictionary. Bright calls for a revival of the genre of “ethnographic dictionaries,” that is works arranged like dictionaries, but with articles consisting of short ethnographic essays, describing an artifact, a category, or an institution within a particular culture. Such dictionaries of course require an immense investment of time, are the logical extreme from bare minimum “vocabularies” mentioned by Michael Noonan.
The tension between elicited and naturally occurring data One of the current controversies among descriptive linguists is the place of data in grammatical descriptions. Typically, grammars have used simple, unadorned examples to illustrate morphological and syntactic patterns. Such examples are commonly obtained by asking a bilingual native speaker to give a translation of a hypothetical utterance in a research language, such as English or Spanish. Examples of this sort tend to be clearer to audiences who are totally unfamiliar with the language being described. However, there is always the risk that such elicited examples may misrepresent the way the language is actually used in spontaneous conversation. Therefore, a tension arises between accessibility, and accuracy. Elicited data is controlled and focused, but may accentuate patterns found in the research language that are uncommon or unknown in spontaneous discourse in the described language. Text material, on the other hand, is uncontrolled, variable, and is likely to present elements and patterns not found in elicited material. Like a dictionary, natural text forces a linguist to be honest, and helps ensure that the grammatical description is a true to the “spirit” of the language, and not just to some theoretical preconception. David Weber’s paper on “the linguistic example” is perhaps the most explicit discussion of the place of linguistic examples in a grammar. This paper provides many concrete suggestions on how to present, format and “frame” linguistic examples, whether elicited or extracted from spontaneous discourse, so as to maximize their effectiveness in a grammar. Marianne Mithun describes how and why there may be a difference between the range of logically possible structures, and the distribution of structures that
Introduction
occur in natural speech. In many cases, native speakers may be able to provide literal translations of constructions in the research language, even when such constructions never occur in spontaneous conversation. Aleksandr Kibrik provides an overview of how descriptive linguistics of underdocumented languages has been practiced in Russia for the past 40 years. In the distinctive team fieldwork technique developed by the Moscow school, elicited data has always played a central role. Kibrik provides a balanced perspective by listing advantages and disadvantages of such an approach. In the Russian tradition, as in North America, texts were originally the centerpiece of a program of language documentation. However, without a thoroughgoing linguistic analysis, achieved via elicitation as well as analysis of written material, texts collected by early researchers were largely unusable to modern day speakers of the languages and to linguists. Tom Payne’s paper argues that elicitation is most appropriate for analysis of “lower level” linguistic material, such as phonology and morphophonemics, and for completion of “inventories” of inflectional, derivational and syntactic possibilities. Text material, on the other hand, is more useful for analyzing “higher level” patterns, such as intonation and syntactic structure.
Grammatical description as “art” as well as “science” A photograph may be a faithful representation of a person’s physical appearance, but is not a very good way to express the “essence” of who that person is; the personality, passions, talents, etc. So a grammar that focusses purely on structural facts may be accurate in a technical sense, while failing to capture something of the “personality” of the language. How does it “feel” to speak the language? Which nuances of meaning are central to the thought processes involved, and which are peripheral? What is this language particularly “good at”? A grammatical description should be more than a “photograph” of a language. Bill Bright argues that language is more than simply one “department” within culture, alongside law, religion, technology, social structure etc. It is the central clearinghouse of culture, within which the transmission, organization and elaboration of all the other departments takes place. Therefore a grammar of a language should be “contextualized” as much as possible. This involves recognizing its place within a larger cultural system, and making explicit references to psychological, cultural and social information where possible and appropriate. Several symposium papers mention the idea that it is impossible to “program” grammar writing. Over the past twenty or thirty years, several “handbooks,”
Thomas E. Payne
“manuals,” “templates”, or “checklists” for grammatical description have been produced. While it is agreed that such material can be useful, and that grammar writers would do well by consulting them, it is also true that there is no single pattern that fits every language. Each grammatical description must be sensitive to the unique characteristics of the language being described, and will grow out of the nature of the language, in combination with the interests, experiences and talents of the authors. Another tension mentioned by several papers is that between comprehensiveness and practicality. On the one hand, all who have an interest in a written grammar hope that it will substantially cover all the patterns that constitute a native speaker’s internalized linguistic knowledge. Therefore broad and deep coverage is a value. There is a danger in short grammatical “sketches” and partial grammars in that they may fail to adequately represent the “spirit” of the language as a whole. On the other hand, no completely comprehensive grammars of any languages have ever been written. A language contains such a vast wealth of patterns and elements, each used creatively by individuals and sub-groupings within the broader language community, that it would be impossible to ever “freeze” an entire linguistic system within the covers of a book. Finally, in his paper on “growing a grammar” David Weber gives us a list of characteristics that the “ideal” linguistic grammar will exhibit: Weber asserts that “A grammar should be corpus-based, data-driven, theory-informed, user-friendly, publisher-compliant, web-deliverable … and an expression of genuine love.” Indeed, no written grammar can be totally “objective,” in the sense of being a mechanistic diagram of the linguistic patterns and elements in use within a speech community. The context that gives rise to any linguistic grammar of value includes the cultural backgrounds of speakers and grammar writer, special features of the language, and special interests of the linguist.
Conclusion The present collection of papers is humbly submitted as a small step toward encouraging the production of good, accessible, detailed grammars of the many underdescribed languages of the world. May this volume be an encouragement to scholars whose curiosity drives them to study languages often considered “marginal” by the international cultural forces of our times, and whose passion incites them to apply their knowledge in service to endangered language communities.
Introduction
References Boas, Franz. 1893. “Notes on the Chinook language.” American Anthropologist 6, 55–63. Boas, Franz. 1894. Chinook texts. Bureau of American Ethnology, Bulletin 20. Washington, DC: Smithsonian Institution. Boas, Franz. 1911. “Chinook.” Handbook of American Indian languages, Part 1, ed. by Franz Boas. Bureau of American Ethnology, Bulletin 40. Washington, DC: Smithsonian Institution. Hale, Kenneth. 1992. “Language endangerment and the human value of linguistic diversity.” Language 68.1.35–42. Krauss, Michael. 1992. “The world’s languages in crisis.” Language 68.1.4–10. Sapir, Edward. 1921. Language: An Introduction to the Study of Speech. New York: Harcourt, Brace, and Company
Contextualizing a grammar William Bright University of Colorado
The grammar of a language is not only un système ou tout se tient; it also lies within a nexus involving time, space, and social relationships. On the dimension of time, descriptive and historical linguistics are not totally separate: native speakers’ knowledge of their language may include recognition of archaic forms and foreign borrowings. As for space, native speakers’ knowledge also includes awareness of geographical dialect variation. On the social dimension, the relevance of sociolinguistic variation is now widely recognized. Linguists who undertake to write grammars should consider all these factors, particularly as regards the interrelatedness between the grammar, the lexicon, and text collections.
1. Introduction When Tom Payne invited me to speak at this symposium, it was not clear to me what might be an appropriate topic for my talk. Later, when Tom suggested “Contextualizing a grammar”, I thought, “Okay, that should be general enough to cover anything I might feel like saying.” When I also saw that my paper would be first on the program, I thought, “Hmm, that probably means that whatever I say will be discussed in more detail by the speakers that follow me” — and a good thing, too. The notion of “context” is a very powerful one, applicable in all areas of human behavior. Many linguists, social scientists, and indeed literary scholars have taken, as an article of faith, the idea that “Context is all.” Well, maybe it’s not literally all, but it’s probably true that no act of language behavior, or any proposed rule of grammar, can be satisfactorily understood without taking context into account. Since this paper of mine is itself an act of language behavior, and since its context is this symposium, it follows that anything I say about the contextualization of grammar should be understood within a context of what I have previously said, and what I am going to say, about the subject — and in the context of this entire symposium, as it will unfold before you in the course of this day.
12
William Bright
A famous Francophone linguist once said that language was un système ou tout se tient ‘a system where everything is connected to everything else’. This has been taken as axiomatic by linguists whose main concern is the structure of particular languages. However, I propose that the system that we call “language” lies within a nexus of larger systems that involve human psychology, neurology, and culture, as well as dimensions of time, space, and social relationship; and that the many connections among all these systems tend to be concentrated in the phenomenon of human language. If we look at a language, and at the grammar which we as linguists may attempt to write for it, we can then raise the question: What kind of contexts are relevant? Incidentally, because so many of the points I’m making here today have been in the “public domain” for years, I will not try to give specific bibliographical references, but I do want to point to one especially important book: Rethinking context: Language as an interactive phenomenon, ed. by Alessandro Duranti and Charles Goodwin (1992). I’ve mentioned the connections that are studied by psycholinguists and neurolinguists. However important these factors are, they are seldom represented in descriptive grammars; and in any case, I’m not qualified to talk about them. Rather, I propose to talk about several other types of connection which are typically studied by scholars who are variously known as anthropological linguists or linguistic anthropologists. Such people have approached language from the standpoint of “culture”, understood to include those aspects of human life which are transmitted by learning rather than genetically — thus culture includes the choice of forks vs. chopsticks vs. fingers for eating our rice, or the choice to communicate in English vs. Chinese vs. Hindi — as contrasted with the choice to walk upright rather than on all fours, which is genetically determined for all human beings. Within culture, language is one department, along with law, religion, technology, social structure etc. But it is not just one department; it is the key element of culture which makes possible the transmission, the organization, and the elaboration of all the other departments. A historical note: The development of linguistics in the US since the early 20th century has involved two important strands. One was the philological tradition represented by Indo-European linguistics, as elaborated in Europe during the 19th century; it focused on written records, on historical development within languages, and on the historical relationships among languages. The other important strand was the tradition of anthropological linguistics, founded by Franz Boas at Columbia University, and continued by his students, including such outstanding figures as A. L. Kroeber and Edward Sapir; emphasis was placed on descriptive rather than historical research, on spoken rather than written language, and on what may be called “exotic” languages, especially those of North American
Contextualizing a grammar
Indians. By the 1930s, these traditions were closely intertwined. The two leading American linguists of the period were Sapir, whose training had been anthropological, and Bloomfield, who came from Indo-European philology; but both were doing historical and descriptive research, on both European and Native American languages. This blending survived the “Chomskyan” revolution in American linguistics which began in 1957. However, a new dichotomy has tended to emerge. Some linguists are commonly labeled “formalists”; they tend to see language as an autonomous object of description, often according to mathematical, computational or logical models, and they are more concerned with universal characteristics of language than with the grammars of individual languages. Other linguists are often labeled “functionalists”; they tend to be concerned with the ways that language is used to organize cognitive, cultural, and social information. Many of them take an interest in the contextualized description of individual languages, including those that have traditionally been unwritten or are otherwise “exotic”. Within this historical framework, it’s clear that anthropological linguistics and functionalist linguistics are closely linked. It’s been said that the goal of anthropological linguistics is the description of languages within the context of their particular cultures. In this view, then, the ideal grammar would not only account for all the internal facts about the structure of a language, but also for all the connections to non-linguistic cultural context. It is hard to imagine how such a very large grammar could be written or published; all grammars in the real world are much more limited in scope. Nevertheless, I suggest that the notion of limited scope should not be embraced and held up as an ideal, as has been done by some “formalist” linguists. It is certainly appropriate for a grammar to show sensitivity to factors of non-linguistic cultural context; and the more it does so, the more “real” the grammar will appear to native speakers and to informed observers of the language. One very valuable way to contextualize a grammar is to link it closely to a dictionary; and the more cultural context can be given in a dictionary, along with grammatical and semantic information, the more valuable the dictionary will be, both for use by itself and for its use in contextualizing a grammar. More about this later.
2.
The historical dimension
A major distinction in 20th century linguistics was that between descriptive, or synchronic grammar, vs. historical, or diachronic grammar. This distinction was embodied in Chomsky’s dictum that a grammar consists of what a native speaker knows about a language. Though we may accept this, it is now perhaps clear that
13
14
William Bright
native speakers may know some historical facts about their languages, which are reflected in synchronic variation. Thus even relatively uneducated people in any society are aware that some forms of speech are “old-fashioned”, such as English thou goest, he goeth. These forms may survive in particular sociolinguistic contexts, such as that of prayer, though in most contexts they have been replaced by you go, he goes. Looking at more recent changes in English, we may note that, during the 20th century, the speech of most Midwestern and western Americans has undergone a merger of the vowel phonemes /a/ and /o/, so that the names Dawn and Don are becoming indistinguishable. But many people are aware that they speak differently in this regard from their parents, or from their children, and it’s even possible to have intergenerational arguments within a family as to whether Dawn and Don are pronounced the same or differently. For an example in the area of vocabulary, consider areas like the southwest US, where Spanish-English bilingualism is common, many English speakers who have no working command of Spanish nevertheless are fully aware that many of the words in their vocabulary are of Spanish origin. This is again synchronic knowledge of a change in English vocabulary which has occurred over time. All such phenomena, then, are historical facts which form part of synchronic knowledge, and they cannot be ignored in a descriptive grammar.
3.
The geographical dimension
Everyone is aware that other people, in other geographical areas, speak regional varieties of their language. Texans know what English sounds like when spoken by Londoners and New Yorkers and Californians. They can understand those varieties, and they can even imitate them to an extent, which means that they have incorporated information about dialect variation into their own mental grammars. In some parts of the world, there is a “standard” or “official” language (or dialect) which is associated with a particular region; and, through schooling, this variety is “superposed” throughout a nation — such as France, or India, or China — and is used alongside local languages or dialects. Thus a large part of the world’s population is effectively bidialectal, bilingual, multidialectal, or multilingual. Some linguists in the mid–20th century spoke of basing a grammar on a single “idiolect”; this was sometimes simply a rationalization for using only oneself as a source of data. But all humans grow up in contact with other human beings, and so we are all, to varying degrees, multilectal. Any descriptive grammar ignores these facts to its detriment.
4.
Contextualizing a grammar
The social dimension
Sociolinguists sometimes divide their field into macro and micro areas. Macro-sociolinguistics deals with issues that affect large social bodies such as nations, where there is a need to make linguistic choices within such contexts as education, law, and bureaucracy; I’ve already referred to such matters. Micro-sociolinguistics, by contrast, deals with issues of communication between individuals or within small groups. Within this context, some types of variation are associated with the identity of the sender (speaker or writer), and in fact enable us to guess at the social position of an individual; people who enjoy high social ranking, or who suffer from low social ranking, frequently show distinctive features of pronunciation (or spelling), of grammatical structure, and of vocabulary. Another type of sociolinguistic variation is that associated with the identity of the receiver (hearer or reader); this is illustrated in languages like Spanish, in which one must choose between the “familiar” pronoun tú and the “polite” Usted, with corresponding verb forms. Here a sociolinguistic distinction is of major importance for the morphology and syntax of the language. Finally, there is sociolinguistic variation associated with situation — the difference between such speech acts as judicial proceedings, political rallies, prayer meetings, and personal conversations is crucial to the choice of appropriate linguistic features. Especially common contrasts of situation are those between formal and informal communication, and between written and spoken language, and these two dichotomies are often correlated; but we commonly find that written language has both formal and informal styles, as does spoken language. All these sociolinguistic variables impinge on every aspect of a grammar; in the case of a written language like English or Chinese, a grammar that described only the written style, or only the spoken style, would be of very limited value.
5.
Grammar, lexicon, and texts
When Franz Boas sent students to do field work on American Indian languages, he expected them to produce a grammar, a dictionary, and a collection of traditional texts, with abundant cross-references between these parts. Although Boas was not always able to wring these three publications from every student, the ideal remained alive. When I was a Berkeley student during the 1950s, starting field work on the Karuk language, my teachers told me that students of my generation also would be held to the Boasian expectations. Some of us succeeded better than others. Although the dream is perhaps now realized less and less often, some aspects of it continue to be worthy of respect. For instance, if I were
15
16
William Bright
advising someone how to write a grammar of a little-known language, I would recommend the following: 1. Examples presented in the grammar should be based as much as possible on naturally occurring utterances, not on strings of morphemes or words which the linguist creates himself, or obtains by direct elicitation from speakers. To whatever extent examples fail to represent natural utterances, they fail to represent the cultural context of the language. Sapir is said to have asked an Ojibwe to say “my skunk, your skunk” etc., in order to test a morphophonemic hypothesis; but the Indian rightly refused, saying “No Ojibwe owns a skunk.” The fact that no Ojibwe owns a skunk is a cultural fact which should not be falsified. 2. Scholars should consider the preparation of ethnographic dictionaries — a genre which is little cultivated in recent years, and indeed one which calls for immense expenditures of person-hours. Such a work is arranged like a dictionary, but each article is a short ethnographic essay, describing an artifact, a category, or an institution of a particular culture. Examples are the work on Navajo published by the Franciscan Fathers in 1916; the Encyclopaedia Mundarica, on the Mundari tribe of India, prepared by John Hoffman et al. (Patna, 1930–41); and the Enciclopédia Bororo, on an Indian people of Brazil, published by Albisetti & Venturelli in 1962. (It’s interesting to note that all these publications were the work of missionaries.) Such books not only stand on their own as ethnographic reference works, but they provide ideal companion volumes for contextually oriented grammars. 3. A grammar should recognize that the boundary between grammar and lexicon is not a sharp one, and that apparently morphological material can convey what we might think of as lexical meaning (Sapir used the term “lexical affixes” in such cases.) In Karuk, there is a set of adverbs that express direction with relation to the Klamath River, which is central to Karuk geography: ka’ ‘upriver’, yu’ ‘downriver’, ma’ ‘away from the river, uphill’, and sa’ ‘toward the river, downhill’. Each of these has derived forms; e.g., from ka’ ‘upriver’ we have káruk ‘a long ways upriver’ and kâam ‘a short ways upriver’. But in addition there is a set of derivational suffixes, added to verb stems, which make additional distinctions, such as -rup ‘hence downriverward’, -rôov ‘hence upriverward’, -uraa ‘hence uphillward’, and -unih ‘hence downhillward’. A given sentence may contain both a directional adverb and a directional suffix, and they must harmonize with each other. Any attempt to make a sharp division between grammar and lexicon in describing such sentences is doomed to leave us ignorant about matters crucial to Karuk discourse.
Contextualizing a grammar
4. This shouldn’t really need to be said, but I would emphasize that a grammatical study of a little-known language should at least cover all the major grammatical patterns of that language. Yet this is not always done. A few years ago I had the occasion to write a book review of a work on the Northern Pomo language of California which focused entirely on certain aspects of grammatical structure which the author found to be of theoretical interest. So we are given information on how certain theories of grammar, current in the 1990s, shed light on Northern Pomo subject case marking or on non-clause-bounded reflexives. What ever happened to the système ou tout se tient? Of course it is useless to look in such a book for information about the typological status of Northern Pomo, or its history, or its ethnolinguistic characteristics, or its relationships to other American Indian languages; and of course it is even more useless to look for particular lexical items, or for evidence of what traditional narratives in Northern Pomo were like. The language is said to have had fewer than ten living speakers, all elderly, in 1987. It seems likely that we have lost the chance to understand what kind of unique window on the world was provided to Northern Pomo speakers by their language.
6.
The personal factor
To conclude, I’d like to suggest that, when field workers write grammars based on work with a fairly small number of speakers, there should be some recognition of who the sources of information were: not just the names of these “language informants”, as we used to call them, or “language consultants”, as many of us prefer to call them now, but also some facts about their personal and social situations, and about their personal relationships with the field workers. Such facts are certainly part of the context of the language, as the linguist presents it to us; and, as we know, “Context is all.”
References Albisetti, César, and Venturelli, Ángelo Jayme. 1962. Enciclopedia Bororo. Campo Grande, Mato Grosso, Brazil: Museu Regional Dom Bosco. Duranti, Alessandro, and Charles Goodwin (eds.). 1992. Rethinking context: Language as an interactive phenomenon. Cambridge: University Press. Franciscan Fathers. 1916. An ethnologic dictionary of the Navaho language. St. Michaels, AZ: Navaho Indian Mission. Hoffmann, John, et al. 1930–41. Encyclopedia Mundarica. 17 vols. Patna, Bihar, India: Government Printing Office.
17
Writing grammars for the community James Lokuuda Kadanya SIL Sudan Branch
A language is a living soul to a community, a source of dignity and prestige for all people. Once it dies, the whole community dies. A written grammar can support a community by encouraging the use of a community language. Sometimes, however, grammars are ways that outsiders "mine" a local community for the outsiders' benefit, leaving the community of speakers with nothing. This paper discusses the need for linguistic grammars from the perspective of the community in which the language is used. Arguments for including local concerns in grammar writing are presented. One approach to writing a grammar that will serve the needs of linguists and the community is partnership. University professors, field linguists, retired field workers and others can work with local linguists to write grammars that will truly meet the needs of all.
1. Introduction Mining can be a very good thing for both the company doing the mining, for the government of the country, and for the communities in which it is taking place. Other times mining can have a number of negative effects. Some cases in Africa come to mind. Many companies, like oil extraction and gold mining companies, do much harm to the environment and to the communities in which mining is taking place. Usually, the topography of the area is left in a state whereby the land is of no use any more. At the worst, the inhabitants are displaced, and in almost all cases they do not receive one benefit from the operation. Some oil companies in Africa are such examples. Others, however, do plough back something into the community in the form of infrastructure such as schools, hospitals, and even efforts to reclaim the wasted land. Considering the same issue as far as writing grammars is concerned, we grammar writers, whether past or present, are guilty of similar things. Some grammars that have been written were mainly like mining that is bent on taking from the community and almost care nothing about the people of the community.
20
James Lokuuda Kadanya
I am attempting to persuade future grammar writers around the world to write for both the benefits of the local communities as well as for the wider international community. The benefits for the local communities should be the primary motivation in this endeavor by all means. This write-up is divided into two sections. The first section will focus on why we should be writing for the local community as well as for the linguistic community. By “the local community,” I refer to the people who speak the language being analyzed. The writer should write a grammar that will benefit the ethnic group. We shall see how we can do that in the discussion. Also, the grammar should be good enough for the linguistic community. One could ask is there a way to write such a grammar? The answer is not an easy one. But it is possible. One can choose to write an all-inclusive grammar in one volume or actually present it in two different volumes. The first option could be cheaper, but will be hard work indeed. The later option is time-consuming, but one will have done something beneficial for the community, and may even be more rewarding personally. The last section consists of suggestions on how best we can possibly achieve our aim of grammar writing more quickly and easily. Due to time constraints, this write-up is not a conclusive attempt to address the issues concerned. At the same time, I am aware of the dilemma many grammar writers find themselves in, and want to share my perspective in order to encourage us to think carefully about the import of our work on communities, and to do our work in a deeper and more responsible way.
2. Encouraging a grammar for the local and linguistics communities A language is a living soul to a tribe or community. Once it dies, the whole community dies. That death has many implications. It may mean the actual disappearance of the speakers from the face of the earth. It could mean that these people may actually be absorbed into other communities surrounding them. In case this absorption is into a similar kind of culture as their previous ones, then it may not have to worry us as much as to when they get absorbed into a culture that is not similar to theirs. This second situation can result in people being ridiculed and called names. This can be psychologically tormenting for them. I have seen many examples of this in my own country, Sudan. When we assist the community in preserving and being proud of their language, what are we really doing? First of all, we are helping them preserve their own culture. Language is like a big factory for the culture. It is with the language that new information and new cultural elements are processed and disseminated.
Writing grammars for the community
Secondly, language is a source of dignity and prestige for all people. There is nothing more thrilling and uplifting than having a sense of dignity, being proud of who you are. Mutual acceptance comes about in a closely knit community in which a common local language is one of the binding elements. A national language seldom accomplishes this function, unless it was the language of the parents and grandparents of the community. In short it is a language that is part of those you can trace in your lineage. How can an outsider, a missionary or a linguist, possibly assist the community in preserving and being proud of their language? It is a very difficult task for an outsider to write a grammar that can be used by both the local and the worldwide audience. Having said that, we can be catalysts in this. As linguists who not only have linguistics as a driving interest, but also the interests of the people at heart, we have to exhaust all our efforts in helping the community appreciate their own language and culture. This is best done and preserved through writing, even the writing of grammars. It is only when all efforts at language preservation have been exhausted can we go ahead and publish grammar findings without any feelings of guilt, though the language may in fact be actually dying. It would be very wrong for anyone to come to any given community, live there long enough to collect data, write up a linguistically fine grammar, receive a degree, get a good name and all kinds of benefits from it and forget about the people. This is not acceptable, no matter how it may be justified. We should cooperate with individuals who show interest in their language. Face them with the fact that their language is disappearing. Who knows what might happen; they may wake up to the challenge and we will find ourselves partnering with God-sent interested people. We have all heard stories of some missionaries and linguists who went to a community of the people whose language was almost dying and the presence of the linguists or missionary helped the people appreciate their language and brought that language back to life again. I think that is a more rewarding experience than to be acclaimed for having written the best grammar in the world and having whatever benefits that follow from that. Another obviously rewarding result of grammar writing is using our findings in educational programs that will affect the lives of the people. When a school is founded and is thriving, people joyfully using books in their vernacular language, that will be more rewarding than words can describe. Having argued for a community-based grammar, it is equally a responsibility to write a good grammar for the wider audience, which may include national and international linguists. The choice one might make is to write a grammar which can include both audiences, the local community and the linguistic community. Indeed, it is almost impossible to do this. But something ought to be done. Two
21
22
James Lokuuda Kadanya
grammars could be written, for the local people and for linguists. The order in which they are to be written, should be between the author and the community. The fact that the grammar is written for linguists does not mean that the speakers of the language are not benefiting – they are! A permanent record has been made for that particular group. If the author describes the needs of the group, the international community may respond to those needs, and in that way, the local people will surely benefit. The linguist should include a good introduction about the people in writing a grammar.
3. How to Achieve this kind of a Grammar This is a difficult work. The effort can never be on the shoulders of the linguist or the missionary. It should be made a concern of all individuals and institutions involved in language planning, including linguistics departments in universities, missionary organizations, national governmental and non-governmental organizations. For the institutions and universities, students interested in linguistics, education and Bible translation worldwide should be asked to write the grammars of those languages, making sure that they are making all efforts to revive the language and are working with the local people in writing grammars. This too is something that is difficult to monitor, but all efforts should be made. At the workshop held at SIL International, on October 15th–25th 2002, David Weber, one of the lecturers during the workshop asked “Is it possible for many of the retired linguists to co-author grammar books with some university students or young linguists?” This question was not really discussed. I think this is a welcome idea because the retired linguists whether retired or active in the field can still contribute a lot. Again if they are not in a position to travel widely and gather data, they can still write grammars, as long the younger linguists can gather enough data for the analysis. On the other hand, many retiring SIL field workers did language analysis long ago, learned the language and collected data, but are not conversant with current linguistic theories, thus unable to write grammars. The students can co-author with such linguists because they know the theories and can easily be supported by their professors in the endeavor of writing grammars for a wider audience. Field linguists may also encourage the people to write the grammars for the community. This partnering authorship can not just be for retired linguists or university professors alone, but even the missionaries and the field linguists themselves may still do this with the local people. This venture may show the local
Writing grammars for the community
people that the linguists really are interested in them and may help to encourage the local writers to write the grammars, even in their local languages.
4. Conclusion My discussion above has been largely about the need to write a grammar for the local people. This is important because our interest should be to help the people whose language we are analyzing. Our aim should always be to think of ways of preserving that language for the many reasons we have seen above. I have also expressed the fact that the grammar should be excellently written for the linguistic community as well. The way one could go about meeting the needs of both audiences depends on both the author and the local people. The last thing we have looked at is how we can really go about helping such grammar writers as much as possible. One possibility is partnering in writing grammars. University professors, field linguists, retired field workers and others can work with local linguists to write grammars that will truly meet the needs of all people.
23
Collective field work Advantages or disadvantages?* Aleksandr E. Kibrik Moscow State University
For forty years, linguistic fieldwork on the minority languages of Soviet and post-Soviet States has been conducted by members of the Philological Faculty of Moscow State University. The present paper describes the “Moscow approach” to team-based fieldwork, including the history, problems, advantages and disadvantages of this unique venture. The conclusion is that the advantages of collective fieldwork outweigh the disadvantages, but that there are several requirements that must be in place in order for the method to be effective. These include a clear organizational scheme, several available language consultants, and supportive attitudes on the parts of all participants. In recent years, a combination of researcher specialization and cross-training in multiple sub-disciplines has become an increasingly important component of the collective field trips.
1. Introduction The goal of this paper is to acquaint the readers with the specific practice of collective field work organized under the auspices of Lomonosov State University of Moscow, Russia. 2. History The unique and innovative program of collective fieldwork began in the summer of 1967, when the department of linguistics1 of the philological faculty organized the first linguistic expedition to Daghestan. The tradition of fieldwork in the linguistics faculty dates back to the 1940s: since that time, students have visited various rural districts of Russia to collect spoken data on Russian dialects. The expedition of the summer of 1967 was different in that it was the first time that students were going to do fieldwork on a completely unfamiliar language.
26
Aleksandr E. Kibrik
The expedition team consisted of three Department teachers (myself among them) and five undergraduate students. Our purpose was to practice fieldwork methods in one of the Daghestanian languages — Lak. The Lak village we visited was high in the mountains. For the students, the intention was to provide practical experience applying theoretical linguistic background to the investigation of an unfamiliar language. For me, it was a completely new activity. When I was a student, I had only participated in one of the dialectological expeditions, so I was as inexperienced in fieldwork as any of the students who went to Daghestan with me. This field trip showed us that it was quite realistic to collect language data by questioning a language consultant in the contact language. It also made us see that it is necessary to be well-informed about the whole range of cross-linguistic variation. Ideally, you, the linguistic fieldworker, should be able to foresee likely combinations, arrangements and interactions among various language phenomena. For successful fieldwork, it is also necessary to master the methods of efficient interaction with the language consultants. This allows you to obtain appropriate data in the shortest time possible. My further activities in subsequent expeditions were largely aimed at achieving these goals. 3. Methodological problems and insights The problem of calculating the parameters of cross-linguistic variation belongs to the realm of linguistic typology. In the 1960s, this subdiscipline was evolving and its generalizations were rather inadequate for practical application. But it was obvious that an unknown language is not something totally new or unexpected and that the empirical data should fit into some universally applicable scheme. In other words, it is wrong to think that an unknown language can present you with just anything. If it seems that the language presents some strange facts, it only means that you have insufficient knowledge concerning the boundaries of cross-linguistic variation; it does not mean that the language is inherently strange, or randomly complex. Unfamiliar features of a language are noteworthy in the sense that they make us see the limitations of linguistic theory and force us to reduce the degree of our typological ignorance. As for individual work with the language speakers, the basics of this technology are universal for any linguistic fieldwork. In the course of fieldwork, you can draw from the international pool of experience in field linguistics and contribute your own observations to it. In my own situation, in the late ‘60s, we relied on such works as Longacre (1964), Samarin (1967), Garvin (1972), as well as classical works of the Russian tradition such as Petr Uslar, and V. Bogoraz. I can also recommend more recent works such as Newman and Ratliff (2001).
Collective field work
Our fieldwork practice is quite specific technologically — it is team research work. Its team character can be traced back to the original purpose of the expeditions, namely, practical training of beginner linguists. After the main methodological difficulties of linguistic fieldwork had been overcome, it became clear that practical instruction should not be the only purpose of the expeditions. We realized we could work on documentation of uncharted languages and obtain some fundamental results as well. This is the way the program of collective/team field linguistics was molded. During the past 25 years, I have led 40 expeditions. We have studied many minority languages and unstudied languages of the former USSR. First of all, we have studied all the languages of Daghestan. Some languages have been studied along with their dialects (more than thirty language varieties have been examined in total). We have also done fieldwork on Abkhaz (West-Caucasian) and Svan (Kartvelian) in the Caucasus, Shugnan (Iranian) in the Pamirs, Tuva (Turkic) in Central Siberia, Alutor (Chukotko-Kamchatkan) in Kamchatka. Lately, my former students have been taking their students to do fieldwork on Tatar, Chuvash, Balkar (Turkic), Mari and Komi (Uralic). In all these expeditions, team fieldwork technology has been tested and improved, and this method has proved viable. Eight grammars (Khinalug, Archi, Tabasaran, Avar, Godoberi, Tsakhur, Bagvalal, Alutor) have been published along with the comparative study of lexicon, phonetics, morphology and syntax of more than twenty languages of Daghestan. Our research program in principle is free from reference to previous research — we approach each language as a tabula rasa. This methodology stems from the training purpose of the expeditions: the students were meant to learn how to work with a language without getting any preliminary training in this language and to make decisions without any prompts. At the same time, it was easier to set the training task in this particular way because in most cases the languages we studied had been documented to a very small degree. Even though some information on phonetics and morphology was available, it was not at all comprehensive and needed to be thoroughly checked. As for empirical data (dictionaries, texts, paradigms) the information that could possibly be used was next to nothing. I am now accustomed to working with languages where there has been no strong tradition of linguistic description, and I can conclude that this is the most favorable circumstance for a researcher. This statement may sound paradoxical: at first sight, it seems easier to study a language when there is a lot of information on it. However, one should bear in mind that any tradition imposes some a priori view of the facts. If you learn some descriptive framework before you are exposed to the facts of a language, the descriptive framework can displace the facts in your mind. It may anesthetize your creative will, thus not allowing you to see the facts as
27
28
Aleksandr E. Kibrik
they are. Later, under the pressure of empirical data, a researcher starts to undergo a painful process of breaking through the previous traditional understandings, but a lot of effort may have already been wasted. If a language has a descriptive tradition, I prefer to do things in the following order. First you collect facts and analyze them. Next, when some tentative hypothetical generalizations are formed, these are compared to the existing description. If they are similar, this is evidence in favor of the truth of the initial hypothesis; if they are different, two independent interpretations compete. Of course, the traditional description might well be much better than your fieldwork hypotheses. But even in this case the fieldworker makes an informed choice in favor of the traditional description after examining the empirical data. Nevertheless, sometimes we did not follow this order — for instance, when we worked on Svan, which is closely related to Georgian. Existing descriptions of Svan followed the descriptive tradition of Georgian linguistics. This framework is but loosely related to modern theory and uses idiosyncratic terminology for predicate-argument structure. It made our work more complicated, because in our minds we had two descriptive models, that could hardly be related to each other. The tabula rasa approach to an unwritten language when you are intending to document it and build up a comprehensive description implies that you have to solve various multidimensional tasks. First, in the absence of literacy, each language expression has to be adequately and consistently transcribed and interpreted. This involves solving various phonetic, phonological and morphological problems, and the decisions have to be taken on the spot, at the initial stage of language documentation. This stage has to be passed as quickly as possible, and at least some preliminary agreements have to be achieved. If this stage takes too long, and agreement on the forms of data documentation and interpretation are constantly changing,2 the research pace slows considerably. Second, you have to identify all grammatical units of the language under study (morphological, lexical and syntactic) and their functions. This task consists of compiling all relevant paradigms and identifying all relevant distributional characteristics of all classes of morphemes, words and syntactic constructions. Third, you have to compile a maximally full lexicon where each lexical entry is fully characterized with regard to its usage. Fourth, you have to record and analyze a representative corpus of spoken texts of various genres that embodies the actual usage of linguistic units in different contexts. So, fieldwork based on empirical data collection is time and effort consuming. But this is not the only and by far not the most significant difficulty in the field. Fifth, when fieldwork is intended to produce a full language description, you have to possess thorough theoretical knowledge of all linguistic domains. Today’s
Collective field work
linguistic theory is obviously developing very quickly and is now a fine-grained scientific field. In the course of study of major languages that have a long linguistic tradition, a detailed expert specialization has been naturally formed. Minor languages are being studied by individual researchers. Each fieldworker studying a minority language really has to be a whole-department-in-one-person sort of researcher! Field work of the philological faculty of Moscow State University has laid the foundation of the team method, which, as I am convinced, has doubtless advantages over the “one language — one fieldworker” type of method. However, before I say more about these advantages, I have to point out the limitations of this approach and some problems related to it.
4. Limitations of the team approach 1. People may think that the team approach is not justified financially, because more fieldworkers require more funds. It is true that field trips by individuals cost less than those of a group. However, I will show that team fieldwork is actually more economical in a number of respects, not the least of which is the fact that time in the field and time spent preparing the final linguistic products is significantly reduced. 2. There is a more important limitation, however. The team approach requires a sufficient number of both fieldworkers and available speakers, or language consultants — in a proportion not less than one language consultant to two fieldworkers. Furthermore, language consultants have to be ready to work regularly. Both conditions sometimes cannot be met for several reasons: there are too few speakers living in the same village or too few speakers can speak the contact language or too few speakers can give up their own business to work as a language consultant full time for several weeks. The team method makes sense when all linguists have the opportunity to work regularly with speakers for as long as is necessary. It is obvious, however, that if this requirement is not fulfilled, an individual researcher cannot work effectively either. We often have had to give up a linguistically promising language project for the practical reasons described above. 3. Some linguists want to possess exclusive rights to a language. This proprietor instinct among those who study minority languages works against team work in the field. I have encountered this phenomenon many times when I found myself treading on territory already occupied by someone else. It is painful for some people to see other researchers doing field work on “his” or “her” language. I strongly object to this sort of ethic and have always ignored it.
29
30
Aleksandr E. Kibrik
There is a more delicate ethical question concerning the problem of intellectual property rights. It is true that in any collective activity it is impossible to make a division of intellectual property rights to the products of the research because many discoveries are made collectively and belong to the group, not to an individual. It is well-known, however, that in any scientific work serious results can be achieved by research teams and not individual researchers, so linguistic fieldwork is no exception. In our fieldwork practice, which I am going to discuss below, we have worked out a mechanism for recording each researcher’s contribution to the team project. Even though there is such a mechanism, each researcher always has to be ready to share his or her contribution with colleagues. The decisive motivation in this activity has to be the overall descriptive goal and not the desire to single out one’s personal part. Generally speaking, team work excludes manifestations of absolute individualism, and therefore is not suited to all researchers. Rugged individualists do not function well in a group fieldwork situation. I believe that any field and activity does not suit just any person, so team field work in this respect is no different than any other job. 4. Through our years of field work it has become clear that descriptive linguistics is a creative process and no immediate decision is final. Moreover, many aspects of interpretation are frequently altered. This concerns transcription and glossing as well as insights into the nature of language phenomena. In team work, each team member has to be informed of the current state of research, otherwise the work is open to all sorts of misunderstandings, which are sometimes quite undesirable. Furthermore, though different people approach the same tasks in different ways, team work requires a certain degree of harmony in people’s actions. In the process of team work on a given language, especially when this work is aimed at description and documentation, disagreement is inevitable in the phonetic transcription, translation and interpretation of language units, in the use of terminology and in the choice of formal approaches. Therefore, to achieve consistency in description and a unified account of linguistic phenomena, one has to coordinate the actions of the team and come to agreement on all possible theoretical and technical problems. This can be achieved only by open discussion, in which the team leader exercises authority only as a last resort. An individual researcher normally does not face these difficulties: he or she does not have to harmonize anything with anybody. It is worth mentioning here, though, that individual work is not inherently free of descriptive inconsistencies and contradictions. I must admit, though, that this property of team fieldwork is one of its core difficulties, though this difficulty can be overcome. One should not forget that in
Collective field work
any teamwork people face similar problems. A soloist has fewer problems playing a piece of music than an orchestra or a choir, but no soloist can replace an orchestra or a choir. So, in field work there are some practical problems, but in the majority of cases they can be solved.
5. Advantages of team fieldwork Now I am going to examine some advantages of the team fieldwork method. 5.1 Time scale The team approach allows researchers to collect empirical language data many times faster than is possible in an individual research program. What normally takes years is done in a few weeks. Our first successful experience of team field work was the Khinalug language. This language is spoken in only one village (Khinalug) high in the mountains in Azerbajdzhan, far from the main enclave of Dagestanian languages. Genetically and structurally, it is also markedly different from other languages of this family. On the one hand, it has preserved many archaic features, that can be traced back directly to the protolanguage; on the other, it has been significantly influenced by Azerbajdzhani because it has long existed in an Azerbajdzhani-speaking environment. The grammatical description of Khinalug is based on data gathered by a team of fourteen people in the summer of 1970. The team consisted of 3 researchers and 11 students. Each researcher was personally responsible for one of the following three domains: phonetics (Sandro Kodzasov), lexicon and texts (Irina Olovjannikova) and morphology (Alexandr Kibrik). The students collecting morphological data were split into groups, and each group was in charge of a particular topic in morphology. Students who had had some fieldwork experience by this time worked on their own or were paired with a newcomer. The fieldwork period itself was six weeks long. After that the three supervisors intensively worked on writing grammar chapters based on the data collected by the group. The final version of the grammar was written by myself and S. Kodzasov, while Irina Olovjannikova prepared the lexicon (about 1000 words) and texts (440 sentences) for publication. In the 1970s, the tradition of interlinear glossing had not yet developed. Still, one third of the sentences in the texts were given interlinear glosses. Then in January 1971 we organized a ten-day expedition to check grammatical hypotheses and phonetic transcriptions; we also intended to
31
32
Aleksandr E. Kibrik
check translations and collect some more paradigms. It took another six months to prepare the final text, and the book was published in 1972 (Kibrik, Kodzasov, Olovjannikova 1972). Right after the work on Khinalug grammar had been completed, we took up the study of Alutor. Alutor belongs to the Chukotko-Kamchatkan language family, it is spoken in a few villages in the North of Kamchatka. Alutor is a clear case of an endangered language; it is spoken by a few hundred people, and young people, as a rule, do not speak it at all. We made three fieldtrips to Alutor. In the first five week expedition in 1971, there were 9 people: 5 students and 4 researchers (Alexandr Kibrik, Sandro Kodzasov, Igor’ Mel’chuk and Serafima Nikitina). Alutor was strikingly different from anything we had seen before. First of all, it fully contradicted our ideas of what should and shouldn’t exist in language. In the following pages I will try to provide examples. During our expedition workshops, we argued a lot about nontrivial phonetic problems, which made transcription very difficult. In particular, we had to determine the status of the [?] sound which was ambiguous. In many cases, its presence was predictable when its function was merely epenthetic: it divided nonvocalized components of consonant clusters. Apart from this function, it could appear within morphological structures where its presence was not stipulated by the standard rules of epenthesis. The formidable verb paradigms were no less complicated. Besides, some of the verb forms inexplicably contradicted the principle of agglutinative additivity,3 and it was impossible to formulate their meanings in usual categorial terms. Propositional semantics was absolutely mysterious. Alutor looked a bit similar to Dagestanian languages due to the ergative sentence structure, but its syntax was unlike anything we knew about ergativity. As before, Kodzasov was in charge of phonetics, while Kibrik, And the students he supervised did grammar. Mel’chuk, being the author of the “MeaningText” model which had been claimed to be universal, was a general consultant. Nikitina studied ethnolinguistics and cultural anthropology. Texts were recorded and a lexicon compiled. Along with this, we started checking texts that had been recorded in 1950s by ethnographer I. S. Vdovin, who kindly gave the texts to us. They were all epic narratives, and by the 1970s this genre had been totally forgotten by Alutor speakers. The work on Alutor documentation continued in the summer of 1972 (4 weeks of fieldwork, 11 team members). In this expedition, we expanded considerably the collection of texts and processed them; at the same time, serious work was done on compiling the lexicon. However, we did not feel satisfied, recognizing that we had failed to comprehend the qualitative “nature” or “spirit” of Alutor. All morphological forms had been collected, but their functions remained, to a large extent, unclear to us.
Collective field work
One difficulty was that we were working on Alutor simultaneously with the Dagestanian project. This presented certain problems for our analytical work in between fieldwork expeditions. It is well known that any period of field work must be followed by thorough analysis of the collected data. It is during this time that additional hypotheses are generated and tested, and generalizations are made on the basis of the collected data. This “homework” made it clear that we needed to make another expedition to Alutor, but we could not go there for the third time until 1978 (4 weeks, 9 team members, 3 students new to fieldwork among them). In this expedition, we focused upon two things — gathering syntactic data and re-checking the texts, lexicon and morphological description. During the years that had passed since the previous expedition of 1972, the number of speakers who could serve as language consultants had been significantly reduced, and we were feeling the constant lack of consultant help. Nevertheless we managed to fulfill the task we had set. For a number of reasons, publishing the Alutor materials was postponed until 20 years later, and the monograph did not appear until the year 2000 (Kibrik, Kodzasov and Muravjeva 2000). It has three parts: the text collection (41 texts, consisting of 1284 sentences altogether), the dictionary (more than 3000 words, 1260 stems) and the grammar, comprising chapters on phonetics, morphology and syntax. An English version of this major work is currently in preparation and should be available soon. The grammar sketch was based on fieldwork reports and spontaneous text examples; it focuses on the typological uniqueness of Alutor. The twenty-year break served us well, in that during this time we accumulated considerable linguistic knowledge. The phenomena that had looked puzzling and unique to us could by that time be explained and interpreted because in the intervening twenty years we had observed similar phenomena in other languages. In particular, it became possible to explain the workings of person marking on the Alutor verb. The reason this mechanism differs from standard agglutinative expression is the principle of reversible markedness which is inherent to the system (see Kibrik 2001a for details). In fact, the agglutinative principle is present but it works hand in hand with a more general semiotic principle of cumulative expression of combinations of grammatical meanings that are cognitively most closely intertwined. This was the reason that some areas of Alutor grammar seemed to diverge from the agglutinative pattern. During the intervening years typological linguistics had progressed to the point where languages previously considered “exotic” could now be described in terms of general typological parameters. Therefore, for our volume we formulated principles of word-by-word morphological analysis, introduced interlinear glossing and checked the texts against the dictionary. In spite of the time span between data collection and publication, it was possible to make do without
33
34
Aleksandr E. Kibrik
additional checks of language data. This I consider to be good evidence for the reliability of the data. A somewhat different documentation scheme was used for Archi. Archi, as well as Khinalug, is spoken in one village only. It had been provisionally classified as belonging to the Lezgian group of Daghestanian languages. Archib, the village where it is spoken, is located in the central part of the mountainous region of Daghestan. Structurally, the language does not deviate from the Standard Average Daghestanian. Our first acquaintance with Archi took place in 1968. It was the first language we studied systematically, though we had a research program which was based on the experience of previous expeditions. The Archi expedition lasted 6 weeks, and consisted of 9 individuals, including 6 students.. The phonetics portion was supervised by Kodzasov and Olovyannokova, and the grammar by Kibrik. Nevertheless, the expedition was mainly for training purposes; we were still working out the collective research scheme and were getting used to the typological parameters of Daghestanian languages. At the same time, this expedition encouraged us to start work on the Khinalug description, which I have already discussed above. The most valuable contribution of the first Archi expedition was the meeting with Dzhamil’ Samedov, an Archi speaker who was a philology student at the university in Daghestan.4 He had a talent for linguistic work and was very enthusiastic about our research. Later it became possible for Samedov to attend Moscow State university. This circumstance played a crucial role in furthering the work on Archi. In 1971, along with the expedition to Alutor, the second expedition to Archi was organized. 15 people, including 13 students, participated. The participants collected morphological data according to a carefully compiled questionnaire. All further work was carried out “at home” together with Samedov, who was no longer a language consultant but a co-author in his own right. I must note, that this was the most effective way of doing collective work on a language, but unfortunately, it was the only opportunity in thirty years of fieldwork. Eight years were spent preparing a major description, co-authored by four people. Kodzasov, as usual, did the phonetic part, Kibrik was in charge of the grammar work, while Kibrik, Olovyannokova and Samedov did the dictionary. Texts were mainly collected by Samedov during his summer trips home and analyzed by Kibrik, And Samedov together. In 1977, the four-volume grammar of Archi was published. The first volume (Kibrik et al. 1977a) contains a very detailed description of Archi phonetics, written by Kodzasov (pp 185–355) and a sketch of the Archi lexicon, including lexical classes (borrowings, toponyms, semantics of noun classes, and ideophonic elements), derivation patterns, and materials for an ethnocultural thesaurus. The de-
Collective field work
scription of the lexicon was co-authored by Samedov, Olovyannikova and Kibrik. The second and the third volumes contain two alternative comparable descriptions of Archi: a taxonomic one (Kibrik 1977b) and a dynamic one (Kibrik 1977c). The taxonomic description was done in the standard structuralist framework, the dynamic one was based on the “Meaning — Text” model (Mel’chuk and Zholkovskij 1970). In addition, a separate volume containing the texts and the dictionary was published. (Kibrik et al. 1977b). The longest team research project was a comparative study of Daghestanian languages. It started in 1973 with the primary goal of collecting and documenting the lexicons of all the Daghestanian languages. About 1000 lexical entries were chosen to go into the dictionary and organized according to the thesaurus principle. These lexemes were to be collected for the majority of Daghestanian languages. In some cases we were able to study one language as spoken in several villages. In those cases, we tried to note variation from one village to another in the dictionary. In other cases, we were only able to study a particular language in one village. In those cases, we explicitly presented it as a particular variety of the language, not as the language as a whole. In Daghestan, as in many linguistically diverse parts of the world, it is difficult to find two villages with identical languages. In addition, the noun and verb morphology of each language (or variety) was studied. For each word, diagnostic forms were recorded, which allowed us to construct full paradigms. We also collected other types of information on each word, for instance, gender for nouns, and argument frames for verbs. The work on the comparative dictionary project was carried out by two or three small groups working simultaneously in different villages. In the period between 1973 and 1982 there were 9 large expeditions to various regions of Daghestan and Azerbajdzhan. Lexical data were collected for 30 speech varieties, representing 22 distinct languages. The overall number of people participating in these expeditions was about 200. The work was supervised by Kibrik, And Kodzasov, and they were the ones who checked and rechecked the recorded data. The lexicon of each speech variety was checked several times in-situ. For the specific purpose of checking, four expeditions were organized in the period from 1978 to 1984, and moreover researchers often had to make checking treks outside the headquarters of the expedition. The work was done in the pre-computer era, and the dictionary had to be rewritten many times on paper and cards, which was quite time-consuming and increased the probability of mistakes. The outcome of this multi-year project was published as a two-volume book (Kibrik, And Kodzasov 1988, 1990). During fieldwork sessions, students were largely responsible for collecting the data and making tentative generalizations. The final descriptive work was done by
35
36
Aleksandr E. Kibrik
more experienced researchers. However, this distribution of roles is not obligatory and depends on the experience and professional capacities of the expedition participants. For instance, one of the co-authors of the Alutor grammar, Irina Muravjeva, was a student when she first started work on Alutor; later, she specialized in Chukotko-Kamchatkan languages and became a major expert in Alutor. So, it is only natural that she became one of the co-authors of the Alutor grammar. Dzhamil’ Samedov started his work on Archi as our language consultant during fieldwork. Later, he became a co-author of two volumes of the Archi grammar. Our first grammatical description to which students successfully contributed as authors was the grammar of Tabasaran (Kibrik 1982). The volume was a result of a 1979 expedition (4 weeks long) to the Northern dialect of Tabasaran (the village of Djubek). 11 people took part in the expedition, including 7 students. It was the third expedition to Tabasaran. Before that we had worked twice on the Southern dialect. Tabasaran is a literary language, which has employed an alphabet based on the Cyrillic script since the end of World War II. The language is spoken in only two districts of Daghestan. It is a typical representative of the Lezgian group of Daghestanian languages. Unlike previous expeditions, during the 1979 Tabasaran expedition, we focused our attention mostly on syntax. Kibrik supervised work on four syntactic topics (syntax and morphology of verbal agreement, causative constructions, complementation, and coordination). The section on attributive constructions was supervised by Ol’ga Boguslavskaja. Kodzasov and Muravjeva worked on phonetics. Each group working on a particular topic later prepared the corresponding chapters in the collective volume. In the 1990s, collective fieldwork became even more efficient, as the level of the students’ research expertise significantly improved. The data from the 1992 expedition into one of the Avar dialects5 (three weeks long) laid the foundation for a volume (Kibrik 1993) in which some students were full-fledged contributors: Elena Kalinina (the chapter on non-verbal predicates), Konstantin Kazenin (the chapter on action nominal constructions), Ekaterina Lyutikova (the chapters on genitive constructions and relativization), and Sergei Tatevosov (the chapter on universal quantifiers). They all had started their expedition careers just a year or two earlier. These formed the core of later expeditions and participated in the description of three additional languages. At present they all supervise students’ fieldwork in their own right. The next expedition, in 1993, saw 17 participants working for 3 weeks on documentation and grammatical description of Godoberi. Apart from students and staff of the Department of theoretical and applied linguistics, there were two external participants: Martin Haspelmath and Andrej Kibrik.
Collective field work
Godoberi is spoken mainly in the village of Godoberi, situated in West Daghestan on the Chechen border. The language belongs to the Andi group of Daghestanian languages, and is one of the least studied languages of Daghestan. Though our fieldwork period was limited, we were able to study phonetics, morphology, syntax, and some problems in semantics. In addition, we compiled a lexicon (about 800 basic words), recorded three texts (about 130 sentences in total) and provided detailed interlinear glosses and translations. During the next year we were involved in preparation of the grammatical description. In the summer of 1994 Kazenin, Tatevosov and Fedorova spent two days in Godoberi resolving problems that had been identified while the grammar was being written. Then there was further work on the text of the grammar, and an English version of the text was prepared. In 1996 the grammar was published. The book had two co-editors: Sergei Tatevosov (linguistic editing) and Alexander Eulenberg (English editing). 13 authors in all contributed to the volume. While the previous publications (Tabasaran and Avar) approached only selected topics, the Godoberi book was a holistic language description. In 1995–1996 two expeditions were organized to the Tsakhur language (19 and 16 participants respectively) lasting 5 weeks altogether. Tsakhur is spoken in the South of Daghestan, under the Great Caucasian Range, and in Azerbajdzhan. Genetically, it belongs to the Lezgian group. There is a description of Tsakhur phonetics and morphology (Ibragimov 1990). This description was used in the preparation course before the expedition. In addition to this, another dialect of Tsakhur was studied during two expeditions in the 1970s dealing with phonetics, morphology and syntax. In the first expedition, the empirical data were collected, and in the second they were checked and expanded. While the final version of the book was being written, new problems arose, so in the summer of 1997 one of our language consultants was invited to come to Moscow. For three weeks he was involved in additional work on the grammar. The Tsakhur project had a significantly broader range of topics than previous team research projects. Along with phonetics, morphology and syntax, we studied discourse organizing principles and propositional semantics; significant attention was paid to compiling a representative text collection (9 texts consisting of more than 700 sentences altogether) and a dictionary. In 1999 the book was published (Kibrik, And Testelec 1999). Fourteen authors contributed to the volume, including four students.6 Finally, the most recent published volume is on Bagvalal. Bagvalal is another Andic language, closely related to Godoberi. No descriptive materials had been published on Bagvalal prior to our research project. This language was studied in the course of three expeditions: in 1994 (16 participants), 1997 (21 participants)
37
38
Aleksandr E. Kibrik
and 1998 (16 participants). The overall duration of fieldwork was 7 weeks. The scheme of fieldwork was no different from those of Tsakhur and Godoberi. The work resulted in a grammatical description (Kibrik 2001a). Its essential component, along with the grammar part, is a collection of 26 texts consisting of 700 sentences altogether, and a dictionary of about 1500 lexical entries. This book was also prepared by three co-editors (Kazenin, Lyutikova, Tatevosov) who were responsible for particular chapters each. As can be seen from this brief survey of nine documentation projects, none of them could have resulted in a significant publication in a comparable time period (and many in principle could not have been accomplished at all) if they had been carried out by individual researchers. 5.2 Balance A balanced application of the collective method takes advantage of the principle of division of labor. Collective work does not just increase the amount of language data collected in a given time period; neither does it simply accelerate the routine work of questioning the consultant and recording his answers, though both of these advantages are important. The team method makes it possible to send each team member to those areas where she can work most effectively with the best possible outcome. In our expeditions participants differed significantly as to their fieldwork experience and linguistic background, so fieldwork was organized hierarchically: more experienced fieldworkers supervise less experienced ones. The depth of the hierarchy can differ from expedition to expedition, but normally the chain consists of the following links: the expedition has the Head, who is in charge of all aspects of fieldwork — designing the research program, including overall and specific goals, harmonizing alternative viewpoints, making crucial theoretical decisions, taking care of the living conditions, etc. Experienced fieldwork researchers report to him; they are responsible for broad topics like phonetics, morphology, syntax, discourse, collecting texts and lexical data, which are the immediate components of an integral language description. Along with research tasks, they may be responsible for technical tasks like computers and software, interaction with the language consultant team, providing food and transport, etc. Research topics are further divided into sub-topics, which can be supervised by experienced student fieldworkers. Students new to fieldwork work on a topic under the supervision of experienced students. A group working on the same topic is called a “cluster.” Natural morphological subtopics include parts of speech (the verb, the noun, the pronoun, the adjective, minor parts of speech), formal morphology and morphosemantics. In syntax, subgrouping includes different construction
Collective field work
types (verbal vs. nominal sentences, complement, adjunct and modifier clauses, comparative constructions etc.) or syntactic processes (agreement, coordination, transitivity increase/decrease, reflexivization, etc.). Each expedition has its own set of topics depending on the research interests of group members, the chosen theoretical framework and the language specifics. In general, all the research topics are hierarchically divided into spheres of interest, and zones of mutual interest to separate groups are established. It is important to note that the topics and subtopics cannot divide the language into non-intersecting zones, and each researcher involved in the work on a more or less narrow topic always has to tread on the ground occupied by colleagues. This double work is not at all excessive; moreover, it compensates for the undesirability of relying on one’s own language competence to check the accuracy of the data and reduces the risk of making mistakes in documentation and interpretation of language facts. Plurality of the researchers’ efforts in the study of the same phenomenon allows a multisided stereoscopic view. It certainly requires effort to harmonize and work out a unified account of each phenomenon. But this requirement in principle makes it possible to achieve more objectivity in description and to escape many ad-hoc, erratic or ill-conceived hit-or-miss decisions, both in data representation and theoretical interpretation. 5.3 Cross-Training The team method makes it possible to achieve a higher level of researchers’ professionalism. In the very beginning of our fieldwork activities, when we mostly investigated traditional areas such as phonology and morphology, the division of the supervisors’ spheres of influence matched the boundaries between these areas. This was the way the Archi and Khinalug descriptions were created. Each of these levels certainly is to a large extent autonomous and requires special professional skills on the researcher’s part. Sound expertise in phonetics is absolutely necessary for a field linguist, since when he or she is not backed by an existing orthography, accurate transcription of sounds is a necessary condition for any linguistic description. In our work Sandro Kodzasov has always been an all-time expert in this area. His unrivaled ability to imitate and perceive all possible kinds of sounds is coupled with wide typological erudition and knowledge of all modern phonetic theories and methods. Due to his participation we were able to find new phenomena in all the languages we studied, and to achieve significant results. Even if phonetics is not a research goal as such (as is the case, for instance, in lexicographic work), it is impossible to ignore it. It is well-known, that individual researchers may spend years struggling with
39
40
Aleksandr E. Kibrik
mysterious phonetic segments and supra-segmental effects if they cannot learn to hear and record them, let alone describe them. As for morphology, particular approaches are required to identify and describe morphological units, which are markedly different from modern syntactic frameworks. A field linguist has to have a good command of morphological descriptive methodologies as well. Certainly, the present specialization of linguistics is so fine-grained that it cannot be projected onto team work in full. Each field worker has to have wide expertise. At the same time, in team work it makes good sense to combine wide linguistic knowledge about variegated problems and deep knowledge about some phenomena in more narrow spheres. We have been trying to achieve this ideal in our team work. We have been convinced by our fieldwork experience that a balance between wide expertise and deep specialization in more narrow domains helps to level the differences in description of a previously undescribed language and languages with long descriptive traditions. 5.4 Synergy Team fieldwork allows each researcher to gain an overall perspective on the language that would be impossible in a solitary research program. All project members are expected to contribute to the collective body of knowledge developed in real time. In our expeditions, daily workshops are an effective mechanism for integrating individual views on the language. The goal of these workshops is constant monitoring of new information gathered in the process of individual work with language consultants. In this way each researcher has access to knowledge beyond the limits of data that he or she collects. Workshop procedures can differ greatly, from short “hot news” reports to collective work on text analysis. Texts are treasuries where we naturally observe the full range of the language machinery. A detailed analysis of each expression occurring in a text leads to explanations about the way it works and makes it possible to update all team members in the current state of knowledge. Sometimes a language expression may seem mysterious to some team members, but is quite familiar to a researcher who has already encountered it in earlier individual work and found an explanation for it. If some text phenomenon is understandable to no one, it becomes clear which “cluster” is responsible for further work on it. It naturally follows that this cluster should make a presentation on the problem at one of the later workshops. Workshops often show that different people understand the same language phenomena in different ways. So, workshops become the ground for discussions,
Collective field work
where an optimal descriptive decision is chosen and a single, unified account of the language system is worked out. Collective work on texts is also useful because it gives us an opportunity to feel the fabric of the natural language. Another advantage of work with texts is that texts present statistical distributions of frequent and rare language phenomena. The most frequent structures are likely to belong to the core of the language, which means they have to be learned as soon as possible. On the other hand, since they are so common, team members tend to lose interest in them, and the focus of discussions shifts towards less frequent phenomena pertaining to the language periphery. Understanding and interpretation of these peripheral phenomena makes it possible to better understand and even re-interpret some of the basic phenomena. In practice, by the end of the second week of field work, team members are analyzing texts very quickly, and often stumble on fascinating new facts which add zest to workshops. In the workshops, each researcher daily adds to knowledge gained during sessions with a consultant. By getting to know the findings of colleagues all researchers significantly broaden their knowledge of the language. It helps streamline individual work with language consultants and makes this work more effective. For instance, after a three hour session with a language consultant, during the workshop, each fieldworker learns as much as could have been learned in thirty additional hours of consultant work. Workshops exponentially broaden linguistic knowledge, and after a week’s fieldwork, each team member possesses as much information as a lone fieldworker would have collected during many months of fieldwork. It is perhaps the most powerful resource of fieldwork analogous to an assembly line. It transforms linguistic fieldwork from craftwork into linguistic industry. 5.5 Encouraging the habit of professional inclusivity Often, in the course of work on a minority language a linguist finds him or herself in a situation of enforced professional isolation. An independent field linguist studying an endangered minority language can feel as lonely as the language being studied, especially if there are no other outsiders who have immediate knowledge of the language. The field linguist has no one to provide advice, no one with whom to discuss findings or problems. On the contrary, team linguistic fieldwork builds a group of colleagues; their professional interaction creates a favorable creative environment and becomes a powerful psychological incentive. This aspect of any research activity can hardly be overestimated.
41
42
Aleksandr E. Kibrik
Team fieldwork fosters a lively social circle, and encourages a vibrant attitude of discovery. For many students it is an attraction, a treat, a romantic goal. It is both a school of linguistics and a school of life.7 Therefore, supervisors always have more applicants than can be accommodated on any given field team.
6. Concluding remarks I realize that for many field linguists team field work is something unusual and even impossible. For my part, I can hardly imagine field work in isolation. Mostly this is because I have a very good idea of the advantages offered by collective work. Of course, all of our expeditions were not equally productive. As I have said, the first three expeditions were for undergraduate students. In these expeditions, the conception of team field work was being developed. After this, it became possible to prepare the description of Khinalug in a year. Since then, the structure of team work has been constantly changing, depending on many factors. In addition to a well-structured script for fieldwork sessions, human factors are of no less significance. Eventually, the most crucial considerations are high professionalism of the core participants, the right division of functions and everyone’s enthusiastic involvement in the project. Under such circumstances, the advantages of the team method can be realized in full. The peak of methodological maturity of our expeditions was in the 1990s: in 7 years, we completed major descriptions of three languages (Godoberi, Tsakhur and Bagvalal). This was possible only because a reliable team of colleagues was formed. On this team, the founders of the team method worked hand in hand with capable young researchers (Kazenin, Kalinina, Lioutikova, Tatevosov) who had become rooted in this method since the time they were undergraduates. In summary, the team method works, in spite of the disadvantages I have mentioned, when there is a synergetic team, on which everyone is cross-trained in the jobs of others and at the same time are qualified and designated to perform some task that no-one else can do.
Notes * Research for this paper was supported by RGNF (Russian Science Foundation for the Humanities), grant 04-04-0215a. 1. To put it more precisely, the department of structural and applied linguistics (DSAL). In 1992 it was renamed as the Department of theoretical and applied linguistics (DTAL).
Collective field work
2. I don’t mean technical changes: these do not require reinterpreting data that had been collected earlier. 3. “Agglutinative additivity” refers to the notion that paradigms involve standard morpheme values, e.g. Number, Case, Person of Agent and Patient, and the meaning of any string of morphemes is a sum of the meanings of its parts. Furthermore, in agglutinative paradigms, each grammatical meaning has a unified marker. Alutor, though a typical agglutinative language, in several respects deviates from this principle in its polypersonal conjugations. 4. At that time there was only one university: Dagestanskij gosudarstvennyj universitet. 5. Avar has the highest number of speakers in Daghestan and has a literary norm. However, there is much variety in its dialects; the dialect we studied was not very similar to literary Avar. 6. See Polinsky (2002) for a review of this volume. 7. You can read more about daily life of fieldworkers in V. Borshchev’s diary (Borshchev 2001).
References Borshchev, V. B. 2001. Za jazykom (On language). Moskva: Azbukovnik. Garvin, P. 1972. On linguistic method. The Hague: Mouton. Ibragimov, G. X. 1990. Caxurskij jazyk (Tsakhur). Moskva: Nauka. Kibrik, A. E. 1977a. The methodology of field investigations in linguistics (setting up the problem). The Hague-Paris: Mouton. Kibrik, A. E. 1977b. Opyt strukturnogo opisanija archinskogo jazyka (Structural description of Archi). Vol. 2. Taksonomicheskaja grammatika (Taxonomic grammar). Moskva: Izdatel’stvo MGU. Kibrik, A. E. 1977c. Opyt strukturnogo opisanija archinskogo jazyka. (Structural description of Archi). Vol.3. Dinamicheskaja grammatika (Dynamic grammar). Moskva: Izdatel’stvo MGU. Kibrik, A. E. (ed.) 1982. Tabasaranskie etjudy (Tabasaran essays). Moskva: Izdatel’stvo MGU. Kibrik, A. E. (ed.) 1993. The noun phrase in the Andalal dialect of Avar as spoken at Sogratl. Eurotyp Working Papers. Theme 7: Noun phrase structure. Working paper No. 18. Konstanz. Kibrik, A. E. (ed.) 1996. Godoberi. München and Newcastle: LINCOM Europa. Kibrik, A. E. (ed.) 2001a. Bagvalinskij jazyk: Grammatika, teksty, slovari (Bagvalal: Grammar, texts, dictionaries). Moskva: Nasledie. Kibrik, A. E. 2001b. “Anomalies” of cross-reference marking: The Alutor case. Morphology 2000, ed. by S. Bendjaballah, W. U. Dressler, O. E. Pfeiffer, and M. D. Voeikova, 199–212. Amsterdam: John Benjamins. Kibrik, A. E. 2001c. Opyt morfologicheskoj rekonstrukcii kognitivnoj strukrury (na materiale sfery lichnogo dejksisa v alutorskom jazyke) (Toward the morphological reconstruction of cognitive structure: evidence from personal deixis in Alutor). Philology N6.121–131. Vestnik MGU. Kibrik, A. E., and S. V. Kodzasov. 1988. Sopostavitel’noe izuchenie dagestanskix jazykov. Glagol (Comparative studies of Daghestanian languages. Verb). Moskva: Izdatel’stvo MGU.
43
44
Aleksandr E. Kibrik
Kibrik, A. E., and S. V. Kodzasov. 1990. Sopostavitel’noe izuchenie dagestanskix jazykov. Imja. Fonetika (Comparative studies of Daghestanian languages. Noun. Phonetics). Moskva: Izdatel’stvo MGU. Kibrik, A. E., S. V. Kodzasov, and I. A. Muravjova. 2000. Jazyk i folklor aljutorcev (Language and folklore of the Alutor people). Moskva: Nasledie. Kibrik, A. E., S. V. Kodzasov, and I. P. Olovjannikova. 1972. Fragmenty grammatiki xinalugskogo jazyka (A partial grammar of the Kinalug language). Moskva: Izdatel’stvo MGU. Kibrik, A. E., S. V. Kodzasov, I. P. Olovjannikova, and Dzh. Samedov. 1977a. Opyt strukturnogo opisanija archinskogo jazyka (Structural description of Archi). 1. Leksika, fonetika (Lexics, phonetics). Moskva: Izdatel’stvo MGU. Kibrik, A. E., S. V. Kodzasov, I. P. Olovjannikova, and Dzh. Samedov. 1977b. Archinskij jazyk. Teksty i slovari. (Archi: Texts and dictionaries). Moskva: Izdatel’stvo MGU. Kibrik, A. E. and Y. Testelec (eds.) 1999. Elementy grammatiki caxurskogo jazyka v tipologicheskom osveshchenii (Studies in Tsakhur: a typological perspective). Moskva: Nasledie. Longacre, R. 1964. Grammar discovery procedures: Field manual. The Hague: Mouton. Mel’chuk, I., and A. K. Zholkovskij. 1970. Towards a functioning ‘meaning-text’ model of language. Linguistics, an international review 57: 10–47. Newman, P. and M. Ratliff (eds.). 2001. Linguistic fieldwork. Cambridge: Cambridge University Press. Polinsky, M. 2002. The marvels of Tsakhur. Linguistic Typology 6.403–22. Samarin, W. 1967. Field linguistics. New York: Holt, Rinehart, and Winston.
Grammars and the community Marianne Mithun University of California, Santa Barbara
The audience for a grammatical description is an important consideration for anyone involved in descriptive linguistics. Potential grammar users include linguists, the interested public, and members of the communities in which the language is spoken. An awareness of the target audiences is necessary in shaping the grammar to meet varying needs. It might, for example, affect the choice of topics to be discussed, the organization and style of the presentation, the depth of detail to include, the use of technical terminology, and the nature of exemplification. It is not yet clear whether one grammar can serve all potential audiences and purposes. Whether it can or not, however, there is a good chance that any grammar will eventually be pressed into service for more than one. This paper offers some suggestions based on the author's experience with Mohawk communities situated in Quebec, Ontario, and New York State.
1. Introduction — Who reads linguistic grammars? A useful point to consider when planning a grammar is its intended audience. Potential grammar readers might include linguists, interested laymen, or, increasingly, members of the communities in which the language is spoken. Since these groups approach grammars from a variety of backgrounds and with a variety of purposes, an awareness of the target audience can be useful in shaping the grammar to meet their needs. It might, for example, affect such decisions as the choice of topics to be discussed, the organization and style of the presentation, the depth of detail to include, the use of technical terminology, and the nature of exemplification. It is not yet clear whether one grammar can serve all potential audiences and purposes. Whether it can or not, however, there is a good chance that it will eventually be pressed into service for more than one. This likelihood is increased in the case of endangered languages, where further descriptions may not be possible.
46
Marianne Mithun
None of the three audiences mentioned above, linguists, interested laymen, or community members, is itself a homogeneous group. Among linguists, probably the most avid grammar readers are typologists, scholars engaged in comparing particular structures across large numbers of languages. Their primary goal is often to check certain predetermined features and correlations. Does the language show basic verb-initial, verb-medial, or verb-final word order? Does it show preposed or postposed relative clauses? Other linguists consulting grammars might be engaged in constructing models of a particular area of language structure, such as coreference relations between noun phrases and pronouns in different clauses. Such theoreticians may hope for detailed accounts of all logically possible combinations of relevant structures within a particular domain: subject pronouns in preposed subordinate clauses (Because he had left, John missed seeing Bill), object pronouns in preposed subordinate clauses (Because John had seen him, Bill asked Sam to wait …), etc. Comparative linguists, interested in how languages change over time, might consult a grammar in order to see how the language described compares with others that are genetically related or areally contiguous. They might hope to find systematic and explicit lists of correspondences to those other languages, as well as discussions about the sources of forms and constructions that can be traced within the language itself. Still other linguists might be working on grammars themselves. These readers may be more interested in learning about the unexpected possibilities: unusual distinctions languages might encode, or previously unidentified structural generalizations that could characterize large numbers of expressions. Reading a grammar can alert them to categories and patterns to investigate in the language they are documenting. A more general lay public might include individuals who are simply interested in the community in which the language is spoken and its culture. It might include anthropologists and historians hoping to gain an understanding of traditional patterns of thinking. It might include translators whose goal is to produce materials for community use or to bring material from the community to a broader audience. It might include persons from outside the community who hope to provide various support services. Such community workers might wish for information that might help them to gain a rudimentary speaking ability in order to facilitate communication. Finally, an important and growing audience for the grammar might come from within the community in which the language is spoken. Among these readers might be language teachers, that is, speakers charged with passing on the traditional language to children or adults who do not know it or do not know it well. There might be curriculum planners who are constructing lesson plans and teaching materials for the community. There might be community members who
Grammars and the community
are simply intrigued by the beauty and intricacies of a system of which they were not previously conscious. Or, increasingly, there might be descendants of speakers, eager to discover and appreciate this central aspect of their heritage. Important functions of grammars directed at these audiences might be to show the overall structure of the language in its own terms, and to document its richness, particularly of what makes it special. (For an additional perspective, see the contribution by James Kadanya in this volume.) The potential audience for a good grammar is thus likely to be diverse, with a variety of backgrounds, skills, interests, desires, tastes, and purposes. It will also be a changing one. Topics of interest to linguistic theoreticians are notoriously ephemeral. The hottest issues, those generating the most excited discussion at one moment, are often the quickest to become outmoded. Even more important is the potential for change within the community audience. Not long ago relatively few community members were very interested in the work produced by linguists. When everyone knew the traditional language well and used it as a primary means of communication, their linguistic knowledge was often taken for granted, and the humble attempts of an outsider to record and describe a fragment of it could offer little excitement. Increasing globalization, however, has meant increasing worldliness, sophistication, education, and curiosity on the part of members of many communities. Speakers and their descendants are increasingly aware of the complexity and value of their own heritage. A good grammar can even heighten their appreciation of it. If grammars are to be consulted by such a varied and ever-changing audience, it might be useful to begin discussion of how grammar writers can best serve the diverse needs of their readers now and in the future. Among the issues for discussion are the relation between grammar writing and linguistic theory, the use of technical terminology, the quantity of detail to include in descriptions, the choice and format of examples, and the nature of the data that provide the foundation for the description.
2. Grammar writing and linguistic theory An ongoing question for grammar writers is the optimal balance between theoretical abstraction and language-specific substance. It is clear that a sound knowledge of the categories and structures that have already been observed to occur in other languages is essential to the effective analysis and description of a new language. If one has never heard of ergativity, for example, a tremendous amount of time can be wasted trying to find subjects and objects in a language without them. A
47
48
Marianne Mithun
familiarity with ergative patterns, on the other hand, can allow the researcher to identify such patterns quickly and, furthermore, alert him or her to note how far the patterns extend. Are they confined to lexical noun phrases or do they extend to pronouns as well? Do they appear in both main and subordinate clauses, or just one? Do they appear in all tenses, aspects, and moods, or just some? Antipassive constructions frequently play a significant role in languages with ergative patterns. The theoretically sophisticated researcher will be on the lookout for antipassives in a language with ergative constructions, and the syntactic and pragmatic roles they might play in connected speech. At the same time, a certain amount of judgment is called for in determining the extent to which the shape of a grammar should be driven by current theoretical concerns. Because of the modern history of linguistic theory, with its strong roots in Europe and North America, the theoretical issues most heavily discussed have tended to involve structures found in English and, to a lesser extent, related languages. The syntactic structures of written English in particular have defined the foci of a substantial proportion of theoretical work to date, though fortunately there is increasing consideration of a wider range of languages. If a grammar is based primarily on translations of the English sentences that underlie current theoretical issues, the potential contribution of that grammar to both the linguistic and local communities is diminished. We will be deprived of some of the most theoretically interesting aspects of the language under consideration, those which would allow us to broaden our theoretical perspective in the most interesting ways. We will also fail to document what makes this language special, a record of the particular culture that shaped it. Linguistic typologists often express frustration at the gaps they find in grammars. They would like each grammar to contain not only a specification of the categories and structures that exist in the language, but also a specification of those that do not. This desire is certainly understandable: the goal of typology is the comparison and classification of recurring features or constructions across large numbers of languages. The work raises deeper questions about whether languages are fundamentally the same, with only incidental differences, or profoundly different. These discussions will probably continue throughout the life of the discipline. But the issue calls for reflection about the level of categories and structures worthy of mention. Should we specify for every language whether or not it has a passive construction? Double wh-movement? Prepositions? A perfect? An ablative case? A potential danger in over-inclusiveness is that of shaping the description of a littleknown language in terms of the structures currently recognized in better-known languages. One alternative that has been suggested is a large index at the end of
Grammars and the community
the grammar, listing all occurring and non-occurring structures, with pointers to the pages on which the occurring structures are discussed, and explicit labeling of non-occurring ones. A closely-related issue is the extent to which the grammar should reflect the full range of logically possible structures, as opposed to the distribution of structures that occur in natural speech. In many cases speakers can supply literal translations of English constructions in their language even when these constructions rarely if ever occur spontaneously. The elicited translations may fail to occur for a variety of reasons. They may be pragmatically self-contradictory, for example, or other constructions may normally be used in their place. Certain constructions may be used only with particular lexical items in natural speech. A theoretician could be chagrinned to find no example of a particular logically possible construction, and no specification of its ungrammaticality, but the grammarian might wonder whether an elicited translation is truly part of the language and belongs in the permanent record. Again, such decisions must be made on a case-by-case basis. Awareness of the logical possibilities can prompt the grammarian to search for the full range of pertinent examples in the database, so that crucial examples are not inadvertently omitted. Creating structure, however, can provide a distorted view of the intricate system of systems that is the language. On this issue, community concerns might motivate greater fidelity to actuallyoccurring speech. In many cases, local languages are being replaced by global languages in stages. Skilled speakers use the traditional languages in fewer and fewer contexts. Succeeding generations control smaller inventories of rhetorical styles, complex constructions, and vocabulary. In such situations, faithful documentation becomes increasingly important. A record of what is actually said by skilled speakers in a variety of situations, when they are choosing what to say and how to say it, can provide a priceless record not just of relative clause structures, for example, but also of what was said, of how experience was segmented into concepts and how these concepts were combined, of how speakers interacted with each other. Each example taken from spontaneous speech provides an illustration not just of the point it is meant to illustrate, but of many more aspects of the language and language use. Such documentation can also provide answers to theoretical questions we do not yet know enough to ask. It can be our best hope for serving future readers from all backgrounds.
49
50
Marianne Mithun
3. Terminology It has sometimes been suggested that the actual technical terminology used in a grammar is insignificant, so long as all terms are defined. Definitions are certainly important, both for the casual, uninitiated reader and for the theoretician interested in knowing the exact concept intended by the author. But the choice of terms can also be important. A number of sometimes contradictory factors can enter into the optimal choice. If the grammar is to be accessible to the largest possible audience, unnecessary technical terminology can be detrimental. In any field, technical terminology plays both intellectual and social roles. Intellectually, it can permit the refinement of our understanding of a concept. But socially, it can distinguish members of the ‘in-group’ from all others. In-group terminology can constitute a barrier to understanding and discourage general readers. (It can also be short-lived, so that work incorporating it can appear outmoded quite quickly.) Where technical terms are necessary, several considerations are at stake. Linguistic typologists often express frustration at the incommensurability of grammars if terminology is not uniform. Someone researching subjunctives, for example, might prefer that every grammar include a section labeled ‘Subjunctives’, with either a description of the forms or a statement that they do not exist in that language. As the kinds of linguistic categories that occur have become better known and understood over the past half-century, much common linguistic terminology has become widely established. It is important for the grammarian to know what terms are in general use, and exactly how they are commonly understood. Inventing a new term for a category that is common cross-linguistically and already known under another name is counterproductive. At the same time, if incommensurate categories are forced too quickly into a single terminological box, we can lose an important value of the grammar: the opportunity to appreciate the potential richness of language variation. Two other factors can enter into the choice of technical terminology where it is necessary. One is the existence of terminology in previous studies of the language or related languages. If there is a strong tradition, for example, of using a certain term for a particular grammatical prefix, the grammarian might think twice before choosing a different term. If a change is made, it is helpful to specify the equivalence. To lessen frustration for grammar users within the speech community, of course maximal transparency of terminology is extremely important. All else being equal, a term like ‘past’ is more transparent than a term like ‘preterite’. It can sometimes be useful to include the form of a marker when mentioning the term, as in ‘the s‑Repetative’. Learning new terminology requires an investment of time
Grammars and the community
and patience for anyone, particularly for those who are not in the habit of reading grammars. For this reason, it can be important to make careful choices the first time, in order to minimize the trauma of revision. The linguist might suddenly decide that a particular suffix would be more appropriately called an Inceptive than an Inchoative. In the end, however, changing an established term might not be worth the risk of alienating the audience.
4. Questions of quantity Different audiences will certainly vary in the amount of detail they would like to find in a grammar. Furthermore, the preferences of individual grammar users can change over their lifetimes. We expect a grammar to specify the basic patterns that characterize the language. But just how much detail should be supplied? Should the patterns be simply stated, or should they be explained where this is possible? Should apparent irregularities be discussed? Should examples be simple, brief, and few, or should they be elaborate and numerous? Should full inflectional paradigms be included, even where they are predictable by general rule, or are general rules sufficient? Should forms be illustrated in the context of larger stretches of speech? Should gaps in usage be discussed? The best answers to such questions will undoubtedly be different for different communities. Some points for reflection are offered here from the experiences of a consortium of six Mohawk communities situated in Quebec, Ontario, and New York State. Beginning in the early 1970’s community members noticed that children were no longer learning the language at home. Speakers began offering language classes in community schools, but they soon realized that their skill at speaking the language did not fully equip them to pass on the language during the limited classroom time at their disposal. They felt they could benefit from some technical knowledge of the structure of the language and teaching techniques, so they set about establishing university certificate and degree programs in these areas. Community orthographies were established, and teachers began to learn about the complex structures of their language, making conscious what had previously been largely unconscious. They realized that if they were to teach Mohawk effectively, with the goal of passing on Mohawk ways of viewing the world, their classes should not be organized according to the curriculum used for teaching French. Mohawk teachers did not teach grammar as a formal subject, but their lesson plans were based on a fundamental understanding of Mohawk in its own terms. At this stage, emphasis in teacher-training programs was on the basic, recurring, regular patterns specific to the Mohawk language.
51
52
Marianne Mithun
Over time, language programs for both children and adults blossomed in all of the communities. Immersion programs, in which students learn all of their subjects in Mohawk, were established and have continued with great success. Teachers became increasingly sophisticated in their conscious understanding of the intricacies of the language. Curriculum committees were formed to produce materials for the classrooms. At a certain point, community members involved in these programs realized that they would like a comprehensive reference grammar of the language, complete with fine details about the sounds of the language, dialect differences, word structure, sentence structure, discourse structure, and style, as well as explanations of the seeming exceptions that crop up repeatedly as they prepare lessons. The issues considered below come from our experiences in planning and working on this grammar. 4.1 How much detail should be included with descriptions? All grammars should of course outline the basic structures of the language. Mohawk, for example, contains three different constructions comparable to the possessive constructions of languages like English. If teachers are to present coherent lessons, they need to be aware of the three constructions and perhaps present them at different times. (1)
Three Mohawk possessive constructions: Inalienable possession k-kahrà:ke Alienable possession ak-hnà:ta’ Interpersonal relationships rak-hsótha
my eye my purse my grandfather
How much more should be said in the grammar? Certainly the situations in which each construction is used should be specified, and the form of each should be described. Mohawk teachers, who are quite sophisticated, are comfortable with at least as much description as below. Inalienable possessive constructions are used for most, but not all, body parts. They are formed from an inalienable possessive prefix referring to the possessor (k‑ 'my'), followed by a noun stem identifying the possession (‑kahr‑ 'eye'), a noun suffix (‑a’), and an ending (=ke) meaning 'place'. (2)
Inalienable possession kkahrà:ke k-kahr-a’=ke 1sg.inalien.possessor-eye-noun.suffix=locative.nominalizer ‘my eye (place)’
Grammars and the community
Alienable possessive constructions are used for most other possessions. They are formed from an alienable possessive prefix referring to the possessor (ak‑ 'my'), a noun stem identifying the possession (‑hna’t‑ 'purse'), and just a noun suffix (‑a’). (3)
Alienable possession akhnà:ta’ ak-hna’t-a’ 1sg.alien.possessor-purse-noun.suffix ‘my purse’
Most kinsmen and other associates are identified in terms of relationships rather than possession. Both parties in the relationship are specified in the pronominal prefix. Senior members are expressed first, junior members second. (4) Kinship terms a. rakhsótha rak-hsót=ha m.sg/1sg-be.grandparent.to=diminutive ‘he is grandparent to me’ = ‘my grandfather’ b.
riiaterè:’a rii-atere’=a 1sg/m.sg-have.as.grandchild=diminutive ‘I have him as a grandchild’ = ‘my grandson’
When both members are of comparable rank, a reciprocal construction is used. (5)
Reciprocal kinship terms ontiara’sè:’a onki-ar-a’se’=a 1.exclusive.du-reciprocal-be.cousin.to=diminutive ‘we two are cousins to each other’ = ‘my cousin’
In addition to the major question of just how much detail to include in the description, there is the issue of how much additional commentary should be provided. Some explanation can certainly be interesting, particularly if it is accurate, and it can make the structures easier to understand and remember. Relating particular structures to deeper generalizations can be useful for some readers. The Mohawk inalienable possessive prefixes resemble the agent pronominal prefixes that appear on verbs (I jumped, I grabbed him), with certain minor, systematic differences in form. The alienable possessive prefixes resemble the grammatical patient pronominal prefixes that appear on verbs (I slept, he grabbed me), again with certain minor, systematic differences in form. When these parallels between the pronominal
53
54
Marianne Mithun
prefix paradigms are pointed out, it significantly reduces the overall learning time for the reader. 4.2 How much detail should be included about form? As in many languages, the shapes of prefixes and suffixes in Mohawk can vary according to the sounds adjacent to them. The possessive prefixes show different shapes before different sounds. (6)
Allomorphy of possessive prefixes k-kahrà:ke my eye ke-neri’tstà:ke my navel k-a’kotarà:ke my heel t-ia’tà:ke my body
ak-hnà:ta’ akè-:sere akw-atháhsteren à:t-iare’
my purse my car my pants my pouch
The Mohawk communities show dialect differences as well. (7) Dialect variation k-ia’tà:ke my body t-ia’tà:ke my body
(Ohswé:ken, Taientané:ken, Ahkwesáhsne) (Kanehsatà:ke, Wáhta, Kahnawà:ke)
A traditional value in linguistics has been economy of description. A simple rule has been more highly valued than a complex one. No more examples have been deemed necessary than those required to prove the point. It is not clear that this kind of economy is the best answer for a useful grammar, particularly for community members. Additional examples might be of several kinds. In describing allomorphy, for example, one could first provide a full paradigm, like that below. (8)
Possessive paradigm ke-nontsì:ne se-nontsì:ne ka-nontsì:ne ie-nontsì:ne ra-nontsì:ne
my head your head its head, her head one’s head, her head his head
teni-nontsì:ne iakeni-nontsì:ne seni-nontsì:ne keni-nontsì:ne ni-nontsì:ne
you and I, our heads he or she and I, our heads you two, your heads they two (animals or women), their heads they two (males) their heads
tewa-nontsì:ne iakwa-nontsì:ne sewa-nontsì:ne
you all and I, our heads they and I, our heads you all, your heads
Grammars and the community
konti-nontsì:ne rati-nontsì:ne
they (animals or women) their heads they (males) their heads
To show patterns of allomorphy and dialect variation, additional paradigms could be provided. (The initials to the right below identify dialects.) (9)
Inalienable possession with a‑ stems k-ahsi’tà:ke my feet s-ahsi’tà:ke your feet w-ahsi’tà:ke its feet, her feet ion-hsi’tà:ke one’s feet, her feet r-ahsi’tà:ke his feet
ti-ahsi’tà:ke ki-ahsi’tà:ke iaki-ahsi’tà:ke iati-ahsi’tà:ke tsi-ahsi’tà:ke ts-ahsi’tà:ke ki-ahsi’tà:ke ti-ahsi’tà:ke i-ahsi’tà:ke
you and I, our feet O, T, Kw, Ks, W you and I, our feet A he or she and I, our feet O, T, A he or she and I, our feet Kw, Ks, W you two, your feet O, T, A you two, your feet Kw, Ks, W they two animals or women, their feet O, T, A they two animals or women, their feet Kw, Ks, W they two males, their feet
tew-ahsi’tà:ke iakw-ahsi’tà:ke sew-ahsi’tà:ke kon-hsi’tà:ke ron-hsi’tà:ke
you all and I, our feet they and I, our feet you all, your feet they all (animals or women) their feet they all (males) their feet
To convey the full pattern, still more paradigms would be necessary, displaying forms with noun stems beginning with other vowels and consonant clusters. Alternatively, the basic paradigm could be followed by rules or statements about phonological processes or alternations, in the style preferred by the grammar writer. The form of the Mohawk neuter-zoic prefix, for example, used for animals and for some women and girls, appears as ka-, w-, or i-. (This i is the orthographic convention for the palatal glide [j] before a vowel. Nasalized vowels are written as digraphs en and on.) (10)
Neuter.zoic inalienable possessor its, her ka- before consonants and the vowel i ka-i > ken (en = [˜ә]) w- before vowels a, e, en w > i before vowels o, on (i = [j], on = [õ])
55
56
Marianne Mithun
The alternations described here represent a variety of types. The dissimilation of the back glide w to a front glide i [j] before back vowels is a regular, recurring alternation that can be seen throughout the grammar. The fusion of the vowels a and i to the nasal vowel en (phonetically a nasalized mid-central vowel) can be seen just within the pronominal prefix paradigms. The shift of ka‑ to w‑ before vowels a, e, and en is not a regular phonological alternation at all. It is unique to this form and suppletive. The grammar writer describing this allomorphy must decide whether to state these alternations together in the same part of the grammar, individually in separate areas of the grammar, or both. For the Mohawks, the best solution appears to be a combination, even if certain regularities are stated more than once. Full sets of paradigms are crucial. The grammar writer may have assimilated the intricate sets of alternations and be able to recreate them, and of course speakers can produce the forms without reflection, but the full sets of paradigms allow readers to see the patterns for themselves. They provide teachers and curriculum designers with examples. And they provide non-speakers with forms they can be confident are correct, incorporating all of the complex alternations that interact. At the same time it is useful for many readers to see a summary of the forms as in (10), whether in that format or some other. It provides helpful generalizations to adults learning the language or teachers and curriculum planners assembling lessons that will allow students to absorb patterns. Finally, particularly if the grammar is meant to serve linguists as well, it is appropriate to repeat the individual alternation patterns in the various appropriate sections of the phonological description. 4.3 How many examples? As noted, scholarly linguistic analyses typically provide only enough examples to illustrate the point under discussion. Similarly, language patterns are sometimes easiest for the non-linguist to see in one or two simple examples. The Mohawk first person singular pronominal prefixes k‑ and ak‑ are followed by an epenthetic e before certain consonants (n, r, w) and consonant clusters beginning with glottal stop. This point can be made by a statement of the distribution and an illustrative example as in (11). (11)
Mohawk prefix allomorphy: the simple view khnia’sà:ke akhnià:sa’ k-hnia’s-a’=ke ak-hnia’s-a’ 1sg.inal.poss-throat-ns=locative 1sg.al.poss-throat-ns=locative ‘my throat’ ‘my collar’
Grammars and the community
ke’nionkserà:ke ke-’nionkser-a’=ke 1sg.inal.poss-toe-ns=locative ‘my toe’
ake’niónkseri ake-’nionkseri 1sg.al.poss-toe-ns=locative ‘my onion’
The grammar writer could stop with these examples, or provide more, as in (12). (12)
Mohawk prefix allomorphy: the more extensive view k-konhsà:ke ‘my face’ k-ken’kwarà:ke ‘my forehead’ k-hsonhkarà:ke ‘my lip(s)’ k-hsineko’tà:ke ‘my ankle(s)’ k-hiohsà:ke ‘my elbow(s)’ k-hsinà:ke ‘my leg(s)’ k-hna’tshà:ke ‘my rump’ k-ho’kwà:ke ‘my buttock(s)’ k-tsi’erà:ke ‘my nail(s)’ k-hsiahontà:ke ‘my stomach’ k-ihnà:ke ‘my skin’ k-entskwe’nà:ke ‘my chest’ k-ahsi’tà:ke ‘my foot/feet’ k-ahontà:ke ‘my ear(s)’ k-ahiakwirà:ke ‘my toe(s)’ k-ahsi’tò:kon ‘sole of my foot’
ke-nontsì:ne ke-nho’kwà:ke ke-nawirà:ke ke-nahsà:ke ke-nia’kwà:ke ke-ronhkwe’nà:ke ke-rhiotshà:ke ke-weionhkarà:ke ke-’nionhsà:ke
‘my head’ ‘my cheek(s)’ ‘my teeth’ ‘my tongue’ ‘my throat’ ‘my back’ ‘my chin’ ‘my thumb’ ‘my nose’
ke-na’wasà:ke ke-nhoskwà:ke ke-na’ahtà:ke ke-nhonhrò:kon ke-nekwen’tà:ke ke-nentshà:ke ke-ratà:ke ke-’rhiotshà:ke ke-’nahsà:ke
‘my eyelid’ ‘my jaw’ ‘my rib(s)’ ‘my armpit(s)’ ‘my belly’ ‘my arm(s)’ ‘my heel’ ‘my chin’ ‘my tongue’
As in the case of paradigms, more examples do allow readers to see patterns for themselves. They can provide teachers and curriculum designers with useful material, as well as second language learners. They can also provide a check on combinations of rules and details of allomorphy. 4.4 Should apparent irregularities be discussed? It was noted earlier that the possession of body parts is usually expressed in Mohawk with an Inalienable possessive construction. But some body parts appear only with Alienable possession. (13) Mohawk body parts with Alienable possession ake-nónhkwis ‘my hair’ ake-nekwénhsa’ ‘my blood’ akw-atstiéhseri ‘my kidney(s)’ ak-htsinonhiatà:ke ‘my vein(s)’
57
58
Marianne Mithun
Furthermore, some Mohawk nouns for body parts can appear in either Inalienable or Alienable possessive constructions. (14) Mohawk body parts with either k-tsi’erà:ke ak-tsi’erà:ke ra-nontsì:ne rao-nontsì:ne
‘my fingernail’ ‘his head’
For community readers learning about the structure of their language for the first time, it might be easier not to hear about such exceptions while they are attempting to make sense of the overall generalization. They might prefer to hear about just the forms that contribute to the general pattern. If the seeming exceptions are pervasive, however, they will appear soon enough. Teachers and curriculum designers will bump into them as they try to devise coherent lesson plans. If exceptions are to be included, there are again options. The choice will depend on the pervasiveness and transparency of the seeming exceptions and the preferences of the readership. One option is simply to list exceptions at the end of the description. The Mohawk chapter on possession could list body part terms that appear in Inalienable possessive constructions, then those that appear in Alienable possessive constructions, then those that can appear in both. A second option is to offer possible explanations for the differences. Again, an important consideration is the quality of the explanation. Good, valid explanations can make the patterns easier to understand, assimilate, and appreciate. The Mohawk body parts that appear with Alienable possession are not a random set. Most are internal organs like veins and kidneys, not ones normally visible to one viewing a live person or animal, and not ones that the owner normally has conscious control over. Mohawk speakers have suggested that body parts like arms and legs are extensions of the owner in a way that veins and kidneys are not. If someone hits my arm she hits me. Asked about hair, speakers note a sense of separation and lack of control. The Mohawk body parts that appear in both Inalienable and Alienable possessive constructions, like eyelashes and fingernails, are also not a random set. They occur both attached to and separated from their owners: one can lose an eyelash or cut off a fingernail. They tend to occur in Alienable possessive constructions when they are separated from their owners, and in Inalienable constructions when attached. The final example with the head is especially revealing. Possession of a head is normally inalienable. The Alienable construction raonontsì:ne ‘his head’ came from a story about a Headless Horseman who carried his head under his arm.
Grammars and the community
4.5 Should examples illustrate the use of forms in context? Another consideration is whether words illustrating morphological structure should be shown in the larger syntactic or discourse contexts in which they occur. For beginners and anyone else unacquainted with the language, it is obviously easier to see morphological patterns when just the word is provided, as in the examples above. The paradigms in (9) and the lists of terms in (12) allow speakers to see generalizations they would miss in more elaborate examples. If the purpose of the grammar is to provide as full a description as possible of the language, however, examples of the construction in a larger context can be important as well. The words described may not fit into natural speech in the same way as their English translations. Their use may be dependent on certain features of the larger context that may not even be apparent to the grammarian at the outset. For full documentation, then, at least some examples from the spontaneous speech of skilled speakers can be useful. If such examples are to be included, one might consider how much grammatical information about the example should be included. Is a simple sentence translation sufficient? Would a word-by-word gloss help? A full morphological analysis? For examples like those in (15) and (16), for example, one could provide just the first lines given here (the sentence as spoken) and last lines (the free translation); or the first (as spoken), fourth (word-by-word translation), and last (free translation); or the first (as spoken), the second (morphological parsing), third (morpheme-by-morpheme gloss), and last (free translation); or all five. (15)
Possession in context: Sha’tekenhátie’ Phillips, Kahnawà:ke, speaker p.c. Iáh ò:ni’ ne énska tekhé:kén ne: akhwá:tsire’ iáh o’ni’ ne énska te-khe-ken-’ ne ak-hwatsir-’ not also the one neg-1sg/3-see-prf the 1sg.inal.poss-family-ns not also the one did I see the my family And not once did I see my family.
(16)
Possession in context: Margaret Edwards, Ahkwesáhsne, speaker p.c. Ionkwara’se’okòn:’a ionkw-ar-a’se’=okon’a 1.exclusive.pl-reciprocal-be.cousin=distributive we all are individually cousins to each other My cousins
kiótkon thatinatà:re’s. tiotkon t-hati-nat-a-hr-e’-s always cislocative-m.pl.agt-visit-joiner-purposive-stative-distributive always they come to visit here and there were always coming to visit.
59
60
Marianne Mithun
Mohawk speakers feel that such examples are crucial for documenting the language in its full glory. The morphological structure of the language is sufficiently complex that it will not be obvious to most readers, whether or not they are speakers themselves, so the analysis is useful. (Not all readers, linguists or non-linguists, will read a grammar from cover-to-cover, memorizing each form along the way.) Printing the medial analysis lines in finer type allows those who are not interested in such detail to pass over them more easily. Each example can also provide information about not just the particular construction being described at that point, but about many other points of grammar. 4.6 Should frequency be addressed? As mentioned earlier, a recurring criticism by some grammar readers is the absence of overt statements about structures that do not occur in the language. Most grammar writers attempt to present the language in its own terms. To what extent should the structures of English or other well-known languages shape the description of the target language? At least in part, such decisions will depend to some extent on the frequency of particular constructions cross-linguistically. A more subtle issue is the relative frequency of a construction within the language described. A construction that is central to one language and common in natural speech may be marginal in another and extremely rare. Quantifying the difference precisely is nearly impossible in most situations; it would require an extensive corpus of speech in a variety of genres, somehow weighted to reflect daily usage. But frequency can be an interesting fact about a form or construction. Again the Mohawk possessive constructions provide an example. Apart from kinship terms, possessive constructions are surprisingly rare in spontaneous Mohawk speech. It can be useful to discuss the kinds of constructions that are used in their place. Nouns are comparatively rare In Mohawk speech. In situations where English speakers would use sentences with independent possessed nouns, Mohawk speakers often use verbs alone. To say I cut her hair, for example, Mohawk speakers do not use an inalienably possessed noun her hair. They use a verb into which the noun stem for hair has been incorporated. The affected person, the one whose hair was cut, is not identified by a possessive prefix. Instead, she is a core argument of the clause, the person directly affected. She is represented by a pronominal prefix on the verb. (17)
Mohawk affected person: Wa’khenónhkwahre’ wa’-khe-nonhkw-a-hre-’ factual-1sg/f.sg-hair-joiner-cut-prf ‘I haircut her = I cut her hair.’
Grammars and the community
Verbal constructions are also used in place of alienably possessed nouns. Instead of a sentence with possessed noun phrase, like They plowed my garden or I turned off my light, a verb is used. The person indirectly affected by the plowing and the darkness is identified not by a possessive prefix on a noun, but by a pronominal prefix on the verb. The garden and the light are specified by nouns incorporated into the verb, and the indirect affectedness is indicated by a benefactive suffix. (18)
Indirect affectedness: Awenhráten Deer, Kahnawà:ke, speaker Wa’tionkwatonhontsahríhten. wa’-t-ionkw-at-onhontsi-a-hri-ht-en-’ factual-dv-3pl/1sg-middle-earth-joiner-break-causative-ben-prf ‘They broke the ground for me = they plowed my garden.’
(19)
Indirect affectedness: Leatrice Beauvais Kahnawà:ke, speaker p.c. Sok wa’katà:swahte’. sok wa’-k-at-a’sw-a-ht-’ so factual-1sg.agt-reflexive-go.out-joiner-causative-prf ‘So then I extinguished for myself = So then I turned off my light.’
In fact in many cases, verbal constructions with incorporated nouns are used in Mohawk where sentences with independent possessed nouns would be used in English. (20)
Verbal construction: Josie Horne, Kahnawà:ke, speaker Kaieríthon nì: ióntiats. Kaieríthon n=ì:’i ionk-iat-s (name) the=I indefinite/1sg-call-habitual As for me, they call me Kaieríthon = My name is Kaieríthon.
(21)
Verbal construction: Sha’tekenhátie’ Phillips, Kahnawà:ke, speaker Iáh ki’ nowén:ton nahò:ten teken’ not actually ever anything not is There was never actually anything
tha’tewakatonnhatierónnion. tha’-te-wak-atonnh-atieron-nion’ contr-dv-1sg.pat-life-be.strange-distributive.stative I was life-strange unusual about my life.
61
62
Marianne Mithun
5. The view from the data Simple examples can frequently provide the clearest illustrations of grammatical points. But the language that occurs in spontaneous conversation and narrative is rarely that simple. Often the only way to avoid all extraneous complexity is to construct examples for the point to be illustrated. It can be hoped that speakers are always involved in this process, either in drawing up sets of examples themselves to illustrate the patterns, or in providing translations of key sentences drawn up by the grammar writer. For many purposes, constructed examples provide an accurate view of the language. In some domains, however, even the best speakers produce material that falls short of capturing the essence of the language. Material based consciously or unconsciously on translations from the encroaching language can be shaped by that language. Simplification can also filter out elements that represent the heart of the language. It is interesting to compare some textbook materials devised by an excellent Mohawk speaker with her spontaneous speech. Now an elder, she was raised by her grandparents. She taught Mohawk for years and attended courses in Mohawk linguistics. The materials she developed, consisting primarily of lists of vocabulary and sentences, are an extremely valuable resource. At the same time, it is interesting to observe some of the subtle ways her written materials differ from her normal speech. The written material is not technically ungrammatical; it is simply, on occasion, different from what is usually said. (In examples (22) and (23) below, the first and third lines are from her original manuscript. The middle line, with wordby-word translation, has been added here to make the examples easier to follow. In the spontaneous examples transcribed from conversation, each line represents an intonation unit or prosodic phrase.) (22)
Textbook example by Mohawk language teacher Istèn:’a wa’ehní:non akotià:tawi. my mother she bought dress My mother bought a dress.
(23)
Textbook example by Mohawk language teacher Rake’níha rawé:ka’s sewahió:wane’ tewà:ia. my father he likes apple pie My father loves apple pie.
(24)
Same speaker in casual conversation Né: ki’ na’ ni’ enkathrória’te’ né ki’ nà:’a n=ì:’i en-k-at-hrori-a’t-e’ the just guess the=emphatic.I fut-1sg.agt-middle-tell-causative-prf
Grammars and the community
the guess myself I will cause myself to discuss What I would like to talk about is
tsi niió:re’ tsi ni-io-or-e’ so partitive-n.pat-cover-stative so as it is far how unusual
niwakenhnhò:ten ni-w-akennh-o’t-en partitive-n.agt-summer-be.a.kind.of-stative such a kind of summer this summer has been.
(25)
Same speaker in another conversation about golf Speaker A Iáh tho niió:re’ thiahón:we. iáh tho ni-io-r-e’ th-i-ahon-w-e not there partitive-n.pat-be.far-stative contr-transloc-opt-n.agt-go not there so it is far would it go there It won’t go that far.
Speaker B Tsi ní: tsi tsi ni-io-ht tsi so partitive-n.pat-be.so.stative so so so it is so The way,
wáhi’ shes ahshwà:’eke’ tho’k niió::re’ wáhi’ shes a-hs-hwa’ek-’ tho=’k ni-io-r-e’ right used.to opt-2.sg.agt-hit-prf there=just prt-n.pat-be.far-stative right used to you would hit it just there so it is far you know you used to hit it so far that
kwah seronhkè:nek enhsatkáhtho’ kwah se-ronhkehnek en-hs-at-kahtho-’ really 2.sg.agt-struggle fut-2.sg.agt-middle-see-prf really you struggle you will see it you were really struggling to see
tsi niió:re’ tsi ni-io-r-e’
tsi tsi so so
ò:ia o-hi-a’ n-other-ns other
ieká:ien’. ie-ka-ien
nòn:wa. n=onhwa the=now now
63
64
Marianne Mithun
so partitive-n.pat-be.far-stative transloc-n.agt-lie so so it is far it lies there to see how far away it was. (laughter)
Speaker A Iáh ki’ nòn:wa ken’k thiekaién:ta’s. iáh ki’ n=onhwa ken=’k th-ie-ka-ient-a’-s not just the=now just=only contr-transloc-n.agt-lie-inch-habitual not just the now just only does it come to live over there Now it doesn’t even go over there.
Speaker B Tóka’ ni’ khé: tha’kaién:ta’ne’. tóka’ ohni’ khé:ken th-a’-ka-ient-a’-ne’ maybe also instead contr-factual-n.agt-lie-inch-prf maybe also instead it just comes to lie And it might just land right here.
5.1 Lexical categories and idiomaticity One difference between the constructed and spontaneous Mohawk is in the kinds of words that occur. Constructed example (22) contains 1 predicate and 2 nominals; example (23) contains 1 predicate and 3 nominals. In the spontaneous examples in (24) and (25), however, which are much longer, there is not a single nominal. (The examples were not chosen to illustrate this difference.) What is expressed in nouns and complex noun phrases in English is more typically expressed with verbs in Mohawk. Constructions with independent nouns are not ungrammatical; they certainly do occur. They are simply comparatively infrequent and pragmatically specialized. Multiple lexical noun phrases within a clause, like those in the constructed examples, are particularly rare. If example sentences are consciously or even unconsciously based on English, the distribution of lexical categories such as nouns and verbs can be affected. Idiomatic ways of speaking, an integral part of the language, are easily lost. A sample of the kind of expression that may fail to appear can be seen by comparing the textbook example in (26) with a similar request taken from spontaneous conversation in (27). Both are grammatical. The idiomatic expression in (27) is simply less likely to appear in constructed examples. (26) Textbook request Enwá:ton en-w-at-on
ken enkatshó:ri ken en-k-atshori
onòn:tara o-onon’tar-a
Grammars and the community
fut-n.agt-middle-be.possible Q it will be possible Q May I have some soup? (27)
fut-1sg.agt-sip n-soup-ns I will eat soup
Spontaneous request in conversation Enwá:ton ken enkatathnekáhrhahse’ en-w-at-on ken en-k-atat-hnek-arh-ahse-’ fut-n.agt-middle-be.possible Q fut-1.sg.agt-rfl-liquid-serve-ben-p it will be possible Q I will liquid serve for myself May I have a cup of coffee?
5.2 Particles Another striking difference in the kinds of words used in the constructed and spontaneous examples is the density of particles: small, morphologically unanalyzable words such as wáhi’ ‘you know,’ kwáh ‘quite,’ and ki’ 'just.' There are no particles in the constructed examples in (22) and (23). In the spontaneous examples in (24) and (25), there are more than three times as many particles as verbs. One of the clearest marks of talented speakers is their extravagant and lively use of particles. Older, admired speakers use them exquisitely, while younger speakers who spend more time in English use them noticeably less often. The particles, which are notoriously difficult to translate, serve a variety of functions. Some mark syntactic constructions, but many others shape discourse, contribute innuendo, provide humor or surprise, mark presupposition, suggest the basis on which a statement is made, involve the listener, or convey politeness. Speakers are rarely conscious of the functions of particles or of their pervasiveness, due in part to their small size and their broader discourse and extra-linguistic scope. When helping to transcribe and translate recordings of Mohawk speech, speakers often simply fail to notice them. Because the particles rarely have clear English equivalents, and because the kinds of meanings they convey are often not concrete, speakers tend to omit them in careful, written Mohawk. Since their functions are so subtle, they are difficult to explain and are seldom taught in language classes. But they can be the heart and soul of the language. Constructed examples are certainly clearer without the distractions of particles. At the same time, if all examples in the grammar are constructed, a major richness of the language will not be documented.
65
66
Marianne Mithun
5.3 Styles of interaction Documentation of spontaneous conversation can provide a record of how people talk to each other, how they interact. Constructed examples may or may not capture these traditions. The textbook request for soup in (26) and the spontaneous request for coffee in (27) show approximately the same general structure. But numerous particles that appear throughout conversation never appear in constructed sentences. An example is wáhi’ or wáhe’, loosely translated as you know or right?. It rarely occurs in textbook examples, but it can be seen in the third line of the spontaneous example in (25) You know you used to hit it so far that … Differences in style pertaining to setting and politeness can be seen by comparing the requests for a name in (28) and (29). That in (28) is perfectly grammatical and straightforward. It would not be appropriate in many contexts, however. The spontaneous request in (29) is more courteous: less direct and mitigated with particles. (28)
Textbook example Nahò:ten iesá:iats? what one calls you What is your name?
(29)
Spontaneous request Enwá:ton’ kati’ ken n ní:se’ ahsatatenà:ton it will be possible just Q also yourself you would name yourself Would you also give your name.
5.4 Word order The constructed and spontaneous examples also differ in their constituent order. The order in the textbook examples is a consistent replication of English SVO (Subject-Predicate-Object): My mother bought a dress; My father loves apple pie. This is not an ungrammatical order in Mohawk. All major orders are possible under appropriate conditions. But it not the predominant order. Mohawk constituent order is not based on the syntactic roles of constituents, such as subject or object. It is instead fully pragmatic, used by speakers to manipulate the flow of information. Speakers highlight important, newsworthy information by placing it early in the clause, and background more accessible or secondary information by placing it later. The power of this pervasive device is lost in the constructed examples.
Grammars and the community
5.5 Syntactic structure: Conjoined arguments When speakers are accustomed to writing the encroaching language, the very process of writing can bring out syntactic structures from that language. The prevalent SVO order in Mohawk textbook examples is perhaps due at least in part to this phenomenon. Its effect can also be seen in conjoined structures. The constructed example in (30) is not ungrammatical strictly speaking. It is quite unlike usual Mohawk style in several ways, however. It is heavy with independent nominals: your son, my children, movie. It has only one particle, the syntactic conjunction and. It shows SVO word order. And it contains a very heavy agent expression consisting of conjoined nominals, both newly introduced in this sentence: your son and my children. (30)
Conjoined structure: textbook example Tièn:’a tánon’ kheien’okòn:’a ronaterohrókhon teióia’ks. your son and my children they have gone to watch movie Your son and my children have gone to the movies.
A more usual way of identifying joint participants is with a structure like that in (31). The full set of actors is identified by the pronominal prefix ionki‑ we two on the verb. The speaker is obviously one of the participants. The other is specified by name after the verb. (31) Conjoined structure: spontaneous conversation … teiontiahthenno’khónhne Charlotte. te-ionki-ahthenno-o’k-h-on-hne Charlotte dv-1du.pat-ball-hit-purposive-stative-past (name) we two had gone to hit balls Charlotte Charlotte and I had gone to play golf.
5.6 Pragmatically marked syntactic constructions Even the small samples of conversation given here show that Mohawk sentences rarely consist simply of a subject and predicate. The speaker opened a new topic of discussion in (24) above by saying What I would like to talk about is how unusual this summer has been. She accomplished this with series of particles and embedded clauses. (24)
Topic introduction: spontaneous conversation Né: ki’ na’ ni’ enkathrória’te’ it is just guess myself I will cause myself to discuss What I would like to talk about is …
Such structures rarely appear in constructed examples.
67
68
Marianne Mithun
6. The shape of the grammar The best shape for a grammar will ultimately depend on the needs it is destined to serve. Most grammars will be consulted by a variety of users. In many cases grammars are being consulted more and more by members of the communities in which the language is spoken. This audience is likely to be not only quite diverse, but also changing rapidly. Where the language is endangered, spoken well by ever fewer people, readers are likely to be becoming increasingly sophisticated and eager for knowledge about their heritage. For the Mohawk grammar project we have arrived at a format geared to serve this evolving readership and to document as much as possible of the richness of the language. The presentation is layered. Each topic is introduced with an overview consisting of just the basic facts, illustrated with a few simple examples. Following the overview are subsections with additional detail, full paradigms where these are pertinent, and parsed examples from spontaneous connected speech, with representation of all communities. Additional subsections may contain information on related topics not of interest to all readers. Readers who are just beginning their study of the language, or who simply want an overview before plunging into a particular topic, can read the basic overviews alone. In the description of the sound system, for example, the basic overview simply lists each distinctive sound with the symbol representing it in the standard community orthography and an example. The overview is followed by more detailed discussion of the phonetic properties of each sound, then a description the phonetic properties of stress, tone, and vowel length and their patterns of occurrence. Another section describes the intonation patterns characteristic of larger syntactic constructions. An additional section that might not be of interest to the casual reader traces the history of transcription practices, primarily in missionary documents, so that those interested in consulting earlier written records of the language will have a guide to equivalences. Another section provides a brief sketch of the cognates of Mohawk sounds in related languages. In the description of the verbal morphology, each prefix and suffix is described in a separate section. Each section opens with the basic meanings and forms of the affix under discussion, its position within the template, and one or two pairs of examples for each of its uses. If a verb can occur both with and without the affix, an example of each is provided with just a free translation. Additional subsections provide examples of the affix in context drawn from spontaneous speech, discussion of its formal and semantic behavior in combination with other affixes, and any special idiomatic uses and combinations.
Grammars and the community
Other chapters follow the same general format. Some deal with larger grammatical structures, such as enumeration and quantification, question formation, and complex sentences. Others deal with particular semantic domains, such as kinship, color, and place names. All contain brief introductory overviews of the structures under discussion which can be read on their own, plus additional sections providing fully parsed, naturally occurring examples of the structures and terms in context. The layered approach could prove overwhelming to communities just beginning to be interested in the inner workings of their language. For the Mohawks, it is hoped that this format can provide an introduction for beginners, a resource for more sophisticated users such as language teachers and curriculum planners, and a record of the language, in all its grammatical and stylistic complexity, as spoken by the elders who still use it so eloquently.
69
From parts of speech to the grammar* Pamela Munro University of California, Los Angeles
Making dictionaries is a vital aid to completing a full grammatical analysis of a language, particularly if the dictionary requires the specification of the part of speech for each entry. English (or “universal”) parts of speech may not be relevant in all languages, as can be shown by structural comparisons of “adjectives” in San Lucas Quiaviní Zapotec vs. Chickasaw and of “adpositions” in San Lucas Quiaviní Zapotec (and Mixtec) vs. Chickasaw. Each language will present its own structurally determined inventory of parts of speech, relevant for its own grammatical facts. Thus, discovering the actual parts of speech of a language is a critical part both of dictionary making and of grammar writing. These aspects of linguistic analysis are crucially interconnected. Were a language ever completely “grammatical,” it would be a perfect engine of conceptual expression. Unfortunately, or luckily, no language is tyrannically consistent. All grammars leak. – Edward Sapir, Language, ch. II
This is a very personal paper about some ways in which carefully identifying parts of speech can be important for a linguist writing descriptive grammars. I have learned this lesson through years of very rewarding but often difficult work on dictionaries. I believe that dictionary work can elucidate many features of grammatical analysis and reveal problems that must be dealt with that might not have been noticed otherwise. What I discuss here is personal because for the most part I will talk about things I have worked on myself. All aspects of linguistic analysis are necessarily interconnected. Basic phonological and syntactic analysis must be begun before almost anything else, and textual study is obviously important too, since many times constructions appear in texts that never show up in simple sentences. Work with a variety of speakers from different backgrounds obviously helps make one’s analysis truer and broader. But dictionary work can also make an important contribution, for the simple reason that in making a dictionary one has no license to ignore anything.
72
Pamela Munro
My work on dictionaries has perhaps made me more sensitive than many linguists to the need to classify and specify the structural role and thus part of speech of every word. In my experience straight syntacticians are very free to simply ignore difficult or annoying words, but a dictionary maker cannot do this (and be honest), nor can he or she ignore the many vexatious conflicts (or “leaks”) between such categorization and semantic intuition. Figuring out the answers to this kind of puzzle can make a grammar better — more complete, certainly, but also, often, richer and deeper. Dictionaries reveal hidden grammatical secrets of a language that may be overlooked or discounted during other types of analysis. Good dictionaries make — or at least lead to — good grammars. But Sapir’s famous comment about how grammars leak is worth remembering, since close examination of lexical data often will turn up many things that are hard to explain. Stressing the contribution that working on dictionaries can make to grammatical analysis was of course particularly appropriate for the audience at this symposium, because SIL has produced so many wonderful dictionaries (especially, as I know, for languages of Mexico), in many cases before grammars of the same languages were written. Making dictionaries helps in grammatical analysis, and in fact in the absence of dictionary work a grammatical description is very likely to miss important things. Do dictionary makers need to know grammar? Of course, but even if they don’t, makers of good dictionaries will learn about it while doing their job. The different ways I will focus on here in which work on the lexicon can aid grammatical analysis and description are interrelated. Creating a large body of data on lexical items (and their parts of speech!) and how they behave can lead to a better understanding of sentence structure and can help reveal how productive various syntactic and morphological processes are in the grammar. And being forced to look carefully at and decide on an analysis for each lexical item helps one realize the tricky parts of the grammar that must be dealt with if one is to be honest. I will begin this paper by briefly considering the notion of parts of speech. Next I’ll talk about some cases that help us test the boundaries among these. At the end of the paper I will discuss part of speech classification vs. analytical abstraction, and propose an expanded list of parts of speech for linguists to be on the lookout for.
1. Parts of speech: An introduction Their conference was put an end to by the anxious young lover himself, who came to breathe his parting sigh before he set off for Wiltshire. Catherine wished to
From parts of speech to the grammar
congratulate him, but knew not what to say, and her eloquence was only in her eyes. From them, however, the eight parts of speech shone out most expressively, and James could combine them with ease. – Jane Austen, Northanger Abbey, ch. 15
Languages vary in terms of how many parts of speech they have, and linguists need to consider carefully exactly how a language works in order to understand how to classify its words. And a linguistic classification of parts of speech will be different from a traditional grammatical one, even though “part of speech” is not the same as the linguist’s “lexical category”. The eight parts of speech of English referred to in the quotation from Jane Austen at the beginning of this section are nouns, pronouns, verbs, adjectives, adverbs, prepositions, conjunctions, and interjections. As I’ll discuss in Section 4 at the end of this paper, this list would probably not be the one a linguist making a dictionary or writing a grammar would use to classify the words of English, however, and almost certainly is not the most complete or appropriate list of parts of speech likely to be encountered in other languages. Probably the best linguistic classification of English into “parts of speech” would have more actual categories, a topic I’ll return to in Section 4. My view of what dictionaries should do in terms of classifying words into parts of speech and finer distinctions has changed and evolved over the last three major dictionary projects I’ve worked on. These projects involved Chickasaw, a critically endangered Muskogean language spoken in Oklahoma (Munro and Willmond 1994); Wolof, the national language of Senegal and Gambia, a member of the West Atlantic branch of Niger-Kordofanian (Munro and Gaye 1997); and San Lucas Quiaviní Zapotec (SLQZ), a Valley Zapotec language spoken in the Tlacolula District of Oaxaca (Munro and Lopez, et al., 1999). I have not always appreciated the usefulness of part of speech identification. In the introduction to my first important dictionary, which I did with Catherine Willmond of her language, Chickasaw, I did not spend enough time discussing parts of speech. In our introduction (1994: xi), we say, A Chickasaw noun is defined with an English noun…. A Chickasaw verb is defined with an English verb (given in the infinitive “to” form) whose subject matches that of the Chickasaw verb….
Definitions for other classes (adverbs, interjections, and so forth) work similarly. And that’s it. The reason for this cavalier attitude (as I would now, alas, describe it) was that Chickasaw, like many other languages of North America, does not really have many words that are not either nouns (interpreting this class to include pronouns) or verbs. Adverbs are an interesting variety of word (some are nominal in form, some clausal), and there are certainly some interjections, but
73
74
Pamela Munro
this language does not have separate classes of adjectives, prepositions, or (for the most part) conjunctions, and quantifiers (like adjectives) are verbal. So my having observed that there were just two major classes of Chickasaw words (noun-type words and verb-type words) was useful, but was hardly sufficient. My statement was certainly of some interest typologically (a point we’ll return to later), but it did not advance the analysis of the grammar. However, my knowledge of the grammar of Chickasaw does let me formulate a list of parts of speech for that language. The task of identifying a list of parts of speech is important for both dictionary makers and grammar writers, for two reasons. First, it’s something many readers expect to see,1 so why not include it? Second, the need to determine the part of speech of every new word encountered maintains a useful analytical tension. I would recognize a different list of parts of speech for each of the three languages whose dictionaries I just mentioned: (1) Parts of speech in three languages Chickasaw: nouns, pronouns, demonstratives, verbs, interjections, particles. Wolof: nouns, pronouns, determiners (demonstratives and articles), verbs, quantifiers, adverbs, prepositions, interjections, particles. San Lucas Quiaviní Zapotec: nouns, pronouns, demonstratives, verbs, adjectives, quantifiers, adverbs, prepositions, interjections, particles.
The lists in (1) immediately reveal some crucial typological differences among the three languages. – Although Wolof and SLQZ are much more similar in their part of speech inventories, Wolof, like Chickasaw, does not have a special syntactic class of adjectives; in both languages, “adjectives” are a type of stative verb. – In both Wolof and SLQZ, quantifiers behave differently from other modifiers, while in Chickasaw they are much more like ordinary (stative, adjectival) verbs. – Also unlike the other two languages, Chickasaw does not seem to have a class of adverbs. Morphologically speaking, Chickasaw has both simple adverb words that seem to be grammatically nouns and more complex adverb expressions that can be analyzed as clauses, such as onnakma ‘tomorrow’, literally ‘when it is the next day’, which contains the verb onna ‘to be the next day’ and the irrealis different-subject subordinator -kma, as in (2):2 (2) Onna-kma chi-pisa-l-a’chi. dawn-irr.ds 2sII-see-1sI-will ‘I’ll see you tomorrow’, (more literally) ‘I’ll see you when it is the next day’
From parts of speech to the grammar
(So what does this mean? Should we claim that Chickasaw onnakma is not functionally an adverb? This is an interesting and ultimately extremely important question that I won’t attempt to answer here, one that may shed light on the whole notion of the category of “adverb”, but which seems to me to be outside the realm of basic grammatical description.) The first two lists in (1) do not appear in the Chickasaw and Wolof dictionaries (though there is discussion that could lead one to extract them, pretty much, especially for Wolof); I included the above list in the SLQZ dictionary mainly because I felt that that language’s unusually large number of syntactically and morphologically distinct parts of speech in this language was worthy of comment for an American indigenous language. Each of the lists in (1) is augmented with the dread word “particles”. What is a particle, exactly? What are the grammatical characteristics of particles? “Particle” is a cop-out term, but a useful one, which I’ll return to again in Section 4. The number of separate parts of speech one recognizes (whether or not we count particles!) depends on a number of factors. My analysis in (1), for any target word, is based on consideration of the criteria listed in (3): (3) Some criteria for determining part of speech – syntactic factors: what other types of words a target word can combine with, in what types of phrases; – morphological factors: what affixes the target word is used with; – semantic factors: what the target word means, and how it contributes to the meaning of the phrases it is used in.
Each of these factors is important. English adjectives, for example, have both predicative and attributive syntactic uses and can be used with the comparative and superlative suffixes -er and -est. (4)
Characteristic syntactic and morphological examples of English adjectives The house is big. (predicate use) It’s a big house. (attributive use) bigger, biggest (comparative and superlative forms)
Semantically, adjectives generally describe some quality of a noun (such as the bigness of a house). But this semantic characteristic of adjectives, which may be all some people retain from long-ago grade school grammar classes, is not a sufficient criterion for identifying adjective status. One of the underdocumented speech varieties for which I make dictionaries is American undergraduate slang (e.g., most recently, Bonds et al. 2001). I’ve found that students asked to identify the part of speech of slang words, even advanced linguistics majors, will often report that sucks and bit in sentences like those in (5) are adjectives:
75
76
Pamela Munro
(5) Examples of college slang from Bonds et al. (2001) I have to work all weekend. That sucks. I saw a horrible movie last night that bit.
True, we can paraphrase sucks and bit with is terrible and was really gross — but the fact that these words are used with the verbal morphology and occur with verbal, rather than adjectival, syntactic distribution proves that they are verbs, not adjectives. Of course this is a trivial piece of analysis for linguists. But it provides a remarkably convincing demonstration when presented to students with no extensive background in — and sometimes an active antipathy toward — grammar. This type of example helps students understand something about the structure of dictionary definitions (which must follow what I call the “substitution principle” in my slang classes) and often shows them that grammar can be a lot more fun than they realized. Now, the slang words suck and bite used in (5) are not prototypical verbs (especially as the concept of “verb” is usually presented in school grammar lessons), since they have nothing to do with actions. Each of the three identifying criteria named in (3), when considered carefully, will identify many subtypes3 (or “exceptions”): for English verbs, for example, this might include verbs with non-prototypical semantics (like stink, a standard example that works like the slang verbs suck and bite), verbs (like bite) with non-prototypical morphology, and verbs that occur in nonprototypical syntactic patterns (like auxiliaries or modals). Here’s another example. My student slang collaborators have identified a special use of nouns as address terms or vocatives. Here are a few examples from Bonds et. al. (2001): (6) baby (addr. used by males to male friends) chief (addr. used by one male to another, especially by one who provides some service) girlfriend (addr. used by a female to a female friend) mang (addr. used to a male) nerd (addr. used affectionately to and by females)
Although many of these words look like familiar standard or slang nouns, according to my most recent class these words are used by fluent slang speakers in the meanings given only in direct address, not for nominal reference. A very similar class of words occurs in SLQZ (Munro and Lopez et al., 1999):
(7) cagwe’t 1. man, buddy, guy, dude (addr. used to a man); 2. man!, dude! (excl.) erre’eh (addr. used to someone who could be addressed with inf., primarily someone younger)
From parts of speech to the grammar
gu’x: sibling’s spouse’s male relative, parent’s sibling’s spouse’s male relative (addr. used by men) maa girlie, little girl, young lady (addr. used to a girl or young woman, almost always younger than the speaker) paa sonny, little boy, young man (addr. used to a male, usually a boy, almost always younger than the speaker)
In terms of part of speech, such forms are probably best analyzed as nouns, and speakers would generally identify them as nouns. But speakers report that they cannot be used, as normal nouns can, as the subject or object of sentences (although the first Zapotec word is also used as an interjection, as shown). Thus, for the dictionary and the grammar it makes sense to have a special label for these words (like the “addr.” used in the shortened dictionary entries in (6–7). Subclassification of lexical data is very important both for grammatical analysis and in terms of making dictionary entries that are maximally informative and helpful, but I do not have space to discuss this important notion further today.
2. Some tricky cases In this section, I’ll describe two pairs of cases where analysis of part of speech is disputed or difficult, with significant consequences for the grammar. In 2.1 I’ll discuss verb/adjective issues and in 2.2 noun/adposition issues, in each case considering data from Zapotec (and Mixtec, in 2.21) vs. Chickasaw. I will return to the general question of the typological relevance of the analyses for which I will argue and their consequences for the grammar in Section 3. 2.1 Verbs or adjectives? In this section, I’ll discuss two cases, one where one class of verbs and the class of adjectives must be distinguished in the grammar and one in which adjectives function as a subclass of verbs. In both cases, a variety of syntactic features distinguish the words in question. 2.11 SLQZ adjectives and Neutral aspect verbs4 Most Zapotec languages5 have two prefixes of the shape n- (alternating with zero for the same meaning), occurring on stative forms that function either as verbs or as adjectives, which have been analyzed as the same morpheme in various grammars. The two n- prefixes may well have a common origin, but the morphological forms they mark are synchronically distinct in the syntax of SLQZ. Crucially, the
77
78
Pamela Munro
two items differ in part of speech — the adjectives are “adjectives”, and the neutral verbs are “verbs” — as shown by a variety of different grammatical features. Zapotec languages indicate “aspect” by verb prefixes, with time and mood reference inferable through context, the use of adverbs, and syntactic construction.6 SLQZ has a rich set of these prefixes, as illustrated by the verb stems in (8).7 (Zapotec verb morphology is complex, and some stems — like the perfective of ‘puts on a shirt’ — are suppletive.) (8)
Habitual: ra’ahcw ‘puts on (a shirt)’ Perfective: gwu’aht ‘put on…’ Irrealis: ga’acw ‘will put on…’ Subjunctive: nya’ahcw ‘(if…) had put on…’ Progressive: caya’ahcw ‘is putting on…’ Definite: za’ahcw ‘will surely put on…’ Neutral: naa’cw ‘is wearing…’
The Neutral prefix n- varies with m- (before some b’s) and (before most consonants) zero. (9) presents a few examples of SLQZ Neutral stems.8 bèe’b is located on § neut. of rbèe’b ‘gets put on; rides’ bèe’cy is wearing (pants, a skirt) § neut. of rbèe’cy ‘puts on (pants, a skirt)’ de’èi’by is wrapped up § neut. of rdeèi’by ‘gets wrapped up’ gui’ìi’dy is stuck in § neut. of rguììi’dy ‘gets stuck in’ nagya’ihsy is asleep § neut. of ra’ihsy ‘sleeps’ nu’bi’izh is sick-looking § neut. of ru’bi’izh ‘gets sad’ nyi’ùu’ is shut, is closed (of a door, book, mouth, eye); is shut up, is shut in, is imprisoned, is in jail; is off, is turned off (of a light or faucet) § neut. of rìu’ ‘gets shut up; gets shut; gets turned off ’ zèèi’by is hanging § neut. of rzèèi’by ‘hangs’ zucàa is in a leaning position; is leaning against § neut. of rzucwàa ‘leans against’ zuu is standing, is located (standing) § neut. of rzuh ‘stands’ (9)
Cognates of the Neutral prefix n-, and the verb forms they mark, are called “Stative” in many Zapotec grammars (e.g. Black (2000: 51) for Quiegolani). Indeed, SLQZ Neutral verbs generally express states (often resultative, usually durative) (10a), and most Neutral verbs cannot express a temporally bounded or punctual event, in contrast to verbs in aspects like the Perfective (10b) or the Habitual (10c). (10) a. Nài’ r-cah gài’ n-yi’ùu’ ru’uh wrraahlly. yesterday hab-ring five neut-close mouth yard ‘The gate was closed yesterday at five o’clock [that’s the way it was then]’
From parts of speech to the grammar
b. Nài’ r-cah gài’ b-ìu (*n-yi’ùu’) ru’uh wrraahlly. yesterday hab-ring five perf-close (neut-close) mouth yard ‘The gate got closed yesterday at five o’clock [someone closed it then]’ c. R-cah gài’ r-ìu’ (*n-yi’ùu’) ru’uh wrraahlly. hab-ring five hab-close (neut-close) mouth yard ‘The gate gets closed at five o’clock [someone does this every day]’
There are problems with the “stative” label, however, since some Neutral verbs don’t seem semantically stative.9 The most clearly non-stative Neutral verb is nnah ‘says’10 (11), which is used to report a punctual event (apparently an achievement in the sense of Dowty (1979)), not a state:11 (11) “A z-a’=a’,” nnah Gye’eihlly nài’. “already def-go=1s” neut.say Mike yesterday “I’m going,” said Mike yesterday’
In contrast, most SLQZ adjectives are undeniably stative semantically. Like Neutral verbs, SLQZ adjectives have a prefix n- (often m- before b) or are unprefixed. SLQZ predicate adjectives are used with or without one of several copulas, as in (12a) and (c) versus (12b), respectively. The copula may itself include the Neutral prefix n-, as in (12c):12 (12) a. b. c.
N-cweeby nàa rraady ‘The radio is new’ adj-new cop radio N-cweeby rraady ‘The radio is new’ adj-new radio N-gàa’ah n-aa’c=ëng ‘It’s green’ adj-green neut-be=3s.prox
As attributive modifiers following nouns, as in (13), some adjectives lose the nprefix: (13) rraady cweeby radio new
‘new radio’
Some native and borrowed adjectives with and without n- are given in (14).13 (The examples, excerpted from our dictionary, give an idea of the complexity of adjective syntax: the adjectives are marked according to whether their predicative use is with a copula {C}, without a copula {P}, or either way {P/C}. Attributive (attr.) forms of n-initial adjectives without the n- prefix are also listed in these entries.) (14) a. Non-loan adjectives that don’t start with n-: me’eu dirty, filthy {P/C}
79
80
Pamela Munro
la’as thin, skinny {P/C} qui’rìa’ah provocative, sexy looking; vain {P/C} xniaa red {P/C} zyuùa’ll tall; long {P/C} b. Non-loan adjectives that do start with n- (or its variant m-): mbiihahz dry {P/C; attr. biihahz} na’ahzhy wet {P/C} ncweeby new {P/C; attr. cweeby} ndaàa’ hot (inan. subj.); (of the weather); hot (of a place) {P/C}; “hot, warm” (culturally out of equilibrium with something “cold” (nahll); of a food or natural phenomenon) {C} nte’u brown; grey {P/C; attr. nte’u, te’u} c. Loan adjectives without n-. arrie’t loose, easy, promiscuous (of a woman); daring, adventurous (of a man) {P/C} [< Sp. arrecho] biieb smart; alert {P/C} [< Sp. vivo] lo’oc crazy, insane {C} [< Sp loco] rrye’cw rich {P/C} [< Sp. rico] traba’jw hard, difficult (of work; of (life, things in) a location) {C; no attr} [< Sp. trabajo] d, Loan adjectives with n-:14 nlo’oc hard to talk to; mean (of a person); apt to charge, badly behaved (of a bull) {P} [< n- + Sp. loco nsua’ll blue {C} [< n- + Sp. azul] nto’onn stupid, foolish {P/C; attr. to’onn, nto’onn}; mean-spirited, not nice {P/C; attr. nto’onn} [< n- + Sp. tonto]
Thus, Neutral verbs and adjectives both have generally stative meanings and are marked with segmentable n- prefixes or are unprefixed. Moreover, all Neutral verbs and many adjectives can be related to inchoative verbs. (8) illustrated the relationship of Neutral n-aa’cw ‘is wearing (a shirt)’ to inchoative r-a’ahcw ‘puts on (a shirt)’ within a complete aspectual paradigm; additional Neutral/inchoative Habitual verb pairs were given in (9). (15) shows the relationship of some adjectives to inchoative Habitual verbs: (15)
Adjective la’as ‘thin, skinny’, r-la’ahs ‘loses weight, gets thin’ Adjective m-biihahz ‘dry’, r-bihahz ‘gets dry’ Adjective n-a’ahzhy ‘wet’; r-a’ahzhy ‘gets wet’ Adjective n-cweeby ‘new’, r-cweeby ‘gets renewed’ Adjective n-daàa’ ‘hot’, r-daàa’ ‘gets hot’
From parts of speech to the grammar
Perhaps on the basis of such evidence, Black (2000) equates the n- on Quiegolani adjectives with the Stative (Neutral) prefix. Her excellent grammar declares that “the Stative aspect marker is usually found on the adjective in these [predicative] constructions” (p. 52).15 In SLQZ, however, the apparent similarity of Neutral verbs and adjectives is misleading. While Neutral/inchoative verb pairs like those in (9) represent parts of a paradigm of related verbal forms, adjective/inchoative verb pairs like those in (15) occur only irregularly. Moreover, Neutral and adjective forms themselves are syntactically distinct in a variety of ways, following from the most basic difference between them, which relates to part of speech: Neutral forms are verbs, while adjectives are not; Neutral forms behave like verbs, and adjectives behave like adjectives. – SLQZ adjectives (even those that are normally used without a copula in sentences like (12b)) require a copula (in the Irrealis) to express a simple future reference, as in (16): (16) N-daàa’ g-a’c nazh:ih adj-hot irr-be today
‘Today will be hot’
In contrast, Neutral verbs cannot cooccur with a copula (17a). A simple future meaning must be expressed with a different aspect from the Neutral, such as the Irrealis (17b): (17) a. * Li’ìi’by g-a’c te’ihby sinnydoor làa’iny bùunny… neut:get.tied irr-be one sash in person (for ‘A sash will be tied around one’s stomach…’) b. Y-liìi’by te’ihby sinnydoor làa’iny bùunny… irr-get.tied one sash in person ‘A sash will be tied around one’s stomach…’
– In complex constructions that require the use of the Irrealis (for instance, in complements of r-càa’z ‘wants’),16 Adjectives appear with an Irrealis copula, as in (18), while Neutral verbs cannot be used (19), but must be replaced by Irrealis verb forms: (18) Que’ity r-càa’z=eng g-a’c=ëng n-to’onn. not hab-want=3s.prox irr-be=3s.prox adj-stupid ‘He doesn’t want to be stupid’ (19) R-càa’z=ëng g-aa’cw=ëng (*n-aa’cw=eng) cotoony. neut-want=3s.prox irr-put.on=3s.prox (neut-put.on=3s.prox shirt ‘He wants to be wearing a shirt’ (literally, it seems, ‘He wants to put on a shirt’)
81
82
Pamela Munro
– A great many adjectives can also be used as adverbs (modifying verbs or occasionally sentences): (20), for example, shows an adverbial use of the adjective n-dàa’ ‘loose, slack’ (modifying a Neutral verb!). However, no Neutral verbs can be used as adverbs. (20) Ua’s n-dàa’ n-ài’y=ëng zùu’dy=ëng. very adj-loose neut-put.on.skirt=3s.prox corte=3s.prox ‘She has her corte [traditional wrap-around skirt] put on loosely’
– Neutral verbs cannot be used as postnominal attributive modifiers,17 the way adjectives can (as in (13), for example), but must occur in relative clauses,18 as in (21). (21) a. nchààa’ nih zuubih dish rel neut.crack b. (* …nchàa’ zuubih) * …dish neut.crack
‘dish that is cracked’ (for ‘cracked dish’)
Clearly, both the noun plus adjective phrase in (13) and the noun plus Neutral relative clause phrase in (21a) are constituents: when they occur in first position in a sentence, for example, second-position clitics like the epistemic modal clitic =zhyi’ must follow both types of phrases without interrupting them, as in (22).19 (No other position of the clitic is acceptable.) (22) a. b.
Rraady cweeby=zhy=ëng ‘It must be a new radio’ radio new=mod=3s.prox Nchààa’ nih zuubih=zhy=ëng ‘It must be a dish that’s cracked’ dish rel neut.crack=mod=3s.prox
However, there are several syntactic differences between these two types of complex noun phrase that show that (as indeed we would expect) the relative clause containing the Neutral verb is not as closely in constituency with the noun it modifies as the adjective is.20 – Noun plus adjective versus noun plus Neutral relative structures differ in terms of their possessed forms. Possessed nouns in SLQZ may begin with a prefix x:(x:a- before consonant clusters) and are followed by a possessor noun or clitic pronoun,21 as in (23): (23) a. b.
x:-nna’an=a’ poss-mother=1s x:a-rraady=a’ poss-radio=1s
‘my mother’ (nnàaan ‘mother’) ‘my radio’
From parts of speech to the grammar
The possessor must follow a noun plus adjective phrase, as in (24a); it may not appear directly following the noun: (24) a. x:a-rraady cweeby=a’ poss-radio new=1s b. * x:a-rraady=a’ (n-)cweeby poss-radio=1s (adj-)new
‘my new radio’ (for ‘my new radio’)
With a noun modified by a Neutral verb in a relative clause, however, the speaker has two options, as in (25): (25) a. x:a-nchàa’ nih zuubì=a’ ‘my cracked dish (my dish that is cracked)’ poss-dish rel neut.crack=1s (25) b. x:a-nchaà=a’ nih zuubih ‘my cracked dish (my dish that is cracked)’ poss-dish=1s rel neut.crack
These examples show that although both phrases may be treated as constituents (just as was indicated by clitic placement) there is a potential syntactic break between the noun and relative clause modifier that is not present in the noun plus adjective sequence. – Noun plus adjective versus noun plus Neutral relative structures differ more dramatically in their diminutive forms.22 SLQZ has a productive diminutive clitic =e’eh (=(‘)iny after a vowel) that is added to a noun to indicate that that noun’s referent is small or cute, as in (26). (26) a. b.
rraady=e’eh radio=dim nchàa’=iny dish=dim
‘cute/little radio’ ‘cute/little dish’
The diminutive clitic follows a noun plus adjective, as in (27a). An adjective (whether or not it begins with n-) cannot follow a diminutive noun within a noun phrase (27b): (27) a. rraady cweeby=e’eh radio new=dim b. * rraady=e’eh (n-)cweeby radio=dim (adj-)new
‘cute/little new radio’ (for ‘cute/little new radio’)
However, the diminutive clitic cannot follow a noun plus modifying relative clause, as in (28a). The diminutive clitic must appear on the noun, followed in turn by the relative clause, as in (28b):
83
84
Pamela Munro
(28) a. * nchààa’ nih zuubih=’iny (for ‘cute/little dish that is cracked’) dish rel neut.crack=dim b. nchàa’=iny nih zuubih ‘cute/little cracked dish’ dish=dim rel neut.crack
– Adjectives and Neutral verbs (with or without an n- prefix) may cooccur in the same clause — with a predicate adjective and a Neutral copula, as in (12c), or an adjective functioning as an adverbial modifier of a Neutral predicate, as in (20). The cooccurrence facts indicate that these two types of word have a different function in the synchronic grammar. – Finally, as noted earlier, adjectives are adjectives and Neutral verbs are verbs. One feature of this difference in lexical category is that the class of adjectives is constantly being augmented by Spanish (and possibly other) borrowings, while the class of verbs, and thus of Neutral verbs, is not being increased in this way, since SLQZ does not borrow verbs directly.23 Thus, while the two SLQZ n- prefixes have some similarities, the syntactic behavior of the forms they mark — Neutral verbs and adjectives — are quite distinct. Thus, adjectives and Neutral verbs have differeant parts of speech within the synchronic grammar of SLQZ.24 2.12 Chickasaw adjectival verbs In Chickasaw, unlike SLQZ, there is no separate class of adjectives; “adjectival” concepts are expressed adjectival verbs that function just like other verbs. In this section, I’ll justify this statement, showing that Chickasaw really has no part of speech “adjective” and that adjectival concepts are expressed by members of the class of verbs, as shown by a variety of morphological and syntactic properties. Chickasaw is a language with an active system of verb agreement. Chickasaw has three classes of agreement markers for first and second person verb arguments, which are presented in Table 1.25 Table 1. Chickasaw Agreement Markers class first person singular second person singular first person plural second person plural
I -li ishiihash-
II sachipohachi-
III (including dative im-) a+mchi+mpo+mhachi+m-
Class I markers are used for most agentive or volitional intransitive subjects and for almost all transitive subjects; class II markers are used for many intransitive subjects (often non-agentive or non-volitional) and for most transitive objects:
From parts of speech to the grammar
Chickasaw thus has a morphologically active agreement system. As the table shows, Class III markers, which are used for dative, benefactive, and various other objects and subjects, are segmentable, since they include the dative prefix im-. Class I and II markers are added directly to bare verb stems, as in (29–32). For example, chompa ‘buy’ (29) is a transitive verb that takes a class I subject and a noun object (specified or not); malli ‘jump’ (30) is an active intransitive verb that takes a class I subject; and tikahbi ‘be tired’ (31) is a non-active intransitive verb — a canonical “adjectival” verb — that takes a class II subject. Finally, halili ‘touch’ (12) is a transitive verb that takes a class I subject and a class II object. (29) chompa chompa-li
‘buy’, ‘he/she buys it/them’, ‘they buy it/them’ ‘I buy it/them’ etc. – class I subject, noun object
(30) malli malli-li
‘jump’, ‘he/she/it jumps’, ‘they jump’ ‘I jump’, etc. – class I subject, intransitive
(31) tikahbi sa-tikahbi
‘be tired’, ‘he/she is tired’, ‘they are tired’ ‘I am tired’, etc. – class II subject, intransitive
(32)
‘touch’, ‘he/she/it touches it/him/her/them’ ‘I touch it/him/her/them’ ‘he/she/it touches me’, ‘they touch me’ ‘I touch you’, etc. – class I subject, class II object
halili halili-li sa-halili chi-halili-li
Such inflected verb words can all be used as complete sentences.26 As the first example lines in (29–32) show, verbs without first or second person affixes can be interpreted as having third person arguments (there are no third person markers in the chart in Table 2). Markers from class III (glossed here as units combined with the dative prefix)27 replace the dative prefix im- on a verb, as illustrated in (33–34).28 The intransitive stative verb in-takho’bi ‘be lazy’ (33), for example — another prototypical “adjectival” verb — takes a dative subject and, thus, class III marking when its subject is non-third person. The transitive verb i-hollo ‘love’ in (34) takes a class I subject and a class III object. (33) in-takho‘bi ‘be lazy’, ‘he/she/it is lazy’, ‘they are lazy’ an-takho‘bi ‘I am lazy’, etc. — class III subject, intransitive (34)
i-hollo i-hollo-li a-hollo chi-hollo-li
‘love’, ‘he/she loves him/her/them’, ‘they love him/her/them’ ‘I love him/her/them’ ‘he/she loves me’, ‘they love me’ ‘I love you’, etc. — class I subject, class III object
A crucial feature of Chickasaw agreement is that it is not fully predictable, either syntactically or semantically (Munro and Gordon 1982). While the verbs
85
86
Pamela Munro
exemplified in (29–34) and indeed the overwhelming majority of Chickasaw verbs follow the basic semantic principles outlined here (or the slightly different ones discussed in studies like Payne 1981), a great many other verbs do not. There is no reason (other than convention) why the stative verb toklo ‘be two in number’, for instance, should take class I “active” marking, nor why the verb issikopa ‘act mean, be mean’, which can have either a volitional or a nonvolitional interpretation, should consistently take class II “non-active”/”non-volitional” marking. Although most Class II markers index intransitive subjects or transitive objects, as shown above, there are transitive verbs, such as banna ‘want’ or nokfónkha ‘remember’, whose subjects are Class II. Dative arguments are similarly problematical. Class III prefixes may index canonical datives or benefactives, as in in-taloowa ‘sing to, sing for’, and some class III arguments, like the object of i-hollo ‘love’, could be considered semantic experiencers, but the subject of in-takho’bi ‘be lazy’ does not seem like an experiencer. While semantic principles categorize the basic system, the agreement features of many verbs must be lexically marked. In particular, although the majority of stative verbs with meanings comparable to those of English adjectives take Class II marking for non-third person subjects, others are marked with Class I or Class III marking. Thus, there is no particular reason why píhhihcha ‘be very short’ should take Class I marking and tilofa’si ‘be short’ Class II marking, or why yáppalli ‘be lackadaisical’ take Class I marking and lhakcha ‘be lethargic’ take Class II marking, or on the other hand for why tikahbi ‘be tired’ should take Class II marking and im-aalhlhi ‘be exhausted’ take Class III marking, or why abika ‘be sick’ take Class II marking and in-chokmishto ‘be healthy’ take Class III marking.29 (Another plug for dictionary work: it’s hard to really appreciate the range of agreement possibilities in a language like Chickasaw without extensive lexical study.) Although Chickasaw adjectival verbs can mark agreement with their subjects in any of three ways, in other ways these verbs work just like other verbs. Below I briefly review a number of features of Chickasaw verbs, including adjectival verbs, exemplified by a selection of both semantically regular verbs (which follow the agreement class marking generalizations just discussed) and other verbs. – Nominal subjects of Chickasaw verbs of all classes are marked with nominative case (with the suffix -at), as in (35a). This is true for subjects of adjectival verbs like yáppalli ‘be lackadaisical’, tikahbi ‘be tired’, and in-takho’bi ‘be lazy’, just as for subjects of other types of verbs of all three classes are, as shown in (35b). (35) a. Hattak-at malli. man-nm jump
‘The man jumps’ (class I subject)
From parts of speech to the grammar
b.
Hattak-at yáppalli. Hattak-at toklo. Hattak-at chompa. Hattak-at halili. Hattak-at tikahbi. Hattak-at banna. Hattak-at in-takho’bi. Hattak-at i-hollo.
‘The man is lackadaisical’ (class I subject) ‘There are two men’ (class I subject) ‘The man buys it’ (class I subject) ‘The man touches it’ (class I subject) ‘The man is tired’ (class II subject) ‘The man wants it’ (class II subject) ‘The man is lazy’ (class III subject) ‘The man loves it’ (class III subject)
– As suggested by the translations in examples (29–34), there is no distinction between third person singular and plural in the Chickasaw pronominal inflectional system. However, there is a third person plural subject prefix hoo- that can optionally appear on verbs of any inflectional class with third person plural subjects, including adjectival verbs, as in (36). (36)
Hoo-malli. Hoo-yáppalli. Hoo-chompa. Hoo-halili. Hoo-tikahbi. Hoo-banna. Hoo-in-takho’bi. Hoo-i-hollo.
‘They jump’ (class I subject) ‘They are lackadaisical’ (class I subject) ‘They buy it’ (class I subject) ‘They touch it’ (class I subject) ‘They are tired’ (class II subject) ‘They want it’ (class II subject) ‘They are lazy’ (class III subject) ‘They love it’ (class III subject)
(There are a variety of other syntactic properties shared by Chickasaw subjects of all agreement classes, including adjectival verbs, as discussed in Munro (1999).) – Adjectival verbs use the same morphology for tense-aspect, modality, and questions as other verbs. (36a–b) illustrate the use of the remote past suffix ‑ttook, (37a–b) the incompletive modal suffix ‑a’chi, and (38a–b) the question suffix ‑taa: (36) a. b.
Malli-ttook. ‘He jumped (long ago)’ (class I subject) jump-rem Yáppalli-ttook. ‘He was lackadaisical (long ago)’ (class I subject) Toklo-ttook. ‘There were two of them (long ago)’ (class I subject) Chompa-ttook. ‘He bought it (long ago)’ (class I subject) Halili-ttook. ‘He touched it (long ago)’ (class I subject) Tikahbi-ttook. ‘He was tired (long ago)’ (class II subject) Banna-ttook. ‘He wanted it (long ago)’ (class II subject) In-takho’bi-ttook. ‘He was lazy (long ago)’ (class III subject) I-hollo-ttook. ‘He loved it (long ago)’ (class III subject)
87
88
Pamela Munro
(37) a. b.
Mall-a’chi. jump-inc Yáppall-a’chi. Tokl-a’chi. Chomp-a’chi. Halil-a’chi. Tikahb-a’chi. Bann-a’chi. In-takho’b-’achi. I-holl-a’chi.
‘He’s gonna jump’ (class I subject)
(38) a. b.
Malli-taa? jump-Q Yáppalli-taa? Toklo-taa? Chompa-taa? Hali-taa? Tikahbi-taa? Banna-taa? In-takho’bi-taa? I-hollo-taa?
‘Does he jump?’ (class I subject)
‘He’s gonna be lackadaisical’ (class I subject) ‘There will be two of them’ (class I subject) ‘He’s gonna buy it’ (class I subject) ‘He’s gonna touch it’ (class I subject) ‘He’s gonna be tired’ (class II subject) ‘He’s gonna want it’ (class II subject) ‘He’s gonna be lazy’ (class III subject) ‘He’s gonna love it’ (class III subject)
‘Is he lackadaisical?’ (class I subject) ‘Are there two of them?’ (class I subject) ‘Does he buy it?’ (class I subject) ‘Does he touch it?’ (class I subject) ‘Is he tired?’ (class II subject) ‘Does he want it?’ (class II subject) ‘Is he lazy?’ (class III subject) ‘Does he love it?’ (class III subject)
In many languages, predicate adjectives are used with a copula. A copula is used in Chickasaw with predicate nouns (in sentences with non-third person subjects or marked tense-aspect or modality), as in (39), but cannot be used with stative verbs (40): (39) a. b.
Alikchi’ sa-ya. ‘I’m a doctor’ doctor 1sII-be Hattak-at alikchi’ a-ttook. ‘The man was a doctor (long ago)’ man-nm doctor be-rem
(40) a. * Yáppalli sa-ya. ‘I am lackadaisical’ (for Yápalli-li.) be.lackadaisical 1sII-be lackadaisical-1sI b. * Tikahbi sa-ya. ‘I am tired’ (for Sa-tikahbi.) be.tired 1sII-be 1sII-tired c. * In-/An-takho’bi sa-ya. ‘I am lazy’ (for An-takho’bi.) dat-/1sIII.dat-be.lazy 1sII-be 1sIII.dat-lazy
Chickasaw and languages like it have SOV basic word order but do not follow the “harmonic” pattern characteristic of “typical” head-final languages, since attributive modifiers follow the noun, as in (41):
(41) a. b. c.
From parts of speech to the grammar
Hattak toklo yamm-a pís-li-tok. man be.two that-acc see-1sI-pt Hattak tikahbi yamm-a pís-li-tok. man be.tired that-acc see-1sI-pt Hattak in-takho’bi yamm-a pís-li-tok. man dat-be.lazy that-acc see-1sI-pt
‘I saw those two men.’ (class I subject) ‘I saw that tired man.’ (class II subject) ‘I saw that lazy man.’ (class III subject)
This order is a reduced form of the typical relative clause order, because “adjectives” in Chickasaw are verbs, and must appear (just like SLQZ Neutral verbs, and other more prototypical Chickasaw verbs) in relative clauses30 (Munro 1985), as in (42): (42) Hattak malli yamm-a pís-li-tok. man jump that-acc see-1sI-pt
‘I saw that man who jumps’
Thus, the Chickasaw words that translate English adjectives are indeed verbs, and Chickasaw has no part of speech “adjective”. 2.2 Nouns or adpositions? In this section I discuss two cases where similar relational words have been analyzed variably as nouns or as pre- or postpositions. Once again, the correct analysis depends on the syntactic behavior of the words in question in each language. 2.21 Zapotec and Mixtec body part prepositions31 The use of body part words in the expression of location, as in the sentences from San Lucas Quiaviní Zapotec in (43–46), is an areal feature of Mesoamerica (Stross 2002). (43) Bèe’cw zùub ni’ih me’es. ‘The dog is sitting under the table.’ dog neut.sit foot/under table (44) Gw-eh dehts wrraahlly. perf-go back/behind corral
‘Go on the other side of the fence.’
(45) B-ìe’ny=ënn gaan te’ihby rràady loh rrie’f. perf-do=1p winning one radio face/in raffle ‘We won a radio in the raffle.’ (46) M-nnàa’az gue’ehcy me’es ch-oo’nn-nèe=n=ëng perf-grab head
la’n-yu’uh.
table irr-go.1p-with=1p=3s.prox stomach/into-house
‘Grab the head of the table and we’ll take it into the house.’
89
90
Pamela Munro
The words ni’ih ‘foot, lower leg’, dehts ‘back’, lohoh ‘face’, làa’iny ‘stomach’, and gue’ehcy ‘head’ boldfaced in these sentences are basic body part terms. In their basic use, such words are always followed by a possessor: thus, ni’ih me’es in (43) could be translated as ‘the foot of the table’ or ‘the table’s foot’. In these sentences, however, the meaning of the first four words is more like that of English prepositions, and the words following them are more like prepositional objects than like possessors. This is indeed how Lopez and I classified these uses in our dictionary (1999), in contrast with body part senses such as that of gue’ehcy ‘head’ in (46), where gue’ehcy me’es indeed does mean ‘the head of the table’ (though in other sentences gue’ehcy can be used to mean ‘on’ or ‘at the top of ’), or the literal use of ni’ih ‘foot’ as the subject of sentence (47). (47) R-ahc ni’=a’. 32 hab-hurt foot=1s
‘My foot hurts’
Are the relational uses of the body part words in (43–46) in fact prepositions, or are they something else? The syntax and semantics of such expressions, and thus how they should be classified in terms of part of speech, is the topic of an important Masters thesis by Brook D. Lillehaugen (2003). The metonymic extension of the meanings of body part words in Otomanguean languages has received some attention in the literature (for example, in the important studies by MacLaury (1989) and Hollenbach (1995)), and Lillehaugen’s study of these words in SLQZ and closely related languages (2001, in preparation, 2003) has considerably enriched our coverage of their uses in the current version of the dictionary. Sometimes the new meanings are transparent (e.g. ‘back’ > ‘behind’), sometimes less so, as the examples here show. Hollenbach’s and Lillehaugen’s work raises the important question of the degree to which the original metaphor involved in such developments is active and always accessible by the speaker. There is some difference of opinion on the syntactic status of body part words in sentences like (29–31), however. MacLaury (1989: 120) has claimed that body part locatives in Ayoquesco Zapotec … are not prepositions, because there is no justification for setting them apart from their primary classification as nouns. Unlike English prepositions, they are identical in form to the nouns applied to body organs, their use in syntax is optional, they only add specificity to other locative expressions, they do not complicate syntax, they do not denote direction, and they do not mark grammatical relations as do case markers.
From parts of speech to the grammar
While these statements may be appropriate for Ayoquesco, Lillehaugen (2003) argues, they cannot be applied to SLQZ and other similar Valley Zapotec languages. A sentence like (43),33 as Lillehaugen notes, contains a locative complement that is a prepositional phrase, not a noun phrase. One indication of this is the fact that a simple noun phrase cannot replace ni’ih me’es ‘under the table’ in sentence (43), as shown in (48): (48) * Bèe’cw zùub me’es. dog neut.sit table
Thus, zùub ‘is sitting’ takes as complement not simply any noun phrase referring to a location, but a prepositional phrase: as Lillehaugen argues, it cannot be the case that the use of words like ni’ih here is optional or only for the purpose of adding specificity to an existing locative. The question as to whether the use of body part locatives “complicates syntax” is difficult. Since most languages probably have adpositions of some kind, and since in fact most Zapotec languages have some non-body part prepositions,34 the analysis of words like ni’ih as prepositions does not seem to me to complicate the syntax. Indeed, if these words were not prepositions one would need to develop an ad hoc explanation for the facts in (43) vs. (48). The crucial question is that of what syntactic role the body part phrases play in the sentences in which they appear, and it’s for this reason that deciding on their part of speech is relevant for the grammar. Typically, languages can add locative or other prepositional phrases to a sentence fairly freely regardless of the lexical subcategorization of the verb of that sentence. Thus, in a Zapotec sentence like (49) — (49) Ca-gye’eht=ëng dehts yu’uh. prog-play=3s.prox back/behind house ‘He’s playing in back of the house.’
— the verb r-gye’eht ‘plays’ is intransitive: a sentence with this verb cannot typically take a noun complement. But a locative phrase like dehts yu’uh ‘in back of the house’ can be added freely to such a sentence as an adjunct phrase to tell where the event of playing takes place. There is a clear syntactic difference between a phrase like dehts yu’uh used as a noun phrase (‘the back of the house’) (50a) and as a prepositional phrase (‘in back of the house’) (50b): (50) a. Ca-cwààa’ Jwaany coloory dehts yu’uh. prog-throw Juan paint back house ‘Juan is painting the back of the house.’
91
92
Pamela Munro
b. Ca-cwààa’ Jwaany coloory me’es dehts yu’uh. prog-throw Juan paint table back/behind house ‘Juan is painting the table in back of the house.’35
The SLQZ phrasal verb r-cwààa’ coloory ‘paints’ (literally ‘throws paint at’) is transitive: it must be used with a lexical object. Thus, in (50a) dehts yu’uh functions as a noun phrase expressing the direct object of the verb: dehts is a part of the house, and yu’uh is its possessor. In (50b), however, the object of the verb is me’es ‘table’. Dehts yu’uh is not part of the core argument structure of sentence (50b): rather, the speaker adds this phrase in order to specify the location where the action occurs. Note that if dehts yu’uh were a simple noun phrase rather than a prepositional phrase, that would mean that sentence (50b) included four noun phrases: Jwaany, coloory, me’es, and dehts yu’uh. As far as I know, there are otherwise no sentences in SLQZ consisting of a verb and four separate noun phrases. The analysis of dehts yu’uh as a noun phrase vs. a prepositional phrase thus has consequences for the description of possible sentence structures in this language. I have not succeeded in finding a case of actual structural ambiguity resulting from the two interpretations of body part locative phrases in SLQZ, but Lillehaugen and I recorded the following one in the closely related Zapotec language of Tlacolula de Matamoros:36 (51) Ca-tìi’a’=na làa’iny ydòòo’. prog-paint=3s stomach/inside church ‘He’s painting the inside of the church.’ / ‘He’s painting inside the church.’
Here, as Lillehaugen (2001) notes, ambiguity arises according to whether làa’iny ydòòo’ is a noun phrase (direct object) or an adjunct prepositional phrase. The two interpretations of (51) and the contrast between (50a–b) show that if the body part phrases cannot be interpreted as prepositional the syntax will certainly have to be complicated in terms of specifying prepositional phrase-like structures for certain noun phrases. It’s true, as MacLaury suggests, that body part prepositions prototypically encode locational rather than directional meanings. However, sentences like (44) and (46) above contain directional rather than locational uses of body part prepositional phrases, so it is not the case that such uses are impossible. Moreover, SLQZ body part locatives also are used to express nonspatial relational concepts. The question of whether prepositions in any language can “mark grammatical relations as do case markers” is beyond the scope of this paper, but certainly many uses of SLQZ lohoh ‘face’, in particular, are similar to non-locative meanings of dative and other case markers in many languages, in sentences such as the following:
From parts of speech to the grammar
(52) Loh Jwaany b-zì=a’=ih. ‘I bought it from Juan.’ face/from Juan perf-buy=1s=3s.prox (53) B-zhùu’azh=a’ gueht loh bèe’cw. ‘I tore up the tortilla for the dog.’ perf-tear=1s tortilla face/for dog (54) Zyuùa’ll=ru’ Rrodriiegw loh Lia Oliieb tall=more Rodrigo face/than Ms. Olivia ‘Rodrigo is taller than Olivia.’
As Hollenbach (1995: 177) notes, “nouns prototypically refer to things, while prepositions express relations”. Examples like (52–54) show that a “referential shift” has occurred with the SLQZ Valley Zapotec body part words, with the result that in this relational use they are no longer nouns. The classification of the Valley Zapotec body part locative and relational words as prepositions is in line with the analysis in Black’s recent grammar of Quiegolani Zapotec: “these prepositions are usually body part terms” and “the preposition lo ‘face’ is required with indirect objects” (2000: 46). Pickett and Black (1998: 84–86), while identifying Isthmus Zapotec body part words in sentences comparable to those discussed here as prepositions, nonetheless refer to non-body part words like ne ‘con’ (‘with’) or pur ‘por’ (‘because of ’; from Spanish) “preposiciones verdaderas” (“true prepositions”).37 The notion that the body part locative and relational words are not “true prepositions” probably reflects an idea that the “preposition” classification is inappropriate for words that have such an obvious nominal use and source, recalling MacLaury’s statement that “unlike English prepositions, they are identical in form to the nouns applied to body organs”. But many languages have cases of homophony (and indeed semantic and historical relationship) between words of different lexical categories. The English word down, for example, can be used not only as a preposition (as in down the street), but also as a verb (He’s going to down the beer), a noun (fourth down), and an adverb (The plane went down), all of which are related; this does not interfere with classifying their parts of speech differently. The (avian) body part word down is not related to the locative/directional down set, but (as Lillehaugen suggests) there are certainly English prepositions, such as behind and inside, that have the same form as body part words. Another group of Otomanguean languages in which body part words are used in the expression of location and other relations is the Mixtec family. Differences in the linguistic analysis of body part locatives can be seen in two descriptions of the Lowland Mixtec language of Silacayoapan, for example.38 Like SLQZ and Tlacolula Zapotec, Silacayoapan Mixtec has body part locatives, as in (55–57) and other relational uses of body part words, as in (58):39
93
94
Pamela Munro
(55) cuahan da jata yúcu went he back/behind mountain
‘He went behind the mountain.’
(56) Quíji rí tixi yíto sleep it stomach/under tree
‘It (an animal) is sleeping under the tree.’
(57) íin.coo da nuu ita sitting he face/on grass
‘He is sitting on the grass.’
(58) ndítúhún nde saha ñūū nde ‘We are talking about our town.’ con.discuss we.ex foot/about town our.ex
(As Shields (1988: 318) notes, “locative adjuncts are often found in sentences that are metaphorical in nature; in such sentences they do not necessarily refer to a spatial entity”. Lillehaugen would (I assume) question the degree to which active metaphor was actually involved in the uses seen in a sentence like (58).) In their nontechnical Gramática Popular, North and Shields (1986 [1976]: 35–36) identify the boldfaced words in the examples above as prepositions, but they are referred to as “adverbial noun phrases” in the more technical description by Shields (1988: 369–371, 404). “Adverbial noun phrases” seems like a more appropriate description than (simply) “noun phrases” (since adverbs, like prepositional phrases, typically may be used as adjuncts more freely than ordinary noun phrases), and a careful grammatical description will certainly differentiate these uses,40 since they have important consequences for sentence structure.41 2.22 Chickasaw applicative marking and relational nouns42 According to the online SIL Glossary of Linguistic Terms (Loos et al. 1999), “An adposition is a cover term for prepositions and postpositions. It is a member of a closed set of items that occur before or after a complement composed of a noun phrase, noun, pronoun, or clause that functions as a noun phrase, and form a single structure with the complement to express its grammatical and semantic relation to another unit within a clause”. This definition supports identifying the locational and relational uses of the Zapotec and Mixtec body part words discussed in Section 2.2 as true prepositions. However, I would argue that the definition does not go far enough, since the crucial feature of the Otomanguean prepositional phrases that confirms that they are not, in fact, noun phrases is the fact that they can be freely added to sentences that are otherwise structurally complete (perhaps this is actually what the Glossary means by “expressing a grammatical relation to another unit within the clause”). Let’s consider a case where there are “relational nouns” whose meanings are very similar to that of the Otomanguean body part prepositions, but whose phrases function not as adjuncts, but as complements. In contrast to English, Zapotec,
From parts of speech to the grammar
or Mixtec, Chickasaw has no prepositions, no postpositions, and no oblique case markers. All nominals that would be the objects of adpositions in more typical languages must be licensed in a Chickasaw sentence by applicative affixes on the verb, appearing as arguments rather than syntactic obliques. We saw in (35) above that the subjects of Chickasaw sentences are marked with the nominative suffix -at. Example (59a) shows a Chickasaw object noun marked with the accusative suffix -a. Object nouns may also be unmarked, as in (59b). (An unmarked object noun must appear immediately before the verb; cf. Munro (1999).) (59) a. Ihoo-at bala’-a chompa. woman-nom beans-acc buy
‘The woman buys beans.’
(59) b. Ihoo-at bala’ chompa. woman-nom beans buy
‘The woman buys beans.’
Chickasaw is a language with very strict lexical transitivity. A transitive verb like chompa ‘buy’ always takes exactly two arguments, a subject and an object (though they need not appear overtly, as in (60)); no additional nominals can be added to any sentences containing the simple verb chompa. (60) Chompa. buy
‘She/he buys it/them; They buy it/them’
Thus, a locative, comitative, dative/benefactive, or other semantic oblique can appear in a sentence like those in (61) only if the verb includes the appropriate applicative prefix: (61) a. b. c.
Ihoo-at Albertsons-a bala’ aa-chompa. woman-nom Albertsons-acc beans loc-buy ‘The woman buys beans at Albertsons.’ Ihoo-at i-hattak-a43 bala’ ibaa-chompa. woman-nom dat-man-acc beans cmt-buy ‘The woman buys beans with her husband.’ Ihoo-at chipot-a bala’ in-chompa. woman-nom child-acc beans dat-buy ‘The woman buys beans for the child’
Each of the sentences in (61) has three arguments, the original subcategorized subject ‘woman’ and object ‘beans’, plus an additional argument whose appearance is licensed by the (boldfaced) applicative prefix on the verb. The added argument is generally more salient in the sentence than the original object (since the speaker feels it’s worth adding), and typically appears immediately after the subject, though
95
96
Pamela Munro
numerous other word orders are possible. The applicative prefix thus functions to increase the valence of the verb to which it is added — thus, chompa is a two-argument verb, but aa-chompa, ibaa-chompa, and in-chompa are three-argument verbs. Semantic obliques can be added in a similar way to original intransitive verbs like ‘jump’ and ‘be tired’, as in (62) and (63): (62) a. b.
Ihoo-at kasbi-a aa-malli. ‘The woman jumps in the yard’ woman-nom yard-acc loc-jump Ihoo-at chipot-a i-malli. ‘The woman jumps for the child’ woman-nom child-acc dat-jump
(63) Ihoo-at i-hattak-a ibaa-tikahbi. woman-nom dat-man-acc cmt-be.tired ‘The woman is tired along with her husband.’
The sentences in (62–63) are transitive, with two arguments each: the original subject, plus the added semantic oblique, which functions just like any other syntactic object. The three Chickasaw applicative prefixes whose use was just exemplified are members of a set of eight applicative markers, seven prefixes (in addition to aalocative, ibaa- comitative, and im- dative, these include a- ‘against’, imaa- ‘from’, okaa- ‘in’, and on- ‘on’) and an instrumental proclitic, isht. The applicative verbs these markers derive work like any other verb as regards agreement and other morphological features: in particular, pronominal prefixes (from Table 1 above) generally precede them on the verb, as shown in (64):44 (64) a. b.
Ihoo i-hattak-a bala’ ish-ibaa-chompa. woman dat-man-acc beans 2sI-cmt-buy ‘You buy beans with the woman’s husband.’ Ihoo-at bala’ sa-baa-chompa. ‘The woman buys beans with me.’ woman-nm beans 1sII-cmt-buy
In cases like (61a–c) the semantic “object” of the added applicative is separated from that applicative by the original object noun ‘beans’, so the applicative and its object do not form a linear constituent. In fact, both nominal objects in any Chickasaw ditransitive sentence might be viewed as syntactically equivalent. The first of a sequence of two Chickasaw objects is normally case-marked accusative with the suffix -a, as in (61), and the second is left unmarked. The order of the nouns (though not the case marking) can be reversed, however. An alternative to (61c), for example, is (65), with the original (patient) object, rather than the applicative (benefactive) object, receiving accusative case marking:
From parts of speech to the grammar
(65) Ihoo-at bala’-a chipota in-chompa. woman-nom beans-acc child dat-buy
In addition to the applicative prefixes illustrated in (61–63), Chickasaw has a class of relational nouns, postposition-like words such as anonka’ ‘inside’, most of which are used to further specify the precise location of locative objects. (These, like the applicative prefixes, are boldfaced below.) Sentence (66) is syntactically entirely parallel to (61a), but with a more specific locative or directional object, aachompa’ anonka’ ‘the inside of the store, the store’s inside’. Aachompa’ anonka’ is casemarked just like any other noun, and functions in (66) as a nominal argument just like Albertsons in (61a) (thus, whether or not the relational noun is present, the verb must have the locative applicative prefix aa-). (66) Ihoo-at aachompa’ anonk-a bala’ aa-chompa. woman-nom store inside-acc beans loc-buy ‘The woman buys beans in(side) the store.’
The applicative verb aa-chompa means ‘buy at’ in both (61a) and (66); thus, a more literal translation for (66) would be something like ‘The woman buys beans at the inside of the store’, with aachompa’ anonka’ filling the same role as the locative object Albertsons in (61a). A relational noun phrase can also be used as a subject in non-locative sentences like (67): (67) Aachompa’ anonk-aat litiha. store inside-nom be.dirty
‘The inside of the store is dirty’
Relational nouns like anonka’ may be analyzed syntactically as inalienably possessed nouns (semantically quite similar to the Zapotec body part words). Others include pakna’ ‘top’, nota’ ‘bottom’, and ashaka’ ‘rear, back’. Crucially, a phrase like aachompa’ anompa’ ‘inside the store’, just like the simpler expression aachompa’ ‘store’, can be used only in a sentence whose verb is subcategorized for a nominal or pronominal locative argument, as in (66),45 or as a subject (67). Because of the structural features of Chickasaw, such locative phrases (like all Chickasaw noun phrases) can only appear as complements, never as adjuncts. A sentence like those in (68–69) is thus impossible: (68) a. * Ihoo-at aachomp-a bala’ chompa. woman-nom store-acc beans buy (for ‘The woman buys beans at the store’) b. * Ihoo-at aachompa’ anonk-a bala’ chompa. woman-nom store inside-acc beans buy (for ‘The woman buys beans in(side) the store.’)
97
98
Pamela Munro
(69) a. * Ihoo-at bala’-a aachompa’ chompa. woman-nom beans-acc store buy (for ‘The woman buys beans at the store.’) b. * Ihoo-at bala’-a aachompa’ anonka’ chompa. woman-nom beans-acc store inside buy (for ‘The woman buys beans in(side) the store.’)
Thus, regardless of word order, neither the simple location noun aachompa’ ‘store’ nor the relational noun phrase aachompa’ anonka’ ‘inside the store’ can be added as an adjunct to a sentence with the simple transitive verb chompa ‘buy’ if that verb does not contain the locative prefix aa-. Nicklas (1979: 103) analyzes the cognates of the Chickasaw relational nouns in the closely related language Choctaw as postpositions, but these words are not postpositions in Chickasaw.46 The function of Chickasaw relational nouns like anonka’ is simply to specify more precisely certain locations expressed with ordinary nouns. There are no uses of these words or any simple noun phrase as adjuncts (non-complements, non-subjects or non-objects). There is no sense in which Chickasaw relational noun phrases (relational nouns plus their following possessor) “express [a] grammatical…relation to another unit within a clause”, in the words of the SIL Glossary. In contrast, MacLaury’s statement that “their use in syntax is optional, they only add specificity to other locative expressions, they do not complicate syntax, they do not denote direction, and they do not mark grammatical relations as do case markers”, although it was not accurate for SLQZ, is completely appropriate for the use of the Chickasaw relational nouns!
3. Parts of speech, typology, and analytical abstraction This type or plan or structural “genius” of the language is something much more fundamental, much more pervasive, than any single feature of it that we can mention, nor can we gain an adequate idea of its nature by a mere recital of the sundry facts that make up the grammar of the language. – Edward Sapir, Language, ch. VI
In Section 2, I examined two pairs of cases in which there might be some doubt about or analytical confusion between parts of speech, considering data from Zapotec (and Mixtec) and Chickasaw. First (in 2.1) I considered two cases involving adjectival notions: – In 2.11 I presented evidence that SLQZ adjectives must be distinguished from a certain class of SLQZ verbs, despite their formal and semantic similarity, on the basis of different syntactic behavior.
From parts of speech to the grammar
– In 2.12 I argued that Chickasaw has no separate class of adjectives, since all adjectival concepts are expressed by verbs, and since these adjectival verbs work similarly to other verbs in the language. Next (in 2.2), I considered adpositional notions. – In 2.21, following Lillehaugen (2001, 2003) and Hollenbach (1995), I showed that the relational use of body part words in various Zapotec languages and Mixtec is sufficiently different from the use of these words as nouns to justify classifying them as (“true”!) prepositions. – In 2.22 I examined the expression of adpositional relations in Chickasaw, showing that that language has no pre- or postpositions, and that members of its class of relational nouns are simply nouns, not postpositions. One interesting consequence of these arguments is to confirm the surprisingly short list of Chickasaw parts of speech given in (1) at the beginning of this paper. Chickasaw simply does not have a class of adjectives or a class of adpositions, although there are obviously ways to express the concepts languages like Zapotec or English convey with adjectives or prepositions. This is an important typological observation, since word order typologies typically pay a lot of attention to the order of noun and adjective and adposition and noun. When asked about these typological features of Chickasaw, I typically fudge and say that Chickasaw has N Adj order, since the elements that express adjectival ideas within a (complex) noun phrase do indeed follow the noun, as in (41). The second point is harder, though, since Chickasaw has no adpositions or adpositionlike elements — the closest thing to adpositions in Chickasaw is not the relational nouns but the applicative markers (since it is these which allow the expression of locative and other adjuncts), which are verb prefixes, not independent words. Most typological classifications would probably say that Chickasaw had N P order — the language is otherwise very strictly head-final, and it is tempting to consider that relational nouns fill the role of postpositions. But I hope I showed in 2.22 that they are not, in fact, adpositional at all — so it is very difficult to classify Chickasaw with respect to this typological characteristic. It is periodically fashionable in syntactic theory to treat bound morphemes as if they were independent words. Within such a paradigm, case markers can be analyzed as adpositions, so that, for example, a language with case suffixes can be considered postpositional. A consequence of this view might be to analyze the Chickasaw applicative markers as adpositions that have migrated to a position before the beginning of the verb of their clause. This kind of incorporation certainly can occur — for example, in Pima, a Uto-Aztecan language of Southern Arizona, postpositions may either appear in
99
100
Pamela Munro
a constituent with a noun or before the verb of their clause (Munro 1989). In the Pima sentences (70a) and (71a), for example, the postposition daam follows its object ’u’us; in (70b) and (71b), daam appears in front of the verb, discontinuous from the object. (70a) is an intransitive sentence. (70b) looks more like a transitive sentence — it has a complex (phrasal?) verb, daam mee, independent noun phrases for subject and non-subject (note that when a noun phrase is not sentence or PP initial, regardless of grammatical function, it appears with the article heg).47 The situation is similar for the transitive sentence (71a) and its variant (71b) in which there are two independent non-subject phrases. (70) a. b.
Kalit ’a-t mee ’u-’us daam. car aux-perf run red-stick on ‘The car ran over the sticks.’ Kalit ’a-t daam mee heg ’u-’us. car aux-perf on run art red-stick
(71) a. b.
Kalit ’a-n-t melc ’u-’us daam. car aux-1s-perf drive red-stick on ‘I drove the car over the sticks’ Kalit ’a-n-t daam melc heg ’u-’us. car aux-1s-perf on drive art red-stick
But the difference between the Chickasaw and Pima cases is considerable. Not only does Pima have both adpositional and incorporated structures, but in the incorporated structure pronominal object markers (the only bound pronominal markers on the Pima verb) remain prefixed to the verb, between it and the incorporated postposition, suggesting that the postposition is a recent addition to the verbal complex. Someone who wished to believe that all languages must have adjectives might well propose that adjectival verbs in Chickasaw contained a “non-pronounced” copula, a “light verb” to serve as the bearer of agreement, tense, and other verbal inflection. Thus, by this view, the reason for the incompatibility of adjectival verbs with an overt copula (as shown in (40)) would be that these sentences already have an (incorporated) copula, and do not need another. The fact that some adjectival verbs can occur with Class I and Class III agreement (40a, 40c), rather than the Class II agreement characteristic of the copula (39c) might pose momentary problems but could probably be handled by a determined analyst. But such abstraction, though it aids the cause of “universal grammar” by suggesting that apparently different structures are more similar than they might at first appear, obscures the important facts that constitute the true “structural ‘genius’” of a language (in the sense Sapir used these words in the quotation at the beginning
From parts of speech to the grammar
of this section). If we say that Chickasaw really has adjectives and postpositions, we miss part of the essential quality of Chickasaw. My suggestion that identifying the actual parts of speech of each language be a critical part of grammar writing does not mean that people who are drawn to abstract syntax cannot show us how all languages can be analyzed with similar rules and underlying structures. But before they do that, I believe, they should tell us about the basic overt characteristics of the words of that language — just as dictionary makers (who as I observed earlier are forced in some ways to be more honest than some syntacticians) must.
4. A linguist’s list of parts of speech Now I will come back to the “eight parts of speech” mentioned by Jane Austen and listed in any traditional grammar of English: nouns, pronouns, verbs, adjectives, adverbs, prepositions, conjunctions, and interjections. Although I am arguing for a traditionally oriented return to part of speech analysis as an important stage in writing a grammar (or dictionary!), I believe that this list of parts of speech, while possibly sufficient for descriptions of familiar European languages, is likely to be inadequate as a starting point for linguistic analysis. The most serious problems arise with the traditional category “adjective”, which, according to careful classifications, includes the following types of words in addition to normal adjectives of the type discussed in Section 2.11:48 – – – –
articles (such as a and the) possessive adjectives (such as my, his, and so on) demonstrative adjectives (such as that and this) quantifiers (such as all, three, and maybe any and which)
In English, all of these items (normally) precede a modified noun, so it is barely possible to justify analyzing them all as “adjectives”. I will not consider possessive adjectives further here, since I do not have extensive data for any language that uses an independent word for this function. I will also assume provisionally that articles and demonstrative adjectives may be analyzed, at least some of the time, as members of a category of determiners, although it seems likely that there might be problems with this view. There is certainly evidence, however, that both quantifiers and this category of determiners are syntactically very different from ordinary adjectives.
101
102
Pamela Munro
In Zapotec, for example, quantifiers precede the nouns they quantify, while adjectives follow the nouns they modify, as in example (72) from SLQZ: (72) tyo’p rraady cweeby two radio new
‘two new radios’
Moreover, in many languages (including English!) quantifiers have special structural properties (such as “floating” to a position before the verb). Determiners similarly may occur on a different side of the noun from adjectives. (73) is an example from Tolkapaya Yavapai, a critically endangered Yuman language of central Arizona, in which demonstratives precede nouns but “adjectives” (actually, stative verbs used as modifiers, much as in Chickasaw) follow them (Munro 1985): (73) vyaa vqi hamany this woman young
‘this young woman’
These features lend support to the view that neither quantifiers nor determiners are simply subtypes of a general category of adjectives — a judgment that will certainly come as no surprise to students of syntactic theory or semantics. And at this point, finally, I come back to the question of particles. “Particle” is, as I said at the beginning of this paper, a linguist’s cop-out. Here is one definition of the term, for English (from the Oxford Companion to the English Language (McArthur, ed., 1992): A word that does not change its form through inflection and does not fit easily into the established system of parts of speech. Among individual words commonly so classed are the negative particle not (and its contraction n’t), the infinitival particle to (to go; to run), the imperative particles do, don’t (Do tell me; Don’t tell me) and let, let’s (Let me see now; Let’s go). There is also a set of adverbial and prepositional particles that combine with verbs to form phrasal verbs (out in look out; up in turn up) and prepositional verbs (at in get at; for in care for). The term pragmatic particle is sometimes used for words that play a role in maintaining discourse and are also known as fillers and discourse markers: oh, ah, well, yes, no, actually, anyway.
The SIL online Glossary (Loos et al. 1999) adds one more important characteristic of particles: a particle “typically has grammatical or pragmatic meaning”. The English examples presented in the Oxford Companion definition do not really illustrate the kinds of difficulties that lead linguists to use the term “particle”. Not and the particles of verb-particle constructions (like up in turn up) could be analyzed as adverbs,49 do and let are certainly verbs, and the discourse markers
From parts of speech to the grammar
listed above can be considered either interjections or adverbs. English does not seem to have a serious problem with particles. But — as just one more problematical example — what about languages like SLQZ that have sentence-final question particles, as in (74)? (74) Ncweeby nàa rraady èee? new cop radio Q
‘Is the radio new?’
Èee is an invariable word that does not change its form and has grammatical or pragmatic meaning (it, or another less common alternant, is required in all yes-no questions in SLQZ). It does not seem to fit into the established system of parts of speech: it is not a noun, not a pronoun, not a verb, not an adjective, not an adverb, not a preposition. The two most likely traditional identifications for a word like èee are interjection and conjunction. Typically, though, interjections are not as fixed in their position as SLQZ èee, and generally they can be used alone, which èee cannot. In some syntactic analyses èee might be identified as a complementizer, which should make it a conjunction. From the time of the earliest performative analyses in the 1960s and 70s, complementizers like that and whether have been taken to be diagnostics of the illocutionary force of an associated clause. In a verb-initial language like Zapotec, we would expect question markers, like other complementizers, to precede their clauses. Lee (2002), for example, explains the unexpected word order in SLQZ sentences like (74) by proposing that the clause questioned with èee (ncweeby nàa rraady, in (74)) is a left-dislocated topic that has moved in front of the question marker. On the other hand, SLQZ is a language with no clear complementizers whose conjunctions — including chih ‘when’, tye’nn ‘because’, and the relativizer nih — are all invariably clause-initial, and thus very different from èee. Thus, until we have a better identification, “particle” seems like a good part of speech label for the question marker èee. Therefore, I would add “particle” to the list of parts of speech that we may have to recognize in at least preliminary grammatical analysis, which thus includes, at least, first Jane Austen’s basic eight:
noun pronoun verb adjective adverb preposition interjection conjunction
103
104
Pamela Munro
and then the following
determiner quantifier particle
Not all languages, as we have seen, will have all these parts of speech. And as I noted at the beginning of this paper, subdividing some of these (particularly verbs!) according to syntactic and other properties is an important task for dictionary and grammar making. And clearly if we can find a way to do without the dread class of particles, that is a desideratum. But trying our best to find out the part of speech of every word — and the morphological and syntactic arguments in favor of this classification — will help us understand language better and make it easier for us to describe in our grammars how to “combine them with ease”, as Jane Austen’s James did.
Notes * I am grateful to many people whose contributions have helped to shape the ideas I present here. The ones I must mention most prominently are my collaborators on the dictionary/grammar study projects I discuss in detail, Catherine Willmond for Chickasaw (Munro and Willmond 1994) and Felipe H. Lopez for San Lucas Quiaviní Zapotec (Munro and Lopez, et al., 1999), as well as the other speakers of these languages whose input is acknowledged in the dictionaries named; I’m also grateful to Dieynaba Gaye for Wolof (Munro and Gaye 1997), my undergraduate collaborators in the study of UCLA slang (most recently in Bonds et al. 2001), Virgil Lewis and the late Etheleen Rosero for Pima; the late Molly S. Fasthorse for Tolkapaya Yavapai (Munro and Fasthorse in preparation), and to Olivia Martínez, Rodrigo Garcia, and Michael Galant for input on Zapotec. In addition, I thank Larry Gorbet and Brook Lillehaugen, as well as Cheryl Black and other members of the symposium audience, for very helpful discussion. 1. The more experience I have, the more I respect certain traditional aspects of dictionary makeup and format that I might have scorned earlier. Most of these really are included for a reason! 2. Chickasaw data is presented in the orthography of Munro and Willmond (1994). Note that nasalized vowels are underlined. In general I will not comment on rule-governed morphophonemic variation. The abbreviations used in my glosses include acc : accusative, addr : term of address, attr : attributive, ben: benefactive, C : used predicatively with a copula, cmt = comitative, com : completive, con : continuative, cop : copula, dat : dative, def : Definite, dim : diminutive, dist = distal, ds : different subject, ex : exclusive, excl : exclamation (i.e., interjection, in Bonds et al. 2001), hab : Habitual, inc : incompletive, inf : informal, irr : Irrealis, loc : locative, mod : modal, neut. : Neutral, nm : nominative, P : used predicatively without a copula, perf : Perfective, poss : possessive, prox : proximate, pt = past/perfective, Q : question marker, red: reduplication rel : relative pronoun, rem : remote past, Sp : Spanish. The Chickasaw pronomi-
From parts of speech to the grammar
nal agreement classes are identified with I, II; person and number are indicated with 1, 2, 3; s, p. = marks a clitic boundary and a period separates parts of a complex gloss. In each of the languages I discuss here, morphophonological changes occur in many words. These are certainly a concern of the grammar, but I will not address them here or comment further on such changes in the examples. 3. Levin (1993) is an outstanding example of such minute classification. Most basic descriptive grammars will not be able to present this level of detail, of course. 4. This section is based on Munro (2002). Input from John Foreman and Ananda Lima was especially helpful. 5. Serious classifications of the Zapotecan languages place the number of languages anywhere from five to 10 (Terrence Kaufman’s current proposal; Natalie Operstein, p.c.) to the Ethnologue’s 57 (Grimes et al., 1996). Resolving this disagreement is not important for this paper, so I will use the term “language” to refer naively to any recognizably different speech form (which, in the Tlacolula District, would mean that every pueblo has a different language). The situation for Mixtec languages, which I come to later, is similar. 6. For more on the syntax of tense in SLQZ, see Lee (1999). 7. The orthography used in this paper is described in Munro and Lopez et al. (1999); many of the examples in this paper are from this source or from Lopez and Munro (eds., in preparation).. Most of the SLQZ aspect prefixes have several variant forms. (A “stem” is a verb containing an aspectual prefix without added adverbial or pronominal clitics; a “theme” is a stem minus its prefix.) Only the Habitual (r- everywhere) does not. The dictionary lists verbs in the Habitual stem (translated in the English simple present), with cross-references to other aspectual stems; such Habitual stems are translated with an English simple present verb. 8. I consider that the unprefixed verbs in (2) contain a zero marker of Neutral status because SLQZ aspectual verb stems are otherwise prefixed, and because such verbs are used in the same context and with the same syntactic restrictions as the prefixed verbs. 9. It’s for this reason that I chose the unhelpful name “Neutral” for the n- prefix and the SLQZ verbal form on which it is used. Suggestions for a better name are welcome! 10. Although nnah begins with n (actually the fortis nn), this consonant is part of the verb’s theme. The related habitual is r-nnììi’. 11. I’m very grateful to Eric Jackson for extensive discussion of the nature of stativity (cf. Jackson 2002). 12. SLQZ has three copulas, nàa (as in (8a)), naa’c (as in (8c)), and nu’uh. Naa’c and nu’uh are Neutral verbs related to the Habitual verbs rahc and ru’uh). SLQZ adjectives appear to be lexically specified for which copular verb they can be used with. Nàa can also be analyzed as a Neutral verb and arguably also contains the Neutral prefix n-, but this is harder to demonstrate, since the paradigm for this verb is defective. 13. In the case of adjectives that start with n plus a vowel that don’t have an alternant in which the n is missing, it is of course not possible to tell if these include a segmentable n- prefixed to a vowel-initial theme that does not occur elsewhere or (as I will assume) whether they are single morphemes beginning with n.
105
106
Pamela Munro
14. The use of n- on borrowed adjectives (14d) is quite uncommon. It seems likely that the n-prefixed loan adjectives were borrowed earlier than the loan adjectives without n. This is supported by the case of lo’oc vs. n-lo’oc. Probably the original use of this word in Zapotec was to refer to bulls (an imported concept, unlike all the other meanings for either word). 15. Cheryl Black has suggested to me that it may be best to assume that the marker in question is the same in both cases (at least for Quiegolani). I address here only the question of whether an adjective with an n- prefix should be regarded as a Stative verb. 16. I.e., in modal auxiliary or modal same-subject complement verbs (Munro and Lopez et al. 1999: 19). 17. There are apparent rare exceptions to this general claim. For example, bàany is used as an attributive modifier meaning ‘alive’ (e.g. in bèe’cw bàany ‘living dog, dog that is alive’); the same form is the Neutral verb corresponding to the Habitual verb rbàany ‘lives, survives’. Such relationships are not regular and are probably best analyzed as distinct uses of the words in question. 18. There are a variety of complex noun phrases containing very short relative clauses that do not work like the syntactically formed noun plus relative structures I discuss below. 19. In sentences with longer relative clauses the clitic may follow the head. It’s possible that it is the modifying function of the Neutral verb ‘is cracked’ in (22b) that determines the clitic position. 20. Such differences suggest that a noun plus adjective sequence, at least in some cases, should be analyzed as something almost like a compound, an idea that has been discussed for Isthmus Zapotec by Pickett (1997). This idea is supported by native speaker intuition in many cases (as Pickett has also noted), as well as by the fact that the n-less attributive forms of adjectives beginning with n- in citation form — such as cweeby in (13) — cannot be used alone. In addition, a few adjectives (all color terms) have irregular suffixed n-less attributive forms, though these do not appear to be fully productive: ngaàa’ts ‘yellow’ (attributives ngaàa’ts, -yaàa’ts), nga’as ‘black’ (attributives nga’as, — yàa’as), and ngàa’ah ‘green’ (attributives ngàa’ah, -ya’ah). 21. Here I exemplify only clitic pronoun possessors; the facts are identical if a noun possessor appears in the same (nonaffixed) position. There is an alternative periphrastic possessive construction that supports the same arguments presented here. As (23b) shows, rr is phonologically a cluster in SLQZ. 22. The significance of the facts involving the use of the diminutive was suggested by John Foreman. 23. Verbs are not directly borrowed into Zapotec. Rather, Spanish infinitives may appear as unmarked complements of verbs such as ruhny ‘does’ or rahc ‘is’, for example (thanks to Kevin Terraciano and James Lockhart for stimulating my thinking on these constructions). 24. More work is necessary to determine whether and how far the analysis argued here can be applied to the cognates of the Neutral and adjective forms are in other Zapotec languages. As I mentioned earlier, these are often treated in other grammars and dictionaries as though they are the same item, but generally these sources do not give enough detail on their use for one to really
From parts of speech to the grammar
test how different they are. My work on Macuiltianguis Zapotec suggests that in that language the facts — though perhaps messier — are ultimately similar to those in SLQZ, and briefer work on San Juan Guelavía Zapotec (a language much more similar to SLQZ) suggests the same. 25. Names for the three agreement classes, I, II, and III, follow Munro and Gordon (1982). 26. Case-marked independent pronouns can be added, but are rare except in emphatic contexts. 27. The phonologically conditioned variants of the class III/dative prefixes may include nasalized vowels, which are difficult to segment into pronominal element and dative marker. The m of the dative prefix assimilates to a following stop; a nasalized vowel replaces the Vm of the prefix before glides, nasals, and fricatives. 28. Traditionally (e.g., by Nicklas 1972), the im- prefix has been analyzed as a third person dative marker. Following Ulrich (1986), I regard im- as the sign of the dative or a morphological indication of class III marking, but not as a third person prefix. Like class I and II agreement, then, class III agreement uses a bare (though derivationally complex) stem as the third person form. Ulrich (1986) and Munro (1993) provide a number of arguments against considering the dative marker (or, by extension, any applicative) as marking third person. Crucially, these prefixes appear in many contexts without third person reference, which is inferred only in the absence of other person indicators; in some cases, such unmarked forms have first or second person reference. 29. There are of course historical and other reasons for many of such seemingly arbitrary assignments, some of which are discussed in Gorbet and Munro (2002b). It is likely, however, that speakers do not access these every time they use these words. 30. Chickasaw relative clauses are head-internal; see Gordon (1987). The language has a variety of relative structures one of whose features is the omission of nominative marking on subject heads, as in (42). 31. As will be clear, this section owes a great deal to the work of Brook Lillehaugen and to productive discussion with her. 32. This is actually the verb ‘be’ seen, for example, in (12c) above (so (47) could apparently be literally translated as ‘My foot is’). I have not figured out if this is some kind of metaphor or a homophonous verb. 33. Sentences (43), (48), and (51) are cited from Lillehaugen (2001), but I have slightly updated spelling and glosses. In particular, I give both nominal and prepositional interlinear glosses (separated by a slash) for the body part words. The prepositional meanings are those that are appropriate for the particular sentence, since many of the body part words can be translated by a number of different English prepositions. 34. In fact, there are several classes of prepositions in SLQZ: in addition to the body part prepositions just exemplified, these include non-body part native prepositions like caàa’n ‘along, by’ and gahx: ‘near’ (as well as the genitive preposition x:tèe’n), as well as prepositions borrowed from Spanish such as cëhnn ‘with’ and pahr ‘for’.
107
108
Pamela Munro
35. Unlike the English, the Zapotec sentence is not ambiguous: this does not refer to ‘the table in back of the house’ (as opposed to another table), but specifies the location of the event of painting. 36. Analysis of the phonology and pronominal system of Tlacolula Zapotec is as yet incomplete, and this sentence is cited in the SLQZ orthography. Thanks to Roberto Antonio for teaching us about Tlacolula Zapotec. 37. Even Hollenbach (1995: 177), in her discussion of Mixtec body part prepositions, refers to “genuine prepositions” that don’t have a body part source! 38. I am grateful to Barbara Hollenbach and Joanne North for graciously sharing information and materials on Mixtec with me. The description here, based on Shields (1988) and North and Shields (1986 [1976]), reflects the speech of San Jerónimo Progreso in the Silacayoapan District of Oaxaca. Marcus Smith and I have been studying a closely related variety of Mixtec with Timoteo Mendoza of San Mateo Tunuchi in the Juxtlahuaca District for several months; I thank Smith for extensive and very productive discussion and Mendoza for his generous teaching, which has helped me understand the Silacayoapan data discussed here. Silacayoapan Mixtec has been written several ways; the examples I cite here are in the orthographies of the original sources. 39. (55–57) are from North and Shields (1986 [1976]: 36); English translations of the glosses and sentences are mine, though I have added body part translations preceding a slash, as in the Zapotec examples above. (Actually, the Spanish translations provided by North and Shields for words like jata in (55) (‘detras’) and tixi in (56) (‘debajo’) are not in fact full prepositions (they are missing de).) (57–59) are from Shields (1988): 318, 336–37). Unlike North and Shields (1986 [1976}), Shields (1988) gives only body part translations for the locative nouns, regardless of their use. I have added prepositonal translations following a slash, and have made changes in format to follow the other examples in this paper. 40. Shields’s list of the meanings of the body part locatives (p. 404) regrettably does not consistently distinguish nominal and prepositional uses: for example, tishi ‘stomach’ is given only the added translation ‘underside of ’ (a possessed noun) rather than (prepositional) ‘under’ (as in (56)). 41. There may be syntactic features of body part prepositional phrases in Silacayoapan that are not shared by other noun phrases. Hollenbach notes (1995: 181) that Mixtec prepositions can be stranded when the prepositional object is fronted for focus, but that possessed nouns cannot be. 42. This section is based on the first section of Munro (2000); cf. also Gorbet and Munro (2002a, b). 43. A dative prefix is used on nouns to indicate alienable nominal possession. 44. There is some relatively more obscure data I will not discuss here that might be adduced in support of the postposition incorporation hypothesis I will discuss in Section 4.
From parts of speech to the grammar
45. Certain verbs, such as aya ‘go’ and many positional and locational verbs, take locative or directional (direct) objects; these verbs don’t require applicative prefixes to license a locative or directional object. 46. They are not postpositions in Choctaw either, but I will not cite data to prove this here. 47. There is one important difference between incorporated postpositional objects and ordinary direct objects in Pima, however. The inanimate plural postpositional object ‘u’us can optionally trigger the appearance of the third person plural object prefix ha- on the postposition, In contrast, a direct object prefix goes on the verb itself. 48. I have synthesized the list of atypical adjective types from a variety of sources, including Palmer (2002) and Hardy (2002) — special thanks to Don Hardy for reminding me about the Austen quote which I used in Section 1. 49. This sort of particle is sometimes considered an “intransitive preposition”.
References Black, Cheryl A. 2000. Quiegolani Zapotec syntax: A principles and parameters account. Dallas: SIL International and University of Texas at Arlington. Bonds, Ashlee M., et al. 2001. U.C.L.A. slang 4. UCLA Occasional Papers in Linguistics 22, ed. by Pamela Munro. Dowty, David. 1979. Word meaning and Montague grammar. Dordrecht: Reidel. Gorbet, Larry, and Pamela Munro. 2002a. Directionality and affectedness: Semantic extension in Chickasaw applicatives. Proceedings of the 50th anniversary conference of the Survey of California and Other Indian Languages. Reports from the survey 12, ed. by Conathan, Lisa and Teresa McFarland. Berkeley: Department of Linguistics, University of California. Gorbet, Larry, and Pamela Munro. 2002b. Chickasaw applicative subjects. LASSO Annual Meeting, Pasadena. Gordon, Lynn. 1987. Relative clauses in Western Muskogean languages. Muskogean linguistics, UCLA Occasional Papers in Linguistics 6. ed. by Pamela Munro, 66–80. Grimes, Barbara F., et al. 1996. Ethnologue. Dallas: SIL International. www.sil.org/ethnologue/ Hardy, Donald. 2002. Parts of speech. http://www.engl.niu.edu/dhardy/grammarbook/program/ Hollenbach, Barbara E. 1995. Semantic and syntactic extensions of body part terms in Mixtecan: the case of ‘face’ and ‘foot’. International Journal of American Linguistics 61: 168–190. Jackson, Eric M. 2002. The stative s- morpheme in Pima. M.A. thesis, UCLA. Lee, Felicia A. 1999. Antisymmetry and the syntax of San Lucas Quiaviní Zapotec. Ph.D. dissertation, UCLA. Lee, Felicia A. 2002. Three question markers in San Lucas Quiaviní Zapotec. Ms. Levin, Beth. 1993. English verb classes and alternations. Chicago: University of Chicago Press. Lillehaugen, Brook Danielle. 2001. The syntax and semantics of body part prepositions in Valley Zapotec. Handout.
109
110
Pamela Munro
Lillehaugen, Brook Danielle. 2003. The syntactic and semantic development of body part prepositions in Valley Zapotec languages. Presented at the Workshop on American Indigenous Languages, Santa Barbara. Lillehaugen, Brook Danielle. 2003. The categorial status of body part prepositions in Valley Zapotec. U.C.L.A. MA thesis. Loos, Eugene E., et al., (eds.). 1999. Glossary of linguistic terms. Dallas: SIL International. http:// www.sil.org/linguistics/GlossaryOfLinguisticTerms/Index.htm Lopez, Felipe H., and Pamela Munro (eds.). In preparation. X:a Mo’od Bìe’d Ra Bùunny Sann Lu’uc Lohs Aa’nngl: Zapotec immigration narratives [How people of San Lucas came to Los Angeles]. MacLaury, Robert E. 1989. Zapotec body-part locatives: Prototypes and metaphoric extensions. International Journal of American Linguistics 55: 119–154. McArthur, Tom, (ed.). 1992. The Oxford companion to the English language. Oxford: Oxford University Press. Accessed at http://www.xrefer.com/entry.jsp?volid=28 Munro, Pamela. 1985. A new ‘harmony’ within type 24. LSA Annual Meeting, Seattle. Munro, Pamela. 1989. Postposition incorporation in Pima. Southwest Journal of Linguistics 9:108–27. Munro, Pamela. 1993. The Muskogean II prefixes and their significance for classification. IJAL 59: 374–404. Munro, Pamela. 1999. Chickasaw subjecthood. External possession, ed. by Doris L. Payne and Immanuel Barshi, 251–89. Amsterdam: John Benjamins. Munro, Pamela. 2000. The leaky grammar of the Chickasaw applicatives. Proceedings from the main session of the Chicago Linguistic Society’s thirty-sixth meeting, vol. 36–1.285–310, ed. by Okrent, Arika and John P. Boyle. Chicago: Chicago Linguistic Society. Munro, Pamela. 2002. Aspects of Stativity in Zapotec. LASSO Annual Meeting, Pasadena. Munro, Pamela, and Molly S. Fasthorse. In preparation. A Tolkapaya dictionary. Ms. Munro, Pamela, and Dieynaba Gaye, 1997. Ay Baati Wolof: A Wolof dictionary (Revised Edition), UCLA Occasional Papers in Linguistics 19. Munro, Pamela, and Lynn Gordon. 1982. Syntactic Relations in Western Muskogean: A Typological Perspective. Language 58: 81–115. Munro, Pamela, and Felipe H. Lopez, with Olivia V. Méndez, Rodrigo Garcia, and Michael R. Galant. 1999. Di’csyonaary X:tèe’n Dìi’zh Sah Sann Lu’uc (Dictionary of San Lucas Quiaviní Zapotec / Diccionario Zapoteco de San Lucas Quiaviní). Los Angeles: Chicano Studies Research Center. Munro, Pamela, and Catherine Willmond. 1994. Chickasaw: An analytical dictionary. Norman: University of Oklahoma Press. Nicklas, T. Dale. 1972. The elements of Choctaw. Ph.D. dissertation, University of Michigan. Nicklas, T. Dale. 1979. Reference grammar of the Choctaw language. Durant, OK: Choctaw Bilingual Education Program, Southeastern Oklahoma State University. North W., Juanita, and Juliana Shields W. 1986 [1976]. Gramática popular del Mixteco: Mixteco de Silacayoapan. Third printing. México, D. F.: Instituto Lingüístico de Verano. Palmer, Gail. 2002. Adjectives. http://users.ece.gatech.edu/~gpalmer/ece8020/presentations/ Adjectives.ppt. Payne, Doris L. 1981. Chickasaw agreement morphology: A functional explanation. Studies in transitivity, ed. by Sandra A. Thompson and Paul J. Hopper. New York: Academic Press.
From parts of speech to the grammar
Pickett, Velma. 1997. When is a phrase a word? Problems in compound analysis in Isthmus Zapotec. Presented at the UCLA American Indian Linguistics Seminar. Pickett, Velma B., and Cheryl Black. 1998. Gramática popular del Zapoteco del Istmo. Juchitán, Oaxaca, and Tucson: Centro de Investigación y Desarrollo Binnizá A.C. and Instituto Lingüístico de Verano A.C. Shields, Jäna K. 1988. A syntactic sketch of Silacayoapan Mixtec, Studies in the Syntax of the Mixtecan Languages, vol. 1, ed. by Bradley, C. Henry and Barbara E. Hollenbach. Dallas: Summer Institute of Linguistics and University of Texas at Arlington. Stross, Brian. 2002. Languages of Middle America. http://www.utexas.edu/courses/stross/ant322m_files/langfeat.htm Ulrich, Charles H. 1986. Choctaw morphophonology. Ph.D. dissertation, UCLA.
111
Grammar writing for a grammar-reading audience* Michael Noonan University of Wisconsin-Milwaukee
Thousands of languages are currently in danger of extinction without having been adequately documented by linguists. This fact represents a tragedy for communities in which endangered languages are spoken, for linguistics as a discipline and for all of humanity. One major role of the field of linguistics is to describe languages accurately and thoroughly for the benefit of all concerned. This paper presents the results of an informal survey of major users of grammatical descriptions and gives lists of dos and don’ts for those contemplating a descriptive study of one of the many endangered languages of the world. Concrete suggestions are provided that will help grammar writers produce user-friendly, thorough and useful grammatical descriptions.
1. Introduction I’ll begin this paper with a recitation of a set of familiar and depressing facts. Of the approximately 7000 languages spoken today, the following can be said with regard to demographics, all gleaned from the Ethnologue (Grimes 1996): – Fully 50% of the human race has as their first language one of just 10 languages; – 52% are spoken by fewer than 10,000 people; – 28% are spoken by fewer than 1,000. Long term estimates concerning language vitality vary considerably, the most pessimistic being that of Michael Krauss (1992) who predicts that by the end of the present century up to 95% of the world’s languages will be moribund or extinct. The linguists associated with the Volkswagen Foundation’s DOBES project opt for the rather more optimistic prediction that 60–70% of the world’s languages will have died out by that time. Most everyone would probably agree that by the middle of this century, there will be less than 2000 languages spoken by integral
114
Michael Noonan
communities, that is by communities in which both children and adults speak the language, and many of those languages will be changed in fundamental ways through contact with languages of national and global importance. All of this would be discouraging enough if the world’s languages were adequately documented. But here too the situation appears grave. We’re confronted with the following, all of which are my [probably overly optimistic] estimates: – We have full-scale grammars and dictionaries and abundant textual material for perhaps 500 languages; – We have grammatical sketches or short grammars, and dictionaries for perhaps another 2000 languages — many of these are of poor quality; – We have only rudimentary documentation [word lists, a few sentences, perhaps a paper on some aspect of the grammar] for another 2500 languages; – We have little useful grammatical or lexical data for about 2000. So, not only are most of the world’s languages inadequately described, but given the rate at which languages are becoming moribund or extinct, we are engaged in a race against time [and against war, poverty, and the good and bad effects of globalization] to preserve as much of the world’s linguistic heritage as possible. In running this race, we face, among others, the following three major obstacles: the standard of grammar writing is not uniformly high, there is not a lot of funding to support grammar writing, and the profession does not sufficiently value or support the writing of grammars. This paper will be concerned primarily with the first of these problems, namely the establishment of higher standards for grammar writing. As regards the second problem, funding, there have been many encouraging developments recently, and I will have nothing further to say about this issue. About the third issue, the evaluation of grammar writing within the profession and the professional support provided to grammar writers, I will make a few comments at the end of this paper.
2. Standards for grammar writing — full-scale grammars When you teach the craft of writing to students, one of the things you try to impress upon them is the importance of taking into account the audience to whom they are addressing their writing. The success of a piece of writing is usually measured by its reception. One has to know who will be reading the thing written and what their needs and expectations are in order to assess whether a work is successful. Reference grammars have two sorts of audiences. The first consists of those who use a grammar to help them learn to speak or write the language described
Grammar writing for a grammar-reading audience
therein, and the second consists of those who consult a grammar to obtain information for typological or theoretical studies, i.e. professional linguists. The first sort of audience will most likely be learners of languages taught in formal classroom settings, languages which are used by people with political or economic power and which have significant literary traditions. Linguists, of course, may also be interested in obtaining information about these languages, but linguists are also likely to be interested in gathering information about languages which are not taught in formal classroom settings, which have little or no tradition of literary production, and which are not spoken by people with political or economic power — that is the great majority of the world’s languages. In fact, the primary audience for the grammars of such languages are linguists, and those writing grammars of languages of this sort should write their grammars with the needs and expectations of professional linguists in mind. One should also bear in mind that many of the grammars written in the next few decades will someday be the only sources of information about the languages they describe — not just for linguists, but also for the communities that currently speak them. The transition from vital, to moribund, to extinct can happen surprisingly quickly. For example, when I started working on the Nar-Phu language of Nepal in 1996, the language seemed quite vital — protected, it seemed, by the very hardships faced by the people who spoke it and the remoteness of their home region. By 2001, the young people had switched almost completely to the national language, Nepali, and when speaking Nar-Phu, many had difficulty recalling NarPhu words, in particular the special and distinctive honorific vocabulary, and were unable to produce the correct [i.e. historic] tones for many words they could recall. So, in preserving languages, it is important to record them not just before they die, but while they are still spoken by integral communities and while they still preserve that which made them distinctive representations of the human mind and spirit. The Nar-Phu that the younger generation will remember will not be the same language as that of their forebears. With the preceding as background, we should be aware that when we are writing grammars of those languages which will likely be moribund or extinct by the end of the century — that is, the great majority of the world’s languages — that we are writing for the ages. So, we must make sure that what we are doing reaches for a very high standard. What follows is a set of prescriptions for grammar writers intended to inform them about matters of form and content which would help make their grammars meet the needs and expectations of linguists and help give their grammars lasting value. In putting this list together, I consulted with a number of linguists — major users of grammars — who graciously took the time to contribute ideas on what
115
116
Michael Noonan
a good grammatical description should be or should contain. These linguists are, in alphabetical order: Sasha Aikhenvald, Balthasar Bickel, Bernard Comrie, Bill Croft, Bob Dixon, Matthew Dryer, Mike Hammond, Martin Haspelmath, Larry Hyman, Ian Maddieson, Edith Moravcsik, and Randy LaPolla. Where one of the these linguists has suggested to me one of the points below, s/he is identified. The list of twenty-eight dos and don’ts is organized informally into three general categories: "user friendliness", "descriptive adequacy", and "comprehensiveness". User friendliness: make your grammar one that is easy to use and obtain information from. 1. Where possible, avoid theory-specific terminology and use instead "basic linguistic theory" as a source of terms: Experience has shown that grammatical theories have a short shelf-life. Special terminology and notational conventions employed by new theories are appropriate for journal articles but not for grammars, which ought to have a shelf-life longer than any given grammatical theory. Libraries are littered with virtually unusable grammars employing the very specialized conventions of extinct theories. Grammar writers should try, therefore, to write grammars for the broadest possible audience of linguists, employing only notational conventions and terms that trained linguists of any theoretical persuasion could reasonably be expected to understand, using terms drawn from what has recently come to be called "basic linguistic theory". Where new or theory-specific terms are introduced, they should be carefully defined. [Dixon, Hammond, Haspelmath, Maddieson, Moravcsik] 2. Provide a detailed index and table of contents: The index and table of contents of a grammar are important tools, especially useful when they are sufficiently detailed and properly organized. Though both serve to help grammar readers find information, they serve somewhat different functions. The index provides an alphabetical listing of topics, whereas the table of contents provides a schematic overview of the topics covered. The latter can be very important where the terms used by the grammar writer differ from those used by [or expected by] the grammar reader: the grammar reader may well find the information s/he needs quickly with a logically organized, detailed table of contents. In addition, in organizing an index, some conventions employed by Martin Haspelmath in his Lezgian grammar should be followed, namely noting in boldface type the main entry [or entries] for a given feature, and noting the absence of a given feature directly in the index with the dash. So, for example, one could indicate the absence of tone as follows: tone: — . [Bickel, Comrie, Croft, Haspelmath]
Grammar writing for a grammar-reading audience
3. The text should be divided into numbered and titled sections and subsections, and there should be ample cross-referencing within the text: Within the text itself, the following conventions make the text easier to use and should be encouraged: the text should be divided into numbered and titled sections and subsections, and, where appropriate, there should be crossreferencing between parts of the text where related issues are discussed; important terms should be highlighted by boldface type and/or by placing the terms in the margins; and finally, section numbers and/or titles should be placed in the margins as headers. [Moravcsik] 4. Provide plenty of examples: Properly glossed and translated examples are a necessary accompaniment to descriptive statements in all components of the grammar. Care should also be taken in the choice of examples: made-up sentences are appropriate for the presentation of information about basic grammatical structures, but examples should be drawn also from naturally occurring discourse. Further, the grammar writer should avoid using examples which are drawn from translations of foreign texts since such sentences may have been influenced by structures in the source language. [Aikhenvald, Maddieson] 5. Provide interlinear morpheme translations [glosses], as well as translations of the whole, for all examples: The necessity for this should be obvious, but too often this courtesy to the grammar reader is ignored or applied inconsistently. There should be a conveniently accessible list of all abbreviations and symbols used in morpheme glosses and all other parts the grammar and dictionary. 6. A typological sketch, consisting of no more than three to five pages, should be included at the beginning of the grammar: A short typological sketch is useful to the grammar reader because it helps the reader put the more detailed information contained in the grammar proper into appropriate context. The sketch should outline prominent features of the phonetics and phonology; should provide information about basic word order patterns, the presence of basic nominal and verbal inflectional categories, the presence of concord classes or classifiers, special lexical features, and so on; and should briefly describe if the language is head- or dependent-marking and if [and to what degree] the language exhibits properties of ergativity, inverse marking, split intransitivity, and so on. [Aikhenvald] 7. The absence of a feature should be noted just as reliably as the presence of a feature: Too few grammars explicitly note the absence of commonly encountered grammatical features in the languages they are describing. Grammar readers cannot always be certain, therefore, whether the lack of discussion of a particular feature results from its absence in the language or from the fact that the grammar writer simply did not discuss the matter. Since the grammar writer cannot reasonably be
117
118
Michael Noonan
expected to note the absence of any of the full set of possible grammatical features, s/he will have to choose carefully which features to note the absence of. The absence of any feature which might be expected on areal, genetic, or typological grounds should be noted [e.g. the absence of concord classes in a Bantu language or ejective consonants in a language of the Pacific northwest], as should features which are commonly encountered worldwide [e.g. tone]. [Aikhenvald, Bickel, Comrie] Descriptive adequacy: Make your grammar one in which topics are described in adequate detail, in a manner that reveals important generalizations, and with adequate backing. 8. The grammar writer should prepare him/herself by obtaining a good background in general typology and the linguistics of the relevant language family: A sound background in general typology is imperative for grammar writers; it is also important for the grammar writer to acquire a sound knowledge of the linguistics of the relevant language family and/or speech area and to consult works dealing the typology of features prominent in the language under investigation [e.g. works on ergativity for works on languages exhibiting ergative structures, works on classifier systems for languages having classifier systems, etc.]. [Aikhenvald, LaPolla, Maddieson] 9. In explaining phonetics, standard IPA characters should be used: Standard IPA phonetic symbols, not "practical orthography" symbols [even where these are sanctioned by tradition for a given speech area or language family], should be used in presenting information about the phonetics of a language. Even when IPA symbols are used, they should be accompanied by a statement describing their value [which could be done by reference to a standard work such as the IPA Handbook]. When non-IPA characters are introduced for use in the practical orthography for the language, their relation to the IPA characters used to describe the phonetics should be made transparent. [Hammond, Maddieson] 10. To the degree possible, detailed instrumental documentation of the phonetics should accompany descriptive statements: Descriptive statements of the phonetics of a language should be accompanied by instrumental documentation of at least the following features: acoustic characterization of vowels in formant space, measurement of the acoustic durations of vowels and consonants in a controlled set of data [so that some aspects of the duration patterns of the language can be quantitatively described], and for tone or pitch accent systems the shapes of F0 patterns associated with the contrasts. Any typologically unusual phonetic features should be described together with appropriate instrumental documentation. [Hammond, Maddieson]
Grammar writing for a grammar-reading audience
11. Provide a full description of the segmental and suprasegmental contrasts and an explanation of the basis for arriving at them: A grammar should contain a full description of the segmental contrasts and an explanation of the basis for arriving at the analysis. In particular, the analysis must demonstrate contrast between segments [in the form of minimal pairs] and present arguments in favor of [or against] the unitary vs sequential analysis of complex elements. In addition, the analysis must include careful descriptions of suprasegmental contrasts: tone [if phonemic], stress, and intonation. Intonation and stress may have implications for morphology and syntactic structure. [Maddieson] 12. Provide a description of distributional patterns of the elements of the phonology: A grammar should contain a discussion of the basic distributional patterns of the elements of the phonology in terms of syllables, words, and any other units which seem to control, or be marked by, different distributional possibilities. Minimally, canonical syllable structure, possible onsets and codas, and nucleus-margin restrictions should be described. [Maddieson] 13. Provide full paradigms showing combinations of all relevant morphemes, not just lists of affixes: In morphologically complex languages, full paradigms, arranged in some accessible form [e.g. as tables] should be provided and not just a list of the affixes. When the language has complex morphophonemic processes, there should be a sufficient number of paradigms provided to illustrate the various combinatorial possibilities. [Hyman] 14. Define grammatical categories used in the grammar: When using grammatical category labels, the grammar writer should make clear what grammatical tests are being used to define those categories. This applies especially to labels for grammatical roles [e.g. subject, direct object, etc.] and word classes [e.g. adverb, adjective, etc.]. [Aikhenvald, Croft] 15. The choice of labels for grammatical features is not as important as a thorough presentation of the facts: Grammar writers should not worry too much about whether a label for a given feature is the most apt. More important is a thorough description of the phenomenon. [Croft, Haspelmath, LaPolla] 16. It’s better to admit ignorance about a grammatical feature than to say nothing: Saying nothing about a particular grammatical feature leaves the matter open to incorrect interpretations, for example that the feature does not exist in the language. When the grammar writer lacks sufficient information about an issue, s/he should not be afraid to admit this, using perhaps a statement like the following: "There is insufficient data at this time to determine conditions under which this form is used." Further, it is important to emphasize that grammar writers should
119
120
Michael Noonan
present data about grammatical phenomena that their theories either cannot account for or are not concerned with: technical papers in journals can be narrow in their scope, but reference grammars must endeavor to describe the entire language system. [Comrie, Hammond, Haspelmath] 17. Existential statements are not sufficient: In writing grammars, it is not sufficient simply to note that the language has some feature X; one must also describe under what conditions that feature is used. For example, noting that Chantyal has a plural suffix ‑ma and then limiting the discussion to a few examples is not enough: an adequate description will also say under what conditions the suffix must be used and under what conditions it cannot be used. [Haspelmath] 18. Indications of frequency should be provided where appropriate: Indications of the frequency of grammatical constructs can provide useful information. These include segmental frequency counts [based on texts or the lexicon] and indications of whether a given morphological or syntactic feature is rare or marginal [e.g. lexically limited]. [Bickel, Maddieson] 19. In general, it’s best to describe morphology with a form-to-function orientation and syntax with a function-to-form orientation: It’s probably impossible [or at least impractical] to write a grammar which is exclusively form-to-function [as the Bloomfieldians attempted in principle] or function-to-form [as Comrie-Smith grammars attempted in principle]. The best compromise seems to be to write the morphology primarily form-to-function and the syntax primarily function-toform. The reason for this arrangement is that form is very easy to recognize in morphology, but is rather less so in syntax. This does not mean that function-toform should be entirely absent from the morphology, nor form-to-function from syntax. For example, alongside a form-to-function discussion of the meanings of verbal affixes, it would also be useful to have a function-to-form discussion of TAM categories, describing how they are expressed in the language. And in syntax, in addition to function-to-form chapters on nominal modification, interrogation, expression of negative senses, etc, it would also be very useful to have a form-tofunction chapter on the uses of various word order possibilities. [Haspelmath] 20. A vocabulary consisting of all the lexemes which occur in the grammar is a necessary component of a good grammar: Grammatical descriptions longer than a sketch should contain a vocabulary consisting of all the lexemes used in the grammar and accompanying texts. These lexemes should be properly labeled for word class or concord class [if appropriate] and, for grammatical morphemes, there should be an indication of the section in the text where the item is discussed. Where variant
Grammar writing for a grammar-reading audience
forms of a lexical item are found in the grammar and texts, the variants should be noted in the vocabulary. [Dixon, Aikhenvald] 21. A good collection of texts should be included with the grammar: A good collection of texts with morpheme glosses and translations is of great importance for a successful grammar. The texts provide illustrations of the constructions described in the grammar, but unlike the examples in the grammar itself, textual examples show the constructions in context. Further, grammar writers cannot be expected to write about, or even know about, all possible grammatical constructions. A good collection of texts can permit later analysts to gather information about constructions that the grammar writer, for whatever reasons, did not discuss or did not discuss in much detail. There should be at least twenty to thirty pages of texts. [Dixon] Comprehensiveness: make your grammar one which is comprehensive in scope. 22. The grammar writer should consult survey questionnaires and wellregarded grammars to make sure that important topics are not missed: The most detailed survey questionnaire that is generally available is the Comrie-Smith [Lingua Descriptive Series] questionnaire, which has formed the basis for a number of grammars. This function-to-form questionnaire [see above] can be profitably consulted by the grammar writer, as can well-regarded grammars, of which several lists are available. [Comrie, Dryer, LaPolla] 23. The grammar should contain information about genetic and areal affiliations of the language: A description of the genetic affiliation of the language should be provided, along with information about the linguistic area and possible areal influences, where such information is known. This can be fairly brief unless there are special issues to be discussed [e.g. controversies concerning genetic affiliation]. [Aikhenvald] 24. The grammar should contain information about the sociolinguistic context: There should be a short description of demographic and socio-cultural facts relating to the language being described, including the number and geographical distribution of speakers, the demographics of language use [e.g. whether the language is spoken only by adults and not children], the degree and nature of multilingualism in the speech community, and the degree of literacy and access to education. In addition, the author should reveal how data about the language was obtained and should provide information about the native speakers who served as language consultants.
121
122
Michael Noonan
25. Grammars should contain ample references to previous scholarship on the language and its culture: The bibliography should contain ample references to previous scholarship on the language; if the language has not been the subject of much previous scholarship, the list of references should be comprehensive. Previous scholarship on the culture of the speakers of the language should also be noted. 26. Grammars should contain good descriptions of the phonetics and the phonology, as well as of the morphology and the syntax: Some grammars include little more than a list of phonemic or "practical orthography" symbols and only a brief description of the phonology with just a few examples. Proper descriptions of the phonetics and the phonology of a language are just as important to the success of a grammar as proper descriptions of the morphology and syntax. And here too there should be a plenitude of examples, narrowly transcribed. [Hammond, Hyman, Maddieson] 27. A good dictionary is an powerful adjunct to a good grammar: As noted, vocabularies, consisting minimally of the lexical items employed in the grammar, should be included with any grammatical description longer than a sketch. The existence of such a vocabulary, however, does not diminish the value of a good, fullscale dictionary, which, apart from any other uses it may have, is also a powerful adjunct to a grammar. A good dictionary comes with lots of example sentences from which a great deal of useful grammatical information can be gleaned even about topics that the grammar writer/dictionary creator has no knowledge of or interest in. For example, very few grammars, even good grammars, contain much information about clausal complementation. Typically there will be discussions of complement-types with some example sentences that hint at the distribution of these forms, but very rarely will the grammar contain much explicit discussion on the distribution of complement-types vis-à-vis complement-taking predicates, or contain a sufficiently large number of examples to allow the analyst to infer the distribution. A good dictionary, on the other hand, can be a goldmine of information about such topics. 28. Where practical, audio and video recordings should be made of various language genres: Where practical, audio and video recordings of a variety of language genres and cultural activities should be made. A subset of these should be annotated, provided with appropriate commentary, and deposited in an archive where they may be studied by scholars. [A number of such archives are now in existence.] Further, photographs and drawings can be very useful in documenting information about cultural artifacts, the natural environment, and botanical and zoological vocabulary.
Grammar writing for a grammar-reading audience
Points 27 and 28 are desiderata rather than requirements for successful grammars. However, with regard to point 28, it’s well to keep in mind that over the years the materials to be included in “good” grammars, i.e. ones that would fully meet the highest expectations of their day, have progressively expanded with advances in the field and in technology. At this time, we can’t anticipate the degree to which, say, intonational information, ordinarily not transcribed in published discourses, will be considered crucial to syntactic analysis. It’s best to archive audio and visual records of speakers, if at all feasible. Needless to say, not everything that a grammar should contain is noted in the points above since the list focuses mostly on those things that are left out or not done with some regularity. This accounts for the fact that the prescriptions relating to phonetics and phonology are much more specific than those for morphology and syntax, reflecting the generally lower standards of training of field linguists in phonology and, in particular, phonetics. The standard of grammar writing detailed above raises a pretty high bar, and I’m the first to admit that a grammar can be very useful and still not conform to all of the prescriptions listed. However, the more that are observed, the more useful the grammar will be.
3. Standards for grammar writing — sketches and short grammars The list of dos and don’ts already given applies, of course, only to full-scale reference grammars and not to grammar sketches, which necessarily will fail to conform to many of the prescriptions on the list, in particular those placed under "descriptive adequacy" and "comprehensiveness." Grammar sketches and short grammars have, of course, a legitimate function. In a perfect world, we would have full-scale grammars for every language, and grammar sketches would serve as convenient summaries of the facts. In our very imperfect world, often all we have for many languages — and maybe all we will ever have — is a sketch or short grammar. These shorter works, then, should be prepared with the same care as full-scale grammars. Of the list of dos and don’ts, the following are applicable even to sketches and short grammars. The numbering of the prescriptions is preserved from the original list: User friendliness 1. Where possible, avoid theory-specific terminology and use instead "basic linguistic theory" as a source of terms. 2. Provide a detailed index and table of contents.
123
124
Michael Noonan
3. The text should be divided into numbered and titled sections and subsections, and there should be ample cross-referencing within the text. 4. Provide plenty of examples. 5. Provide interlinear morpheme translations [glosses], as well as translations of the whole, for all examples. 7. The absence of a feature should be noted just as reliably as the presence of a feature. Descriptive adequacy 8. The grammar writer should prepare him/herself by obtaining a good background in general typology and the linguistics of the relevant language family. 9. In explaining phonetics, standard IPA characters should be used. 11. Provide a full description of the segmental and suprasegmental contrasts and an explanation of the basis for arriving at them. 12. Provide a description of distributional patterns of the elements of the phonology. 13. Provide full paradigms showing combinations of all relevant morphemes, not just lists of affixes. [applies to short grammars, not sketches] 14. Define grammatical categories used in the grammar. 15. The choice of labels for grammatical features is not as important as a thorough presentation of the facts. 16. It’s better to admit ignorance about a grammatical feature than to say nothing. 18. Indications of frequency should be provided where appropriate. 19. In general, it’s best to describe morphology with a form-to-function orientation and syntax with a function-to-form orientation. Comprehensiveness 22. The grammar writer should consult survey questionnaires and well-regarded grammars to make sure that important topics are not missed. 23. The grammar should contain information about genetic and areal affiliations of the language. 24. The grammar should contain information about the sociolinguistic context. 25. Grammars should contain ample references to previous scholarship on the language and its culture. Of the twenty-eight prescriptions on the list, all but 6, 10, 17, 20, 21, 26, 27, and 28 apply to sketches and short grammars as well; 13 applies to short grammars, but not necessarily sketches. Needless to say, the more of these additional features are incorporated into a grammar or even a sketch, the more useful the it will be.
Grammar writing for a grammar-reading audience
4. Community support I’ll close with a few remarks on how the community of linguists can support grammar writing. Among professional linguists, grammar writing is subtly discouraged by the way in which grammars are evaluated for purposes of hiring, tenure, and promotion. At research universities in particular, published articles on theoretical matters are valued more highly than published grammars or grammar sketches. Professional organizations could help change this situation by, for example, raising awareness among linguists concerning the urgent need for grammar writing at this time and by encouraging grammars as topics for dissertations. The latter could be accomplished by the creation of special dissertation fellowships for grammar writing and by skewing the selection criteria for existing dissertation fellowships in favor of grammar writing. Professional organizations such as the LSA or ALT could also establish awards for the best published grammar of the year, thereby raising the profile of grammar writers. Further, the community of linguists could provide support for grammar writing through the establishment of consultancy services for qualified grammar writers. This service is of greatest importance in the area of phonetics, where the standard of training throughout the profession is probably the lowest among those skills that are required for writing grammars. Finally, online publication of grammars and dictionaries has a number of advantages over paper publication: online grammars and dictionaries can easily be updated and revised, thereby increasing the likelihood that they will conform to the prescriptions discussed earlier. They can also be made available to a wider audience [especially if access is free] than is possible with paper publication. And lastly, online, or at least electronic, publication can facilitate the addition of audio and visual materials to the written text of the grammar. There are two problems with online publication. The first is that, in many cases, it is not evaluated as highly as paper publication for purposes of hiring, tenure, and promotion. If the venues for online publication meet the same standards as high quality paper publication — that is, meet high standards of editorial review — then they should be evaluated equally. Again, professional organizations have a role to play here in encouraging academic departments to accept high quality online publication.1 The second problem relates to the relative impermanence of electronic and online publication media. Here too, professional organizations can play a role in arranging for the archiving of grammars published electronically in longer-lived formats.
125
126
Michael Noonan
Notes * I’d like to thank Edith Moravcsik and an anonymous reviewer for helpful comments. In addition, I’d like to thank the linguists, named below, who provided me with their insights on grammar writing. Needless to say, all the shortcomings of this paper are mine alone. 1. Online publication has a number of virtues, but it also poses a number of challenges both for academic departments evaluating candidates for hiring, promotion, performance raises, and so on and for editorial boards. For example, one of the virtues of online publication is that it allows for easy revision of texts. For editorial boards, a question arises as to how much revision should be allowed, how it should be monitored, and how much revision is allowable before the work must be reviewed again. Different standards might well be applied to journal articles and reference works such as grammars. In the case of grammars, I would suggest a fairly liberal standard for revision, drawing the line at changes that affect the framework or mode of analysis. In such cases, the revised grammar would have to undergo another review process.
References Grimes, Barbara F. et al. 1996. Ethnologue. Dallas: SIL International. Krauss, Michael. 1992. The world’s languages in crisis. Language 68(1):4–10.
A grammar as a communicative act or What does a grammatical description really describe? Thomas E. Payne University of Oregon and SIL International
A descriptive grammar is a document that is intended to communicate important and complex patterns of knowledge within a speech community. Previous conceptions of linguistic grammars have emphasized a metaphor in which the internal grammar of a language is perceived as a “machine” that transforms thoughts into language structure. Written grammars were conceived as schematic representations of such machines. The present paper argues for a different metaphor. Because a grammar is primarily an act of communication, a good grammar possesses qualities known to hold of other communicative acts. The paper invokes recent work in linguistic pragmatics to help potential grammar writers conceptualize and present their work as an efficient and engaging act of communication.
1. Introduction For many years, linguists have implicitly embraced a metaphor for grammatical description in which a written grammar is viewed as a logical machine that replicates the internal grammar of a language. The internal grammar itself consists of all the implicit knowledge speakers possess relative to the language they speak, and is conceived as a system of rules and regularities that takes something as input (“lexical structure,” “semantics,” “intentions,” etc., depending on the model) and produces (or “sanctions”) all the well-formed sentences of the language as output. Linguistic grammars are written according to certain “models,” i.e., toy versions or representations of the real rules and regularities that constitute internal grammars of real languages. The best models are those that allow the most inclusive yet economical account of observed linguistic behavior. From this point
128
Thomas E. Payne
of view, a written grammar is more similar to a computer program, or a schematic diagram of an electronic circuit, than an esthetically gratifying depiction of some real-world phenomenon. In this paper I would like to argue for a different metaphor for grammatical description. I would like readers to consider a written grammar as a communicative act, performed in a context for a particular audience. I will be using insights and examples from the growing body of literature on the nature of communicative acts (Austin 1962, Searle 1970, 1975, Grice 1968, 1975, 1981, Sperber and Wilson 1995, and others) to illustrate that good written grammars, even many of those that have been written according to formal models, always display many of the characteristics known to hold true of communicative acts in general, but which are inconsistent with the grammar-as-machine metaphor. By acknowledging this fact, I hope to free field linguists to produce grammatical descriptions that are more consistent with more familiar modes of communication, and thus more stimulating to write, and engaging to read than those written according to the grammaras-machine metaphor. Furthermore, grammatical descriptions that are conceptualized as communicative acts are more likely to endure for future generations than are those written according to transitory mathematical models. What I am proposing is really nothing new. Grammars have been written with audiences in mind for many years. Good linguists write good grammars in spite of the models that dominate the field at any given point in history. What I am arguing against is an ideal that underlies much linguistic scholarship, yet is never fully realized in practice. I believe that if the ideal can be exposed for what it is — an impossible and insidious dream — it will free many ordinary working linguists from the burden of self-doubt and paralysis that comes from unrealistic expectations of what a grammar “should” be. As a communicative act, a written grammar has more in common with nonfiction literature than with a computer program. It is more similar to a story than to a schematic diagram. This alternative conceptualization of the task of creating a written grammar has profound implications for the organization, content and development of a linguistic grammar.
2. The “grammar-as-machine” metaphor Perhaps since Bloomfield (1933), and certainly Nida (1951), the task of syntactic description has most prominently been conceived as one of specifying the workings of a logical “machine” that takes a particular set of things as input and transforms them into empirically observable speech patterns. Sometimes the input is
A grammar as a communicative act
“meaning” (Generative Semantics), “base component” (Generative Grammar), or any number of other model-specific notions. The output, of course, is the complex patterns of sounds that people use in everyday communication. Since the output is all that can be observed directly, the input, as well as the algorithms that apply to it, must be inferred by the grammarian. The idea was that all the structures in a language could be explicitly “represented” in mathematically precise formulas. If the formulas were written with enough care and precision, any possible utterance in a language could be accounted for by the formulas, and all impossible utterances would be disallowed. Much as a map “represents” the arrangement of streets, landmarks and geographic features of a city, a grammar was supposed to represent all the linguistic patterns that are available to speakers engaged in discourse in that language. There are several problems with this metaphor. First of all, no grammar has ever been written that achieves the ideal of representing all of the rules and regularities of the internal grammar of any language. It remains an idealization, something to be working towards, never fully accomplished, but always left for future investigation. Grammatical descriptions can range from three page articles, to 1000+ page tomes. Yet even the latter do not provide enough information to allow a computer to do even a fraction of the work native speakers do every day in expressing and interpreting communicative acts. A grammatical description is always an approximation. It is what a stick diagram is to a real person — a map to a real city. While a map may be useful, it will never capture the “essence” of a city. It will not begin to describe the special features and sensations that make a city unique, the different characters of the various neighborhoods, the hopes, goals and frustrations of the city’s inhabitants who must negotiate the streets of the city as part of their everyday lives. A map is no more than a stark, idealized representation of certain rather rigid and mundane aspects of a city. Second, the grammar-as-machine metaphor ignores communicative function altogether. Just as a city is much more than the arrangement of its streets and highways, so a language is much more than the arrangement of its phrases and clauses. It is primarily and incontrovertibly a tool used by people to communicate with one another. In the extreme case, grammars written according to the machine metaphor need not even consider the fact of human communication. Meaning, pragmatics, sociolinguistics and cultural considerations are not necessary to construct the algorithms that make up the grammar-as-machine. If a semantic consideration determines the applicability or non-applicability of some rule, a “feature” can be added to the form in question in order to ensure that it performs in the model as its counterpart does in the internal grammar. Indeed, early texts on syntactic analysis all use permutations of meaningless symbols (circles and squares, Nida
129
130
Thomas E. Payne
1951, or alphabetic variables, Chomsky 1957) as illustrations of basic syntactic phenomena. Real language is assumed to be subject to the same kind of logical interactions and permutations that govern abstract, mathematical systems. Third, grammars written according to the machine metaphor tend to be unintelligible to anyone who has not had the same training in the particular theoretical model as the author of the grammar. Since there is no consideration of the reader of a grammar-as-machine, no attempt is made to make the information comprehensible to future generations. The emphasis is on precise representation of regularities and patterns, rather than clear communication of valuable information to an interested reader. Theoretical frameworks in linguistics are notoriously shortlived. What is currently in vogue, will tomorrow be anachronistic. Much good linguistic work of past decades remains largely inaccessible to modern scholars simply because the frameworks employed have gone the way of the dinosaurs. In the following subsections I will describe some “myths” that I believe many of us ordinary working linguists implicitly or explicitly believe regarding writing a grammar of a language. Many of these myths can be dispelled if a written grammar is conceptualized as a communicative act rather than as a logical machine. 2.1 The myth of inclusiveness One dangerous myth surrounding the task of grammatical description is the notion that a written grammar must be all-inclusive. This idea is rooted in the myth of grammar as machine. A grammar is supposed to include all of the patterns and regularities in the internal grammar of native speakers. As mentioned above, there is no such thing as a complete grammar of a language, even among those that presume the “grammar as machine” metaphor. A language is a constantly changing and varying set of overlearned behavioral patterns. Even the most elaborate mathematical formalisms cannot begin to represent all the creative ways that language users can employ the conventionalized patterns to express unique and nuanced ideas. The patterns themselves are conventionalized to varying degrees (see, e.g., Langacker 1987). An innovative pattern that proves useful will be used again and again, thus fixing itself, over time, in the collective memory of the language community. Such patterns, and their varying degrees of conventionalization, cannot possibly be captured between the covers of a book. I have known many fieldworkers who have felt paralyzed when it comes to describing the grammar of a language because they don’t know where to start. They can’t say anything until they can say everything. Unfortunately, the point will never come when you can say everything, so you might as well say what you know.
A grammar as a communicative act
If a written grammar is viewed as a communicative act, the author is liberated from the myth of all-inclusiveness. It has been observed by many researchers (e.g., Grice 1981, Sperber and Wilson 1995 inter alia) that sentences are only partial representations of speaker intentions. Much of what is communicated via language is understood via inference. Consider the following actual conversational exchange between a 16 year old girl and her 45 year old father. The setting is the kitchen of their home. The daughter is standing at the sink, and the father is reading the newspaper at the table: (1) a. Daughter: Who left the eggshells in the sink!?! b. Father: Yes, it’s annoying when people don’t clean up after themselves, isn’t it? c. Daughter (indignantly): I clean up after myself!
While a grammar-as-machine may be able to parse each of these sentences, and a formal semantics may be able to assign unique sentence-meanings, such representations fall far short of capturing the obvious communicative intentions involved. While utterance a. may be in the syntactic form of the question, the intonation and the context make it clear that the proposition being expressed is something like “I am annoyed that someone left eggshells in the sink.” This hypothesis is supported by the father’s echoic response. Rather than simply provide the name of a person, which would be a reasonable answer to the overt question, he responds with a rather loaded question of his own. Again, this “question” is not really a question at all, but a thinly veiled attempt to make a “teaching moment” out of the interaction, undoubtedly rooted in previous discourse between father and daughter. Of course the daughter picks up on this immediately, and in good teenage fashion responds with a quick rejection of the father’s insinuation. This example of how communicative acts communicate more than what is actually expressed “in the words” is the normal way that people interact with one another. This does not strike us as “bizarre” use of language at all. It is the normal way that humans go about the business of communication. From this point of view, a written grammar conceptualized as a communicative act is liberated from the myth of having to account for every nuance of meaning and usage observed in the data set. There is no limit to the creative ways that speakers can use the tools their language provides to express the concepts they choose to express. If the corpus on which the written grammar is based includes natural discourse, it will contain features that are not conventionalized and may not even be describable in linguistic terms. The goal of any written grammar is to provide an account of the conventionalized patterns that are common to the speech community. Much language in use transcends these conventionalized patterns, and thus may not be amenable to linguistic analysis.
131
132
Thomas E. Payne
2.2 The myth of absolute precision I must be careful in this section to head off a potential criticism that I am arguing for vagueness or sloppiness in grammatical description. This is far from the case, and I need to clarify that from the outset. I believe linguistic researchers should be as precise and explicit as possible, given the circumstances. Insofar as possible, grammatical evidence should be presented for every claim. Precision is definitely a value in any grammatical description. What I am arguing against is the myth that it is possible or even desirable for grammatical descriptions to be absolutely precise representations of the categories and regularities of the internal grammar of the language being described. The reason is that a written grammar is a communicative act. All communicative acts are approximations. A classic example of this fact is a sentence like:
(2) Salu wrote all over the living room wall.
Such a sentence is not considered communicatively deficient at all, even though there is a lot of vagueness about it. First, most living rooms have more than one wall, so the fact that the specific wall is not identified constitutes a lack of precision. Also, the notion of “all over” is rather vague. What does the speaker mean by “all over the living room wall”? Does the image evoked by such a sentence necessarily specify that every square centimeter of space on the wall was covered with Salu’s writing? Of course not. Yet this sentence communicates because the idea of “all over” is commonly interpreted as hyperbole referring to “an annoying amount” of writing. Also, “the living room wall” identifies the intended referent closely enough for the communicative purposes at hand. There is no need to specify which living room wall was involved, and to do so would potentially distract the audience by calling attention to a feature that has no obvious relevance. Like all other communicative acts, written grammars are approximations — context-dependent abbreviations of the full range of complexity in a language. Here are a couple of examples of how the myth of absolute precision is routinely exposed in the writing of linguistic grammars: 1. Greenberg (1954) provided a way of explicitly calculating the morphological typology of a language. If Greenberg’s system is applied to a language, numerical values can be assigned that will tell readers exactly how agglutinative and how polysynthetic it is, in comparison to other languages. However, such calculations are so complex, and the payoff so small in terms of the amount of valuable information obtained from the result, that few linguists have chosen to include applications of Greenberg’s methodology in grammatical
A grammar as a communicative act
descriptions. If a linguistic description mentions morphological typology at all, terms such as “somewhat agglutinative,” “very polysynthetic,” “extremely fusional” are usually employed. This level of detail is considered “relevant enough” for average readers of a grammatical description. Readers do learn something useful about the general morphological typology of a language without burdening author and audience with the complex and unrewarding task of running Greenberg’s calculations. 2. The importance of word classes (or “parts of speech”) has seldom been questioned in discussions of what should be included in a grammatical description of a language. How can you even begin to describe a language if you don’t have a clear idea of what the building blocks of that language are? Every grammatical description must at least make mention of nouns and verbs, and probably adjectives, adverbs and some kinds of particles as well. Yet, as I hope to demonstrate, classes such as “Noun” and “Verb” are convenient approximations. They are imprecise generalizations that help readers understand something important about every language, but which do not directly correspond to any fixed categories in even one language. Word classes are typically defined in terms of grammatical properties. For example, a typical linguistic description of a “noun” is “a word that can be the subject or object of a clause.” However, if you investigate the grammatical properties of a number of words, you soon find that the lexicon of any language is not divided into clear, mutually exclusive classes. There is in fact a continuum between the major classes of Noun and Verb, and there are many subtly different sub-classes at various points on that continuum. Each subclass possesses a “cluster” of grammatical properties that may or may not have any obvious relation to one another. Furthermore, subclasses may even consist of only one member. Take for example, verb roots with the ‑ing suffix in English. Are they nouns, are they verbs, or do they belong to some other word class? Nouns have the property of being able to serve as subjects or objects of clauses. Verbs with ‑ing also have this property: (3) a. subject: b. object:
Walking is good for you. I like walking.
Nouns can also take descriptive modifiers (the red car), and their arguments (possessors) are expressed via genitive case pronouns (my car). Again, verbs with ‑ing pass both of these tests: (4) a. descriptive modifiers: ‘slow walking’ b. genitive case pronouns: ‘his walking’ (*he walking)
133
134
Thomas E. Payne
Even though walking passes both of these major tests for nounhood, other tests reveal that it is not the best example of a noun. For example, ability to pluralize, take determiners and to take a wide range of descriptive modifiers are all properties of prototypical nouns in English. It is rare or odd-sounding for verbs with the ‑ing suffix to display these properties: (5) a. pluralization: ‘?many walkings’ b. determiners: ‘?the walking to school’ c. descriptive modifiers: ‘?red/little/pretty/fine walking’
Also, verbs suffixed with ‑ing sometimes can take adverbial modifiers that nouns cannot: (6) a. ? I like slowly walking. b. I like walking slowly.
c. *I like slowly cars. d. *I like cars slowly.
This morphosyntactic property makes the form walking seem a little more like a verb. Notice that the cluster of grammatical properties displayed by ‑ing nominalizations cannot be attributed to “abstract” nouns. Abstract nouns like sincerity possess some, but not all of the properties possessed by walking: (7)
a. b. c. d. e. f. g.
Subject: Sincerity is a virtue Object: I question his sincerity Modifiers: Full sincerity Genitive pronouns: Your sincerity Pluralization: *Many sincerities Determiners: The sincerity of your beliefs is evident. Adverbial modifiers: *Truly sincerity/sincerity truly
Neither can this cluster of properties be attributed to “mass” nouns, rather than count nouns, as though walking were a kind of mass noun: (8)
a. Subject: Water is good for you. b. Object: I like water. c. Modifiers: Good/clean/green water. d. Genitive pronouns: Your water e. Pluralization: ?Many waters f. Determiners: The water is cool. g. Adverbial modifiers: *slowly water/water slowly (trying to refer to water as a noun, as in *I like slowly water, rather than the verb to water)
Neither can this particular cluster of properties be consistently attributed to verbs with the ‑ing suffix. For example, some verbs with ‑ing take plurals more easily
A grammar as a communicative act
than others: his many failings vs. ?his many eatings. This fact puts failing closer to the noun end of the continuum than eating. These subtle differences among the behaviors of various forms are probably not available to the fieldworker faced with thousands of forms, each potentially exhibiting a cluster of from zero to about 10 nominal properties and a similar number of verbal properties. To exhaustively categorize every word according to the particular cluster of verbal and nominal properties it exhibits is a potentially never-ending task. The point of a grammatical description is to help readers understand how particular constructions function within the grammatical system of the language. It is clear that a detailed ranking of structures according to their relative noun or verbhood would be of limited use in accomplishing this task. Given the observation above that such a task would also potentially be of unlimited complexity, it is not likely that many field linguists would attempt such a ranking. Therefore the concepts of Noun and Verb stand as imprecise approximations that nonetheless are precise enough to be useful in expressing important grammatical concepts. 2.3 The mythological analogy between phonology and syntax Along with a mathematical conceptualization of grammatical description, came a deep respect for hierarchy. This regard for hierarchy is, of course, well grounded — hierarchical structure certainly does exist in language, and provides an important dimension of functionality. But like physicists engaged in a holy quest for a “unified theory” of the behavior of subatomic and celestial bodies, linguists have repeatedly tried to apply insights gained at one “level” of structure to other levels, with varying degrees of success. For example, the notion of the phoneme was a revolutionary concept in the 1930s. It opened up a whole range of applications of structuralist theory to the social sciences. Because it was such a useful concept, and was so instrumental in establishing linguistics as a legitimate academic discipline, the phoneme became the archetype for linguistic analysis in general. Morphemes and allomorphs became the morphological analogs of phonemes and allophones; syntagmemes and allotagmemes became their syntactic analogs. For example, Nida (1951) explicitly argues for the parallel between phonology and syntax by using the following example of complementary distribution in syntax: (9) I am. You are.
The claim is that since am appears after I and are appears after you, this is a case of complementary distribution. Since the two forms, am and are, are in
135
136
Thomas E. Payne
complementary distribution, that means they must be “allotagmemes” of the same “syntagmeme.” They must underlyingly be the same unit, the verb be. This is the extent of the argumentation. But a bit more thought reveals that am and are are not just contextually determined variations of the same unit, the way [p] and [ph] are contextually determined allophones of one phoneme. Am and are have different meanings. Am does not appear just after I, and are is not excluded after I: (10) I certainly am coming. Am I needed? Mary and I are coming.
So clearly the difference between am and are is not just blind “allotagmemic variation” that is independent of meaning considerations. These verb forms are conditioned by the meanings of first person singular, versus plural, respectively. This is conceptually and formally a very different matter than allophonic variation. Furthermore, beyond such cases of simple agreement phenomena, how can the principles of contrast and complementary distribution be applied to syntax? The possible systematic syntactic structures of any language are recursive and unlimited, whereas possible phonemes are limited by a relatively fixed inventory of features. Structuralist grammars often provide lists of “sentence types” that are supposed to be analogous to phoneme inventories. There is no doubt that there are construction types in languages, but distinguishing them from one another is a very different analytic task than identifying phonemes and allophones. For example, from the point of view of communication, a negative clause certainly contrasts with the identical affirmative counterpart, as in the following: (11) a. I’m coming. b. I’m not coming.
But what does contrast mean to a syntax that is based on a phonological analogy? Some structuralist grammars may treat these as containing two emically contrastive verb phrases — negative verb phrases and affirmative verb phrases. Others would treat them as variations of one verb phrase type that contains an “optional” slot for a negative marker. If 11a and b are considered as contrastive, what about the following: (12)
a. b. c. d.
I’m coming. I’m definitely coming. I’m thinking of coming. I’m a poor wayfaring stranger who is looking forward to coming.
A grammar as a communicative act
If contrast and complementary distribution are used in determining sentence types, then there are exactly as many sentence types as there are possible sentences in a language. 2.4 Problems with a hierarchical approach to grammatical description Not only is the analogy between phonology and syntax faulty because of different analytical techniques required, but also because it encourages a tendency to organize grammars into hierarchical “levels.” A strict adherence to a hierarchical organization of a grammatical description is less than optimal for both languageinternal and cross-linguistic reasons. 1. Language-internal reasons. Languages consist of interlocking subsystems that involve regular alternations at multiple levels. If each level must be described independently, then the coherence of some of these systems is lost. For example, past tense in English is commonly expressed morphologically, usually by the addition of a suffix, ‑ed, while the future tenses are expressed syntactically by addition of the auxiliaries will or gonna (going to). Because of the requirement that levels of analysis be kept distinct in a hierarchical model, past tense would have to be treated in the word-level chapter while future tense would be treated in the phrase level or syntax chapter. Thus the notion of a unified “tense system” would be lost. Secondly, recent research has shown that the boundary between morphology and syntax is in fact quite arbitrary, and need not necessarily have any implications for linguistic description. For example, in Yagua (Peba-Yaguan, Peru) there is a set of forms that function syntactically with whatever follows but which attach morphologically to the item to the left: (13) Sa-suutá-ra míjay. 3SG-wash-INAN mosquito.net ‘He/she is washing the mosquito net.’
That the ‑ra morpheme functions syntactically with the object mijay and is not an “agreement” marker on the verb is illustrated by the facts that (1) nothing can intervene between ‑ra and mijay and (2) ‑ra attaches to anything that comes after the verb, as in the following: (14) Sa-súúta waturá-ra míjay. 3SG-wash childless.woman-INAN m.net ‘The childless woman is washing the mosquito net.’
That the ‑ra morpheme really does attach to the previous word, and is not a free particle, is evidenced by the stress pattern. Penultimate stress is the normal pattern
137
138
Thomas E. Payne
for words in Yagua. Note that in 13 the presence of ‑ra triggers stress on the previous syllable. In 14 sasúúta takes the normal stress pattern and waturára undergoes the stress shift. These, and many other similar examples from around the world, plus the commonly observed fact that word boundaries in many languages are quite difficult to determine, suggest that morphology and syntax are not universally distinct structural domains. 2. Typological (cross-language) reasons. It is well known that the same or similar jobs are accomplished via very different kinds of systems in individual languages. Many readers of grammatical descriptions are interested in making crosslanguage comparisons. For example, one researcher may be interested in tense systems, another in deixis, etc. A grammar written as a communicative act will take such potential readers into consideration by organizing the topics according to such functional systems, rather than by a strict hierarchical template. A grammar written from a strictly hierarchical perspective may treat some deictic operators in the section on noun morphology, others in the section on verb morphology, and still others in a distinct section on demonstratives. In none of these sections is there likely to be any observations or hypotheses concerning how the various deictic operators function as a system that expresses deictic notions in discourse. There may be no way for a reader interested in deixis to scan the table of contents or index and find all the information needed on the topic. As mentioned earlier, actual languages tend to be organized into functional systems, or “families” of structures that conspire to express a range of related functions, e.g., tense/aspect, mode, valence, deixis, semantic roles, participant reference, topicality, etc. Even as the members of these families are not always paradigmatically related to one another at the same hierarchical level in a single language, so whole systems operate at different levels from one language to the next. Therefore, a grammar organized according to such systems, rather than structural hierarchies, is more faithful to the actual structure of the subject matter.
3. Drawbacks to a strictly functional approach to grammatical description So far I have argued that the traditional conceptualization of a written grammar is an idealization based on certain myths and misconceptions. Most of those myths and misconceptions stem from an overemphasis on language form in grammatical description. Indeed the machine metaphor itself can be understood as a result of an over-commitment to the idea of form as primary in linguistic description. Machines are blind to meaning. A good grammar-as-machine should be able to
A grammar as a communicative act
operate blindly on formal input to produce formal output, without the benefit (or handicap) of the very human concern with meaning. Lest I have made my case too strongly, let me caution against throwing formal analysis out entirely, and resorting to the opposite extreme. The opposite of a purely formal grammatical description would be a purely functional one. What would such a description look like, and has one ever been written? Form-driven grammatical description is consistent with the view prevalent in pre-structuralist linguistics and anthropology that each language should be analyzed on its own terms. Linguists should have no expectations about what they “ought” to find upon investigating a previously unstudied language. Languages were thought to vary randomly and in unpredictable ways. Furthermore, linguists who are not native speakers of the language being described are prone to impose categories from their own internalized linguistic system and to be blind to categories and distinctions that their own system lacks. This danger certainly was realized in a very obvious way in linguistic descriptions that appeared before the advent of structural linguistics. For example, many of the grammars of Amerindian languages written between 1500 and 1900 described the realizations of the seven cases of Latin, even though in fact the languages lacked morphological case marking altogether. Like these early descriptive grammars, an extreme functional grammatical description would run the risk of imposing distinctions that are not made in the language being described. For example, if we are determined to find plural marking in every language in the world, we will find it, even if it always has zero realization! Observable structure is the main indicator of functions that are important to speakers of any language. Purely functional distinctions are seldom sufficient. For another example, a “passive” construction is often defined functionally as one that “downplays” an AGENT and “upgrades” a PATIENT. Well if that is the definition of passives, then would the following English sentences all be passives? (15)
The glass broke. (Downplaying the AGENT who broke the glass.) Some guy broke the most beautiful vase in the world. These jeans wear easily. John underwent surgery. As for black-eyed peas, I can’t stand them. Black-eyed peas is what I can’t stand.
These all can be construed as somehow “downplaying an AGENT” and/or “upgrading a PATIENT.” Without clear structural guidelines, there is no way to constrain the range of sentences that might be construed as fulfilling a given function. For this reason, I would like to argue for a balanced formal/functional approach to linguistic description, to which I turn in the next section.
139
140
Thomas E. Payne
4. The solution: A balanced formal-functional approach The type of grammatical description that I would like to recommend is one that employs a form-first approach for those areas of grammar that are the most controlled, systematic and rule-dominated, and a function-first approach for those areas that tend to cross-cut structural levels. The controlled, systematic and rule-dominated parts of language are often best approached with an emphasis on elicited data. These would include: 1. Phonology (excluding intonation). 2. Morphophonemics 3. Inventory of derivational morphology (which derivational categories apply to which roots, etc.) 4. Inflectional inventory (determining the range of inflectional possibilities for person and number “agreement”, case marking, etc.) 5. Pronoun inventory (isolating the entire set of free pronouns or pronominal clitics) 6. Lexical inventory (acquiring the words for a large number of culturally significant things and activities) Notice that in elicitation there is an emphasis on obtaining inventories of various expression types. This is because languages typically employ a small number of forms in text, though many more forms are possible. For example, a declarative sentence with a second person subject is very rare in texts. This is because people don’t often inform other people concerning activities of the person spoken to, e.g., You are baking bread. Questions are much more natural in such a context. Nevertheless, a description of the language would be incomplete if the second person declarative forms were missing. Elicitation is essential to the completion of paradigm charts. Often the meaning of a particular morpheme or construction is not clear until the entire range of possibilities that could replace it is identified. Entire paradigms are rarely obtained by inspection of texts. The same observation can be applied to syntactic constructions. For example, whether a particular transitive construction is a passive or an ergative depends at least partially on whether there exists a corresponding “active” construction. Similarly, the precise function of Subject-verb-Object word order may not be apparent until minimal pairs with Verb-subject-object order are obtained. Text data may exhibit other orders, but in examples extracted from texts, there are usually enough other formal differences that the precise contribution of word order to the observed semantic differences is obscured. True minimal pairs are usually obtainable only through elicitation.
A grammar as a communicative act
The more pragmatic, semantic and subtle parts of language are best analyzed via a large body of naturally occuring text, supplemented by elicitation where necessary. These would include: 1. Intonation. 2. Constituent order. 3. Inflectional morphology (determining the precise functions, including tense/ aspect/mode). 4. Voice (alignment of grammatical relations and semantic roles of verbal arguments) 5. Sentence level particles (evidentials, validationals and pragmatic highlighting particles). 6. Clause combining (including relativization, complementation, adverbial clauses and clause chaining) 7. Lexical semantics (determining the nuances associated with various lexical choices, including derivational morphology and pronouns). 8. Pragmatically marked structures, such as clefts, questions, etc.
5. Conclusion If a grammar is a communicative act, it will have several characteristics not possessed by mathematical machines. First, like all communicative acts, its communicative effect will be derived from the interaction of the information presented, and the context available to the audience. Just as an utterance like I have a headache, can have any number of communicative effects, depending on the context in which it is used, so a written grammar will have different effects in the context in which it is used. While authors of any piece of literature, including a grammatical description, cannot possibly envision all the possible communicative effects their work might have, authors do have intended communicative effects in mind. These intended effects will guide the writing of the grammar at various points. The intended communicative effects will depend on the intended contexts in which the grammar will be used. For example, a grammar that is written as an instruction manual for how to speak a language will have a different form than a grammar written as a reference work for linguists. Even two pedagogical (teaching) grammars will be different depending on whether the author envisions the context of use to involve a trained teacher or not. Grammars for native speakers are different than grammars for the international community. Grammars for schoolchildren are different than grammars for university scholars. It is very important to consider the audience and the
141
142
Thomas E. Payne
context of use when one is writing a grammar. A grammar is not an objective “representation” of a language divorced from any particular use. It is a communicative act, performed and interpreted in a context.
References Austin, John L. 1962. How to do things with words. Oxford: Clarendon Press. Bloomfield, Leonard. 1933. Language. New York. Holt, Rinehart and Winston. Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton and Co. Greenberg, Joseph H. 1954. A quantitative approach to the morphological typology of language. International Journal of American Linguistics 26.178–94. Grice, H. Paul. 1968. Utterer’s meaning, sentence-meaning, and word-meaning. Foundations of Language 4. 1–18 Grice, H. Paul. 1975. Logic and conversation. Syntax and semantics 3: speech acts, ed. by Peter Cole and Jerry L. Morgan. 41–58. New York: Academic Press Grice, H. Paul. 1981. Presupposition and conversational implicature. Radical pragmatics, ed. by Peter Cole. 183–98. New York: Academic Press. Langacker, Ronald W. 1987. Foundations of cognitive grammar. Volume 1: theoretical prerequisites. Stanford: Stanford University Press. Nida, Eugene A. 1951. An outline of descriptive syntax. [mimeo. 115 pp.] Glendale, CA: Summer Institute of Linguistics, Inc. Searle, John R. 1970. Speech acts: an essay in the philosophy of language. London: Cambridge University Press. Searle, John R. 1975. Indirect speech acts. Syntax and semantics 3: speech acts, ed. by Peter Cole and Jerry L. Morgan. 59–82. New York: Academic Press. Sperber, Dan and Dierdre Wilson. 1995. Relevance: communication and cognition [2nd Edition]. Oxford: Basil Blackwell Ltd.
A typology of good grammars Keren Rice University of Toronto
Writing a grammar is an exhilarating and exhausting experience. All grammars must meet high expectations. This paper reports on a study of qualities sought in grammars based on a survey of reviews of grammars. The study reveals that linguists expect a grammar to be comprehensive, clear, and accessible to all. It should also contain careful argumentation and a wealth of data that is appropriate, authentic, and meticulously checked. Writers of grammars must ask themselves a variety of questions before they begin, including what the goal of the grammar is, who the audience for the grammar is, how much time there is to write the grammar, and what the language reveals about its grammar. The paper then discusses and evaluates specific examples of good grammars of various types.
1. Introduction When I was invited to participate in the Symposium on Grammatical Description of Undocumented Languages, Tom Payne suggested that I talk about my perspective as an editor, focusing on what makes a grammatical description “sellable.” What does an editor, or reviewer, look for in a grammar? What are the salient characteristics of the best grammars you have reviewed? I responded that I did not think that I could talk about this alone, because I did not think that there was a single template for a good grammar. We corresponded about titles, and ended up with “A typology of good grammars” — “grammar” was fairly obvious to me; “good” was, on first thought, somewhat less so. “Typology” was included because, I believed, there is more than one type of good grammar. This article is an attempt to understand these different terms, especially “good” and “typology,” with respect to grammars. I begin by introducing the reader to my work on grammars. I then examine some of the factors that are, I believe, required of a good grammar based on a study of reviews of grammars. Following this I look at some of the factors that lead to the word “typology” as well as “good” being part of the title of this article.
144
Keren Rice
2. A little background I began doing field work on Slave [slevi], an Athapaskan language of northern Canada, in the early 1970’s, for my Ph.D. dissertation. The following years were a time of great change in the Canadian north, and the federal government was interested in building an oil and natural gas pipeline through much of the area in which the language is spoken. There was tremendous concern about the human and social dimensions of this, and the government responded in several ways, including holding hearings in the communities to gauge community response to the proposed pipeline, and supporting research on the languages and the status of the languages in the area in which the pipeline would be built. I worked under contract, starting in 1976, on the research for what eventually became a grammar of the Slave language. In the first few years I prepared a preliminary grammar of one dialect of Slave and a dictionary of that same dialect. I conducted intensive field work between 1976 and about 1984, working with many speakers in a number of communities. This work culminated in a grammar of Slave, published in 1989. I have been involved in curriculum development, teacher training, and dictionary work on Slave as well over the years, and have continued to do theoretical research on Athapaskan languages, including comparative work. In thinking back about writing the Slave grammar, my memories are at two extremes, exhilaration and exhaustion. Some quotes are perhaps of interest on both of these. First, Valentine, in the introduction to his 2001 Nishnaabemwin Reference Grammar, vividly conveys the excitement and pleasure of writing a grammar: Writing a grammar is a profoundly humbling experience. Languages are almost unbelievably complex and represent the richest traditions that we as human beings possess. What remotely compares with them, whether we are considering the massively intricate and fluid physical gestures involved in the articulation of sounds, or the systems behind the thousands of possibilities of distinct expression in the verb system of a language such as Nishnaabemwin? A language is a natural object with a beauty and a capacity to inspire awe on the order of Niagara Falls or Lake Superior, if we take the time to appreciate it. Writing a reference grammar provides the enjoyment of thousands of hours of careful scrutiny, though at the same time one realizes acutely the truth of Michael Krauss’s statement that a hundred linguists working for a hundred years could never get to the bottom of a single language. Nishnaabemwin is a language exceedingly rich in structure, inviting many levels of analysis — it is an inexhaustible source of pleasure and challenge for its students. (Valentine 2001:xxxi)
In a recent book on dictionaries, Frawley, Hill, and Munro 2002 begin their introduction with what they term lexicographic war stories. Some of these war stories
A typology of good grammars
are as relevant to the writing of grammars as they are to dictionaries, and get at some of the sometimes overwhelming difficulties in writing a dictionary. • how difficult it was to know where to begin; • how, after finally beginning, the work was both glacially incremental and wildly circuitous; • how the project went on and on and the book kept expanding from a modest list of words and glosses to something like a cultural encyclopedia what must be the lexicographic version of Congress’s feared “mission creep”; • how some things turned out to be absolutely unachievable, no matter how clever the lexicographer; • how, for some at least, the project had back-burner status because those who promoted and rewarded employees wanted more short-term and ostensibly theoretical results; • how, when the work was finally completed and published, some people were unimpressed, and even mean-spirited and critical “What?! You forgot to include … ???”
While grammar writing and dictionary writing are different in many ways, many of the statements above reverberate with the writer of a grammar as much as they do with the compiler of a dictionary — the difficulty of finding a starting point, the length of the project, the inability to get the material wanted, the apparent lack of short-term progress, the near inevitability of errors, the sense of how much is not there rather than how much is there. Frawley, Hill, and Munro go on to ask “Why do it?” They respond to this question about lexicography with something similar to what Valentine had to say about writing the Nishnaabemwin grammar: “There is something at once both marvelous and practical about producing a guide to the mind, world, and behavior of a group of people. The benefits that accrue from such a handbook — literacy, preservation, history, discovery — only add to the excitement of seeing the published dictionary standing upright on the bookshelf.” (2) All this is to say that, despite the difficulties one faces in writing a grammar, the experience in the end of producing a piece of work that, one hopes, will be of long-lasting value is one that cannot be replaced.
3. On the “grammar” and “good” parts of the title With this background, let me now turn to the task at hand, grammars and the essence of what makes a grammar good. This can be considered from at least two different perspectives. First, what does the grammar include? What topics must be covered in a good grammar, what is optional? This is a question addressed by
145
146
Keren Rice
Noonan in his article in this volume, and I do not concern myself with this to any large extent. Second, how is the grammar presented? What are those more elusive qualities that go beyond coverage of the language that are required to make a grammar stand out? This is the question with which I am concerned here. It is surprisingly difficult to find anything explicit in print about this when one looks at what strikes me as a natural place, information sheets about grammars for potential authors submitting to grammar series. Instead, we generally find much written about what goes into a grammar. These descriptions assume that we know what overall makes a good grammar, and the author requires guidance as to scope and organization, but not to quality. In the following discussion, I take my turn at trying to define what qualities a good grammar has based partly on a study of reviews of grammars, partly on what was required in the Cambridge University Press Grammar Series that R.M.W. Dixon and I edited, and partly on my own observations. 3.1 On the “grammar” part of “a good grammar” As noted above, Noonan’s article in this volume deals extensively with the content part of a good grammar, and I examine this topic only cursorily. Content is, of course, critical to a good grammar; it is simply a secondary concern to me here. A dictionary (in this case The Random House Dictionary of the English Language, 1973 edition) gives a number of definitions of “grammar.” Most relevant is the following; “1. the study of the system underlying the esp. formal features of a language, as the sounds, morphemes, words, or sentences” (614). This definition is, unsurprisingly, not particularly useful in helping us to understand what makes a good grammar (in the sense of a book that contains this information), but rather tells us, in the broadest of strokes, what goes into a grammar. In looking at descriptions of some grammar series, we find something similar: descriptions that focus on content. In a description of the LINCOM EUROPA Descriptive Grammar Series (Linguist List May 1999), information is included about what the grammars should include (phonology/phonetics, morphology, syntax, texts), the number of pages (120), the readership (scholars in general linguistics, especially in typology, students of the language (family)), special concerns (endangered languages, un-surveyed languages, scattered material). These guidelines provide information about what might be included in the grammar and who the audience is. In the announcement of a major grammar series, the Lingua series, later the Routledge Descriptive Grammar series, “need for the production of purely descriptive studies which will provide a wide variety of data about individual languages”
A typology of good grammars
(Comrie and Smith 1977: 5) is stressed. This introduction goes on to say that “the ultimate aim is a framework which would act as a catalyst in the elicitation of all information that could be of interest for theoretical work in the above-mentioned fields [language universals, language typology, or comparative syntax, morphology, or phonology] in terms of any of the modern theories of language” (5). A detailed questionnaire is included, and its topics are syntax (sentence types, structural questions, coordination, negation, anaphora, reflexives, reciprocals, comparison, equatives, possession, emphasis, topic, heavy shift, other movement processes, minor sentences types, operational definitions), morphology (inflection, derivation), phonology (phonological units [segments], phonotactics, suprasegmentals, morphophonology [segmental], morphophonemics [suprasegmental], ideophones and interjections, lexicon. In this introduction, Comrie and Smith focus on aspects of what “good” means beyond content: They say that a grammar must meet “conditions of accessibility, flexibility, and completeness,” recognizing that “with the passage of time we should expect to have to modify several of the assumptions behind the questionnaire in light of fresh data” (page 5). They also require that the grammar be “amply illustrated by examples from the language, with glosses” (page 9). They thus move from the topics to be covered in a grammar to those other factors that contribute to quality. In a survey carried out by a group at Munich several years ago (Linguist List May 1995; summary of results reported October 1995), linguists were asked for their responses as to what a good reference grammar should be. Questions included what a good reference grammar should contain, the role of linguistic theory in a grammar, the kind of justification necessary in a grammar (exemplification, theoretical support, cross-linguistic support), organization of the grammar, and a range of other topics. Their findings are summarized here: A good reference grammar should contain a grammar and a lexicon, not just a grammar, with emphasis on morphology and syntax, and strong examples and texts. A description of language variation is important. A good reference grammar should be descriptive and provide synchronic and diachronic, but primarily synchronic, information; it should not be bound to a particular theory; form and function should be described together; terminology should be suitable to the language described. The preferred order is phonetics/phonology — orthography — morphology — syntax — semantics — pragmatics. While these guidelines are helpful, elaborating the topics that must, should, or could be included in a grammar, they do not go too far beyond this to answer the question of “What are the characteristics of a ‘good’ grammar beyond its content?”
147
148
Keren Rice
3.2 On the “good” part of the title In order to try to identify just what characterizes a good grammar in the sense that I am using this phrase, I decided to undertake a study of reviews of grammars, hoping that they would be helpful in defining what I was thinking about. This was very informative, and I found a remarkable consistency among the reviews as to what makes a grammar work and what does not. In the following, I quote passages from several reviews, chosen randomly, and drawn largely from IJAL, Linguistic Anthropology, and Lingua. While I tried to work from positive reviews, I have include some quotes from negative ones as well as these are often especially revealing of what is required of a good grammar. I have highlighted phrases that point to the qualities that the reviewers found to be particularly important in the grammars. This series of quotes goes on for some pages, and is followed by a summary and discussion. 3.2.1 A survey of reviews of grammars Review of István Kenesi, Robert M. Vago and Anna Fenyvesi. Hungarian. A. Lipták. Lingua. 112.10. 849–854. (Routledge Descriptive Grammar Series) Easy access reference book for the grammar of Hungarian, but it is also more than that: it provides the empirical basis of specific grammatical phenomena like overt movement of contrastive focus items … (854)
Review of Ho-Min Sohn. The Korean language. S.J. Rhee, J. Van de Weijer. Lingua 112.6 481–488. (Cambridge Language Series) To conclude, this book provides a wealth of data and ideas which are presented with unsurpassed clarity. No linguist will want to write anything about the Korean language without consulting this excellent contribution. Moreover, the book is also wonderfully accessible to the interested layman, who will find illuminating discussion and copious illustration of a wide range of interesting topics which are highly relevant for current phonological, syntactic and morphological theory. More languages deserve a book like this. (486)
Review of R.E. Asher and T.C. Kumari. Malayalam. P. Madhavan. Lingua 111.8. 629–638. (Routledge Descriptive Grammar Series) There are plenty of data covering various aspects of the structure and use of the language, enriched further by the thoughtful inclusion of subtle dialect differences. … Profuse cross-referencing further enhances the usefulness of the data. (631) Now I turn to those aspects that I consider as the limitations of the work. As I said earlier, this is meant to serve as an all-in-one source book for any (potential)
A typology of good grammars
researcher. It then is a little ironic that by predetermining the format in minute detail (…) some of the really interesting aspects (from the researcher’s point of view) of the language tend to get left out. This, I think, is an inherent problem for a descriptive grammar. Let me explain this. AK have been scrupulously noncommittal to any particular theory of grammar, their purpose being just that of description. And there lies the rub. There is no atheoretical or theory-neutral research: it amounts to a contradiction in terms. Be it linguistics or any other scientific discipline, research problems are invariably determined by the paradigm one is working is. Take for the sake of illustration, someone working in the framework of generative syntax. Immediately the gaps begin to show. There is a section in the book on quantifiers, where it is noted that the universal quantifier ellaa always has a particle um with it, contiguous or otherwise. But this important piece of information is given in passing, without connecting it to other facts … (631)
Review of Philip Davis and Ross Saunders. A grammar of Bella Coola and Jan van Eijk. The Lillooet language: phonology, morphology, syntax. David Beck. IJAL. 67.1. 2001. 88–92. In its own way, each of these works has had to strike an uneasy balance between the need for comprehensive, detailed, and accessible documentation of a language and the need for theoretically relevant and original linguistic analysis. (88)
Review of Brent Galloway. A grammar of Upriver Halkomelem. Donna B. Gerdts. IJAL 67.3. 2001. 354–356. … the authenticity and robustness of the data make this a very significant and lasting contribution to the description of the world’s languages. … Galloway has not only provided the linguistic community with a detailed description, but he has also given the Sta:lo people the reference material they need to proceed with language maintenance and revitalization projects. His research provides a model for other linguists undertaking the crucial work of documenting endangered languages. (356)
Review of Jean Ormsbee Charney. A grammar of Comanche. Jon P. Dayley. 1996. IJAL. 62.4. 426–428. The prose is clear and the discussion is especially detailed where difficult or problematic aspects arise. Throughout the monograph, each aspect of the language under discussion is amply illustrated with many sentence examples. (428)
Review of Laurence C. Thompson and M. Terry Thompson. The Thompson language. Steven M. Egesdal. 1996. IJAL. 62.1. 120–124. It is remarkable for the great amount of data it presents, and even more remarkable for how easily a user can interpret that data. Their descriptive prose throughout is refreshingly clear and succinct — taut as a bowstring, crisp as a rifle shot. (120)
149
150
Keren Rice
The Thompson Language might be characterized fairly as a “field guide” to NSe‘kepmxcín. It began as a tool for fieldwork and was refined continually over more than a dozen summers in the field. That evolution explains in part its meticulous organization: the Thompsons wanted to be able to find information in it quickly. Related topics were cross-referenced, ever more faithfully and completely over the years, until the present grammar has within it an exceptional concordance of information. Adding to the internal concordance are several other welcome means with which to access the data: a precise table of contents …, lists of abbreviation, … a detailed subject index. (120–121)
Review of Maurice L. Zigmond, Curtis G. Booth, and Pamela Munro. Kawaiisu: A grammar and dictionary with texts. Hansjakob Seiler. IJAL 59.3. 1993. 355–357. It has become fashionable these days to raise one’s voice in favor of so-called endangered languages. The best, it seems to me, that a linguist can do for the preservation of a language approaching extinction is to come up with a description as complete and as informative as possible… (357)
Review of Kashi Wali and Omkar N. Koul. Kashmiri: A cognitive-descriptive grammar. Alice Davison. Anthropological Linguistics. 42.3 2000. 414–417. In the exposition of the syntax, each point of grammar is well supported by a clear example sentence, and morphological categories are presented with short but helpful explanations. (414) The description of the language is well supported by clear, appropriate examples… (416)
Review of Laurence C. Thompson and M Terry Thompson. The Thompson language. Paul D. Kroeber. Anthropological Linguistics. 37.3. 1995. 387–391 The description detailed, and copiously exemplified, includes a good many intriguing remarks on the semantics of forms — inevitably at this stage of the investigation — those remarks are often best taken as suggestive rather than definitive. (388) All these complexities are lucidly detailed. (389)
Review of Jan van Eijk. The Lillooet language. Phonology, morphology, syntax. Paul D. Kroeber. Anthropological Linguistics. 40.4. 1998. 642–644. The book reviewed here does an excellent job of describing Lillooet grammar in a compact, informative, and intelligible manner. (642) Though van Eijk does not waste words, the facts of Lillooet generally emerge in a straightforward and perspicuous fashion. But occasionally the presentation is a bit too compressed, requiring an extra degree of alertness on the reader’s part. (644)
A typology of good grammars
Review of Shahrzad Mahootian, with Lewis Gebhart. Persian. Jila Ghomeshi. Anthropological Linguistics. 41.4. 1999. 574–577 In general, each section contains a pithy, yet clear, statement, illustrated with two or three relevant examples about the construction or phenomenon described by the title of that section. (575) … Such an oversight, while not too misleading for the careful reader, is reflective of a more general tendency on the part of Mahootian to tailor her descriptions around the examples she provides rather than giving more general information … (575) A second problem — one that will arise in writing any grammar — is that decisions must be made as to the status of certain morphemes, words, constructions, etc. However, Mahootian has not included any discussion of how these decisions have been reached or what the areas of contention are. (575)
Review of G. Tucker Childs. A grammar of Kisi: A Southern Atlantic language. W.A.A. Wilson. Anthropological Linguistics. 39.2 1997. 335–336. Every detail of its contents, from the introduction and the phonology to the chapter on embedding, is most painstakingly described — clearly the fruit of much meticulous analysis — yet the work is not always easy to use. (336)
Review of Philip Davis and Ross Saunders. Bella Coola. Lisa Matthewson. Anthropological Linguistics. 41.1. 1999. 120–125. Unfortunately, the introduction is marred by an opaque writing style. As with most parts of this book, I found that two readings were necessary before the points were fully understood. Clarity is further reduced by the fact that several of the technical terms introduced are not well defined. (120) Throughout this work, one is struck by the excellence of the authors’ fieldwork. Care is taken to situate utterances within stretches of discourse… (123) There are a number of factors that make it difficult for the reader to extract the core generalizations proposed in this book. One problem is the large number of typographical errors, some of which involve matters of content. (123)
Review of Veronica Du Feu. Rapanui. Robert Weber and Nancy Weber. Anthropological Linguistics. 41.1. 1999. 142–146. Du Feu does not give the reader a clear presentation of how Rapa Nui handles the grammatical feature under discussion before citing confusing details. For example, Section 1.1.1.1 (p. 14) addresses the question of direct versus indirect speech. It would seem logical that one would explain how each of these kinds of sentences is formed in the language before discussing in what ways they differ, if they do. Instead, since the questionnaire specifically asks about differences between the
151
152
Keren Rice
two, Du Feu enters right into a discussion of details relating to differences between these kinds of sentences before we are even told what they look like. (143) Throughout the book, the spelling of words is inconsistent, as is word division, and glossing. (145)
3.2.2 What makes a grammar “good”? These reviews of grammars demonstrate that a good grammar is much more than a discussion of the phonetics, phonology, morphology, syntax, and so on of a language. These topics provide the content of the grammar, and thus are absolutely necessary to the “grammar” part of the good grammar, but in and of themselves they are not sufficient to make a quality grammar. As mentioned earlier, Noonan, in his contribution to this volume, discusses many of the topics that should be included in a good grammar, and I do not examine these further here. Instead, I focus on these other more intangible qualities that go to make a grammar good. I have highlighted in the reviews some of the major points that, I believe, help us to understand what the essence of a good grammar is. The reviews are of a range of types of languages — well-studied languages, languages with little else written on them but from well-studied language families, languages with little else written on either the particular language or the language family. We find several notions echoing throughout the reviews that help us know what the essence of a good grammar is, and I discuss these now. All are important; the list is not meant to be ranked. First, a good grammar is comprehensive and complete. The reviewers stress that a good grammar includes at least the traditional areas of phonetics, phonology, morphology, and syntax. Again, see Noonan’s article in this volume on this point. Second, the writing style must be clear. Third, examples are of the utmost importance. There should be a wealth of data which is appropriate and authentic. The data should be carefully chosen to illustrate the point at hand. In addition, the data should be meticulously checked for spelling, and for consistency of spelling, glosses, and manner of glossing throughout the book. Fourth, the reviewers agree that accessibility is very important. Accessibility is related to clarity of writing, but has a slightly different focus; the grammar should be easy to use and reach out to the reader. Another way of putting this is that the author of the grammar should place him/herself in the reader’s shoes, attempting to anticipate what the reader needs to understand the grammar. Accessibility is also related to organization: a grammar should have a user-friendly organization which is appropriate for the language.
A typology of good grammars
Argumentation is particularly important in a grammar in order to establish the credibility of proposed analyses. This too needs to be clear, accessible, coherent, and cohesive. In the guidelines that R.M.W. Dixon and I prepared for the Cambridge Grammatical Description Series, we set out some general statements that encompass many of these points. I have listed these general statements below, and have bolded pieces of the description in order to focus on the points relevant to the current discussion. Writing a good descriptive grammar is a difficult task, demanding a sound theoretical background, the ability to pursue a well-founded chain of argumentation, the capacity to recognize an analytic problem and to put forward and compare competing solutions to it, and then to present an integrated analysis of the language. 1. Style A grammar is a piece of scientific description. As with any type of writing, there are different ways to present the analysis. One is to state the facts in a dry manner, so that looking up something in the grammar is like using a dictionary or a cookbook. Another alternative is to think of writing a grammar as telling a story, gradually developing a theme, moving in a natural and integrated manner into the next theme, and so on. A good grammar conveys to the reader some of the excitement the linguist experienced while analyzing the language. Thus, part of the goal in writing a grammar is to make it user-friendly. 2. Organization Within a topic, the presentation should generally begin with simpler patterns and move to more complex patterns. For example, it is usually better to begin the discussion of syntax with simplex sentences than with complex sentences. 3. Examples Each point in the grammar should be illustrated with appropriate examples. … Examples should be taken from texts, …, whenever possible. Length and completeness The grammar should be comprehensive, and should attempt to describe every aspect of the structural organization of a language. It must be kept in mind that there is no such thing as an exhaustive grammar of any language, and not every question about a language can be answered within a grammar. It is good to indicate clearly what can be stated with confidence, those areas about which there is some doubt, and points that are not at present understood. Open questions which require study should be listed. The grammar should be as concise as possible while describing and explaining everything in a clear and accessible manner.
153
154
Keren Rice
The guidelines for grammar series and the reviews of grammars help to elucidate what it is that editors and readers look for in a grammar: clarity, comprehensiveness, conciseness, care, coherence, accessibility, richness of illustration. Basically, what is required of a good grammar is what is required of any good piece of linguistic work: careful and insightful analysis that is well-supported by strong argument and presented in a way that is easy to grasp and can draw in the audience at which the grammar is aimed. 3.2.3 Some additional criteria In this section I raise a few points that do not come through as clearly in the reviews but which are, I believe, nevertheless prerequisites to a good grammar. A few of the reviews mention the importance of discussing language variation. Anyone doing fieldwork quickly learns how much variation there is in speech, both in an individual and in a speech community. Perhaps in an attempt to uncover the systematicity in a language system, it is often tempting to overlook this variation in writing a grammar, and to amalgamate material from different dialects. While this might be legitimate at times, it can lead to the creation of a dialect that is not spoken by anyone. As an illustration, a Ph.D. student in my department who is a native speaker of a particular language listened to a talk on aspects of the phonological system of that language. It turned out that the presenter was, unknowingly, conflating data from two dialects and the particular forms in question could not co-occur within a single dialect. The analysis then, was invalid, as it was based on something that was not in fact a possible dialect. The available materials led to this problem: they combined different dialects and the speaker simply drew from this material. If data comes from a number of dialects, it is incumbent on the author to make clear what the differences between these dialects are so that the user of the grammar does not fall into the trap of creating incoherent language systems. The use of formalism and technical terminology also benefits by some comment. It is very easy in writing a grammar to get caught up in some particular formalism that is current at the time. Consider, for instance, grammars written in tagmemics. I personally find these to be extremely difficult to read because I am not very familiar with the theory. Similarly, while I was trained in generative grammar, I find that grammars that make heavy use of phonological rule formalism (not just slash/dash rule notation, but angled brackets, curly brackets, and the like) often obscure for me something that is really quite straightforward. Thus, in addition to aiming at clarity, conciseness, and the other characteristics discussed above, one should be quite careful about the use of formalism. Things should be said in words as well as in formalism. Just as data must be checked over and over
A typology of good grammars
for accuracy, formalism must be carefully examined to be sure it really is capturing the generalizations that it is designed to capture. Something similar can be said about terminology. Linguistics, like any discipline, uses a considerable amount of technical terminology. The way in which a term is used can change over time, the same term can mean different things in different language families, terms are sometimes invented to define social groups rather than to illuminate what is going on in a language. It is very important to take tremendous care in using terminology and, in general, to define what is meant by a particular term. For instance, in the Athapaskan literature, both the terms “aspect” and “mode” are used, but sometimes aspect refers to A and mode to B and sometimes vice versa. In writing the Slave grammar, I included a chapter in which I tried to define these terms as I would use them so that readers would be clear what was under discussion when they saw one of these words. Obscure words also need to be defined. For instance, in the Athapaskan literature, the term “conjunct” is used to identify morphemes that are closer to the verb stem, draw from a limited phonological inventory, are often syncretic in nature, and are, by and large, inflectional in function. In the Algonquian literature, this same word is used to identify a type of clause that is, for the most part, subordinate to a main clause. Reading the literature of these two language families is extremely difficult for a newcomer to the field, and this is partly because of the use of terminology which is unfamiliar to the reader. These comments clearly fit under the notion of accessibility. 3.3 Summary To summarize, I have tried in this section to identify some of those sometimes elusive qualities that define what it means for a grammar to be “good” — a clear style; completeness; comprehensiveness; wealth of appropriate, carefully chosen data that is carefully explained; compelling argumentation, and consistency. In addition, two even less tangible points are critical. First, to speak anthropomorphically, each language has a story to tell, and it must be allowed to tell it. Second, a grammar is a work that, one hopes, will be used for years to come. A grammar is written for the present, but it also should be written for the future. This means thinking carefully about formalisms and technical terminology and, perhaps most important, about making a presentation that will be understood long after that grammar has been published. The writer of a grammar thus must in many ways be a visionary, able to see far beyond present times. Writing a grammar presents an enormous challenge to the author, but one that can be met, as many of the outstanding grammars written in the past demonstrate.
155
156
Keren Rice
4. What is a “typology” of good grammars? Or how can good grammars differ? 4.1 A series of questions In the previous section, I discussed a number of characteristics of a good grammar. In this section, I turn to the word “typology” and consider some of the ways in which these ingredients of a good grammar can be stirred together and what some of the questions are that enter in to deciding just what the grammar should look like. I begin with a series of questions, put here as direct questions. – – – –
What are your goals in writing a grammar? Who is your audience? How much time do you have to write the grammar? What does the language “say” to you about the grammar?
These are some of the questions that I asked myself through the process of working on the Slave grammar. The set of answers very much shape the form that the grammar ultimately takes. What are your goals in writing a grammar? While every grammar is designed to provide a grammatical description of the language in question, this does not on its own define what the goals are. Is the goal to write a full reference grammar? Is it to write a short grammar to give the flavor of the language without the depth of a full reference grammar — for instance, for the LINCOM EUROPA series? Is it a sketch? Is the goal of the grammar to present the language on its own? Is it designed to be a grammar that easily allows for comparative work between languages of the family? Is it meant to be a grammar to be used for work on linguistic typology, as the grammars of the Routledge series are? Is it designed primarily to further linguistic theory? Who is your audience? To put this another way, what is the shared world of the desired audience? This question relates closely to the one about goals. Is the grammar written for linguists? If so, is it designed mainly for linguists who are experts in the language family? Is it for linguists who might wish to work in the language family, but who do not as yet have any knowledge of the family? Is it for typologists? Is it for speakers of the language? Is it for those wishing to learn the language? For teachers of the language? Very often, grammars are written for a community of speaker/learners of the language. A grammar designed primarily for speakers/learners of the language is not necessarily the same as the grammar designed primarily for linguists, as Callaghan 2002 points out in her discussion of Miwok dictionaries, and as I have found in my own work experience. Depending on the audience, one might use different vocabulary, present different types of
A typology of good grammars
accounts, organize the material in different ways, and so on. It can be an enormous challenge to try to meet the needs of all potential audiences in a single volume. In reality, a grammar of a little-documented language is likely to be used for the full range of purposes. Yet authors should have in mind why and for whom they are preparing the grammar, as this will help with a number of critical decisions. How much time do you have? This will help to determine what your goals are — something very different can be accomplished if you are doing six months of field work or six years of field work, especially on a language for which there has been little or no description. It is important to set one’s sights realistically. It is probably impossible to write a major reference grammar on a language which has not been described based on six months of field work. However, something can be written in this amount of time. It is important to let the reader know what the framework is so that s/he can know how to interpret and evaluate the grammar. What does the language say to you about its grammar? This is the overall question that I consider in the remainder of this article. I will focus primarily on grammars that are designed as reference grammars and are based on in-depth research of a language over some time period. The audience in mind is largely one of linguists. The details vary depending on the goals and the audience, but these questions are, I believe, ones to take into account no matter what. 4.2 How can grammars differ? 4.2.1 Content 1: The major topics First consider the content of a reference grammar. There is little disagreement that a grammar should include information on phonetics, phonology, morphology, syntax, and semantics. More and more, discourse structure is also being discussed in grammars. In general, guidelines for grammars ask for some textual material and a lexicon. Within these general guidelines, however, there is room for much variation. Again, see Noonan’s article in this volume for detailed discussion. 4.2.2 Organization What can be variable about organization? Essentially, assuming that the goal is to write a reference grammar of the language rather than a grammar to fit some predetermined outline, as with the Routledge Descriptive Grammar Series, each language demands its own strategy of presentation. To quote again from the guidelines for the Cambridge Grammar Series: …, compare Dixon’s grammars of Dyirbal [(1972)], Yidiny [(1977)], and Boumaa Fijian [(1988)], where three totally different strategies are employed. Basically, if an analytic decision concerning category X needs to refer to some facts
157
158
Keren Rice
concerning category Y, then the chapter dealing with Y should be positioned before that dealing with X.
The Dyirbal grammar (Dixon 1972), begins with chapters on Australian languages and the Dyirbal language and its speakers. Dixon then looks at word classes, syntax, deep syntax, morphology, phonology, semantics, lexicon, and prehistory; he has appendices on Dyirbal logic, previous work on Dyirbal; he also includes texts, vocabulary, a list of affixes, references, and an index of Australian languages. The grammar of Boumaa Fijian (Dixon 1988) is ordered in a different way and divided into many more chapters. In this grammar, Dixon again begins with an introduction, and then turns immediately to phonology and the word. There are two major overview chapters dealing with syntax — the first on clauses and phrase structure and the second on verbs. These are followed by chapters on pronouns, deictics, predicates, noun phrases, possession, clausal NPs, classifiers, numbers prepositions, time expressions, interrogatives, word derivations, verbs, adjectives, word classes, clause, sentence, complement clauses, the relator me, syntactic organization, previous work on Fijian, texts, references, and vocabulary. While the grammars include similar topics, Dixon organized the books very differently because he felt that the languages demanded different types of treatment. Another example comes from the domain of grammars of Athapaskan languages. There have been remarkably few grammars of Athapaskan languages in recent years; dictionary work is far more prevalent. One grammar of an Athapaskan language is Eung-Do Cook’s grammar of Tsuut’ina (Sarcee). Cook has a discussion of the orthographic system and sound-symbol correspondences near the beginning of the grammar, but leaves the discussion of morphophonemics until the end. Cook (1984:5) is quite explicit about why he does this: “The organization of this book is different from many comparable linguistic descriptions written in a structural framework in that it starts with syntax, proceeds to morphology, and then to phonology. This procedure does not necessarily reflect any theoretical bias, although I believe it would be only natural from the native speaker’s point of view.” In my grammar of Slave, on the other hand, I follow a more traditional path, presenting an overview of major phonological processes near the beginning of the grammar, and then interleaving discussion of morphophonemics into the chapters of grammatical categories, as it is appropriate. The point that I wish to make here is that the language should be presented in such a way that its own character comes out and the needs of its audience are met. However, there is not a single way to accomplish this task. Different linguists may have different ideas about what this is for a particular language, and different goals and different audiences may lead to variation in how a grammar is organized. Thus we will find variation across grammars, even of the same language.
A typology of good grammars
While the overall organization of a grammar will vary from language to language, within a particular topic, the presentation should generally begin with simpler patterns and move to more complex patterns. For instance, in Algonquian languages verbs are normally divided into two major classes, intransitive and transitive, and then each of these classes subdivides into animacy classes. While person marking is quite straightforward with intransitive verbs, the role of a person hierarchy becomes evident when the transitive classes are studied. It is probably a good idea to discuss intransitive verbs before discussing transitive verbs for this reason. In some more transparent cases, one would normally present simplex sentences before introducing complex sentences; affirmative constructions before negative constructions; the sound system as a whole before positional restrictions on where segments can occur. 4.2.3 Content 2: Some more details While grammars are generally thought to include the topics of phonetics, phonology, morphology, syntax, and semantics, there are other topics that they could include. A major area that is found in some grammars is comparative/historical information. Should a grammar include diachronic discussion in addition to the synchronic description of the language? In the Cambridge University Press grammar guidelines, the following is suggested: While the primary purpose of a grammar is to provide a synchronic description of a language, diachronic information should be introduced when appropriate. If enough is known or has been reliably reconstructed of the history of the language, it is useful to provide diachronic explanations for synchronic irregularities, homonymies, and the like. Comparative remarks on what happens in closely genetically related or geographically contiguous languages may also be helpful and instructive.
In writing the Slave grammar, when confronted with problems of page limits, I decided to take out much of the comparative/historical material that I had originally included in it because it seemed to me that this information was secondary to the synchronic description of the language. In some cases, however, it is probably close to impossible to write a grammar without taking historical information into account (e.g., Michif, which is traditionally thought of as being a mixture of Cree and French). 4.2.4 Theoretical perspective There is general agreement that if a grammar is written to be of lasting value, it is important that it not be written in the linguistic framework that is of currency at the time, but that it be presented in what Dixon calls “basic linguistic theory.” As the Cambridge University Press guidelines say,
159
160
Keren Rice
The language analysis should be informed by insights and techniques of linguistic theory. However, the write-up should be in terms of basic linguistic theory, using the terminology that is used and understood by the majority of linguists, and that is relevant for the language … If it is necessary to use an unusual term, it must be defined. A grammar should be a statement about a language, not an argument for a particular theoretical framework.
Two issues come up in this quote. First, the grammar should be informed by theory. This will help make it coherent, and it will allow questions to be asked that might not come up otherwise. But second, theory is not the goal of a grammar. Another work can be written that directly addresses interesting and current theoretical issues, but the role of the grammar is not, say, to critique whether Optimality Theory is superior to derivational theory or to decide how the minimalist account of question formation requires enhancement based on the facts of this language. These are important and valid areas of linguistic research, but they are not the content of a grammar, although they could well inform the research. I will give a few simple examples of what I mean by “informed by the theory” through a comparison of some works on Athapaskan languages. The first comes from what is sometimes called conjugation and sometimes called aspect in the Athapaskan literature. In Athapaskan languages, there are morphemes that are usually identified as n-, s-, and gamma (gh-) based on their reconstructed forms. In my 1989 Slave grammar, I treated these as what I called conjugation markers. The idea was that one of these is listed as part of the basic lexical entry of each verb, and is simply memorized as part of that verb. Between the time of the Slave grammar and my more recent work on word formation in the verb of Athapaskan languages, I learned a lot about aspectual systems, and came to view these same elements in a very different way. In my 2000 book, I treated these as contentful items that I called, following Carlota Smith (who in turn followed Vendler and others) “situation aspect,” marking accomplishments, activities, achievements, and semelfactives. While these items can be lexicalized, overall what I found is that there is a remarkable consistency in aspectual semantics between the verbs that take a particular one of these morphemes. If I were to redo the Slave grammar now, I would treat these morphemes as having meaningful properties overall, with some idiosyncrasy, rather than treating them as entirely idiosyncratic, as I did in 1989. Additional knowledge about language and current theoretical concerns around aspect systems allowed me to view these morphemes in a way that I had been unable to in the past. Another issue that arises in Athapaskan linguistics has to do with the treatment of what is at least historically a concatenation of morphemes before the verb stem. When one looks at many grammars of Athapaskan languages, one finds the verb
A typology of good grammars
described as consisting of a template. For the discussion here, the important part of this template is as follows (I use the terminology in Rice 2000).
situation aspect — viewpoint aspect — 1/2 subject — voice/valence — stem
My concern here is the status of the situation aspect through voice/valence portion of the verb. Some typical forms in the imperfective, taken from Navajo (Faltz 1998), are presented below. I follow Faltz’s terminology here in naming the paradigm: Imperfective-mode conjugation (216) sg pl 1 sh iid 2 ni oh 3 zero For instance, in what Faltz calls the Imperfective-mode conjugation, the form of the second person singular subject is ni. n-Imperfective conjugation (416) sg pl 1 nish niid 2 ní noh 3 form is variable (Ø, í, ee, yí) depending on what precedes it In the n-Imperfective conjugation, the second person singular subject has the form ní, differing from its form in the imperfective-mode conjugation by the presence of a high tone. s-Imperfective conjugation (417) sg pl 1 shish siid 2 sí soh 3 zero Here the second person singular has the form sí, without the nasal component seen in earlier forms. How are these patterns to be treated in a grammar? In Rice 1989 I treated the parallel Slave forms of the n-imperfective and s-imperfective conjugations as bimorphemic. The first segment, n- or s-, I called a conjugation marker and the consonant the subject marker (sh-1sg, iid 1pl, ni- or high tone 2sg. oh- 2pl., with other complications, and ignoring the vowels). While I considered these sequences to be morphologically complex, I did not try to assign any meaning to the first part, but identified it as a conjugation class marker, as discussed above.
161
162
Keren Rice
In Rice 2000, I treated the first part of these spans as a marker of situation aspect, with n- marking an achievement and s- an accomplishment. This makes semantic sense in that n- occurs with verbs that occur at a point in time (e.g., arrive) and s- with verbs that have duration and an endpoint (e.g., make sg. object). A very different treatment is offered by Faltz in his book The Navajo verb, A grammar for students and scholars. The morphophonemics of this span of the verb is complex. This can be seen in the imperfective forms above, but becomes even more clear in the perfective paradigms. Here we see the morphemes called classifiers, labeled voice/valence above, entering in to play a role in the form that this span has. s-Perfective conjugation (417) zero/barred-l classifiers plain-l/d classifiers sg dpl sg dpl 1 sé siid sis siid 2 síní soo síní sooh 3 z or s s Notice here that the form of the aspect–subject complex differs depending on which of the classifiers is present. This is true in what Faltz calls the s-perfective conjugation, given above, and in what he labels the n-perfective conjugation as well, shown below. n-Perfective conjugation (418) zero/barred-l classifiers plain-l/d classifiers sg dpl sg dpl 1 ní niid iish niid 2 v´íni/ííní/yíní oo v´íní/ííní/yíní nooh 3 ní depends on what precedes Here we see two things. First, segmentation of the complex into morphemes is more difficult, especially in the n-perfective forms where an n- is not always present as the first element. Second, the first person singular forms in particular look quite different depending upon the voice/valence marker in the verb: with imperfective forms and plain-l/d classifier perfective forms, the first person singular is easily identified as a voiceless sibilant, but this phonological unit is not even present in the zero/barred-l classifier perfective forms. Something similar can be said for the second person dual/plural forms. It is much more difficult to segment these chunks and maintain a constant correspondence between sound and meaning. Faltz thus treats these units as single chunks, without worrying about the number
A typology of good grammars
of morphemes that are involved, and does not try to break them up into component parts. These various ways of approaching the material are all reasonable, and it is difficult to dismiss one as opposed to the other. In 1989, I took a very traditional approach to Athapaskan verbal morphology, and simultaneously decided to employ a basic principle of one meaning/one form, thus building much that I later came to view as historical rather than synchronic into my synchronic grammar. This was highly influenced by the generative phonology framework prevalent at the time. Further, I decided that a principle of recognizing recurrent partials was important, but did not really look at meaning. Later, I recognized a meaning component; thus the change in terminology from conjugation to situation aspect (and from classifier to voice/valence). Further, I decided that I had gone overboard in trying to maintain the principle of one meaning/one form, and I would now introduce allomorphy to account for the different forms of the first person singular and second person dual/plural. This is done in a different theoretical context for looking at the relationship between phonology and morphology, one in which a theory of allomorphy has found a role. Faltz makes explicit his reasons for choosing the descriptive framework that he did; he writes that “I realized that the mode inflections would be more lucidly described taking an approach familiar from European languages, namely, by defining sets of different subject prefixes for different modes. Thus, the reader will find no “mode markers” in this book — rather the combinations of the classical mode markers with the classical subject prefixes are themselves viewed as the synchronic subject prefixes which differ, therefore, not only from mode to mode but also from conjugation type to conjugation type” (xi). Faltz goes on to suggest that the subject marking systems for the various modes become easier to recognize, that formal properties of the inflectional system emerge in a natural way, and the dependence between perfective mode subject prefixes and the classifier can be seen as an instance of the kind of familiar relation exemplified by the dependence of imperfect subject suffixes on conjugation class in Spanish, or the dependence of future subject suffixes on conjugation class in Latin and Irish (xi). To return to what I would do now if I were to revise the Slave grammar, I would use a combination of these approaches. First, I would treat the n- and s- in the verbs above as a separate component, as I believe that this leads to insight into the aspectual richness of the verb. Second, I would treat the subjects as having special allomorphs in certain perfective paradigms. This treatment comes from what I think is an increased understanding on my part of what the semantics of the verb is.
163
164
Keren Rice
I will give one more example of how an analysis is informed by theoretical assumptions. In Holton’s 2000 dissertation on Tanacross, another Athapaskan language, he argues that stems can end in either voiceless unaspirated or voiced stops/ affricates. In light of comparative Athapaskan studies, this is a rather surprising conclusion since voicing is not usually thought to be an underlying property of stops/affricates in these languages, but rather stops/affricates can voice intervocalically. Holton argues that the vowels that can follow these voiced stops/affricates are excrescent rather than underlying vowels synchronically, and therefore the consonants must be treated as voiced underlyingly rather than predictably. If Holton’s primary focus were comparative/historical, he would treat these vowels as underlying; synchronic considerations led him to a different analysis. What we see in all of these examples is that linguistic theory informs linguistic description, and changing theory can change a description. Depending upon the theory that one adopts, very different descriptions may be accorded to a language. 4.2.5 Terminology Something else can be seen in the examples above. There are several terms that are used in the Athapaskan literature that are undergoing some rethinking these days. We have seen situation aspect vs. conjugation marker. A parallel case is voice/valence vs. classifier, also mentioned above. In both of these cases, the latter term is the one used in the Athapaskan literature. Part of the reason for this, I believe, is that in many cases it is difficult to figure out why the particular morpheme in question is present rather than some other one — thus, idiosyncrasy is the focus. However, I have tried to turn this on its head, and focus on the regularities, recognizing that nevertheless there are idiosyncrasies. Again, we see different assumptions present: with a mixed system which is sometimes regular and predictable and other times irregular and unpredictable, what do we take as the basis for the description? I now lean towards bringing out the regularities, while recognizing the unpredictable cases, and in using terminology that highlights the predictable nature; in the Slave grammar I used terminology that reinforced the unpredictable, idiosyncratic nature of things. 4.2.6 Data presentation Data can be presented in different ways in grammars. I begin with the writing system adopted. The Cambridge University Press guidelines say the following: Authors should use either IPA symbols or a practical orthography with IPA equivalents unless general practice in the family strongly mitigates against this.
A typology of good grammars
If IPA or some standard phonetic tradition is used, the system needs to be commented on. If one of the standard systems is not used, it is extremely important that the symbols be interpreted. For instance, in the Athapaskan language family, symbols like {d} are used for voiceless unaspirated stops and {t} for voiceless aspirated stops. This is common in North American orthographies, but I have seen phonological treatments on North American languages where the analysis ultimately must be rejected because the source did not make clear that these symbols have non-standard values. As another example, vowel symbols are notoriously easy to misconstrue. The symbol {i} is used in both Navajo and Slave, but in the former it represents a far laxer vowel than it does in the latter. As a phonologist who is interested in such details, I find it frustrating to go to a grammar and find that the range of vowel qualities is not described, but simply a statement along the following lines: {i} has the value /i/. Moving beyond transcription systems, presentation of data also varies from grammar to grammar. While the three-line presentation is probably the most common (line 1: data in language, line 2: morpheme-by-morpheme gloss, line 3: English translation), some have adopted a four line system where the orthography without morpheme breakdown is given on the first line and the morpheme-bymorpheme breakdown on the second line. The Cambridge University Press guidelines make the following comment: “Each example should be given in three lines, the word or sentence, with hyphens separating morphemes (when it is possible to specify morpheme boundaries), interlinear gloss, and translation.” There are cases in which morpheme-by-morpheme breakdowns may not be necessary. This may be the case in syntax, where word-by-word breakdowns may be sufficient, or only partial morpheme breakdowns (e.g., to highlight material that is relevant to the syntax). It may also be the case in the presentation of paradigms where the user can pick out the differences on their own. I find I still have problems with morpheme-by-morpheme breakdowns in Slave, not because I don’t know the meanings but because I’m never sure where to put the hyphens. Sometimes the writer of a grammar makes a decision that goes against the idea of morpheme-by-morpheme breakdowns. Valentine 2001 comments explicitly on this. A typical example from Valentine is shown below. Zhashkoonyan gye zhaangweshwan wnisaawaan daawewaad. ‘They kill muskrats and mink and sell them.’ (AM15.7) Zhashkoonyan na 3obv ‘muskrat(s)’; gye av ‘and’; zhaangweshwan na 3obv ‘mink’; wnisaawaan vta ind 3pProx>>3obv ‘ANpl kill Anobv’ daawewaad vaio conj 3pProx ‘(CONJ) ANpl sell (Y).’
Valentine makes the following introductory remarks:
165
166
Keren Rice
Each example sentence is introduced with a specific reference number, and below the sentence, there appears a free English translation. At the end of the free translation, there is usually a reference code, here (AM15.7) providing the source of the example. Below the free translation is a box containing a word-by-word analysis of each word in bold face, followed by an italicized annotation of its grammatical properties, followed by a word-level gloss.… Overall, the glosses introduce many useful innovations in the characterization of Algonquian meanings. … Linguistic researchers may be disappointed to see that morpheme-level segmentations of examples are rarely provided. At a conference held in Thunder Bay, Ontario, in 1996, a steering committee of Nishnaabemwin speakers explicitly requested that such details not be included, as it was felt that they interfered with the flow of the presentation, and contributed to what is sometimes called the “intellectual mining” of aboriginal languages and cultures. (pages xxxii–xxxiii)
Even more detail is sometimes needed in a grammar. The Cambridge University Press guidelines include the following advice. “It may be necessary to provide background that will enable the reader to contextualize the example. … It can also assist in understanding a grammatical point if the context in which a sentence was produced is described.” In the Slave grammar, I included discussion in the text to help clarify the difference in meaning between two common complementizers, talking about the kinds of questions that one might ask in order to elicit a response with a particular complementizer. 4.2.7 Summary Grammars can vary in many ways, and I have discussed a number of these ways in this section. The author of a grammar has many questions to think about as s/he begins the process and as s/he works through the process of writing the grammar. Most important probably are the goals and the audience, but a host of issues demand attention of the grammar writer as s/he sets out on the process of writing a grammar and as the work continues.
5. Some example grammars I have so far talked about “good”, “grammar”, and “typology,” focusing on what I have called the “essence” of a good grammar beyond its content, and ways in which grammars can vary. I am sure there are many more ways in which a good grammar can be put together, and I have not offered a typology in any real sense, but rather given a sense of variation. In the last part of this article, I look briefly at three works with different goals. The first is not a full grammar, but a sketch of
A typology of good grammars
Chipewyan by Fang-Kuei Li. The second is the Slave grammar. The third is a recent grammar by Rand Valentine, on Nishnaabemwin, an Algonquian language. “Chipewyan” by Fang-Kuei Li is a sketch of Chipewyan, an Athapaskan language, published in 1946. It appears in a book of sketches of several Native languages of North America. Li states that the sketch is based on fieldwork done in the summer of 1928 in Fort Chipewyan, Alberta, Canada; Li had already worked with other Athapaskan languages. The article is twenty-five pages long. The discussion of phonology includes the consonant system (inventory, phonetic description, distribution in terms of syllables and morphological positions), the vowel system (discussion of the phonemicization and variation), tones, syllables, various morphophonemic alternations (e.g., voiced and voiceless fricative, simplification of double consonants, lenition processes, assimilation, nasalization), morphology (word classes, noun, possessive prefixes, possessed forms, vocative suffix, postpositions and their objects, postpositional suffixes, verbs including discussion of the semantics of verb stems, verb stem variation for aspect and the form and function of the verb prefixes; verb suffixes, independent particles, and word order (probably half a page!). Even more than fifty years later, Li’s sketch is the first thing that I recommend to someone who asks me for an introduction to an Athapaskan language. What makes it so valuable? It is a lucid, precise, and detailed description of the phonology and morphology of the language. While there is a sense in which it is not complete, it opens the door to every topic in the phonology and morphology of these languages that has been recognized as important over the years. Each time I return to the sketch, I find something there that I had not realized was present, so it has continued to be a rich resource for me over the years. I find it accessible, although it is not wordy, given it length. Examples are copious and clearly illustrate the point that is being made, terminology is easy to understand. I have hoped to write something like this that would retain the essence of what Li did, while adding more syntax and representing a somewhat more current perspective on the languages after the passage of fifty years. Other than adding syntax, it is difficult to go beyond Li’s work. The Slave grammar represents a much deeper view of an Athapaskan language. This was a multi-year project, covering several dialects of Slave as spoken in the Northwest Territories, Canada. The grammar was designed to be a reference grammar with as in-depth a description as possible. I began doing fieldwork in 1973; the majority of the field work for this book was done by 1980, although I continued to add to the grammar until shortly before it was published, in 1989. It is thus the product of many years of fieldwork, analysis, and writing. When I began work on the grammar, and at various stages along the way, I spent considerable
167
168
Keren Rice
amounts of time in the library reading grammars. I took as my model a grammar of English by Quirk, Greenbaum, Leech, and Svartvik 1974. This grammar was far beyond anything that I would ever be able to produce for Slave, but its depth and comprehensiveness made it stand out to me as something that I could emulate. The Slave grammar is organized into five major sections, introduction (overview of language, typological characteristics), the sound system (overview, consonants, vowels, development, tone, major phonological rules, constraints (morpheme, surface, syllable)), categories, model (the lexicon), syntax (word order, pronouns, coreference, conjunction, comparison, negation, commands and hortatives, tense/aspect/mode, yes/no questions, alternative questions, direct questions, indirect questions, right extraposition, topicalization, complements, complementizers, direct and indirect discourse, relative clauses), and texts, followed by references and an index. I also included a detailed, section-by-section table of contents. This multi-dialectal grammar probably has more depth and breadth of material available than for any other Athapaskan language as yet, although I suspect (and hope) that the grammars being written today will go far beyond it, partly because it has served as a model to some of those working today on grammars. I tried to achieve both breadth and depth in the grammar. Coverage was important to me, as was accessibility. In terms of coverage, over my years of using this grammar, I have found that much is there. Yet there are areas in which I really would like to know far more than is in the grammar. This is especially true of syntax and semantics, but is also so to some degree true of morphology. I single out these areas because these are the areas in which linguistic theory has probably evolved the most in the past thirty years, and there are many questions that I have now that I do not think I could possibly have thought of when I was doing the fieldwork. Some of this information is available in material that I have, but much requires additional fieldwork. In terms of accessibility, I tried to use several different ways to make the grammar user-friendly (detailed table of contents, index, cross-referencing), and I tried to design it so that chapters could be read relatively independently of other chapters. The Slave grammar is the product of many years of intensive work, and yet it remains only the barest beginnings of achieving an understanding of the Slave language. The final grammar that I want to discuss is Valentine’s grammar of Nishnaabemwin. This reference grammar has two major goals. First, Valentine designed it to be a reference grammar of the language. Second, Valentine’s audience explicitly includes speakers of the language as well as linguists, and he deliberately incorporated into the book material that would belong in a textbook called something like “The structure of Nishnaabemwin.” As Valentine says, “This grammar is designed to be useful to a variety of readers, including researchers, Nishnaabemwin
A typology of good grammars
teachers, and any diligent student of the language” (xxxi). Further, “One reviewer pointed out that this grammar is actually a compound work, consisting of an introduction to linguistics as well as a grammar. This I have done, again, to accommodate my intended primary audience, those interested in teaching the language, who typically lack extensive linguistic training” (xxxxiii). Valentine is certainly not the first to follow this path, but he is, I believe, particularly successfully in how he does it. Valentine covers a range of topics in the grammar. The introduction is followed by sections on phonology and orthography, parts of speech, noun inflection, verb inflection, derivation, nominals and pronominals, verbs and their semantic roles, being and becoming, location and manner, time and circumstance, quantity and degree, and syntax and communicative function. In addition, there is a glossary, references, and an index. No texts are included, although the copious examples are drawn largely from texts. Note that many Nishnaabemwin texts are available already, making the need to include them in a grammar somewhat less than with languages where there is little or nothing available. Because Valentine’s goals are different, I would like to end this section with a detailed quote from his grammar on stress and syncope (51–52). This quote shows clearly how Valentine tries to weave together a description of a complex pattern in Nishnaabemwin, stress and syncope, with instruction about the basic building blocks of a stress system. Many Algonquian languages share similar rules of stress patterning. The stress patterns that are presented in this section show similarities with those found in Delaware, Potawatomi, and Menominee, all sister languages of Nishnaabemwin. We have already seen that the consonants and vowels of words are organized into syllables. Syllables are also organized into larger groupings called metrical feet. By “metrical” I mean having to do with rhythm patterns of pronunciation. In Nishnaabemwin a metrical foot is a grouping of two syllables together, in which the first syllable of the foot is weaker, and the second syllable is stronger. This pattern can be represented as follows:
Foot
Syllable weaker
Syllable stronger
The system here is quite simple: a standard foot consists of two syllables, the first of which is the weaker member and the second the stronger. To see how this system works, we will use a few words as they are pronounced in Minnesota Ojibwe, closely related to Nishnaabemwin, and show how the stress rules relate the pronunciations of Nishnaabemwin and Ojibwe. Consider the word gidagoshin,
169
170
Keren Rice
meaning “you arrive” in Minnesota Ojibwe. If we organize the syllables of this word in patterns of feet with weak and strong components, we arrive at the following specification:
F
W g i d
F
S W a g o sh
S i n
This word consists of four syllables organized into two metrical feet. When we compare this Minnesota Ojibwe word to its Nishnaabemwin counterpart, we make an interesting discovery. The Nishnaabemwin pronunciation (and spelling) is often gdagshin, which lacks some of the vowels of the Minnesota Ojibwe. If we examine the two pronunciations, we find that the missing vowels in Nishnaabemwin are exactly those which happen to fall in weak metrical positions.
F
W Niashaabemwin: g __ Minnesota Ojibwe: g i
F
S W S d a g __ sh d a g o sh
i n i n
So the Nishnaabemwin word is actually identical to the Minnesota Ojibwe, the only difference between the two words being that in Nishnaabemwin metrically weak vowels are de-emphasized to the point of deletion, while in Minnesota Ojibwe they are not.
Of course, there are consequences of putting into one book material that most often is divided between an introductory linguistics textbook and a grammar. The biggest consequence perhaps is length. The Nishnaabemwin grammar is very long. What I found in using this reference grammar is that, at least from the perspective of a linguist, the introductory linguistic material did not detract from the book in any way whatsoever. It is integrated in such a way that it serves to add to the quality of the book. Whether the book succeeds for its intended audience, the speakers of the language, is a question that I am unable to answer.
6. Summary Writing a reference grammar of a language is a monumental task, one that usually becomes larger and larger as time goes on, as it is a task for which there is no logical endpoint. Further, there is no cookbook for how to write a grammar. The essentials are that it provide good coverage of the structures of a language
A typology of good grammars
in a way that is true to the language and accessible to the user. A good grammar depends on the strengths that the author brings to it — to provide an interesting and well-grounded analysis, to use her/his knowledge of language and linguistics to the fullest, to write in an interesting way, to be true to the language. The best of grammars have the characteristics that we expect in any good piece of work. It is a humbling experience to write a grammar; and writing a grammar is a task that is in many ways never really done even though an author puts closure to it. I personally found it to be the hardest task, but at the same time the most rewarding task, of my career. Grammar writing is not for everyone, but for those who thrive on it, it is an experience not to be missed.
References Callaghan, Catherine A. 2002. Writing a user-friendly dictionary. Making dictionaries: Preserving indigenous languages of the Americas, ed. by William Frawley, Kenneth C. Hill and Pamela Munro. Berkeley CA: University of California Press. 312–321. Comrie, Bernard and Norval Smith. 1977. Lingua Descriptive Studies: questionnaire. Lingua 42. 1–72. Cook, Eung-Do. 1984. A Sarcee grammar. Vancouver: University of British Columbia Press. Dixon, R.M.W. 1972. The Dyirbal language of North Queensland. London: Cambridge University Press. Dixon, R.M.W. 1977. A grammar of Yidiny. New York: Cambridge University Press. Dixon, R.M.W. 1988. A grammar of Boumaa Fijian. Chicago: University of Chicago Press. Dixon, R.M.W. and Keren Rice. No date. Cambridge Grammatical Descriptions. Guidelines for potential contributors. Faltz, Leonard. 1998. The Navajo verb. A grammar for students and scholars. Albuquerque NM: University of New Mexico Press. Frawley, William, Kenneth C. Hill, and Pamela Munro. 2002. Making a dictionary: ten issues. Making dictionaries. Preserving Indigenous languages of the Americas, ed. by Frawley, William, Kenneth C. Hill, and Pamela Munro. Berkeley: University of California Press. 1–22. Holton, Gary M. 2000. The phonology and morphology of the Tanacross (Athabaskan) language (Alaska). Ph.D. dissertation. University of California at Santa Barbara. Li, Fang-Kuei. 1946. Chipewyan. Linguistic Structure of Native America. Viking Fund Publications in Anthropology 6. New York: Viking Fund. Quirk, Randolph, Sidney Greenbaum, Jeffrey Leech and Jan Svartvik. 1974. A grammar of contemporary English. London: Longmans. Rice, Keren. 1989. A grammar of Slave. Berlin: Mouton de Gruyter. Rice, Keren. 2000. Morpheme order and semantic scope: Word formation in the Athapaskan verb. Cambridge: Cambridge University Press. Valentine, J. Randolph. 2001. Nishnaabemwin reference grammar. Toronto: University of Toronto Press.
171
Thoughts on growing a grammar* David J. Weber Summer Institute of Linguistics
Grammars take time; they require data development; they must be planned; and their writing may span many years. Linguists should “grow” a grammar through the course of their involvement with the language. This paper gives practical advice about this process, such as the creation and management of a data corpus, the acquisition of useful intellectual tools, the choice of an effective authoring environment, the formation of habits and disciplines that contribute to successful grammar writing. A grammar should be corpus-based, data-driven, theory-informed, user-friendly, publisher-compliant, web-deliverable, …and an expression of genuine love.
1. Growing a grammar Grammars are not the sort of thing one can just sit down and write. They take time; they require data development; they must be planned; and their writing may span many years.1 1.1 The basic idea The concept of growing a grammar is simply that: preparing the ground, planting a seed, and nurturing the plant until it comes to maturity. • •
You might start with a brief sketch of the grammar. You might write some data-rich work papers that could be incorporated into the grammar.
174
David J. Weber
• •
•
You might write a paper for a class or workshop you attend. Later you incorporate it (with any necessary adjustments). You use a chapter of your growing grammar as the basis for a publishable paper, adjusting as needed to make a good paper. In the process you discover ways to improve that chapter of your grammar, especially as you get feedback on the paper. You might go into a graduate program, where for almost every course you will write a paper. The professors give you great feedback! So eventually you incorporate those.
Perhaps you decide to turn your growing grammar into a thesis or dissertation. Again, that’s great because your committee members will give you lots of feedback. • • •
You keep on growing this grammar, adding, improving, and so forth. You get feedback from one source and another, wherever you can get it. You may want to rework parts into conference papers. This nets you more feedback. You may go on to rework the papers for journal publication. At some point you decide to publish the grammar. Perhaps you want to publish a book; perhaps you want to publish on the web; perhaps both. But at this point publication should not be a big hurdle.
When is the grammar finished? Perhaps never! Since the grammar of a language is so vast, the ideal would be to make the written grammar a “living” document, something to which others, perhaps in future generations, could add. Linguists, especially those who do field work, should “grow” a grammar (as well as a lexicon) through the course of their involvement with the target language. They should start writing early on and add regularly throughout their involvement, as they learn more and more about the language. Eventually the day comes when the entire grammar should be published. 1.2 Growing a grammar on the web There could be tremendous advantages to growing a grammar on the web, either on the internet or on an institution’s intranet: •
•
It would provide some essential services like backup and archiving, and provoke the early resolution of some issues that are sometimes deferred too long, such as issues of ownership. Those responsible for monitoring the progress of the project (e.g., an administrator of the sponsoring institution) could do a better job, since having the developing grammar on the web makes it directly inspectable by those who have the appropriate permission.
•
• •
Thoughts on growing a grammar
It would facilitate publication, especially the electronic publication of all or of parts. If the grammar is grown on an institution’s intranet, then publication might be as simple as moving it to a public site. Advisors/consultants could help grammar writers better: • A consultant could read the growing grammar and give feedback, perhaps from the other side of the globe! • To prepare for helping a grammar writer, a consultant could familiarize herself/himself by reading the growing grammar. If for one reason or another a linguist were not able to continue writing the growing grammar, the project could be continued by someone else. This might be someone assigned by the project’s sponsor. This, of course, assumes that the contributions of each is appropriately acknowledged.
1.3 Soil and climate I once asked several graduate students from a leading university in the United States, “Do you think it would be possible for you to write your dissertation as a co-authorship with a retired field linguist, someone who has learned to speak the language and knows a lot about it but who has not written much about it?” They responded negatively, but not because of co-authorship (which is, of course, a challenge) but because graduate students at their university are discouraged from writing grammars for their dissertations. The reason, they said, is that those who write grammars are disadvantaged in the workplace, that the jobs go to those who have written something more theoretical. One of the students said, “Isn’t this ironic; [our university] has been known for its grammars.” In the past many linguists went to the field with the idea that they would provide a broad description of a language, including a phonology, a grammar, a vocabulary or dictionary, and perhaps even a brief ethnography. To provide these, they expected to have a deep and protracted involvement with the language and its speakers. Some of these were doctoral students intent on documenting a language as a dissertation project. In the last half of the previous century this conception suffered considerable erosion. Pressure and prestige have shifted to the publication of journal articles, not books, and to the theoretical. Descriptive pursuits and their products came to be despised as belonging more to natural history than science. But, of course, good language descriptions, both ones that are broad and ones that go deep, are crucial for linguistics, especially because so many languages are rapidly moving toward extinction. So we need to address questions like the following:
175
176
David J. Weber
1. What can we do to motivate linguists to write grammars? 2. How can we foster positive attitudes toward grammar writing in the discipline? 3. How can we foster more positive attitudes toward grammar writing in linguistics departments? Are certain institutional changes need? 4. How can professors motivate their students to write grammars? 5. How can we get more resources directed toward the writing of grammars? 6. Should the linguistic curriculum be modified to give grammars a more prominent role? Are students adequately trained to write them? 1.4 How grammars can “happen” Grammars can be written in many different ways, with many different combinations of people. Here are some possibilities: • •
•
•
A lonely individual may slog it out. The grammar may be written as a dissertation, in fulfillment of a degree requirement. (This was the case for my Huallaga Quechua grammar.) This has the advantage of ensuring solid feedback by informed linguists, that is, the committee members. A grammar may be started in a class (say a field methods class) or a grammarwriting workshop. For example, I led such a workshop in Colombia shortly after finishing my graduate work at UCLA. The goal was that each participant would write a “grammar sketch” of the language s/he was studying. I walked the participants through basic typology. For every topic the participants explored that topic in the language about which they were writing. At the end of each week they had to hand in a (very preliminary!) write up. At the end of ten weeks, they had sketches and I went home. Then a wonderful thing happened: the participants kept writing! With encouragement from their administrator and help from a resident consultant, they turned those sketches into grammars, most of which have now been published. Co-author with a retired or soon-to-retire field linguist, one who knows a great deal about the language but is not in a position to write a grammar. This could rescue a lot of valuable knowledge that would otherwise be lost to posterity.2 Who might co-author with such a person? Here are some possibilities: • one who has already written a grammar so knows the ropes. For example, I am co-authoring a grammar of Bora (Witotoan) with Wesley Thiesen. Wes did extensive analysis of the Bora language, lived with his family
Thoughts on growing a grammar
among the Bora and learned to speak Bora fluently, published a brief grammar and a substantial dictionary for the Bora people, and cooperated with Bora speakers in translating the New Testament into Bora. Through all this he gained an amazing knowledge of Bora. The grammar we are writing is our attempt to frame this knowledge in a way that will be useful for others, both now and in the future. • a consultant-in-training, possibly mentoring under a more experienced person who has already written a grammar • a graduate student. Yes, a graduate student, possibly one who co-authors the grammar as a thesis or dissertation to meet a degree requirement. This would not be without some challenges. Perhaps the most difficult is the current climate in which departments discourage their students from writing a grammar as a dissertation. But with the growing concern about endangered languages, professors ought to consider this possibility. Another challenge is seeing that the contribution of each person is properly acknowledged. Unless the person who has developed the data — possibly at great personal expense — is convinced that his/her contribution will be fairly acknowledged, s/he is unlikely to share it. Another challenge for the co-authors is to build a relationship of mutual respect, cooperation, and confidence. This is particularly important because linguists who have remained long in the field may have been long out of touch with linguistics, so their analyses may reflect perspectives long since abandoned by the field at large. Un-time-warping “legacy” descriptions: There are old grammars and fragments of grammars, both published and unpublished, that deserve to be picked up, dusted off, and published in a modern way, perhaps best delivered on the web. This, of course, requires familiarity with the theories and notations from the time when the write-up was first made. This enterprise is more interesting if the description can be enhance, as by adding further data gathered from texts written in the language, insights gained from closely related languages, or insights resulting from better theories of linguistics. • I believe that in the future we will see grammars written by working groups composed of people with diverse backgrounds: linguistic, cultural, educational, and so forth. I believe that — in this mix — native speakers of the target languages will play a much greater role than they did as “informant,” “language resource person,” or whatever politically-correct label is applied to this critical role. •
177
178
David J. Weber
By the way, consider the decipherment of Linear B, the script “used for writing Mycenaean, an early form of the Greek language.” (Wikipedia).3 It was an incredibly difficult intellectual challenge. Many people had worked on it without making much progress. The key to success was cooperation. This was implemented by Chadwick through a newsletter, which came to tie together the far flung “community” of those working on deciphering Linear B. • Modern technology makes possible some new modes of partnering. For example, e-mail enables long-distance collaborations in a way that was not possible before. It may be that, in some cases, those working together to write a grammar may not ever meet face-to-face. It is now possible for someone to provide consulting services from the opposite side of the globe. • The advantages of cooperation (“partnering”) may mean that institutional boundaries mean less and less. Perhaps a grammar-writing group will include one or more speakers of the target language, one or more foreign linguists, each affiliated with a university or some other institution, national scholars, consultants residing in foreign countries, and so forth. 1.5 Grammar and the lexicon The grammar of a language is intimately tied to its lexicon. (Indeed, some linguists question whether there is a boundary between grammar and lexicon.) A grammar writer need not resolve this thorny issue but must deal with a more limited question: How should a written grammar relate to a dictionary of the language? Here are some matters to consider: • •
•
•
Ideally a grammar and a dictionary would be grown together. Many grammars include a vocabulary or dictionary, usually as an appendix.4 If so, what should be the nature of this? How much lexical information should be included? What is its purpose? Is its intended simply to support the grammar, to facilitate reading texts, or as a serious effort to communicate lexical information? At a minimum the categories used in the grammar and those used in the dictionary or vocabulary should be in harmony. The grammar may present a more refined set of category distinctions than the dictionary, but these should be compatible with those of the dictionary. If not, then the grammar should explain why it differs from the dictionary. On the other end of the scale, the grammar could be written as a dictionary, putting into lexical entries virtually all the morphosyntactic information concerning their behaviors. This would be consonant with frameworks like Hudson’s Word Grammar.
Thoughts on growing a grammar
Various theories do not go to the extreme suggested in the previous point but have, nonetheless, shifted much of the “burden” of syntax to the lexicon. HPSG would be an example. • Harold Gleason once suggested (personal communication) that a grammar could be written as two heavily cross-referenced parts, one that focuses on grammatical generalizations, the other that focuses on lexical items, including their grammatical properties and behaviors. 1.6 Getting your product to market It is often useful to publish parts of the (eventual) grammar before publishing the whole thing. Parts might lend themselves to be published as papers, perhaps with a more theoretical or technical bent than would be appropriate for a grammar. Eventually, of course, you will want to publish your grammar. Learn about potential publishers and what they want. You’ll need to find out if they have a style sheet. You might ask about their editorial policies. If possible, have your grammar read by one or more knowledgeable people. If possible, have it read by someone good at copy editing. Submit it. Don’t hold your breath because the review process can take quite a while. If it is accepted, the publisher may ask for some revisions. Then the ball is back in your court. I believe that in the future more grammars will be published electronically. Rumor has it that web-published books are read by vastly more people than books published with ink on paper. Electronic publication also requires submitting the grammar to a publisher, a review process, editing, and so forth. Fortunately, if you have grown the grammar on the web, the path to electronic publication is already paved.
2. Cultivating information The quality of a grammar is largely determined by the quality of the data on which it is based and which it incorporates. 2.1 Creating a corpus A grammar should be based on a corpus. Ideally this would be a collection of diverse, real instances of language use, both spoken and written.
179
180
David J. Weber
Elicitation can be used to explore topics and supplement areas of the corpus that are thin. However, elicited examples should be used with care. They often lack the ring of authenticity, which is not surprising since they are generally spoken out of context with no communicative intention (other than to inform the interrogator how someone might say something in some hypothetical context). An analysis of a language based on a corpus will be good to the extent that the corpus is representative of the language. To achieve this, a corpus must be carefully built. It should not be a helter-skelter collection of material, which might have many instances of one type of text while lacking others. A good corpus has diversity: texts from diverse speakers or authors (differing by age, sex, social group,…); texts spoken in different contexts,5 formal to intimate, public or private,…; texts of different types (as discussed further below), texts for different purposes (to inform, to persuade, to entertain,…); monologues and conversations; face-to-face communication versus communication separated by space and/or time; planned (often written) and unplanned (spontaneous, usually spoken); ritual and oft-repeated text; and so forth. It is better to build a corpus with many short, diverse texts than with a few long texts. The former are far more likely to have the desired diversity than the latter. If you can make recordings of live, natural text, by all means do so. But don’t despise written texts. I employed a Quechua friend to work for about three months writing essays: essays on agriculture, festivals, marriage, and so forth. One of the most interesting, from a linguistic perspective, was about personal cleanliness! These essays informed me about culture as well as the Quechua language. (The outline of the “Human Relations Area File” is a great source of ideas for topics.) Let’s return to the issue of genre, commenting on a few types: Folktale: Folktales are often the easiest genre to acquire because they are told and retold, so just about anyone can tell one or more, albeit imperfectly. (This is, of course, not always the case. In some cultures folktales are properties regulated by strict rules.) Folktales are interesting because they usually have one or a few main characters interacting with various other individuals or groups. They may be instructive as to reference tracking. However, most folktales are known to all members of the community so they are told not to inform but to entertain. This means that they may be quite aberrant as to how referents are tracked throughout. For example, in Quechua there is a tale much like “Hansel and Gretel.” In the first version I recorded it is impossible to tell how many children were involved; the text simply said children or child and the listener is expected to know that it was a girl and her little brother. But what this text lacks in
Thoughts on growing a grammar
participant reference it makes up in effects: When little brother gets stuck with a knife and butchered, the orator elaborates with great detail how the child squeals, “Ayyyyy”. The art of the performance is in the effects. While it is good to have some folktales in the corpus, it is a mistake to have too many, as this can lead one to get an unreal picture of the language. Personal accounts: Narratives are an important part of a corpus, both third person ones (“Once when John was going to his field…”) and first person accounts (“I went to town and…”). Almost everyone can readily respond to the question, “Have you ever been near death?” or “Have you ever had an accident?” Description: Descriptions of almost anything are useful (how the village is laid out, the description of a particular house, …whatever). A useful technique is to use pictures or photographs to elicit descriptions. These should be chosen with some care, as discussed below, and then presented to the respondent with a request to describe them. The description can be written down or recorded for later transcription. This can be a very instructive type of text to work on in the early stages of learning about a language because the picture contains (in some sense) the semantics of the text that it provokes. In other words, you are seldom in doubt as to what the text says because you can look at the picture, which informs you about the text. Another advantage of working with picture descriptions is that the respondent should have little trouble describing the picture, unless, of course, it is too foreign to his or her culture. Even when describing a past event, say, what happened when s/he was ten, the teller may have difficulty remembering what happened, perhaps making many false starts. But a picture or photograph is right there in view. There is another big advantage to using pictures: You can control to some extent what grammatical elements the description will have by the nature of the picture. A picture of a static object, say of a church, will provoke the naming of parts (steeple) possibly linked by part-whole or spatial expressions (on the top of the steeple is a cross, inside is a bell, off to the right is another smaller tower), and may be rich in participles: plastered, painted, cemented, thatched. By contrast a picture of an event or activity, like a soccer game, is likely to net active verbs, possibly bearing various aspectual markers, presenting different participants in different roles. Procedural: Procedural texts generally have strong temporal organization, being organized around the sequence of events that constitute the procedure.
181
182
David J. Weber
In contrast to narratives, which tend to be agent/actor oriented (“He did this… He did that…” and so forth), procedural text are often patient oriented. For example, in response to, “How do you weave a basket?” the most topical element is likely to be the emerging basket; in response to “How do you butcher a sheep?” the most topical element is likely to be the sheep. Exhortation: Scolding, advice, indeed, any text with which a speaker is trying to convince someone else of something, are particularly interesting. I recorded an incident that happened when Teodoro Cayco, our primary Quechua coworker, was the Justice of the Peace. A man and a woman came, frantic, saying that a certain man had committed suicide. As Justice of the Peace, Teodoro’s job was to interrogate, to ferret out all the facts. I asked for and was granted permission to record the session. I got 45 minutes of incredibly emotionally charged communication, with rapid fire exchanges. The woman was frantic and crying; the man, sometimes defensive and sometimes aggressive. The testimony made great efforts to be convincing, even though it was largely deception.6 Dialogue and conversation: Texts with multiple speakers are an important part of a corpus. They show the joint creation of a “text” through cooperation, turntaking, negotiations regarding the topics to be pursued, and so forth. Conversations can be difficult to analyze. Seldom do people take turns very politely. Particularly when they are with intimate friends, speeches may overlap. In some cultures it is even quite normal for two people to talk simultaneously. 2.2 Organizing the corpus No matter how good the corpus, organization is the key to making the data available for analysis and for incorporation into the grammar. “If you can’t find it, you don’t have it.” The grammar writer must have an effective method for filing data, ideally one that makes it possible to work on any and every aspect of the grammar and easily recover the relevant examples. Linguists now generally expect to do their data management “on” a computer using some facilitating software. Unfortunately I know of no program that both facilitates organizing a corpus and writing a grammar. (I am optimistic that SIL’s Fieldworks program may one day meet this need.) It is important that, whatever data management scheme is employed, it should handle both textual data (as in a corpus) and elicited data. These often complement each other, as when a sentence in a text raises an issue which is subsequently explored through elicitation.
Thoughts on growing a grammar
What worked for me was an exhaustive concordance on slips, created and organized in a way I learned in a workshop from Harold Gleason. I do not know of an automated data-handling scheme that will do all that one can do with a “Gleason” file, and if I had to write a grammar from scratch for another language, I would begin by creating such a file. But, of course, I would now combine this method with the use of computer programs for various tasks. 2.3 Glossing texts Glossed texts are very useful, particularly if the one writing the grammar can not yet read target language texts with a fair measure of fluency. For this reason the arduous task of glossing texts should not be deferred too long. Glossing text is an instructive analytic task. One aspiring to write a grammar should gloss text at least until s/he ceases to feel that the work is adequately compensated by what it teaches. Glossed texts may be important for engaging others, particularly for making it possible for consultants and advisors to help. The texts enable them to quickly get a feel for the language and to thus develop ideas that the consultee (grammar writer) can further consider. Of course, a collection of glossed texts is a prized end product, one that can be included in or published with the grammar. 2.4 Integrating grammar development with translation work When Teodoro Cayco and I were translating the Gospel of Mark we came to the section on the multiplication of loaves, which has a verse that says that the disciples gathered up seven basketfuls of leftover pieces. Teodoro suggested two options for translating this, differing only by the presence or absence of the suffix ‑no7 Without ‑no the sentence means that there were seven baskets and these were full of leftover pieces. With ‑no it means that the quantity collected would fill seven baskets, but without implying that these were actually in baskets. When I heard this I jotted this example and observation on a 3×5 slip. At the end of the day I filed it under the section on the suffix -no ‘like’. A few years later when I was writing the Huallaga Quechua grammar and came to writing about -no, there was the example, along with some similar cases that had come to light over the years. These enabled me to give — I think for the first time — a reasonable characterization of the as though function of the suffix -no.
183
184
David J. Weber
This happened because I was using an effective filing system, one that — over many years — incorporated thousands of examples, bringing together ones that were mutually relevant, ones that, taken together, made generalizations emerge.
3. Becoming a farmer Seize the day …but take the long view!
If you want to grow’em, you gotta become a farmer. To grow a grammar, you have to become a grammar writer. Writing a grammar is likely to span one or more decades, so prepare for the long-term. It is important that authors shape themselves — their attitudes and lifestyles — to make grammar writing a pleasant and meaningful part of their lives. So here’s some advice; take it or leave it.8 3.1 Tools Use good tools and keep’em sharp!
3.1.1 Analytic techniques and theories Becoming a linguist involves learning many analytic techniques: charting data to discover patterns, form hypotheses, and make generalizations; using minimal pairs to identify phonemes or pin down the crucial difference between two constructions; learning to wield Occam’s Razor. Linguistic theories are really useful tools. They can enlighten us about phenomena encountered in a language, and they can make us see their significance to the broader enterprise of understanding Language. 3.1.1.1 Theories may turn work into play Much of what we do as linguists is very similar to what people do as recreation. A lot of what we do can be turned into play. Whether a task is work or play depends largely on our attitude toward it. A lot of what we must do is “problem solving,” i.e., finding an analysis for some phenomenon. Our training is intense in solving problems. If we were to think of these problems as puzzles we might think of solving them as play rather than work. It’s a matter of perspective, not of what we actually do. “Someone will solve a problem if he finds the problem interesting.”9 But what if we don’t find the problems we must solve interesting? How can we make them interesting?
Thoughts on growing a grammar
The main way — I think — is to see the phenomenon in terms of one or more theories. Indeed, one of the main benefits of linguistic theories is that they make data interesting. (This role is so important that the Chomskian tradition has elevated “interesting” to a technical term meaning approximately “bearing on some aspect of the theory”.) So use theory to build interest, which will then, like a Siren, lure you on to a solution. 3.1.1.2 Which theories? Not all theories are created equal. Some are better suited to the grammar-writing enterprise than others. Ones that take typological adequacy as a basic design criterion are particularly useful. These would include: Functional Grammar, Role and Reference Grammar, Construction Grammar, Givon’s version of functional/typological grammar, and so forth. These are all functionally-oriented theories, but there are also formal theories that meet the criterion of typological adequacy (although perhaps they didn’t beat the drum quite as much at the outset): Lexical Functional Grammar, Head-Driven Phrase Structure Grammar, and even, perhaps, Government and Binding, through the attempt to account for the range of cross-linguistic variation through parameters. Another thing: We need broad theories. Relational Grammar was a great theory, one that led to many insights, and one committed to typological adequacy. But it was a theory of a particular piece of grammar, namely grammatical relations. For that, it was the best in its day, but it said little or nothing about other parts of grammar. So its usefulness to someone writing a grammar was limited largely to the domain of grammatical relations. 3.1.2 Outlines, templates and such When I was in my very first years of studying Huallaga Quechua, I looked at many grammars written about other Quechua languages. I paid particular attention to the outlines. I began to write my own outline, incorporating topics addressed in other grammars. Over time it grew into a pretty good outline.10 With some modifications it became the outline for the grammar I wrote as my dissertation. It was the blueprint for the house I built. Grammar templates go a step farther. They provide a schematic grammar, leaving it to authors to fill out the details, provide examples, and so forth. The idea is that an author, by using the template, can write a grammar in less time than it would otherwise require, and that the grammars produced from the template will be quite similar, facilitating comparison of the languages.
185
186
David J. Weber
Grammar outlines and templates are great tools. However, they must be used with caution, lest the author end up pressing the language into the outline rather than structuring the description along lines more natural for the language. 3.1.3 The authoring environment Let us take an authoring environment (AE) as the collection of tools — particularly computational ones — that a grammar writer has at his/her disposal. In addition to the program(s) used for the writing itself, this could include programs for managing data, especially facilitating the exploration of a corpus. So what makes an authoring environment good for writing grammars? The crucial software at present is an editor, with which the user creates and manipulates the document, and a formatter, which turns the document into attractive viewable and printable forms. The editor and formatter may be separate programs (say, an emacs editor to create the document and LaTeX to format it) or the two may be combined in a single program (like WordPerfect and MSWord) with their much-beloved-but-perhaps-overrated what-you-see-is-what-you-get (WYSIWYG) character. General issues about AEs: What are the virtues and liabilities of using a program that combines editing and formatting, versus one that does them with separate programs? As one writes, how much should s/he be concerned about the document’s structure, and how much about its eventual appearance? At first blush, we might think “Oh, authors should not be concerned about the document’s appearance at all; they should dedicate all their energies to thinking about what it says!” But I think this is wrong. I think there are some benefits to integrating formatting with authoring. (One evidence of this is that most document preparation is now done with WYSIWYG editors.) Here’s a good principle: Converge as directly as possible on the final product. I had a friend, Lou Siringo, who was a very successful portrait photographer. I once asked him what he regarded as the key to his success. His reply: “I try to get as much done as possible in front of the camera.” You see, a photograph can be manipulated at various stages, e.g., screening the negative’s image as it is passed to the photographic paper, accelerating the developer’s effect on certain regions, touchup on the final print with pencils, and, of course, in the very first step, taking the picture. Lou’s wisdom was to get the initial image captured on film as close to the final product as possible, thus minimizing all the subsequent stages. I think that is why it is important to integrate formatting with authoring: it enables the author to converge more quickly on the finished product. •
Thoughts on growing a grammar
Another thing: one does not want to wait until the very end to see the product. Robert Rodriquez, director of the movie Desperado and author of Rebel without a Crew wrote: A movie is filmed in bits and pieces, and only later edited into its final form. In the filming process morale can become low, in large measure because the end product is so hard to imagine. To cope with this problem, Rodriquez keeps a trailer in which he can edit film at the site where he is filming. Periodically he takes some of the best footage and creates a short advertisement for the movie. According to Rodriquez, seeing this gives him a real lift. He shows it to the film crew and actors and it boosts their morale. The consequence is that everyone works harder and with greater enthusiasm.
So one way to boost your morale is to finish a chapter and print it out.
•
How is the grammar stored? Can other programs be applied to it? Would one have to go through lengthy input and output procedures to apply another program to the document? This may become more important as the document comes to have more and more language data. This is an advantage of AEs that keep editing and formatting as separate operations. The form produced by the editor must be usable by the formatting program, so may also be usable by other programs. Some editors make it possible to apply another program to the document being edited, or some portion of it, without even leaving the editor. (This is true for most emacs editors.) Advice about AEs: Use an AE that adequately handles the scripts and special characters of (1) the target language and (2) the languages in which the grammar will be written and published. Like it or not, we should all be headed toward unicode, so choose unicode-supportive tools over ones that don’t support unicode. • Use an AE that automatically numbers and manages the numbers of chapters, sections and subsections; table of contents; lists of figures and tables; examples …especially examples! You may have hundreds or even thousands of them, and you do not want to have to manage their numbering yourself. Not only should the AE number these automatically, but it should facilitate cross-referencing. For example, suppose I’m writing in an early section about noun phrases and I want to refer to a subsection of the chapter on relative clauses. I have no idea what number it will be, but I should be able to refer to it in such a way that the software will automatically generate the correct number (something like “15.5.3”) based on the automatic numbering of sections and the reference to it. •
187
188
David J. Weber
•
Use an AE that is good for large projects. Beware of editors that do not handle large, complex documents gracefully.11 • Use an AE that facilitates creating indices because a good (reference) grammar must have a good index. In the AE I use, index entries are automatically generated from index-entry creating expressions that I put into the document at the place to which the index entry should point. For example, in the section on negative questions I might put \index{questions!negative}. From that would be generated the twotiered entry, with “negative” under “questions”, followed by the page number. (Note: Favor AEs that permit multi-tiered index entries.) Above all, abhor AEs that treat indexing as a separate process carried out after the body of the document has been formatted and require the user to manage the number of the page to which the entry refer. Index entries should be generated automatically from instructions embedded in the text so that the user is relieved of this burden. • Use an AE that supports creating multilingual documents. Ideally, grammars are made available in the local language of wider communication, in the national language, and perhaps in one of the world’s major languages (English being the current front-runner). I once wrote a book on Quechua orthography. I drafted parts in Spanish, and parts in English. Eventually, working with an editor who is a native Spanish-speaker, I made a document in which each paragraph was in both English and Spanish. I tagged each language in such a way that, when I wanted to print the Spanish version, the English was suppressed (and conversely to print the English version). • Value long-term utility over initial ease of learnability. Every tool has a “learning curve,” that is, the time and effort required to become proficient in its use. The dominant philosophy shaping much software has been to keep the learning curve as low as possible, even if functionality must be sacrificed. It has also dramatically shaped the authoring environments we use. For example, the first computer I owned came with an editor explicitly designed to be easy to learn. One could learn to use it in about 15 minutes, but it was a bear to live with. It was like working inside a little box! Some of today’s most popular editors are somewhat like that: easy to learn, fairly powerful, but not particularly easy for large, complicated documents like grammars. My advice is to count the total cost. Be willing to climb a steeper learning curve at the start if this investment is likely to pay good dividends down the road.
Thoughts on growing a grammar
By the way, I use an emacs editor.12 These editors are not easy to learn, but they are extremely powerful. To get started, users learn a small core of functions by working through a brief tutorial. Subsequently they learn other capabilities (and their complexities) as the need arises. Most users probably only ever learn a fraction of the available commands. • Use an AE that supports creating structured documents, one that marks units for what they are, not for how they should be formatted. To some extent this is tied to the notion of style sheets. Using a style sheet keeps most of the formatting information out of the document itself, and that is how it should be. • Use an AE that supports the creation of XML. XML makes the document webdeliverable and XML is (arguably) the form most likely to be readable 100 years from now. (Note: I do not mean XML that represents how the document is to be formatted, but XML that represents the structure of the document.) • Use an AE that can become automatic for you. Heavy dependence on a mouse may impede this. Use an AE that is “ergonomic,” that is, one that will not cause you pain (fingers, wrists, backs) and perhaps eventually incapacitate you (as has happened to many computer programming professionals). • I can’t imagine working without regular expressions,13 without having them right at my fingertips in my editor. This allows me to define a search pattern and immediately find all instances of that pattern in one or more files. This is particularly useful for finding examples in a corpus of texts. Virtually all editors have a find-and-replace function. This becomes more useful when one can search for patterns rather than just character strings. It becomes even more useful when one can “capture” parts of the pattern and manipulate the captured parts in the replacement. • A good AE should support backup and archival: • Does your AE automatically write your document to the disk after every so many keystrokes, or after so many minutes? It should! • You should backup frequently, at least once a day. How many man-hours are you willing to lose? Suppose two of you work together at a terminal for five hours, thrashing out a section; if you lose it, you lose ten hours of work. How much would that be worth (based on what you’re each worth per hour)? • Think about archival now! Start thinking metadata, that is, the data about the data that is needed to archive it appropriately and make it known to others. Start thinking of metadata as an essential part of the data.
189
190
David J. Weber
Finally, do not wait for the perfect AE. It does not exist now and it may never exist! Get to work with the best AE you can acquire now, but remember that some day you may move your grammar to a better AE. 3.2 Forming habits and developing disciplines Form the habit or discipline of: •
jotting down observations, especially new words and expressions, novel constructions, and especially what you do not understand! Listen to people using the language, or read texts written by speakers, noting interesting things and jotting them down. Always carry a notebook or 3×5 cards or something to write on. Since important things often catch up to you in the dead of night, keep a notepad by your bed! One paper I wrote,14 for which I received expressions of appreciation from other scholars studying Quechua, was the result of notes jotted down over a period of a dozen years. Every once in a while I’d hear or see something that struck me as relevant to the topic so I’d jot a note and later put it on a piece of paper I kept in my desk for this purpose. Finally, one day the straw that breaks the camel’s back landed on me. I broke down and wrote the paper. Writing it was fun because I had the data at hand. • writing up your analyses, hypotheses, ideas. Share these with coworkers and solicit their feedback. Then rework the paper and present it at a conference. Submit it to a journal for publication. Get it into your growing grammar. • reading linguistics. Period! Read not expecting to understand everything! Sometimes you simply need to move on, and things may later fall into place. I used to feel guilty if I didn’t finish reading a book. I gave that up, … feeling guilty, that is. Now, for many books, I read only the introduction, the first chapter, or perhaps the first half. Or I might read the chapter introductions and summaries. Often it is sufficient to get an idea of what the author is up to without slogging through the whole text. • motivating and helping others; in giving you receive. Adopt the discipline of organization: • •
Organize your data (books, notebooks, papers). Organize your ignorance. Make “Residue” work for you. • Managing ignorance can help one keep chipping away at it.
Thoughts on growing a grammar
• Adequately handling residue will affect the analyst’s attitudes toward the language project; s/he can know where s/he stands and recognize progress. • A good residue file helps the analyst have more productive interactions with consultants, with language teachers, and with co-workers. In my experience, many field linguists claim to have lots of questions, but when given an opportunity to ask them, find it hard to bring them to mind. By contrast, those who have a good residue file have lots of questions, with data they can use to explain the problem to someone else, thus increasing the probability that they will get useful suggestions. • Organize your e-correspondence, that is, your e-mail messages. (Need I say it, use a mail program that allows you to organize your messages, both those you receive and those you send.) • Organize your computer files. This usually entails designing a rational directory structure and developing a knack for inventing good file names (ones that will make some sense a decade or two later). • Organize tasks. But don’t go overboard. Don’t end up spending too much time on organization. Strike the right balance between the work itself and the organization needed to keep it moving forward. For example, if you lose a lot of time searching in your office for books and papers (“I know its here somewhere”) then perhaps it is time to interrupt your work a bit to organize your office. Ever since an ancient penned “Consider the ant…” people have been giving advice about time management. I’m not sure how useful most of it is; lots of seemingly good advice isn’t. And — of course — it is easy to read good advice without putting it into practice. But if it helps you, go for it. On the outside chance that it will help, here are a few bits of advice: Learn to relish finishing. Learn to celebrate accomplishment. This helps pull one along. (But don’t be afraid to start things you might not finish.) • Learn to decompose big tasks into more manageable pieces. That way you can both relish and celebrate finishing more frequently. For example, relish finishing the first draft of chapter you’re working on (or if a chapter is too big, of a section or a subsection). When you finish it, celebrate! Do whatever enhances the emotional high of finishing, whatever will pull you on to finishing the next chapter. • Here’s a bit of wisdom shared by Harold Gleason in a workshop he led in Peru when I was just beginning my work there: Schedule the intellectually most challenging work for when you are sharpest; save the easiest for when you are dull. •
191
192
David J. Weber
I used to get my best work done in the mornings, while in the afternoons I really dragged. So I would schedule intellectually challenging work in the morning; tasks like alphabetizing and filing lexical slips I would leave for the early afternoon (after the mid-day meal when I would feel pretty drained). Likewise, schedule the tasks that need longer, uninterrupted blocks of time for those parts of the day when interruptions are less likely. For example, in the introduction to his impressive, two volume collection of Quechua texts, Swisshelm writes:15 I began to work on Pantoja’s stories. Since there was no other time to work, I did it all between three and six in the morning during almost two years. Those hours are the most quiet for working. [translation mine — DJW]
•
Learn to use the little bits of time. An accomplished pianist attributed his success to a lesson taught him by his teacher: “Don’t wait for an hour to practice. Just use all those free five minute pieces.” Likewise, don’t wait for a solid block of time to do linguistic analysis, language learning, reading, …whatever! Develop strategies for using the five, ten and fifteen minute periods that might otherwise be wasted. This principle also applies in the year. Do we waste the time between Christmas and the New Year because we think it is too short to do anything? When our flight gets postponed or our reservations force us to stay over a day, do we discount the possibility of getting something done that day? • Be opportunistic. It has been ingrained into many of us to (1) form our plans, and then (2) carry them out as best we can. That works pretty well in stable environments where one can pretty accurately project what the next few months or years will be like. But that is not the sort of environment in which many of us work. A better approach in many situations is — keeping our goals ever in mind — jump on opportunities that can advance us toward accomplishing them. I’m tempted to say that a key to productivity is “Do whatever you can, whenever you can do it,” but — of course — that might lead to pandemonium. Some order is necessary, but let’s be sure our plans do not keep us from getting things done. • Learn to defer what you might avoid altogether. It used to irritate me that — in Peru — people so often put off what I, as a North American, would have rushed to do (with “Never put off till tomorrow what you can do today” ringing in my ears). But my years of living in Peru taught me that — in that rather unpredictable environment — by deferring some things, I might avoid them altogether.16
Thoughts on growing a grammar
For example, one week Diana and I were planning to spend Sunday in a Quechua community. We did this often, usually preparing on Sunday morning. That preparation could take us an hour because we would go prepared to sell books, show a video, tape-record a text, eat a lunch, be prepared for various contingencies like flat tires and mechanical problems, and so forth. But that week I said, “This time, let’s get everything ready on Saturday.” We did so, spending at least an hour on it. Then, for reasons beyond our control, we did not go. If we had deferred “to the last minute” as was our custom, we could have avoided wasting that time and energy. • Learn when to fold’em. Sometimes the best way to get something done is to quit, to stop doing it. If you come to an impasse, perhaps you need to go for a jog, go work in the garden, or do whatever “gets your mind off the problem.” I put that in quotes because the mind doesn’t stop working on a problem just because we stop consciously thinking about it. Rather (according to those who study such things) when we stop our conscious thought, mental processing is shifted from the left brain, the more analytic side, to the right brain, the more intuitive side. Often the impasse is gone by the end of, well, in my case, the jog. • Sometimes we need the illusion of progress more than progress itself. So save those projects that give the illusion of progress for the times when you need them the most. • Spend the time. Some jobs simply take a lot of time. If you want to get them done, you must be willing to put in the time. Remember what Kenneth L. Pike is reputed to have said, “God didn’t invent the eight hour work day; that was the labor unions.” • Tom Payne contributed the following advice: Do a little bit on a project each day. Its amazing how much you eventually get done if you stick to this kind of regime. 3.3 Grammar writing and career Field linguists should approach a language with the intention of constructing a broad knowledge of it. Many do, but not all of them record it in an organized way that can be turned into a publishable grammar. The majority — I suspect — fail to do the modest additional work this requires simply because they haven’t put writing a grammar on their agenda and made it part of their lives. So here are some suggestions about building grammar writing into your life: •
Make writing a grammar a personal goal. If you are fortunate writing a grammar may become a comfortable part of your career, but even if it does not, you
193
194
David J. Weber
• •
• •
•
•
may still be able to realize your goal! (For example, Father Swisshelm found time to make major contributions to the study of Quechua in addition to carrying out his priestly duties; see point 3.2, page 437.) Commit to growing in ways that enable you to write a grammar. Shape your interests so that reading relevant literatures (linguistics, creativity, problem solving, …) becomes a pleasure, and so that difficult problems become exciting challenges. See Section 3.4. Build your library accordingly. Some suggestions about what a grammar-writer’s library ought to contain are given in Section 3.6. Attend professional meetings (conferences, workshops, seminars) to expose yourself to a wide range of current ideas and to interact with others who share an interest in the language you are documenting or topics raised by it. Steer yourself toward situations where grammar writing is appreciated. Whether choosing a graduate school in which to study or choosing a university in which to teach, favor those that value documenting; avoid those that despise this as “mere description.” Learn to be a creative analyst; see Section 3.5.
3.4 Choose useful entertainments Most entertainments, we must learn to enjoy. For example, few enjoy their first experiences playing tennis. The frustration, humiliation and exhaustion are rewarded with aching muscles. But after some learning and experience, tennis becomes fun. We might cheat ourselves out of a lot of fun by thinking of tennis only in terms of some initial experience. I used to enjoy reading chess books, i.e., replaying great games with a written commentary explaining what was going on. While I found this entertaining, shortly after starting my work in Peru, it struck me that it wasn’t a useful entertainment. So I looked for a useful alternative. At some point I discovered that reading linguistics could be entertaining. Mind you, at first it was just like Tennis I: frustrating, humiliating and exhausting. But the more I read, the more I enjoyed it. (Much linguistics is written like a narrated game, the struggle between competing analyses and theories.) A while back I had a yen to read a good novel (which I do now and then). Then I thought, “Rather than read a novel in English, why not read one in Spanish?” I read Arguedas’ Yawar Fiesta. At first it was slow, laborious going, not at all pleasant. (It has a lot of “dialect,” like reading Huckleberry Finn.) But soon I was enjoying it, just like reading a novel in English. Reading in Spanish has helped me speak and write Spanish better. I had picked a useful entertainment.
Thoughts on growing a grammar
Another kind of book that I have come to enjoy are books about computer programming. Mind you, when I first started into computing, it wasn’t fun. But I learned to enjoy it. You might say that I willed to learn to enjoy it, and now I really do enjoy it. I remember how much fun I had reading Aho, Kernighan and Weinberger’s The AWK Programming Language. Bently’s Programming Pearls books are also fun and rewarding. I can hear some of you laughing, saying “What a weirdo!” Well, I may be abnormal, but the nature of that abnormality is that I decided to enjoy some pretty non-standard entertainments, ones that were useful. 3.5 Creativity and problem solving Yes, it is possible to learn to be creative, just like it is possible to learn to draw. Yes, some people have a natural knack for it, but that doesn’t mean that the rest of us can’t learn it. There are some good books on creativity and problem solving. One of my favorites is Adam’s Conceptual Blockbusting (Adams 1986). Two classics well worth reading are Beveridge (1958) and Polya (1957). Here’s a final word of encouragement. In my library in Huanuco I have on my shelf a book on decipherment, in the final paragraphs of which the author reflects on a curious fact: that in the history of decipherment, again and again, correct conclusions were reached based on faulty reasoning, false premises, wrong data, and so forth. Well, sometimes it works like that for linguists, too. 3.6 The grammar-writer’s library What should a grammar-writer’s library contain? What should the library of someone who will be helping others write grammars contain? Here are some suggestions: Linguistic: • • •
•
Tom Payne’s Describing morphosyntax (Payne 1997). one or more sample grammars, ones that you might emulate. a dictionary of linguistic terminology. If the language in which you are writing the grammar is not the one in which you received your linguistic education, a bilingual dictionary of linguistic terms may be very helpful. Any available sources on the language being described, on closely related languages and on the language family to which it belongs.
195
196
David J. Weber
Writing and editing: a manual of style that fits your publishing milieu, such as The Chicago Manual of Style (1993) or Strunk and White (1999). I benefited from the very readable van Leunen (1986) By the way, the chapter of The Chicago Manual of Style that deals with indices has been revised and is available as a separate publication Indexes: A page chapter from the Chicago Manual of Style. An advertising blurb says about this: “The revised chapter on indexing offers considerably more coverage of computer-assisted indexing, including its advantages, its risks, and its modifications of the process. It also provides more examples and updates recommendations on style and method.” I mention this because many grammars have inadequate indices. A grammar tends to be large complicated document, and the index is an important tool for making use of it. Too often it seems that indexing is the last task to which the author and/or editor attended, when there is pressure to get the book to market. • books about writing. These are myriad, almost entirely written for people aspiring to write something marketable, like a novel, a short story, or such. Unfortunately, I have not yet found an inspirational book about technical writing. (Is that an oxymoron?) •
There is no pretense that these meager suggestions are adequate. They are included simply to stimulate readers to begin thinking about what books and other bibliographic resources will help them write a better grammar. To conclude, permit me two final comments. First, take whatever opportunities you have to discuss grammar writing with others who have written grammars. The comments and advice I have given are limited by my experience; others are sure to have different perspectives motivated from their experiences. Second, approach the task with joy. Yes, writing a grammar is a lot of work, but mostly it is fun work. If you have the opportunity to write a grammar, rejoice that you are among the fortunate few to whom this privilege and pleasure is given. And posterity will thank you.
Notes * This paper benefited from the comments of various people, among them Albert Bickford and Tom Payne. They are not responsible for my errors.
Thoughts on growing a grammar
1. This paper was originally written for a workshop on grammar writing. I have tried to maintain a measure of informality, e.g., liberally using the second person throughout. 2. By the way, T. Givon once referred to this sort of approach as “secondary field work,” that is, the person doing the writing and some analysis has little or no direct contact with the language, but works with an experienced field linguist, who has had decades of contact and done a lot of basic analysis. 3. Many books tell the story of the decipherment of Linear B but the classic is Chadwick (1950). 4. Of course, many dictionaries include an outline of the grammar, raising many of the same issues. 5. Texts per se have a limited value; what is really needed is text in context. 6. Subsequent investigation revealed that the man had poisoned the deceased. He had been given milk by a neighbor and guessed that it had been poisoned (which he verified by giving some to his dog). He passed it on to the victim, whose land and animals he coveted, hoping to pin the murder on the neighbor who had given him the milk. 7. The example is as follows: a. Ø shunta-ra-n ganchis canasta aypa-j-ta Puchu-j-cuna-ta exceed-sub-pl-obj gather-past-3 seven basket suffice-sub-acc b. -no -sim ‘They gathered up the leftovers, a. seven baskets full.’ b. sufficient to fill seven baskets.’ 8. By the way, I am assuming that you can shape your attitudes, develop habits, and learn disciplines. If you believe that you can not do so, then you probably should not bother to read further. 9. Attributed to Tim O’Reilly; see Dougherty (1992:2, footnote) 10. About 1973 the Ministry of Education determined to publish, for six Quechua languages, six grammar–dictionary pairs. The outline I had developed was adopted with small modifications for those grammars. 11. Because grammars tend to be large documents, there are many reasons to write them as multiple computer files. Some of the world’s most popular editors were originally designed for short documents, like business correspondence, reports, and such. As initially designed, they dealt with each document as a single computer file. If they provided complex features like automatic numbering, cross-referencing, the generation of a table of contents, and support for index generation, these applied only to a single file. With time, to accommodate large documents written as multiple files, the editors were patched for extending these capabilities across multiple files. One example of this is the introduction of support for a “master document”. However, these extensions introduced new complexities and have become notorious for causing problems. 12. There are various emacs editors. The most sophisticated is probably the Free Software Foundation’s GNU emacs.
197
198
David J. Weber
13. If you are not familiar with regular expressions, put your web browser to searching for “regular expression” and it will find scores of sites explaining them. Wikipedia has an excellent article on regular expressions. 14. “El /qa/ del pre-proto quechua” in Weber (1987). 15. Cuentos y Relatos en el Quechua de Huaraz Swisshelm, Germain. Cuentos y Relatos en el Quechua de Huaraz Volume I. Huaraz, Perú: Priorato de San Benito. 1974. 16. This principle can be applied to decisions too. We often stew about decisions that, when everything works out, never had to be made. (Are we fostering the illusion that we are really in control?) We should ask, “Do I really need to make this decision?” and “By what time is it necessary to make this decision?” By waiting, new circumstances may arise that make the answer obvious.
References Adams, James L. first edition 1974, second 1979, third edition 1986. Conceptual blockbusting: A guide to better ideas. Reading, Massachusetts: Addison-Wesley. Beveridge, William Ian Beveridge. 1950. The art of scientific investigation. New York: Vintage Books. Chadwick, John. 1950. The decipherment of Linear B. Cambridge: Cambridge University Press. The Chicago manual of style. 1993. Chicago: University of Chicago Press. Dale Dougherty. 1992. sed & awk. Sebastopol, CA: O’Reilly & Associates. Payne, Thomas E. 1997. Describing morphosyntax: A guide for field linguists. Cambridge: Cambridge University Press. Polya, G. first edition 1945, second edition 1957. How to solve it: A new aspect of mathematical method. Princeton, New Jersey: Princeton University Press. Strunk, William, Jr. and E. B. White. 1999 (fourth edition). The elements of style. Boston: Allyn and Bacon. van Leunen, Mary-Claire. 1986. A handbook for scholars. New York: Knopf. Weber, David. 1987. Estudios quechua: Planificación, historia y gramática. Pucallpa, Peru: Instituto Lingüístico de Verano.
The linguistic example* David J. Weber Summer Institute of Linguistics
Good language descriptions liberally illustrate their claims with examples. The author must select and order examples, and provide accompanying information. The example may include a reference number, the example in multiple forms (phonetic, phonemic, morphemic or morphophonemic, written), brackets and categories, glosses, translation, punctuation, functional annotations, grammatical judgements, subscripts, empty categories, ellipses marking, information about the author and language variety, attention-directing mechanisms, and so forth. Formatting these diverse sorts of information is a non-trivial task; suggestions are given for “best practice.” The delivery of documents on screens (rather than on paper) makes possible some dynamic enhancements such as inspecting an example’s textual context, toggling on/off various types of information, controlling highlighting and conflation.
1. Introduction The linguistic literature is populated by a menagerie of “specials”: tables, trees, maps, HPSG’s rectilinear attribute-value matrices, RG’s curvaceous stratal diagrams, and so forth. Among these, the most important for baseline language descriptions is the example: words, phrases, sentences, and text fragments used to illustrate claims made about the language under consideration. Good language descriptions liberally illustrate their claims with examples, ideally ones drawn from natural discourses of diverse genre.1 Good examples, well deployed, are a major factor in making a grammar good. Think of a grammar like a Museum of Fine Art. The collection is laid out topically in galleries, so we might come to the Gallery of Relative Clauses. Each piece is tastefully framed, and light is provided to bring the best out of each piece. One can stop and ponder, but is eventually drawn from one awe-inspiring piece to another. A museum, of course, is not a warehouse, which may contain an incredible store of pieces, each carefully shelved according to some organizational scheme. A
200
David J. Weber
grammar must be more — much more — than simply an annotated data catalog. But keep in mind that some day the examples may be appreciated more than the author’s fine words giving some clever analysis or theory. With time the claims may become uninteresting, as concerns and perspectives change, but examples remain as near-primary evidence. (Of course, for a long time to come readers will appreciate surrounding text that helps them understand the example, such as a description of the context in which the example was uttered.) Perhaps the most important reason to pay careful attention to examples is that they stimulate and exploit abduction. Abduction is inference to the best explanation: for an array of phenomena one abduces (guesses) possible explanations. Each explanation is “probably true” in proportion to how well it accounts for the array of phenomena, with higher probability for explanations that cover diverse phenomena. An example (a phenomenon) normally follows the claim (explanation) that it illustrates. Nonetheless, it can stimulate abduction; indeed, the examples should be ones from which a reader might have abduced the claim. Further examples can strengthen the claim, especially if these are diverse, either within the example or used in different ways in different contexts. Diverse examples strengthen the claim because diversity makes it more challenging to abduce a better competing explanation.
2. Selection, order, framing, enrichment When writing a grammar, the author must select, order, frame, and enrich examples, as discussed in the following sections. 2.1 Selection On what basis should the author select examples for incorporation into a text? Here are some suggestions: • •
• • •
First and foremost, the examples must illustrate the claim being made. The examples must be sound, preferably ones spoken or written by a native speaker in natural discourse. Such examples should always be preferred to elicited examples. Avoid examples that presume a contrived context, if possible. Choose a set of examples that illustrate a range of uses. Otherwise the reader might think that the claim holds only for a narrow range of uses. For the same reason, choose a set of structurally diverse examples. An example that is highly similar to another adds virtually nothing. Finally, when all else is equal, choose examples that are culturally interesting.
The linguistic example
There are other kinds of concern: 1. Beware of various types of bias. To take one case, the examples of some grammars reflect gender bias: subjects (agents) are more likely to be male than female; females are more likely to appeared as objects (undergoers) or in oblique roles; examples in which an undergoer is adversely affected by an agent (like “X struck Y”) are more likely to have a male agent and a female undergoer than the other way around; and so forth. Likewise, the examples may reflect racial prejudice, or even hatred for members of another human group. Such biases might reflect attitudes held by the speakers of the language under consideration, or at least of those who provided the data. It may be residual in the language’s literature (both oral and written). Or it might simply reflect what speakers regard as note-worthy. A corpus formed from real, natural instances of language use will reflect the prejudices of the speakers. As examples are taken from the corpus and incorporated into a grammar, the grammar will reflect those prejudices. It will do so, that is, unless the author is aware of the problem and exercises good judgement. I am not advocating the imposition of some radical notion of political correctness, but the problem can be reduced by the judicious choice of examples and occasionally by small adjustments (approved by a native speaker). 2. A grammar writer should bear in mind that the examples in a grammar will project an image of the speakers of the language and their culture, one that may be seen around the world (if made available on the web) and by speakers of the language, now and in future generations. So grammar writers should take care not to expose — or inadvertently perpetuate — prejudices and other aspects of the culture that might embarrass its speakers. 3. If you work from a corpus — as you should — watch out for examples that might compromise individuals or groups. Bear in mind that some people love to tell and retell stories that embarrass or damage their enemies. These stories may be ethnographically interesting but quite inappropriate for a grammar, for the reasons just mentioned. 4. Examples should be avoided that could limit the usefulness of the grammar. For example, when a Huallaga Quechua grammar describes the use of na-, a verb devoid of semantic content, to create suspense, a fine example would be: (1) …pullan pagasnaga nasha auquenga. ‘…in the middle of the night the old man did him.’
Following this sentence, for the next 132 words the reader wonders what the old man did to his wife’s lover. Then the text ends with:
201
202
David J. Weber
(2)
Quiquin wañuraycachir auquenga rucsuntash cuchuriycuran. Bulsicacurcur apacun. ‘He himself, the old man, killed him and cut off his testicles. He put them into his pocket and took them.’
This nicely illustrates the phenomena but might keep the grammar from being used in, say, a secondary school.
2.2 Order It is often appropriate to use several examples. When this is so, how should they be ordered? Here are some suggestions: • • • •
Begin with those that best exemplify the claim being supported, those for which the relationship to the claim is most clear. Progress from the simplest (usually the shortest) examples to the more complex ones. Progress from unmarked cases (vanilla) to more marked ones (rocky road). If included at all, put ambiguous cases last. These would be examples for which there is an alternative interpretation that might undercut the example’s support for the claim.
2.3 Framing Think of a cut diamond. Alone, it is valuable and holds a certain charm, but when set in a piece of jewelry, then it becomes truly beautiful. Examples are like that. Examples must be framed. Generally it does not suffice to make a claim and then simply tack on one or more examples. The reader needs help to see how the example illustrates the claim. For example, consider the following: Barasano verbs bear suffixes agreeing with the subject of the clause: (3) sĩg-o ĩa-a-bõ ‘The woman sees.’ one-fem see-pres-fem.sg
Now compare this to the following, which provides help explaining how the example relates to the claim:
Barasano verbs bear suffixes agreeing with the subject of the clause. For example, in (4) the verbal suffix -bõ agrees with sĩg-o ‘woman’: (4) sĩg-o ĩa-a-bõ ‘The woman sees.’ one-fem see-pres-fem.sg
The linguistic example
In (4) the order is claim-help-example. It is sometimes better to put the help after the example. If so, the order would be (1) the claim, (2) some text like “Consider the following example:”, (3) the example, and (4) help explaining how the example relates to the claim. Now a most solemn piece of advice: integrate claims and examples. Some authors present a group of examples followed by various claims about the examples. Those claims may even be embedded in lengthy discussion.2 Avoid this. Try to keep each example as close as possible to the claim it illustrates. 2.4 Enrichment Typically, an example is a text fragment enriched in various ways. Some enrichments are strictly linguistic: The boundaries between morphemes are marked by hyphens. Each morpheme is identified by a gloss (“tag”). A free translation is included, usually in some major language. (For grammars to be delivered in two or more languages, a translation must be given in each of those languages. The software that renders the grammar would display the appropriate translation(s) based on user-defined preferences and the context.) • Structural units may be indicated with brackets. • Categories may be indicated, usually as subscripts to brackets. • Functions (e.g., subject, instrument, source, and so forth) may be given for noun phrases. • Grammaticality judgements may be indicated (*, **, ?, ??). • Subscripted indices may indicate coreference or disjoint reference between referring expressions. • Empty categories may be indicated: Ø, e, t, PRO, pro • Ellipsis marking may indicate that the example is only part of a larger unit. • Explanatory notes may be attached to morphemes or larger units. In an inkon-paper environment these might be rendered as footnotes or endnotes. In an electronic delivery environment they might be activated by the user (by clicking or hovering with mouse). • A digitized recording of the example might be attached to the example. And so forth. • • •
Some enrichments are aids to the reader, particularly to direct attention: •
Attention may be directed to a particular aspect by italics, bolding, underlining, or some other highlighting mechanism.
203
204
David J. Weber
•
Multiple examples may be conflated by means of braces or parentheses to facilitate the comparison between alternatives. • Punctuation may be added. And so forth. Some enrichments serve to identify the example in the context of the document in which it is cited: Normally each example bears a number identifying it in the description. This number is used for referring to the example, either from within the description or from some other document. • An example may employ internal identifiers for alternatives within the example itself; e.g., “See example 44b.” might direct the reader to the second alternative of example 44 (on page 13). And so forth. •
Some enrichments give information about the example (its source, context, etc.): • • • • • •
Speaker: name, age, sex, dialect,… Context: when, where, to whom,… the sentence was spoken. Register: formal, colloquial,… Mode of production: written, oral, recorded, videotaped,… Textual context: a reference to the text from which the example is drawn, and its position in that text. Situational context: the context in which an utterance might be used, as illustrated in (5):
a. -mi (-dir) (5) qam-pis maqa-ma-ška-nki b. -ši (-rpt) c. -či (-cnj) you-also hit-⇒1-prf-2 ‘You also hit me.’ a. situation: I felt you hit me and realized it was you. b. situation: I was unconscious when you hit me, but someone told me that you did so. c. situation: Various people hit me and I surmise that you were one of them.
•
Residence: the document, archive, or collection, … possessing the text from which the example is taken And so forth. Elements of these various types are combined and given visual form, traditionally on a printed page, but now increasingly on a computer screen.
The linguistic example
3. Layout An example must be laid out, that is, its various parts must be located on the page or screen. This may involve the conflation of two or more examples, deciding how to wrap long lines or break examples from one page to the next, and adding enrichments, particularly those that direct the reader’s attention. We discuss these in turn. 3.1 The layout of the major elements There are only loose, unwritten conventions for how the parts of an example are laid out.3 A rather standard form of example has a number (number); the text fragment, with hyphens dividing morphemes (morphemes); morpheme glosses, usually aligned word-by-word but sometimes aligned morpheme-by-morpheme (glosses); and a free translation (‘translation’). These are laid out as follows: (number) morphemes glosses ‘translation’
For example: (6) [[yapya-y]-ta uša-na-n]-ta-ši šuya-ra-yka-n plow-inf-obj finish-sub‑3p-obj-rpt wait-dur-impfv‑3 ‘He is waiting for him to finish plowing.’
In (6) the morphemes are represented phonemically with characters familiar to linguists. There are various reasons for also including the example written with the writing system used by speakers of the language: •
•
•
Linguistically-oriented representations (phonetic, phonemic, morphophonemic) may be inaccessible to speakers of the language whereas including the traditional/conventional writing system may make it easy for them to read.4 Linguists who study a language seriously should learn its writing system so as to be able to benefit from other documents written in the language. This learning can occur simply by seeing the practical orthography along with a more linguistically-oriented representation. The practical orthography may have information not contained in the morphemic representation. For example, in example (10) below the practical orthography represents phonetic detail not indicated by the morphemic form.
205
206
David J. Weber
The most normal place for the practical orthography (writing) is perhaps on the very first line. Because it is primarily for readers who can read and understand it, it is not necessary to align it with the glosses. (number)
writing morphemes glosses ‘translation’
For example: (7)
“¡Ama aywaychu!” nir willashcä. ama aywa-y-ču ni-r wiλa-ška-: no go‑2imp-neg say-SS advise-prf‑1 ‘I told him not to go. (lit. I advised him saying “Do not go!”)’
When examples are short, this sort of layout may waste space. When printed, it can increase the cost. When viewed on-line, it can push relevant text off the screen. Therefore, when space permits, it may be desirable to use alternative layouts. The translation, for example, might follow the morphemic representation: (number) writing morphemes ‘translation’ glosses
For example: (8) Liguiyta yachachimanga. ligi-y-ta yača-či-ma-nqa read-inf-obj learn-caus-⇒1‑3fut
‘He will teach me to read.’
Or the written form may also fit there: (number) morphemes writing ‘translation’ glosses
For example: (9) huk runa ka-ša one man be‑3prf
Juc runa casha.
‘There was a man.’
Short examples might all fit on one line: (number) morphemes (glosses) writing ‘translation’
For example: (10) a. imana-ša-taq (what.do‑3prf-¿?) ¿Imanashataj? ‘What did he do?’ b. imana-šaq-taq (what.do‑1fut-¿?) ¿Imanashätaj? ‘What will I do?’
The linguistic example
If — as I am assuming — the layout of the major elements depends on the available space, and if in a web-based environment column width is under the control of the reader, then the rendering engine should include a component that adjusts the layout depending on the available space and user preferences. 3.2 Conflation Two or more examples may be conflated by means of braces, parentheses or brackets. In some cases only words are conflated; in others the morphemes within a word might be conflated. For example, consider the following, taken from the International Journal of American Linguistics 65:159: (44a) li:-ta-pa:-chi’:-ní:t tasiw caja instr-ingr-belly-tie-pfv rope box ‘The box has been tied up with a rope.’ tasiw caja (44b) li:-ta-maq-chi’:-ní:t instr-ingr-body-tie-pfv rope box ‘The box has been tied up with a rope.’
These could have been conflated as follows: a. pa: belly -chi’:-ní:t tasiw caja b. maq instr-ingr -tie-pfv rope box body ‘The box has been tied up with a rope.’
(44) li:-ta-
There are various reasons for conflating examples: •
• •
It makes the example more readable: without conflation the reader must scan the examples to isolate the parts being compared or contrasted (as discussed regarding example (11) below). With conflation this is immediately obvious. It makes better use of space. Thus, for a printed page it is more economical, and for a computer monitor it allows more context to be kept in view. It may simplify wrapping examples across lines and breaking lines over pages: Without conflation, two or more parallel lines normally wrap or break independently, which means that as the column is narrowed the layout becomes increasingly difficult to read. When two or more examples are conflated to a single line, this is more likely to wrap or break without creating problems (assuming that the portion in braces moves as a piece).
207
208
David J. Weber
Examples are conventionally conflated with either braces or brackets, as discussed in the following sections. 3.2.1 Braces Linguistic Inquiry is now virtually devoid of braces except the characters { and }. (This is probably as the result of making the journal available on-line, thus submitting to the limitations of HTML.) This has a cost; for example, consider the difficulty of reading and the wasted space in (11), which is example 33 from LI 30:658. (“BP” stands for Brazilian Portuguese.) (11) a. Eu encontrei as minhas velhas amigas e (BP) I met the.f.pl my.f.pl old.f.pl friends.f.pl and amigos juntos. friends.m.pl together.m.pl b. Eu encontrei as minhas velhas amigas e I met the.f.pl my.f.pl old.f.pl friends.f.pl and amigos no mesomo dia. friends.m.pl on.the same day. ‘I met my famous old female friends and male friends together/on the same day.’
This could be conflated as follows, both making it easier to read and saving space: (12) Eu encontrei as minhas velhas amigas e amigos (BP) I met the.f.pl my.f.pl old.f.pl friends.f.pl and friends.m.pl a. b.
juntos.
‘I met my famous old female friends and no mesomo dia. male friends a. together.’ on.the same day. b. on the same day.’ together.m.pl
Here are some further examples (from Huallaga Quechua), ones that illustrate both the utility and potential complexities of using braces. In (13) note the subscripts i and j in both the morphemic representations and the translation:5 (13) a. Magarcamaptin jaytashurayqui. a. -pti maqa-rkU-ma- ds -ni hayta-šu-ra-ykij b. *-špa 3p kick-⇒2-pst‑2p hit-up-⇒1 ss a. ‘After hei hit me, hej kicked you. (i≠j)’
The linguistic example
There may be braces within braces, that is, conflation within conflation. (14) conflates four examples. ((14b) and (c) are grammatical while (14a) and (d) are not.) (14) b. Magarcushpan jaytamaran. c. Magarcur jaytamaran. -špa a.*-ø maqa-rkuss b. -n (‑3p) hit-arr c. -ø -r ss d. *-nin (-3p) b,c. ‘After hei hit himj, hei kicked me.’
hayta-ma-ra-n kick-⇒1-pst‑3
Note that in (14) there are no right braces matching the smaller left braces. Is this good practice? Suppressing the right braces may look better, but it might complicate some computational tasks. In (15) a single left brace is matched by two right braces. Again, is this good practice?
(15)
a. -na-:-paq -sub‑1p-pur b. -q pay-ta rika- -sub him-obj see c. -na-:-paq -sub‑1p-pur d. *-q -sub a,b. ‘I came to see him.’ c. ‘I was waiting to see him.’
šamu-ška-: come-prf‑1 šuya-ra-yka-ška-: wait-dur-impfv-prf‑1
The possibilities, of course, are limitless, and this is a problem! It would be nice to have a statement of “best practice” that would gently constrain authors’ inventiveness. Authors must consider the cost of using conflation mechanisms, keeping in mind that readers’ familiarity with the use of braces, indices, and such devices. Use them judiciously. Above all, avoid needless complexity. 3.2.2 Brackets Square brackets are sometimes used in contrast to (curly) braces to signal a correspondence among bracketed elements. For example, consider (16): (16) a. kay-man aywa-mu-n here-goal go-to.here-3
‘He comes here.’
209
210
David J. Weber
b. čay-man aywa-n there-goal go‑3
‘He goes there.’
This might be conflated as in (17), indicating that kay co-occurs with -mu and ‘He comes here.’, while čay co-occurs with the absence of -mu and ‘He goes there.’: (17)
kay here čay there
-man aywa goal go
- mu to.here -n -ø 3
‘He comes here.’ ‘He goes there.’
My impression is that square brackets are used less and less, and I would like to think that they are a thing of the past. Perhaps this is because it is possible to more explicitly express the correspondence of elements using internal identifiers like “a.” and “b.” as in (18): (18)
a. kay here b. čay there
-man aywa goal go
a. -mu to.here -n 3 b. -Ø
a. ‘He comes here.’ b. ‘He goes there.’
So I recommend not using square brackets. Maybe braces, but not brackets. 3.2.3 In-line conflation There are in-line conflations: •
A XAY X A/B/… Y is equivalent to X B Y, which conflates X B Y .
The examples in (19) are from LI 30:545: (19) If/As/When you eat more, you want correspondingly less. If/*As you had eaten more, you would want less.
•
Ø X (*A) Y is equivalent to X Y, which conflates *X Y Y . *A *X A Y
Example (20) is from LI 30:568. (“t” is a trace.) (20) This is the kind of rice that the quicker (*that) you cook t, the better it tastes.
•
*Ø *X Y Y X *(A) Y is equivalent to X Y, which conflates . A *X A Y
The linguistic example
Authors should assess the costs and benefits of such conflations for the intended audience. In language descriptions intended for non-technical audiences — both present and future — it may be wise to limit the use of conflation. (However, see Section 3.4 below.) 3.3 Line wrapping and page breaks When an example must be broken across a page boundary, it is important that this be done at certain points and not at others. For example, the glosses should never be separated from the morphemes to which they correspond. Likewise, when an example is too long to fit on a single line, it must be “wrapped” in a way that does not interpose text between, say, the morpheme decomposition and the corresponding glosses. To break some lines attractively may require hyphenation. For example, for the Spanish version of my Huallaga grammar, both the Quechua written form (practical orthography) and the Spanish translation were hyphenated, that is, “discretionary hyphens” were computationally introduced. Note: hyphenation differs from language to language subject to convention, syllable structure, and even subjective esthetic criteria. Although quite obvious, we should not forget that the space in which an example is rendered depends on the document context: if it is embedded within an item in a list, where each item is indented, then the effective column width for the example is correspondingly narrower. 3.4 Some future possibilities Documents are increasingly published electronically and read on screens rather from printed pages. Technology will progressively enhance the display of documents in ways that are not possible with ink-on-paper delivery. Software could be developed to enhance the presentation of language descriptions; here are some possibilities: inspect the context: Traditionally, what you see is all you get. Although an example might be a fragment of a text, when used in a linguistic description, the reader can not see what precedes or follows it in the original text. In the future, when examples are fragments of online texts, software should allow the user to dynamically inspect the original text surrounding the example. toggle on/off parts: It may be useful to turn off or on the display of certain kinds of information. For example, native speakers may wish to toggle off parts they do not need, such as the morphemic representation, glosses and translation.
211
212
David J. Weber
Linguists not familiar with the language may wish to toggle off the practical orthography, while linguists familiar with the language may wish to suppress the gloss. Users should be able to tailor the display of information to meet their needs and preferences. buttons and hot zones: Buttons could be provided to activate certain kinds of secondary information, e.g., the speaker’s biographical information, the context of use, the example’s “residence,” and so forth. Perhaps if the gloss is toggled off, morphemes could be “hot,” so that clicking on or hovering over them would trigger the display of information about the morpheme: the gloss, the category, perhaps even a lexical entry for that morpheme. enhanced focus mechanisms: Traditionally attention is directed by static effects like bold or italic type, or by underlining. Now it should be possible to use coloring and effects like blinking. It might be useful to have three variants of comparison, one to signal ‘note the similarity of these’, one to signal ‘note difference between these’, and a default for simple comparison. control conflation It should be possible to control conflation, with the default appearance determined partly by the author and partly by the reader. For example, the author may give a conflated form, but a reader may wish to “deconflate” the alternatives to see them as a list of sentences without braces, brackets or parentheses. At present electronic documents can be enhanced in ways like those just mentioned only by people with considerable technical training and skill. Linguists lack the software with which to implement such possibilities in the course of writing a language description. Ideally language data would be managed — and grammars written — in a computational framework that integrates grammar and corpus, with examples existing in the corpus but accessed from the grammar. Examples would not be “taken from” a text but displayed therein. Grammar writers need hospitable authoring environments, with tools that are powerful and flexible, yet reasonably easy to learn and use. Until these are available we labor under the limitations of ink-on-paper.
Notes * This paper draws from a paper presented at the Workshop on Web-Based Language Documentation and Description, December 2000, Philadelphia. That paper is available at www.ldc.upenn. edu/exploration/expl2000/papers/weber/weber.pdf. 1. My grammar of Huallaga Quechua, for example, has over 1700 examples.
The linguistic example
2. Chomsky’s Lectures on Government and Binding has many fine examples! This may be appropriate in a context for which theory is primary, with examples simply providing grist for the theoretical mill. It is quite a different matter for a language description, in which theory is generally a servant to description rather than its master. 3. We need guidelines leading to good practice and curbing individuals’ tendencies toward the idiosyncratic. Grammar writers need a style sheet for examples! 4. Except for material dealing with phonology, English examples use the practical orthography. Readers would be very put off if they had to read English examples in a phonetic, phonemic, or morphophonemic representation. 5. Avoid examples like (13) and (14), obviously elicited and so richly endowed with violence.
213
Index
A abduction, 200 abstract syntax, 101 adjectives, 38, 71-89, 98-106, 109, 119, 133, 158 adpositions, 71, 77, 89, 91, 9495, 99 adverbial modifiers, 134 noun phrases, 94 adverbs, 16, 73-78, 82, 93-94, 101-103, 119, 133 agglutinative additivity, 32 analytic techniques, 184 anthropological linguists, 12-13 applicatives, 94-97, 99, 107, 109-110 archiving, 3, 125, 174 articles, 6, 11, 16, 74, 100-101, 116, 125-29, 152, 157, 166-67, 175, 198 attitudes, 25, 42, 73, 184, 191, 197, 201 authenticity, 149, 180 authoring environments, 186-188 Azerbajdzhan, 31, 35, 37 B backup, 174, 189 Barasano, 202 basic linguistic theory, 116, 123, 159-160 bidialectal, 14 bilingual, 6, 14, 195 body part, body part locatives, 90, 91, 92, 93, 108 body parts, 52, 57-58, 89-94, 97, 99, 107-110
Bora, 176 braces, 204, 207-208, 210, 212 brackets, 154, 199, 203, 207-208, 210, 212 broad description, 175 C case, 43, 57, 86, 92, 96, 98-99, 139-140 Chickasaw, 71-77, 84-89, 94-110 Choctaw, 98, 109-111 clarity, 148, 152, 154 clefts, 141 co-authorship, 175 cognitive, 13, 43, 142, 150 collective fieldwork, 25, 36 community involvement, 4, 21 comprehensive, comprehensiveness, 8, 28, 52, 116, 12124, 143, 149, 152-55, 168 concise, conciseness, 153-4 concordance, 150, 183 conflation, 199, 205, 207-213 conjunctions, 73-74, 101, 103, 168 constituent order, 66 consulting, 178 context contrived, 200 hypothetical, 180 cultural, 6, 13, 16 discourse, 8, 11-16, 59, 68-69, 78, 117, 141-142, 166, 199-200, 211 of examples in a grammar, 204 situational, 204 sociolinguistic, 14, 121, 124 text in, 197 theoretical, 163
copy editing, 179 corpus, 28, 60, 173, 179-183, 201 corpus-based, 8, 173 creativity, 194-195 cross-reference, cross-referencing, 15, 117, 124, 150, 179, 187 cultural, context, 6, 13, 16 information in a grammar, 7, 8, 13, 121-122, 145 preservation, 20, pressure from majority languages, 5, 8 culturally interesting, 200 D Daghestan, 25-27, 34-37 data management, 182 data-driven, 8, 173 decomposition of big task, 191 demonstrative adjectives, 101 derivational morphology, 140, 141 description, grammatical, 1-8, 45, 52-54, 72, 75, 94, 113, 116, 120, 127, 135, 144 contextualized, 13, 37, 38 as discourse type, 181 functional, 138-139 holistic, 37 integral, 38 linguistic, 27-28, 175, 199, 211 traditional, 28, 137, 177 consistency in, 30 morphological, 33, 132-133 taxonomic, 35 phonological, 68, 119, 122, 124, 167
216
Index
descriptive adequacy, 116, 123 descriptive modifiers, 133 determiners, 74, 101-102, 104, 134 diachronic grammar, 13 information in a grammar, 147, 159 dialects, geographical, 11, 14, information in a grammar, 52, 54-55, 148, 154, 204 dictionary, component of language documentation, 2, 15, 33, 38, 175 difficulties in writing, 145 ethnographic, 16 linked to a grammar, 5-6, 13, 71, 117, 122, 178-179 parts of speech in, 71-77, 101 directional, adverbs, 16 meanings expressed via body part terms, 92-93 suffixes, 16 disadvantaged in the workplace, 175 dissertations, grammatical description as, 5, 125, 174-177 diverse (speakers, texts), 180 E editorial policies, 179 electronic documents, 212 electronic publication, 125, 175, 179, 213 elicited data, 6-7, 140-141, 182 examples in a grammar, 16, 180, 200 translations, 49, endangered languages, 8, 32, 68, 113, 149-150 enrichment, 200 ethnographic dictionaries, 16
ethnolinguistics, 17, 32 examples, constructed/elicited (vs. spontaneous), 64-67, 180, 200 formatting of, 165-166, 187, 205-211 use of in a grammar, 16, 47, 51, 54, 66-57, 59-60, 62, 68, 117, 121-122, 147, 152-153, 199-212 spontaneous (vs. constructed or elicited), 33, 64-67, 117, 121, 140, 200 structurally diverse, 200 prejudice in, 201
grammars, descriptive, 1, 12, 72, 105, 139 as dissertations, 5, 125, 174-177 partial, 8, 123-124 sketches, 8, 114, 117, 123-124, 173, 176 pedagogical, 141 reference, 52, 114-123 multi-dialectal, 168 grammar-writing workshop, 176 grammatical patterns, 17 grammatical relations, 90-94, 98, 141, 185 growing a grammar, 8, 173-174
F feedback, 174-176, 190 fieldwork, 7, 33-42, 44, 150-154, 167-168, 175-176 filing system, 184 folktales, 180-181 formal and informal communication, 15 formal theories, 13, 185 formalisms, 130, 154-155 formatting, 186, 187, 189 framing linguistic examples, 200 Franz Boas, 9, 12, 15 full-scale grammars, 114, 123 functional systems, 138 functionalism in linguistics, 13 functionally-oriented theories, 185
H head-final, 88, 99 hierarchical structure, 135 hierarchy, 38, 159 historical development, 12, 93
G gender bias, 201 genius of a language, 98, 100 genre, 28, 60, 122, 180-182, 199 glossing, 59, 117, 165-166, 183 glossed texts, 183 grammar sketches, 33, 123, 125, 176
I inalienably possessed nouns, 60, 97 indexing, 48-49, 116, 168-169, 188, 196 inference, 131, 200 in-line conflation, 212 intellectual property rights, 30 interjections, 73-74, 103-104 intonation, 7, 62, 68, 119, 123, 140-141 inventories, of grammatical elements, 7, 140, of parts of speech, 71, 74 of rhetorical styles, 49, phonological, 155, 167 IPA, 118, 124, 164-165 J Jane Austen, 73, 103-104 K Karuk, 15-16
L language consultants, 17, 26, 29, 41, 121 description, 2, 38, 175, 199200, 211-213 documentation, 1-5, 7, 9, 28 extinction, 1 planning, 22 preservation, 21 use, 5, 121, 179-182, 201 vitality, 113 layout, 205-207 lexical categories, 64, 73, 84, 93 lexical transitivity, 95 lexicon, 3, 15-16, 28, 72, 133, 178-179 lifestyles, 184 linguistic curriculum, 176 examples, 5, 6, 199 theories, 2, 22, 26, 29, 47-48, 116, 147, 159-164, 168, 184185 linguistic typology, 26, 156 local language, 14, 21, 23, 49, 188 locative objects, 97 M metaphor, 94, 127-130 metonymic extension, 90 Mixtec, 71, 77, 89, 93-95, 98-99, 105, 108, 111 morpheme glosses, 59, 117, 121, 165, 205 multidialectal, 14 multilingual, 14, 188 N narrative, 62, 32, 110, 182 national language, 21, 73, 115, 188 native speakers, 5-8, 121, 177, natural discourse, 131, 199-200 naturally occurring utterances, 16
Index
nouns, 73-75, 82-83, 133-135, noun phrases, 91-94 relational, 94-103, O online publication, 125-126 order, of examples in a grammar, 202 of parts of a grammar, 158 organization of a grammar, 45, 137, 147, 152-153, 157-159 Otomanguean language, 90, 93 P page breaks, 213 parts of speech, 71-78, 89-91, 101-104, 133, 169 participant reference, 138, 181 particles, 65-66, 74-75, 102-104, 141 philological tradition, 12 philology, 2, 13, 34 phonemes, 14, 135-136, 184 phonetics, 39, 118, 122 Pima, 99-100, 104, 109-110 Portuguese, 208 positive attitudes, 176 possessive adjectives, 101 possible sentence, 92, 137 postpositions, 89, 94-95, 98-99, 167 practical orthography, 118, 122, 164, 205-206, 213, 215 prejudices, 201 prepositional phrases, 91, 92, 94, 108 prepositions, 90-95, 99 problem solving, 184, 194-195 procedural text, 182 professional meetings, 194 pronouns, 46, 73-74, 133-134 publisher-compliant, 173 Q quantifiers, 74, 101-102, 149 Quechua, 180, 185, 201 question markers, 103-104, 109
R range of uses, 200 recordings, 3, 65, 122, 180, 203 reference grammars, 52, 114, 120, 123, 144, 147, 156-157, 168, 170 regular expressions, 189, 198 relational nouns, 94-103 relational noun phrases, 97-98 residue, 191 S Sapir, Edward, 12-13, 16, 71-72, 98, 100 semantic oblique, 95-96 situation, aspect, 160-164, contextual, 204 sociolinguistic, 15, 17 Slave language, 144, 168 social context, 7, 14, 121, 124 pressure, 5 ranking, 15 variation, 11-15 sociolinguistic variation, 11, 15 sources of information, 17, 115 special characters, 187 speech acts, 15, 142 structuralist, 35, 135-136, 139 structured document, 189 styles, discourse, 15, of interaction, 66 of presentation, 45, 153, 196, rhetorical, 49 sheets, 179, 189 subcategorized, 95, 97 Sudan, 4, 19-20 synchronic, description, 159 grammar, 13, 84, 163 variation, 14 syntactic behavior, 84, 89, 98 syntactic structures, 7, 48, 67, 119, 136
217
218
Index
syntax, analogy with phonology, 135-138 Chickasaw, 98 Chipewyan, 167 comparative, 147, Dyirbal, 158, place in a grammatical description, 120-124, 152153, 165 Slave, 168 Tabasaran, 36 Tsakhur, 37 Tsuut’ina (Sarcee), 158 Zapotec, 77, 79, 90-92 T terminology, 147, 154-155, 160, 164 harmonizing, 30 technical, 45, 47, 50-51, 154-155 theory-specific, 116, 123 texts, 40-41, 140 Alutor, 32-33 as language documentation, 1-4, 7, 15-16, 28, 38, 71, 120-121, 147 diversity of, 180 Dyirbal, 158 Fijian, 158 genres, 180-182 glossing of, 183 in context, 197 procedural, 181-182 Slave, 168
theories of grammar, 17, 149 theory-informed, 8, 173 traditional narratives, 17 translation, of examples in a grammar, 6-7, 30, 48-49, 59, 62, 117, 121, 165-166, 203, 205-206 of foreign texts, 117 typological adequacy, 185 typological classification, 99 U universal, characteristics of language, 13, 147 grammar, 100 quantifiers, 36, 149 unwritten conventions, 205 user-friendly, 113, 152-153, 168, 171, 173 V verb(s), 73-89, 91-92, 95-100, 102, 133-135, 158-160 adjectival, 84 argument frames, 35 Barasano, 202 Chickasaw, 73, 84-89, 95-98 forms, 15 Mohawk, 53, 60-61, 64-65 neutral, 76-83 paradigms, 32-33 phrases, 136, stems, 16 Slave, 158-163 Zapotec, 76-83, 91-92
verb-particle constructions, 102 voice, 150, 161-164 W web, 173-174 web-deliverable, 8, 173, 189 Wolof, 73-75, 104, 110 word classes, 119-120, 133, 158, 167 working groups, 177 writing system, 164, 205 written and spoken language, 15 written records, 12, 68 written texts, 3, 125, 180 X XML, 189 Y Yagua, 137, 138 Zapotec(an) languages, 73, 7778, 91-92, 99, 105-106, 110
In the series Benjamins Current Topics (BCT) the following titles have been published thus far or are scheduled for publication: 12 Dror, Itiel E. (ed.): Cognitive Technologies and the Pragmatics of Cognition. 2007. xii, 186 pp. 11 Payne, Thomas E. and David J. Weber (eds.): Perspectives on Grammar Writing. 2007. vii, 218 pp. 10 Liebal, Katja, Cornelia Müller and Simone Pika (eds.): Gestural Communication in Nonhuman and Human Primates. xiv, 275 + index. Expected September 2007 9 Pöchhacker, Franz and Miriam Shlesinger (eds.): Healthcare Interpreting. Discourse and Interaction. 2007. viii, 155 pp. 8 Teubert, Wolfgang (ed.): Text Corpora and Multilingual Lexicography. 2007. x, 162 pp. 7 Penke, Martina and Anette Rosenbach (eds.): What Counts as Evidence in Linguistics. The case of innateness. 2007. x, 297 pp. 6 Bamberg, Michael (ed.): Narrative – State of the Art. 2007. vi, 271 pp. 5 Anthonissen, Christine and Jan Blommaert (eds.): Discourse and Human Rights Violations. 2007. x, 142 pp. 4 Hauf, Petra and Friedrich Försterling (eds.): Making Minds. The shaping of human minds through social context. 2007. ix, 275 pp. 3 Chouliaraki, Lilie (ed.): The Soft Power of War. 2007. x, 148 pp. 2 Ibekwe-SanJuan, Fidelia, Anne Condamines and M. Teresa Cabré Castellví (eds.): Application-Driven Terminology Engineering. 2007. vii, 203 pp. 1 Nevalainen, Terttu and Sanna-Kaisa Tanskanen (eds.): Letter Writing. 2007. viii, 160 pp.