Asymmetric Events (Converging Evidence in Language and Communication Research (Celcr))

Asymmetric Events

Converging Evidence in Language and Communication Research (CELCR) Over the past decades, linguists have taken a broader view of language and are borrowing methods and findings from other disciplines such as cognition and computer sciences, neurology, biology, sociology, psychology, and anthropology. This development has enriched our knowledge of language and communication, but at the same time it has made it difficult for researchers in a particular field of language studies to be aware of how their findings might relate to those in other (sub-)disciplines. CELCR seeks to address this problem by taking a cross-disciplinary approach to the study of language and communication. The books in the series focus on a specific linguistic topic and offer studies pertaining to this topic from different disciplinary angles, thus taking converging evidence in language and communication research as its basic methodology.

Editors Marjolijn H. Verspoor University of Groningen

Wilbert Spooren

Vrije Universiteit Amsterdam

Advisory Board Walter Daelemans

Leo Noordman

Cliff Goddard

Martin Pütz

University of Antwerp University of New England

Tilburg University University of Koblenz-Landau

Roeland van Hout

Radboud University Nijmegen

Volume 11 Asymmetric Events Edited by Barbara Lewandowska-Tomaszczyk

Asymmetric Events

Edited by

Barbara Lewandowska-Tomaszczyk University of Lodz

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 Asymmetric events / edited by Barbara Lewandowska-Tomaszczyk. p. cm. (Converging Evidence in Language and Communication Research, issn 15667774 ; v. 11) Includes bibliographical references and index. 1. Asymmetry (Linguistics) I. Lewandowska-Tomaszczyk, Barbara. P299.A85A844    2008 415--dc22 isbn 978 90 272 3899 3 (Hb; alk. paper)

2007048316

© 2008 – 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

Table of contents Introduction

vii

Part 1. Event chains and complex events Asymmetry in English multi-verb sequences: A corpus-based approach John Newman and Sally Rice

3

Asymmetries for locating events with Cora spatial language Eugene H. Casad

25

Spanish (de)queísmo: Part/whole alternation and viewing arrangement Nicole Delbecque

53

What does coordination look like in a head-final language? Nayoung Kwon and Maria Polinsky

87

Verb serialization as a means of expressing complex events in Thai Kingkarn Thepkanjana Notional asymmetry in syntactic symmetry: Connective and accessibility marker interactions Katsunobu Izutsu and Mitsuko Narita Izutsu

103

121

Part 2. Subordination, nominalization, modification Subordination in cognitive grammar Ronald W. Langacker

137

Asymmetric events, subordination, and grammatical categories Sonia Cristofaro

151

Asymmetry reversal Frantisek Lichtenberk

173



Asymmetric Events

Transparency vs. economy: How does Adioukrou resolve the conflict? Kaoru Horie, Prashant Pardeshi, and Guy Kaul Relating participants across asymmetric events: Conceptual constraints on obligatory control Klaus-Uwe Panther The Portuguese inflected infinitive and its conceptual basis Augusto Soares da Silva The periphrastic realization of participants in nominalizations: Semantic and discourse constraints Liesbet Heyvaert

195

209

227

245

Asymmetries in participial modification Barbara Lewandowska-Tomaszczyk

261

Author index

283

Subject index

285

Introduction Asymmetric events – An interpretation Barbara Lewandowska-Tomaszczyk

The present volume is an outcome of the theme session Asymmetric Events, conducted at the 9th International Conference of Cognitive Linguistics held in Seoul in 2005. Some of the contributors to the volume, who did not take part in the conference and whose research does not necessarily fall into a classical cognitive linguistic paradigm but is relevant to the topic, have also been invited to contribute. The book introduces the concept of an asymmetric event which, for the first time in linguistic literature, has been identified in a more systematic way and analyzed in a number of different, frequently typologically or genetically unrelated languages such as English, Italian, Russian, Polish, Spanish, Portuguese, German, Japanese, Thai, Cora, Tobaqabita or Adioukrou. The theme covers topics concerning perception, conception and linguistic expression of the material referring to asymmetric events, i.e. two or more events of unequal status in an utterance. Papers in this volume present ways in which a linguistic description of main events in a sentence is different (morphologically, syntactically, discursively) from a description of backgrounded events in different languages and varieties, also at different historical times. The relevant themes deal not only with different types of complementation and nominalization, which can be considered prototypical instances of asymmetries in perception and cognition, but also with asymmetries in coordinate alternative, relative, adverbial, etc. constructions as well as in modifying constructions. The perspective is basically cognitive and functional, even though in a broad sense of cognition and function, and the relevant linguistic phenomena can be expressed in morphology, syntax and discourse in different languages and may have a variety of stylistic and ideological implications. The principles discussed in a number of papers, which govern form-meaning mapping within and across languages, in particular in the domains of subordination and other asymmetries such as e.g. those found in independent and asyndetic clauses i.e. clauses with no overt linking elements, point to the mental organization of perception and conception. One of the goals of the volume is to identify the relationship between the symbolic morphosyntax that is used to code asymmetries in various constructions and the asymmetries of the perception of events and their discourse roles, as expressed by these morhphosyntactic structures. Different researchers propose that the relationship between the more salient events

 Asymetric Events

expressed in main constructions and those whose profiles have been dominated by the more salient ones be perceived in terms of a continuum between constructions which possess autonomous profiles and those whose profiles are reduced in different ways. The authors examine parameters which contribute to the formation of asymmetry both within one utterance as well as cross-sententially, pertaining to system differences between fully elaborated event descriptions and those which are gradually more and more desententialized and lose or lack their assertive force. Cognitive linguists (cf. Langacker (1991:436) argue that e.g. a subordinate clause is describable as a structure whose profile is overridden by that of a main clause. This perspective, originally postulated in Langacker (1987a, 1987b, 1991) to account for a basic distinction between nominal and relational categories, depending on whether they profile a thing (as an abstract entity) or a relationship, has been further developed in a number of publications (e.g. Achard 1998, Talmy 2000). Problems of subordination and complementation as well as nominal, modifying and other constructions of the asymmetric type that refer to such events and their coding have also been analyzed in linguistic literature from various perspectives, e.g. linguistic typology and contrastive linguistics (Croft 2001, Cristofaro 2003) or (cognitive) functional-discourse analysis (Givón 1993, Horie 2000). This volume brings together researchers from various linguistics and cognitive science background and provides an opportunity to present and discuss matters of asymmetries in perception, conception and expression from different perspectives. The papers in the volume are divided into two main parts, Part I Event Chains and Complex Events and Part II Subordination, Nominalization, Modification. In the first part, the authors define their interpretation of the concept of an asymmetric event in terms of different viewing arrangements, a complexity of (chains) of events as well as with reference to certain discourse properties. They identify and analyze the semantic and syntactic patterns which include the instantiation of asymmetries in coordinate structures, serial verbs, spatial language and viewing arrangement, as well as part – whole relations, using different methodologies and including corpus data. Asymmetric events are also identified in the syntactic structure, in opposition to symmetries existing on the notional level. Part II is a discussion of subordination asymmetry in Cognitive Grammar and its relation to the problem of grammatical category in functional pragmatic approaches. Here what is also found is a discussion of the processes of so-called asymmetry reversal. Other authors explore cross-linguistically semantic and syntactic asymmetries in the processes of subordination and modification, and present analyses of infinitival constructions and other nominalizations, as well as of participial modification. The first article of Part I, which includes six papers and deals with chains of events and complex events, is a corpus-based contribution Asymmetry in English multi-verb sequences: A corpus-based approach by John Newman and Sally Rice (University of Alberta). The authors approach asymmetrical events through an examination of multi-verb sequences in English with particular reference to the V and V construction and



Introduction 

degrees of semantic integration of verbal concepts referred to in these constructions. The asymmetry manifests itself in the properties of the verbs as well as in different unfolding of the events of the simultaneous or sequential type. Eugene Casad (SIL and Gdansk University) analyses in his paper Asymmetries for locating events with Cora spatial language an intricate but highly integrated system of spatial language in Cora, a Southern Uto-Aztecan language spoken in Northwest Mexico. The author examines in detail certain asymmetries in viewing arrangement in Cora, where the locative particles and adverbs construe a scene in a subjective manner, and the locative verbal prefixes which construe a more objective viewing arrangement of the scene on stage. Casad discusses a degree of subjectivity accompanying these usages. The next paper of the first part is Nicole Delbecque’s Spanish (de)queísmo: part/ whole alternation and viewing arrangement. Delbecque discusses in her paper a semantic interpretation and variation of the Spanish (de)queísmo and, similarly to Casad’s analysis, conducts a cognitive analysis of the construction in terms of (asymmetric) viewing arrangement, this time though, with reference to part/whole alternation. In their paper What does coordination look like in a head-final language? Nayoung Kwon and Maria Polinsky (University of California, San Diego) note first of all that a construction, which has the form of a coordinate structure can have in numerous languages either all properties of coordination or all properties of subordination. The authors present and analyse such cases in Korean, a language which does not always express a straightforward contrast between coordination and subordination. In the paper Verb Serialization as a Means of Expressing Complex Events in Thai Kingkarn Thepkanjana (Chulalongkorn University) analyses verb serialization processes and the output of verb serialization in the form of serial verb constructions in Thai. The paper accounts for the serializability of verbs by analyzing semantic relationships between verbs in serial verb constructions and shows asymmetries in their combinations. Katsunobu Izutsu (Hokkaido University of Education at Asahikawa) and Mitsuko Narita Izutsu (Sapporo University) argue for a distinction between symmetric- and asymmetric-event descriptions in terms of some cognitive and discourse-related phenomena in their paper Notional Asymmetry in Syntactic Symmetry: Connective and Accessibility Marker Interactions. The authors show that the analysis proposed can account for the Japanese suruto and Korean geureoja. Part II of the present volume includes six papers and covers the topics of subordination, nominalization and modification and shows an asymmetric status of the events described in these constructions. Ronald Langacker (University of California, San Diego) reexamines the treatment of subordination in Cognitive Grammar in his paper Subordination in Cognitive Grammar. Langacker empahsizes particularly asymmetries in prominence, including profiling, trajector/landmark organization, and status, and presents an interaction between these properties. The author refers to the role of what is known as windows of attention in grammar and dynamic processing.



Asymetric Events

Sonia Cristofaro (University of Pavia), the author of the paper Asymmetric events, subordination, and grammatical categories, notes that the concept of asymmetric events can be considered central in many theories of subordination. The present paper discusses the consequences of this view for a definition of subordination as a grammatical category. The author presents data from numerous languages, where asymmetric events are encoded by morphosyntactically different sentence types, and, on the other hand, from other languages in which individual morphosyntactic features may be associated with both symmetric and asymmetric events. Cristofaro provides evidence that asymmetric events are thus not associated with a particular morphosyntactic property in the languages discussed but with a much larger variety of linguistic properties. In his paper Asymmetry reversal Frantisek Lichtenberk (University of Auckland) asks an important question concerning a relationship between a fully unfolded interpretation of a linguistically described scene and its linguistically more compact form and refers to possible interpretations of such processes in terms of a cognitive linguistic approach as presented by Langacker and, on the other hand, in terms of a functional-pragmatic interpretation as proposed by Cristofaro. The author discusses a restructuring of originally bi-clausal structures into mono-clausal ones in an Oceanic language, Toqabaqita, and attempts to answer the question as to what extent the processes observed are a case of asymmetry reversal. Transparency vs. Economy: How does Adioukrou resolve the conflict?, a paper by Kaoru Horie (Tohoku University, Kawauchi), Prashant Pardeshi (Kobe University), and Guy Kaul (Tohoku University), looks at the problem of event asymmetry in still another way. The authors discuss polyfunctionality of grammatical markers, in particular subordinators, in Adioukrou, a Kwa language from the Niger-Congo language family spoken in Côte d’Ivoire, and observe that Adioukrou shows signs of a neutralization of a classical distinction of subordinate clauses into relative, complement, and adverbial clauses. The authors also address the question of cognitive linguistic transparency (isomorphism) and economy and its manifestation in Adioukrou. In his paper Relating participants across asymmetric events: Conceptual constraints on obligatory control, Klaus-Uwe Panther (University of Hamburg) discusses a specific class of hypotaxis referring to asymmetric events, in which the same participant takes part in both the superordinate event and in the subordinate event. The author is particularly interested in an English-German contrastive analysis of participant identity across events, known as cases of ‘obligatory control’, a relation of referential dependence between one argument in a main clause and a non-expressed argument in an embedded non-finite infinitival or gerund clause. Panther contrasts control verbs in English such as the verbs promise, ask, persuade, advise, recommend, thank and their respective counterparts in German, viz. versprechen, bitten, überreden, raten, empfehlen, and danken and discusses in detail control into non-finite action complements. In the paper The Portuguese inflected infinitive and its conceptual basis, Augusto Soares da Silva (Catholic University of Braga) proposes a conceptual analysis of the Portuguese inflected (or personal) infinitive. Soares de Silva sees this phenomenon as



Introduction 

an objective construal of a process type and posits a continuum of event asymmetry. The inflected infinitival clause occupies in such a model an intermediate position between the subordinate finite clause and the infinitival clause. Soares da Silva exploits imagery dimensions that are specific to subordinate constructions in his research as well as the relationship of subjectivity versus objectivity to account for the semantics of the Portugese inflected infinitive. Liesbet Heyvaert (Fund for Scientific Research – Flanders and University of Leuven), author of the paper The periphrastic realization of participants in nominalizations: semantic and discourse constraints, discusses the inclusion of periphrastic participants in deverbal nominalizations and argues that such participants cannot be analyzed as complements or modifiers. Her second claim is that those cases where participants are not included but understood and retrievable from the discourse context are cases of nominal ellipsis. Heyvaert discusses also a class of event nominalizations in which the participants are not retrievable in the discourse context. She argues that such nominalizations are comparable to a clausal type of events, more precisely to non-finite events, which are not temporally grounded and do not necessarily take a Subject-entity. In the last paper of the present volume, Asymmetries in Participial Modification, Barbara Lewandowska-Tomaszczyk (University of Lodz) conducts a contrastive Polish-English analysis of the structural and conceptual asymmetries in attributive and postnominal participial modification. What is also discussed is a relationship between full sentential, reduced clausal and phrasal modification in the two languages. The focus of the paper are asymmetries in the semantic interpretation of different morphosyntactic structures of participial modifiers and their sentence positions, involving prenominal and postnominal participial modifiers in English and Polish. Most of the examples in the paper are taken from natural language, the BNC – for English, and the PELCRA corpus (University of Lodz) for Polish. The authors’ conviction is that the papers in this volume will bring a fuller interpretation to the concept of asymmetric events, which turns out to be one of the essential concepts in language and present it from different perspectives and points of view. Theoretical orientations adopted in the papers should not be considered to be opposite theoretical stances but rather provide a complementarity dimension and serve better illumination of the concept in focus.

Bibliography Achard, M. (1998). Representation of Cognitive Structures. Syntax and Semantics of French Sentential Complements. Berlin: Mouton de Gruyter. Cristofaro, S. (2003). Subordination. Oxford: Oxford University Press. Croft, W. (2001). Radical Construction Grammar. Cambridge: Cambridge University Press. Givón, T. (1993). English Grammar. A Function-Based Introduction. Vols. I and II Amsterdam: Benjamins.

 Asymetric Events Horie, K. (ed.) (2000). Complementation. Amsterdam: John Benjamins. Langacker, R., W. (1987a, 1991). Foundations of Cognitive Grammar vols. 1 and 2, Stanford, Calif.: Stanford University Press. ---(1987b). “Nouns and verbs”. Language 63:53-94. Talmy, L. (2000). Toward a Cognitive Semantics. Cambridge, MA: MIT Press.

part 1

Event chains and complex events

Asymmetry in English multi-verb sequences A corpus-based approach John Newman and Sally Rice University of Alberta

The English V and V construction provides an ideal opportunity to study asymmetry in the properties of the verbs which enter into each of the two verb slots of this construction. This paper explores the asymmetry evident in this construction by utilizing two corpora of spoken and written New Zealand English (the Wellington Written Corpus and the Wellington Spoken Corpus). The most striking asymmetry which emerges from the corpus data is the dominance of motion verbs and change of position verbs as V1 and their absence in V2. The V2 position shows a preference for verbs referring to activities involving a stationary position. The corpus leads us, therefore, to recognize move (in order) to do as the primary meaning associated with the V and V construction. While speakers of English may sense that this meaning is commonly associated with the V and V construction, only a corpus-based study such as this one is able to quantify the degree to which this meaning is, in fact, present in this construction. The paper also considers the nature of the semantic integration associated with the construction. Examples such as go and tell and go and visit, though superficially similar, illustrate different kinds of semantic integration. Coordinated verbs, in general, present a typical grammaticalizing context, exemplified by the try and V construction in English, as well as examples such as go and prove me wrong typical of spoken language.

Keywords: English, Kalam, Kaititj, construction, verb coordination, auxiliary verb, complex matrix, grammaticalization, intentionality, corpus

1. Introduction One way of approaching the study of asymmetrical events is through an examination of multi-verb sequences. In this paper, we focus on a certain type of multi-verb sequence in English – what we will refer to throughout as a ‘V and V construction’, consisting of two verbs linked with and, without further intervening material. This seems a natural way to delimit the scope of asymmetry such that the concept can still be



John Newman and Sally Rice

profitably investigated, albeit in a relatively restricted manner. We will explore asymmetry as it manifests itself in the properties of the verbs occupying the first and second positions of the V and V construction (designated as V1 and V2 respectively), as well as whether the two events are interpreted as unfolding simultaneously or sequentially. An advantage of focusing on such a narrowly defined construction type, consisting of no more than two coordinated verbs, is that it establishes a uniquely clear and simple basis for a study of asymmetry. Furthermore, delimiting the object of study to the V and V construction allows us to readily exploit a corpus-based approach to the collection of the data. The sequence V and V, as an object of study in its own right, attracts little attention in most syntactic approaches to English grammar (though see the discussion of this construction in Francis, Hunston, and Manning 1996: 122–124, as well as the corpusbased studies of the specific go and V construction in Stefanowitsch 2000 and Wulff 2006). Such approaches are, for the most part, concerned with principles which apply as widely as possible rather than with narrowly defined phenomena as an individual construction. Typical of such approaches is Van Valin and LaPolla (1997: 520–523) where V and V is not identified or discussed as such in their treatment of syntactic structures, but as a special case of “Conjunction Reduction”, which applies to many kinds of conjoined clauses and continues a long history within contemporary linguistics of dealing with conjoined structures. The authors illustrate Conjunction Reduction with the example repeated below as (1).

(1) Robin drove out of Phoenix this morning and pro will arrive in Atlanta tomorrow.

Here, Conjunction Reduction applies to the sequence of conjoined clauses and results in a pro element – which is not realized overtly – as the subject of the second clause, to be understood as referring to and coindexed with the subject of the first clause, Robin. While this is a perfectly reasonable way of analyzing (1), it does not accord any special significance to the V and V construction. However, English V and V sequences are more constrained in terms of their morphological possibilities than the example in (1), in which the two verbs appear in a past tense and future tense, respectively, suggests. A quick search of the 10-million-word spoken subcorpus of the British National Corpus (BNC) is instructive in this respect. There are 259 instances of and will in this subcorpus, but in not a single case does one find a sequence of V-past and will V (like drove and will arrive). Instead, the BNC corpus search reveals a tighter integration of the verbs in the V and V sequences, compared with what the literature on conjoined clauses tells us. This proved to be the case in the corpora we used for our study, too, as described in the following section. Although the modal auxiliaries can occur in conjoined structures in our corpora, they only occur conjoined with another auxiliary (can and can’t do, can and will become etc.). There were no instances of sequences of a full lexical verb conjoined with a modal (went and might, came and will, etc.). A full



Asymmetry in English multi-verb sequences

appreciation of this fact can only be achieved through a focus on V and V sequences as an object of study in their own right. For linguists with an interest in grammaticalization, a tighter coordinated construction like English V and V, and its equivalent in other languages, is also of special interest since multiple verbs in close proximity are often a locus of grammaticalization cross-linguistically. For example, it is this kind of structure which figures in the grammaticalized use of the posture verb ‘lie’ in Manam (Austronesian), as shown in (2). (2) i-pile-la-be 3sg.realis-speak-limiter-and ‘He kept talking.’

i-eno 3sg.realis-lie [Manam; Lichtenberk 2002: 280]

The ‘lie’ verb -eno functions as a frequentative or continuative aspect marker, the latter of which is shown in (3), appearing in a conjoined structure with the main verb -pile‘speak’ and -be ‘and’. A delimiter -la- is also part of this construction. Although this Manam example is not a direct equivalent of an English V and V construction, it is a relatively tight structure built around conjoined verbs which has given rise to the grammaticalized use of eno. In English, too, the V and V construction is a site for grammaticalization, as we shall see. Huddleston and Pullum (2002: 1299–1305), in their discussion of coordination, acknowledge the role that lexical items can play in determining different subcategories of coordination. Of most relevance to the present discussion is their subcategory of “Asymmetric Constructions”, which is understood as the coordination of two elements where the two halves cannot be reversed without a change of meaning, as in their example repeated here as (3).

(3) I fell off the ladder and broke my leg.

Since I broke my leg and fell off the ladder would refer to a different situation with the events in a different relation to each other, (3) is asymmetric. While one can understand ‘asymmetry’ in the V and V construction in the temporal or logical sense used by Huddleston and Pullum, we use the term here to refer to the different lexical preferences evident in the V1 and V2 positions and a concomitant restriction on the ways in which the events coded by the V and V construction are interpreted. This way of approaching asymmetry is consistent with our preferred methodology of working with spoken and written corpora. Corpus techniques allow the investigation of lexical usage and pragmatic inference in this construction, though they do not by themselves yield the kinds of results about asymmetry à la Huddleston and Pullum (which, of course, require manipulation of the sentence and decontextualized acceptability judgments by speakers). The questions we address in this study of the English V and V construction are: (1) What kinds of verbs occur in the V1 and V2 positions? (2) How are the two verbs in the V and V construction integrated semantically? (3) Are there specialized meanings which arise in the V and V construction, beyond what the individual meanings of the





John Newman and Sally Rice

verbs might suggest? We begin, however, with a discussion of the corpus-based methodology we employ.

2. Corpus methodology Our approach involves a combination of ‘corpus-based’ and ‘corpus-illustrated’ techniques, as distinguished by Tummers, Heylen, and Geeraerts (2005). The ‘corpusbased’ technique is typified by a strongly data-driven approach, relying upon quantitative analysis applied to a whole corpus; a ‘corpus-illustrated’ approach, on the other hand, utilizes a corpus as a source of examples to help illustrate one’s position. We begin with a corpus-based approach to discovering the types and tokens of verbs appearing in the English V and V construction in Section 3. In Section 4 we use a corpus-illustrated approach to assist us in discussing semantic details of certain sub-types of the V and V construction. Two corpora of New Zealand English were used in this study: the Wellington Written Corpus (WC) and the Wellington Spoken Corpus (WSC). The WC consists of one million words and is designed along the same lines as the Brown Corpus of written American English (1961) and the Lancaster-Oslo-Bergen corpus (LOB) of written British English (1961), but covering the years 1986–1990. The WSC is also one million words and consists of spoken New Zealand English collected in the years 1988 to 1994. Although the Wellington corpora are much smaller than some corpora currently available, such as the British National Corpus, they have the advantage of being wellbalanced in terms of the representation of written and spoken forms of the language, are drawn from the same larger population group (New Zealand), and reflect roughly the same time period. They are also small enough for an analyst to undertake a comprehensive examination of all occurrences of the V and V construction. We identified all and occurrences in the two corpora and then manually identified and tagged sequences of V and V, such as sing and dance, sang and danced, sung and danced, etc.1 We excluded the copula (as in came and was surprised) and auxiliaries (as in can and will become), but we included verb particles as part of the V category (as in come in and wait, came and sat down). Proceeding in this way, we found ourselves working with V and V structures in which each of the two Vs shared the same tense and aspect marking (past and past, present and present, etc.). Table 1 summarizes the frequencies of the V and V construction in the two corpora, broken down by part of speech (following the CLAWS5 tagset).2 While the overall occurrence of V and V is comparable between the two corpora (1375 in the WC, 1. The WC is tagged for part of speech but we preferred to rely on our own line-by-line inspection of and hits to identify flanking verbs. 2. Lancaster University’s CLAWS5 tagset was chosen since it is the well-known part-of-speech coding system used for the British National Corpus.



Asymmetry in English multi-verb sequences

1475 in the WSC), there are marked differences for certain parts of speech or inflectional categories.3 Thus, coordinated infinitival forms (as in to come and go) and base forms (as in they come and go, will come and go) occur more than twice as frequently in the WSC compared with the WC. Table 1.  Frequency of V and V in the Wellington corpora by CLAWS5 tag. Shading indicates the greater frequency of occurrence of a tag class.   VVI (infinitive) VVB (base)

WC

WSC

328 113

692 208

VVZ (3Sg present)

90

64

VVD (simple past) VVG (-ing)

325 248

267 147

VVN (past participle)

271

97

1375

1475

Total

Even before we consider the actual lexical forms that occur in the two V positions in the V and V construction, we are able to establish an interesting difference, i.e.,

900 800 700 600 500 400 300 200 100 0

828 712 587

658

370 275

V1 ALL

V2 ALL

V1 WC

V2 WC

V1 V2 WSC WSC

Figure 1.  Frequencies of lemma types for V1 and V2 in the two Wellington corpora (WC,

WSC) and their frequencies (ALL) in the combined corpus.

3. While the discussion of V and V in Francis et al. (1996: 122-124) captures some key properties of this construction, it nevertheless fails to differentiate the inflectional classes and the relative frequencies of these classes.





John Newman and Sally Rice

asymmetry, in terms of the number of lemma types occurring in each position. We arrive at lemmas by generalizing across the inflectional categories, so that come, comes, coming, and came, for example, count as instances of a single come lemma. We then count all tokens of the come lemma as one lemma type. The more lemma types one finds in a text or wordlist, the more lexical diversity there is. Figure 1 summarizes the results from adding all of the distinct lemma types in each corpus, as well as for the combined corpora. Both overall and in the separate corpora, there are more lemma types in the V2 position than V1. Put another way, the V1 position draws its verbs from a smaller lexicon than does the V2 position. This asymmetry is reminiscent of the discrepancy one finds between the auxiliary and main verb positions in English, with far fewer auxiliary choices than main verb choices.

3. Verbs occurring in the English V and V construction We begin with a discussion of the V1 and V2 sequences and then consider the V1 and V2 positions separately. To gain an overall appreciation of the lexical items involved, we first consider lexical items at the lemma level and postpone discussion of these items by inflectional tag until later.

3.1

V and V sequences

Table 2 lists the most frequent V and V sequences in the WC and WSC corpora. As noted above, the WSC has more tokens and fewer types of V and V sequences, hence the larger number of tokens per type that we see in the WSC sequences in this table. Properties of each verb position will be discussed in following sections, but there are already interesting observations to be made about the sequences presented in Table 2. For a start, we see repetitions in a number of cases (cry and cry, scream and scream, walk and walk) in the WC, but not in the WSC. In general, more verb types are repeated in spoken language, so it is of some interest that this particular kind of repetition occurs more often in the written register, at least amongst the most frequently occurring V and V types. Presumably, modifying devices like really, a lot, and all the time are preferred as ways of conveying greater degrees, or longer duration, of events in the spoken language. Some other lexical bundles in the WC are typical of a more learned or formal style (arrest and charge, construct and test, know and understand). Of interest in a discussion of asymmetry are the pairings of different kinds of antonyms. Clearly, there is the deictic antonymy of come and go evident in both the WC and the WSC – its inverse, go and come, is conspicuously absent. However, there is a variety of other antonym types in the WC, absent in the WSC list in Table 2: rise and fall, live and die, load and unload. The conjunction of such antonyms using and in English would appear, then, to be a feature more associated with written usage



Asymmetry in English multi-verb sequences

rather than spoken language. Our observations here are, of course, based on the most frequently occurring V and V types as listed in Table 2 and one would need to investigate all these types, including the uniquely occurring pairs, to fully research this idea. Antonymous V and V expressions do occur in the WSC (e.g., opening and closing), but not with a token frequency which makes them evident in Table 2. Table 2.  Most frequent lemmatized V and V sequences in the Wellington corpora (WC and WSC) V and V in WC > 2 tokens come and go read and write come and have come and see look and see cry and cry rise and fall go and have go and play scream and scream arrest and charge come and ask come and get come and give come and look come and play come and sit construct and test go and get go and tell know and understand live and die load and unload say and do smile and nod walk and walk

V and V in WSC > 6 tokens 17 9 7 6 6 5 5 4 4 4 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

go and see go and get go and have go and do try and get come and say come and see come and have go and sit come and go go and look wait and see go and visit go and buy come and do come and get ring and say try and do go and live turn and say come and sit come and stay go and find go and put go and work try and find

59 51 43 36 30 27 18 17 15 13 12 12 11 10 9 9 9 9 8 8 7 7 7 7 7 7





John Newman and Sally Rice

3.2

The verbs in position V1

Table 3 lists the most frequent lemmas in the V1 position in both corpora, minus any verb particle (e.g., the up of get up).4 We show the top 10 or so lemmas in each corpus, ensuring that all the lemmas with equal frequency of occurrence (e.g., read, say, and smile in the WC; stop and walk in the WSC) are included. The first and most striking fact in Table 3 is the dominance of the two verbs come and go in both corpora. We have not attempted to differentiate sub-senses of these two verbs in the corpora, but we do recognize that there can be differences in the way the verbs are interpreted, as discussed in Section 4. Even so, there is invariably some degree of physical motion associated with the use of come and go, as used in the V1 position (as opposed to the copula use in go crazy or the fictive motion sense as in the road goes to the mountains).5 We see, then, that the notion of physical motion plays a significant role as the V1 of the V and V construction. Between them, the two lemmas account for 691/1475 (= 46.8%) of all instances of the construction in the WSC, demonstrating a rather special role for these two verbs as the V1 in spoken usage. As we discuss in Section 4, there are some colloquial features which attach to go as V1, in addition to indicating physical motion, and it is these additional properties in the colloquial style that help to explain the preponderance of go in the WSC. Apart from come, go, and walk, there are other verbs that are associated with a change in physical position or state, depending upon the particle that might appear with them: get (up, down), turn (round), stop, sit (down), stand (up). Although we do not report separate results for the particular verb plus verb-particle combinations in these cases, these verbs are predominantly used to indicate a change in position or state, rather than to describe an existing or continuing position or state. As such, they constitute a second, significant class of verb types in the V1 position. Another semantic class that can be established consists of the verbs referring to an activity requiring minimal motion and no change in location: smile, say, read, sit, stand, look, wait. None of these verbs occurs amongst the top three in either corpus, but they are a recognizable subclass within the table. try also occupies a special place in this construction and we return to this verb in Section 4.

4. Our interest in this paper lies primarily in the asymmetry in the V1 and V2 positions of V and V sequences. Consequently, we are not concerned here with whether or not the verbs in this construction occur more, or less, frequently than one would expect, given their frequencies outside of this construction. Wulff (2006: 110-115) discusses the go and V construction in the BNC using the collostructional technique advocated by Stefanowitsch and Gries (2003) which does take into account overall frequencies of the verbs and the construction type in a corpus. 5. This claim needs to be qualified in light of the interpretation of some go and V combinations discussed in Section 4.3, such as go and prove me wrong. Still, the overwhelming majority of instances of go as V1 imply some kind of motion to a new location.



Asymmetry in English multi-verb sequences

Table 3.  Most frequent V1 lemmas in the Wellington corpora (WC and WSC) V1 in the WC (>14 tokens) come go

V1 in the WSC (> 12 tokens) 80 54

go come

462 229

get

26

try

168

turn

24

get

36

stop

20

sit

28

stand

18

turn

26

try

18

ring

21

sit

16

wait

17

read

15

stand

15

say

15

look

14

smile

15

stop

13

others

1074

walk

13

others

433

Total

1375

Total

1475

Figure 2 presents a finer-grained breakdown of the frequencies of come and go by the most common inflectional tags (base, infinitive, and past tense forms) in the V1 position. The pie-charts clearly show the consistent predominance of come and go in the WSC, compared with the WC in these categories. For these parts of speech, come and go together account for more than half of the tokens in the V1 in the WSC, with go making up the larger percentage of occurrences in each case. This distribution shows that the class of motion verbs, consisting of come, go, and any other motion verbs which happen to occur in the V1 position but are not listed in Table 2, accounts for the majority of verbs in the V1 position for these most frequent inflectional categories. What we find in English parallels tendencies observed about ‘come’ and ‘go’ verbs in other languages. Pawley and Lane (1998), for example, report that these two verbs are the most common lexical items in serial verb constructions in Kalam (Papuan), reflecting exactly the same preference as found for English, even if the Kalam facts are not reported in the same quantified way in which we report the English results. The availability of large-scale corpora for major languages, like English, and their relative unavailability for lesser known languages means that it is less likely that observed tendencies will be supported with quantified results. In other languages, ‘come’ and ‘go’ have an even more privileged role to play in serial verb constructions. In Kaititj (Arandic, Australia), these two verbs are the only ones which enter into a serial verb construction (Foley and Olson 1985). We may see here how the probabilistic behaviour of





John Newman and Sally Rice

‘come’ and ‘go’ verbs in English (as supported by our corpus linguistic study) and Kalam corresponds to the categorical behaviour of these same verbs in Kaititj.6

Figure 2.  Distribution of COME, GO and other verbs as V1 in selected parts of speech 6. Manning (2003: 316) refers to the categorical versus probabilistic facts in terms of hard and soft constraints: “The same categorical phenomena that are attributed to hard grammatical constraints in some languages continue to show up as soft constraints in other languages.” We make the same point for English sit, stand, and lie in Newman and Rice (2004).



Asymmetry in English multi-verb sequences

3.3

The verbs in position V2

Table 4 lists the most frequent lexical items in the V2 position. Consistent with the general observations made in Section 2 about lemma types in the two corpora, we see in Table 4 fewer tokens per type in the spoken register. In addition, one can see a greater lexical diversity of V2 types compared with V1 types, again as alluded to in Section 2. The number of tokens in the “others” category, covering all the less frequently occurring V2 types, is also indicative of this greater diversity. The discrepancy in lexical diversity is particularly evident in the spoken register: 837 V2 tokens in the “others” category (Table 4), compared with 433 V1 tokens in this category (Table 3). V2 tokens are spread across a far greater number of lexical items than the V1 tokens are. Turning to the specific details of the V2 position, one immediately notices the absence of come altogether in this table. go does occur, but far less frequently than it does in the V1 position. In the case of its frequency in the WSC, go occurs just 48 times compared with 462 times as V1. Nor do we find other verbs of motion, like walk, in this table. Clearly, change of physical location is not a dominant semantic property of the verbs in the V2 position as it is in the V1 position. On the other hand, the semantic class of verbs referring to activities involving stationary position rather than a change of place is significant: see, look, say, talk, write. The prevalence of see and get in WSC is noteworthy, representing the two largest V2 types, each consisting of roughly the same number of tokens. They reflect quite natural experiential realities relating to the kinds of activities or purposes associated with motion to a location. Table 4.  Most frequent V2 lemmas in the Wellington corpora (WC and WSC) V2 in the WSC (> 16 tokens)

V2 in the WC (>9 tokens) go see

30 19

see get

107 104

take

18

say

86

look

17

have

85

say

17

do

73

talk

16

go

48

have

15

look

25

play

15

talk

25

get

14

sit

24

write

12

made

23

change

10

tell

21

10

work

17

others

give

1182

others

837

Total

1375

Total

1475





John Newman and Sally Rice

Combining the major results obtained from examining the verbs which occupy positions V1 and V2, we are able to establish the major semantic schema associated with the English V and V construction. As we have already observed, human motion from one location to another is the dominant semantic category associated with V1 in the three most commonly occurring inflected categories, i.e., base, infinitival, and past tense forms. And it is human activity at a single location which is the dominant semantic category associated with the V2 position. The sequence of human motion from one location to another followed by an activity at the destination must be considered therefore as the single most frequent conceptual complex encoded by the V and V construction. Semantically, then, English V and V is most strongly associated with the frame ‘movement to a location plus subsequent activity’ or, expressed more schematically, move (in order) to do.

4. Integrating V1 and V2 We illustrate some of the range of semantic types represented by the V and V construction using four specific sub-types. Distinguishing between the sequential versus simultaneous types of concatenation in the V and V construction is a familiar tactic in semantic analyses of the construction (see, for example, the discussion in Francis, Hunston, and Manning 1996: 122–124) and we give some attention to this in what follows. As we shall see, however, categorizing V and V merely as signaling a ‘sequential’ or ‘simultaneous’ type of integration captures relatively little about the semantic nature of their integration. We particularly focus on go and V because go is the most common verb in the V1 position in the spoken corpus – where one might expect a more complete range of uses compared with the written corpus. We discuss three subtypes of the go and V construction which serve to illustrate the variation and complexity in terms of how the two predicates in this construction are semantically integrated. The three sequences are: go and tell, go and visit, and go and prove me wrong. We also discuss the try and find sequence. Clearly, there are other kinds of semantic relationships which can hold between the two predicates of the V and V construction. While the four sub-constructions we focus on do not, by any means, exhaust all the semantic possibilities, they are nevertheless instructive.

4.1

GO and TELL

Examples of go and tell, and other combinations of the same semantic type, are given in (4). (4) a. and so I should just go and tell them you know [WSC] b. if Debbie was going to go and tell them to stop it [WSC] c. it’d never occur to him to go and buy it [WC]



Asymmetry in English multi-verb sequences

d. Let’s go and look at the damage [WC] e. Then I went and stood behind the next tree [WC] The integration of V1 and V2 in (4) seems relatively straightforward: two events occur in chronological sequence, with the same subject referent for both cases, as is the case with all three go and V types that we are considering. We draw attention, however, to another critical component, a component in fact of all three go and V sub-types: the go event is a purposeful event, carried out with the intention of performing the event referred to by V2.7 The term “complex matrix”, as described by Langacker (1991: 4–5), is useful in this context. The term refers to the collection of relevant information about a morpheme drawn from different domains. So, for example, the complex matrix for the noun knife includes references to the typical shape of a knife, a “cutlery frame” consisting of a certain typical arrangement of a knife, a fork, and spoon, and a “cutting” frame conveying the notion of the typical action associated with knife when used as an instrument. Just as there is a complex matrix associated with knife, comprised of information from these various domains, so we may also recognize a complex matrix associated with go. Specifically, the go complex matrix will make reference to both spatio-temporal and mental domains. The former locates the movement of an entity through space and time and, in the case of go, will refer to the movement of an entity away from a deictic centre. This spatio-temporal perspective of go is familiar from discussions in the literature – see, for example, the diagram of go in Langacker (1991: 6). However, another domain in addition to this spatio-temporal domain needs to be recognized: an “intentional” domain representing the realities of human sentience and volition. In this domain, persons may have, and typically will have, intentions to carry out future acts. The complex matrix of go consists, therefore, of the integration of information from both of these domains, the spatio-temporal and the intentional. The end result is not just the uniting of, but indeed the blending of notions of physical motion and psychological purpose inside the V and V construction in English. We offer a very schematic representation of the go and tell construction in Figure 3, adapting some of the notation employed by Langacker (1987, 1991). The oval shapes represent the two predicates of the construction, abstracting away from the detailed relationships which hold within each predicate (e.g., movement of an entity away from a deictic centre in the case of go, the verbal interaction between two entities in the case of tell). tell is placed after go in this diagram reflecting the temporal sequence of the two events. Each predicate contains a subject argument, indicated by the small empty circles. The identity of the subject referents of the two predicates is expressed by the dotted line joining the two circles. The arrow leading from the subject of go to the subject of tell is used to express the fact that the subject referent of go has the intention of engaging in the second event, involving telling something to someone. Although 7. See Newman and Lin (2007) for a fuller discussion of the purposefulness of go compared with other verbs, also based on a corpus study.





John Newman and Sally Rice

it is possible to explicate the semantic relationships between the two predicates in the V and V construction without recourse to any graphic representation, we find it helpful to refer to such graphic representations in order to highlight the key semantic properties of the construction with different lexical choices for the verbs. The representations in Figure 3 and the Figures associated with the discussions in the following sections communicate in a relatively direct way various semantic properties of interest in this construction, e.g., the distinctness or simultaneity of events, the intentionality of the subject referent, and the semantic contribution made by each predicate.

TR GO

TELL

TIME

Figure 3.  Representation of go and tell integrating spatio-temporal and intentional domains

4.2

GO and VISIT

In (5), we see examples of the go and visit construction. (5)

a. b. c. d. e.

I’m going to contact her arrange to go and visit her and so on [WSC] well if I get a ticket to um just go and visit some of my cousins [WSC] and then the next day go up and visit the other two [WSC] I went out and visit the guy I used to work for [WSC] Sharon went and visited the flat [WSC]

The examples in (5) may appear, at first glance, to involve the same kind of integration of predicates as we saw for go and tell, i.e., sequentiality together with intentionality. The combination illustrated in (5), however, presents a different kind of semantics. For a start, consider typical definitions of the verb visit, as shown in (6). (6) a. visit. v.t. to go to see (a person, place, etc.) in the way of friendship, ceremony, duty, business, curiosity or the like. [The Macquarie Dictionary] b. visit. v.t. go or come to see (person, place, etc. or abs.) as act of friendship or ceremony, on business or for a purpose. [Concise Oxford Dictionary]



Asymmetry in English multi-verb sequences 

Note how in these definitions a verb of motion, go or come, is deployed as part of the definition of visit. This seems entirely natural and correct. This is quite different from tell, where the idea of movement to another location is not at all an inherent part of its meaning. Using Langacker’s (1987 :183–189) terms, we could say that the base of the predicate visit includes movement to a location as well as the planned encounter with someone at that location, while the profile focuses just on the encounter. Thus, visit already subsumes semantically the go predicate in go and visit and needs to be represented accordingly, as in Figure 4. In this figure, the circular shape representing visit includes within it the representation of go.

TR GO

VISIT

TIME

Figure 4.  Representation of go and visit integrating spatio-temporal and intentional

domains

Additional examples which, arguably, are similar to the go and visit type of integration are given in (7). (7)

a. b. c. d. e.

are you gonna go and pick Raewyn and Benny up [WSC] they go and see a solicitor [WSC] and went off and disappeared [WSC] I’ll have to go and fetch him, Rosey [WC] The team prays for two hours before they go and return later to pray again. [WC]

In (7a), the idea of picking someone up (as used here) implies a familiar scenario of driving to a location where someone is waiting for that driver and the driver then bringing the person somewhere else in the car. As with visit, pick up makes reference to a larger scenario in which the actual encounter with another person is the focus. In (7b), see is used in a particular way, comparable to visit in that it refers to a larger scenario of making an appointment, traveling to an office, and meeting with a solicitor. In (7c), the act of disappearing already subsumes the motion of the go predicate. In (7d), fetch implies a motion to some place. In (7e), the return predicate already subsumes motion to the location from which one moves back to the original location.



John Newman and Sally Rice

4.3

GO and PROVE ME WRONG

We now turn to examples of the kind illustrated in (8) which we will refer to as the go and prove me wrong type. This subtype of go and V is remarkable because the sense of precipitating motion suggested by V1 can be heavily attenuated if not effectively non-existent. (8)

a. b. c. d. e. f.

go on go and prove me wrong now [WSC] the silly council went and sprayed the side of the banks [WSC] so she went and moved us up to new Plymouth [WSC] and then a day later he went and wrote this poem [WSC] now different people have gone and sold their houses [WSC] the punters they went and paid forty odd dollars [WSC]

The examples in (8) illustrate a usage found mainly in a colloquial, conversational style, as suggested by some of the rhetorical devices which appear in these examples: Go on…!, the silly council, so she…etc. These devices point to a common, breezy narrative style common in conversation. In all of these cases, some kind of movement to a different location may be involved, though more prominent is the intent to carry out the activity predicated by V2. Thus, the sense of motion directed away from a deictic centre seems far less salient than the idea that some kind of non-specified activity is undertaken, sometimes with a hint of determinedness on the part of the subject referent, leading to the subsequent event. Figure 5 represents this aspect of the go and prove me wrong subtype by showing the first predicate as act (non-specific). The intentionality underlying the action of V2 is present, but the idea of physical motion leading to that action is very weak indeed.

TR ACT

PROVE ME WRONG

TIME

Figure 5.  Representation of go and prove me wrong integrating spatio-temporal and

intentional domains



Asymmetry in English multi-verb sequences 

The semantics associated with this type of go and V construction are such that the contribution of go in the V1 position amounts to little more than indicating that an intention or determinedness to proceed with V2 exists, as in She went and moved us up to New Plymouth. As such, the go and V construction in these cases could be seen as a particular kind of purposive construction, suggesting characteristics ranging from intentionality and purposefulness to determinedness and resoluteness on the part of the person undertaking the activity of V2. The evolution of ‘go’ morphemes to purposive markers is well known, as documented in part by Heine and Kuteva (2002: 163–165), who provide a number of examples from different language families of a change of a lexical verb ‘go’ to a functional gram marking ‘purpose’. Their ‘purpose’ label covers a variety of senses and morpheme types relating to a purpose, e.g., a ‘purpose clause marker’ in Tepo (Niger-Congo), a ‘subordinating conjunction of goal, purpose, etc.’ in Rama (Amerind), etc. They also report a sense of ‘speaker determination’ as part of the extension of ‘go’ predicates to purposive markers in some creole languages.

4.4

TRY and V

Table 5 below summarizes the most frequent types of try and V in the corpora, with the construction occurring far more frequently in the spoken register than the written (see Table 1). One can see in Table 5, even in the absence of the larger context, that the construction often occurs with V2 verb types involving activities associated with effort: analyse, assist, conjure, improve, rectify, learn. In many cases, of course, it is difficult to fully appreciate the nature of V2 without the larger context (as in the case of try and get, try and do, try and make) and a closer examination of the context is needed in these cases. The examples in (9) illustrate the try and V construction in both the spoken and written corpora, with additional context. (9) a. they try and intimidate you they try and see where they can get with you [WSC] b. sort of chap is that they always try and clinch their argument by saying [WSC] c. best to avoid that weekend and try and go for the May date [WSC] d. somehow we have got to try and keep control of our members [WC] e. in two minds about whether or not to try and adopt a son [WC]



John Newman and Sally Rice

Table 5.  Most frequent lemmatized try and V sequences in both corpora (WC and WSC) try and V in WC (all tokens) try and tell try and adopt try and analyse try and assist try and blame try and conjure try and find try and get try and improve try and keep try and make try and peek try and rectify try and see try and show try and talk try and write

try and V in WSC (tokens > 2) 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

try and get try and do try and find try and make try and go try and put try and see try and learn try and play try and say try and take

30 9 7 6 4 4 4 3 3 3 3

In the try and V construction, there is a tight semantic integration of the two verbs, similar to want to V. The act of trying, like the state of wanting, is always directed towards some other activity or thing (or possibly thing in the case of wanting). It is possible to construct a kind of sequential combination of repeated trying, followed by execution of the activity aimed for, e.g., She tried very hard, but in the end she was unable to complete the marathon. But in the examples in (9) and other try and V sequences in the corpora, there is no such clear separation of the two halves, semantically. In effect, they are like the try to V construction. Sometimes, try and V seems to carry a nuance suggesting slightly more certainty or confidence about the outcome than try to V. So, for example, We have got to try and keep control of our members seems a little more insistent that the addressees should ensure the outcome than We have got to try to keep control of our members. We represent the try and V construction in Figure 6 showing the V2 predicate as contained within the semantic scope of try.



Asymmetry in English multi-verb sequences 

TR TRY

V

TIME

Figure 6.  Representation of try and v integrating spatio-temporal and intentional domains

The distribution of try and V is heavily skewed both across corpora and across the inflectional categories, as seen in Table 6. It is far more prevalent in the spoken corpus than the written, where it would be prescriptively avoided. Note also its absence in the VVZ and VVD categories altogether. It does however appear in one instance in the VVG category. The particular semantics associated with try and V and its restriction to certain inflectional categories encourage us to think of try in this construction as becoming a kind of phrasal auxiliary marker indicating inception. Table 6.  try and V across corpora and inflectional categories WC try and VVI try and VVB trying and VVG tries and VVZ tried and VVD

16 2 0 0 0

WSC 130 36 1* 0 0

*Ted kept trying and scraping all this paint off the inside

5. Conclusion There are multiple asymmetries evident in the English V and V construction. There is the chronological priority of V1 over V2 in the sequential type of coordination. This is the kind of asymmetry that Huddleston and Pullum (2002) single out. But there are other asymmetries that can best be observed using corpus-based procedures. Such



John Newman and Sally Rice

procedures enabled us to identify more lemma types in the V2 position than in the V1 position (in both spoken and written corpora), a result that is not intuitively obvious and cannot be fully anticipated prior to carrying out a corpus-based study. Another type of asymmetry concerns the semantic profiles of the individual verbs in each position. Even without resorting to a semantic categorization of every single verb in the V and V construction, we succeeded in identifying two motion verbs, come and go, which together account for the majority of V1 tokens in the most common inflectional categories, whereas we found no comparable presence of these verbs (or other motion verbs) in the V2 position. This led us to identify a schematic meaning of ‘human motion from one location to another followed by an activity at the destination’ as the dominant meaning associated with the V and V construction.8 However much one might intuit that some kind of sequentiality is present in the V and V construction, only a corpus-based approach like the one we have followed permits any degree of confidence in formulating this schema. It is not uncommon to encounter discussions of sequential vs. simultaneous types of coordination, as with the V and V construction. This is a reasonable way to proceed in a semantic analysis of the construction, but as our discussion in Section 4 shows, that taxonomy does not reveal all the intricacies in the nature of the semantic integration of the two predicates. Even with the sequential type of coordination, represented by go and V, we found varying degrees of integration of the predicates, from relatively separate events to relatively unified. The peculiar distribution of the try and V construction, limited as it is to just some of the inflectional categories, suggests an auxiliary-like status for try in this combination, similar to the modal-like future marker that going to has become in English in a different construction altogether. Auxiliaries remain the category par excellence of asymmetrical or subordinating events in multi-verb sequences, and two of the most prevalent V1 verbs discussed here, go and try, either have become or seem destined to become auxiliarized. Only a close examination of the larger context in each major sub-type of the V and V construction has allowed us to appreciate these finer details and empirically establish that, despite the use of the “symmetrical” conjunction and, very little symmetry is involved in English V and V sequences.

References Foley, W. A. and M. Olson (1985). Clausehood and verb serialization. In J. Nichols and A. C. Woodbury (eds.), Grammar Inside and Outside the Clause, pp. 17–60. Cambridge: Cambridge University Press.

8. Our result about the prevalence of the meaning ‘human motion from one location to another followed by an activity at the destination’ bears comparison with the result from Stefanowitsch and Rohde (2004) on the prevalence of a goal adverbial phrase with English go.



Asymmetry in English multi-verb sequences  Francis, G., S. Hunston, and E. Manning (eds.) (1996). Collins COBUILD Grammar Patterns 1: Verbs. London: HarperCollins. Heine, B. and T. Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Huddleston, R. and G. K. Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Langacker, R. W. (1987). Foundations of Cognitive Grammar. Volume I. Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, R. W. (1991). Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin and New York: Mouton de Gruyter. Lichtenberk, F. (2002). Posture verbs in Oceanic. In J. Newman (ed.), The Linguistics of Sitting, Standing, and Lying, pp. 269–314. Amsterdam and Philadelphia: John Benjamins. Manning, C. D. (2003). Probabilistic syntax. In: R. Bod, J. Hay, and S. Jannedy (eds.) Probabilistic Linguistics, pp. 289–341. Cambridge, Mass.: The MIT Press. Newman, J. and J. Lin (2007). The purposefulness of going: A corpus-linguistic study. In J. Walinski, K. Kredens, and S. Gozdz-Roszkowski (eds.), Corpora and ICT in Language Studies, pp. 293–308. Lodz Studies in Language, Vol. 13. Frankfurt am Main: Peter Lang. Newman, J. and S. Rice (2004). Patterns of usage for English sit, stand, and lie: A cognitively inspired exploration in corpus linguistics. Cognitive Linguistics 15: 351–396. Pawley, A. and J. Lane (1998). From event sequence to grammar: Serial verb constructions in Kalam. In A. Siewierska and J. J. Song (eds.), Case, Typology, and Grammar: In Honour of Barry J. Blake, pp. 201–227. Amsterdam and Philadelphia: John Benjamins. Stefanowitsch, A. (2000). The English GO-(PRT)-AND-VERB construction. BLS 26: 259–270. Stefanowitsch, A. and S. Th. Gries (2003). Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics 8: 209–243. Stefanowitsch, A. and A. Rohde. (2004). The goal bias in the encoding of motion events. In G. Radden and K.-U. Panther (eds.), Studies in Linguistic Motivation, pp. 249–267. Berlin and New York: Mouton de Gruyter. Tummers, J., K. Heylen, and D. Geeraerts (2005). Usage-based approaches in Cognitive Linguistics: A state of the art. Corpus Linguistics and Linguistic Theory 1.2: 225–261. Van Valin, R. and R. LaPolla (1997). Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Wulff, S. (2006). Go-V vs. go-and-V in English: A case of constructional synonomy? In S. Th. Gries and A. Stefanowitsch (eds.), Corpora in Cognitive Linguistics: Corpus-Based Approaches to Syntax and Lexis, pp. 101–125. Berlin and New York: Mouton de Gruyter.

Asymmetries for locating events with Cora spatial language Eugene H. Casad

University of Gdańsk

In this paper I discuss several salient asymmetries and explore their roles in motivating usages of Cora spatial language to locate events, situations, conditions, relations, people and things. The first of these asymmetries is that of the speaker’s role as observer: in framing his conceptualizations for linguistic expression, he selects a particular viewing arrangement with a specific focus. The second asymmetry is that of locating entities within some field of spatial extension: specific entities are located spatially with respect to some locative reference point. Many Cora spatial expressions are conceptually and morphemically complex and invoke multiple locative reference points which the speaker accesses simultaneously in order to locate the entity that he has in mind. The most high level asymmetry,. the Subject of Perception vs. the Object of Perception contrast relates to these other asymmetries in subtle ways. Certain grammatical correlates emerge from this study that explain the consistency for the adverbial sets of spatials to construe the Subject of Perception in a highly subjective manner. Several grammatical correlates suggest why the locative verbal prefixes construe the Object of Perception in a highly objective way.

Keywords: asymmetric, subjective, objective, viewing arrangement, profile, ground, search domain, point of reference, reference point, spatial terms

1. Introduction The Cora language, a Southern Uto-Aztecan language spoken in Northwest Mexico is characterized by an intricate and highly integrated system of spatial language. This includes a set of sentence initial locative particles, an even more elaborate set of locative adverbs that are oriented to the lay of the land and an extensive inventory of locative and directional verbal prefixes that sketch out in highly schematic terms the physical and geographical settings within which events are realized and conditions are observed (Casad 1977, 1982, 1988, 1993, 1995; 1999; Casad and Langacker 1985, Langacker



Eugene H. Casad

1991). This locating of events, conditions and entities is inherently asymmetric, as Sovorou has discussed in some detail (Sovorou 1993 :8–9, 24–25, 213, fn. 6). Beyond the clear morphological characteristics that separate the sets of Cora spatials, the subjective-objective distinction, discussed by Langacker (1990, 1991, 1999) relates to the usages of these spatials in various ways, some obvious, others not so obvious.1This distinction, as Langacker notes, is also inherently asymmetric (Langacker 1999 :211). In this paper I examine this question and show that, in general, the locative particles and adverbs construe the Speaker in a highly subjective manner in their usages in the grammar and discourse, whereas the locative verbal prefixes are highly objective in their viewing arrangement of the scene on stage.2Nonetheless, there are degrees of subjectivity and objectivity observable in the usages of all three sets of spatials. I find a variety of Langacker’s notions crucial to my analysis. These include the ideas “viewing arrangement”, “profile”, “ground”, “search domain”, “immediate scope”, “maximal scope”, “point of reference” and “reference point”3. In particular, all of these notions reflect asymmetries of various sorts (cf. Langacker 1999  :176), are necessary for exploring in adequate detail the notions “subject of perception” and “object of perception” and are crucial for showing how these two are reflected in specific usages of Cora spatial terms. I begin my discussion with an overview of Langacker’s formulation of the viewing arrangement for language use and then discuss examples that illustrate the contrast between the maximally subjective construal of the subject of perception and the maximally objective construal of the object of perception. As this description will also show, a degree of subjectivity accompanies all of these usages (cf. Langacker 1999 :298; also, Verhagen 1995; Harder 1996).

1. I would like to thank Ron Langacker for his careful reading of an earlier draft of this paper and his insights and suggestions that have lead to this version. In addition, I would like to thank an anonymous reviewer whose comments were also very helpful. I would also like to thank Len Talmy for putting me on the track of David Marr’s experimental work on visual perception, work in which he substantiated clearly the roles of Subjective and Objective standpoints (Marr 1982). Finally, I would like to thank Peter Harder for his comments and encouragement. 2. Some experimental background that validates the subjective vs. objective distinction within the framework of visual theory is discussed in Marr (1982), esp. 1982:34, 51, 68, 81). 3. In this paper, I distinguish between the general term “point of reference” and the specific term “reference point” as detailed in Langacker (1993). A point of reference is a point with respect to which some other entity is situated or calculated. It may often be offstage and implicit and in many cases may be the conceptualizer him/herself. A reference point, in the narrow sense, is onstage and salient. It is part of a reference-point model and is the focused element that the speaker accesses in order to reach another focused element (Ronald W. Langacker, personal communication).



Asymmetries for locating events with Cora spatial language 

2. The viewing arrangement To begin, Langacker notes that the link between an observer and what he observes is intrinsic to the statements that the observer makes about the situation in focus. With respect to actual language use, Langacker equates the observer with the speaker. The speaker’s observational role is that of apprehending the meanings of linguistic expressions (Langacker 1999 :204). This role is asymmetric at a specially privileged level and opens the door for the speaker’s imposing of lower level asymmetries such as the trajectors and landmarks in clause structure for framing his thoughts for communicative purposes. Langacker’s characterization of the prototypical viewing arrangement for language use includes two isomorphic sets of constructs. For the prototypical visual perception situation, he posits the viewer, who is the subject of perception. That viewer has a prototypical viewing stance in which he is facing in a particular direction and has a maximally extended field of vision with an indeterminate periphery and a central region of high visual acuity. Langacker invokes a theater model at this point and states that the central region of visual acuity is the “onstage” region of the overall field of vision. This “onstage” region is the portion of the situation in focus and represents the object of perception (Langacker 1999 :204–205). These elements are represented diagrammatically in Figure 1.4 The notational conventions employed in Figure 1 include the following: (V) = Viewer, (MF) = Maximal Field of Vision, (OS) = Onstage Region and (F) = Focus. In addition, the dash lined arrow running between the Viewer and the Focus in this context represents the perceptual relationship between the Perceiver and the Perceived Entity. This is the prototype for a variety of possible configurations, each of which can be termed a viewing arrangement (Langacker 1999 :205).

F

V

OS MF

Figure 1.  The perceptual arrangement

The conceptual counterpart to the prototypical viewing arrangement of Figure 1 is more general and more widely applicable in semantic analysis. The constructs in this case include the Conceptualizer (C), who is also the subject of conceptualization. The Maximal Scope of the conceptualization (MS) includes the full range of conceptualized 4. I would like to thank Anke Beck of Mouton de Gruyter for permission to reprint Figures 7.1 and 7.2 from Langacker (1999) in this paper.



Eugene H. Casad

content, both central and peripheral. The analogue of the “onstage” region of the visually based viewing arrangement is the set of central notions that we have in focus. In this case, Langacker uses the term Immediate Scope (IS) to designate the general locus of attention. This IS comprises the full set of elements which potentially can be put in focus as the object of perception, the entity that Langacker terms the profile of the conceptualized situation (P) (Langacker 1999 :5). The profile, then, is the specific focus of attention within this general region. The conceptual counterpart to the perception relationship of Figure 1 is the construal relationship, indicated by the dashed arrow that relates the Conceptualizer to the entity he holds in focus. This represents the full range of ways that the speaker has at his disposal for structuring what he has to say and how he is going to say it. In short, the overt inclusion of the construal relationship in this arrangement effectively opens the door for accounting for all of the components of value, meaning, purpose and belief that Pike specified in his work (Pike, 1960, Vol III: 113, 115, 118). The elements of the conceptual viewing arrangement are depicted diagrammatically in Figure 2.

P

C

IS MF From: Ronald W. Langacker (1999), Grammar and Conceptualization, Figs 7.1 (a) & (b), p. 205.

Figure 2.  The conceptual arrangement

3. Maximally subjective construal of subject of perception: In presenting a maximally subjective construal of his/her role, the speaker excludes himself completely from the “on stage” portion of the viewing scene. The profile is the focus of conceptualization and, by definition “is construed with a high degree of objectivity” (Langacker 1999 :297) The profile may be an event, a relation, a person, a thing, or an area. The Cora locative particles and topographic adverbs saliently profile areas within spatial and temporal domains. Morphemically complex, the Cora locatives constitute a striking example of what Langacker calls “Reference Point Constructions.” The reference point in the relationships that these morphemes signal is, of course, a locative reference point. One of the interesting things about the semantics of these elements is the elaborate nature of these



Asymmetries for locating events with Cora spatial language 

constructions: most of them are morphemically complex; each morpheme in the construction signals a distinct locative relationship and in all cases, the subject of perception is construed subjectively, albeit with a variety of construals of the grounding entities.

3. 1

The basic locative particle pair

To illustrate the points made above, I begin with the simple examples of (1)(a and b), which exemplify a basic and widespread semantic contrast of the Cora locative system. (1) a. á pwa’ake there: out outside ‘Off there outside the house’ b. ú či’i-ta there:in house in ‘off there inside the house These two examples illustrate two slightly different locative phrase constructions; (1a) is a locative particle + locative adverb construction, whereas (1b) is a locative particle + postpositional phrase construction. The locative particles á and ú indicate that some entity is distally located with respect to someone’s position, whether Speaker or other person, and that the distally located entity is to be found either within an indefinitely expanded exterior region, as in the case of (1a), or within a clearly enclosed area, as in the case of (1b). In what follows, I intend to characterize only a and u; the roles of pua’aké and či’i-ta are not characterized in detail, rather their citation here is simply to illustrate typical situations which are construed as either ‘outside’ or ‘inside’ ones. In each case, the scenario includes an area that we can characterize as the “search domain” of the locative particle (Hawkins 1985 :168). This search domain summarizes the totality of locations where the Target of conceptualization might be situated (Langacker 1999 :173–4). It is indicated overtly in Figures Three and Four and elsewhere when essential to the discussion. The respective scenarios are represented diagrammatically in these two Figures. In Figure 3, two entities are shown in an inherently asymmetric stative relationship set at a distal distance from the ground, i.e. the conceptualizer. The first of the two is the locative landmark, indicated by an ‘LR’ in these diagrams. This locative landmark is the reference point for locating the second entity in the stative relationship, i.e. the Target of Conception = Trajector, indicated by a ‘TR’, The distance relationship in Cora is organized into three successively more inclusive spatial areas which I label ‘proximal’, ‘medial’ and ‘distal’, respectively. The proximal area is construed as the speaker’s location, whereas the medial area is viewed as taking in the speaker’s point of visual reference and the distal area is conventionally understood as being out of range of the speaker’s neighborhood (Casad 1982 :81–2).



Eugene H. Casad

LR

TR MED D

C=R

Figure 3.  Cora á ‘off yonder’

In Figure 3, the trajector is shown outside of the second largest circle in the diagram, labelled ‘MED’. As suggested by the label on the circle, this broad area subtends the neigborhood of the Locative Reference Point ‘LR’, whereas the postion of the conceptualizer, who is the default point of reference (C = R) for the relationship, is shown as being on the margin of the viewing arrangement. The neighborhood of ‘LR’ is the inner circle, which in turn has its neighborhood subsumed by the MEDIAL distance region. This divides the area of immediate scope into ‘near’ and ‘far’ regions vis à vis the Reference entity. Following Langacker’s notation, the conceptualizer’s role as a non-focal point of reference, indicated by the circle marked ‘C=R’ is completely off stage and the broken arrow linking C=R with LM indicates the path the conceptualizer follows for making mental contact with both the Locative Landmark and the Trajector of the conceptualized locative relation (Langacker 1999 :174, 177, 179, 182, 197, 236, 395). The scenario associated with u ‘inside’ is depicted diagrammatically in Figure 4. á ‘off yonder’

LR C=R

TR D

MED

Figure 4.  Cora ú ‘off yonder in there’

Figure 4 presents a diagram of ú ‘off yonder in there’. The conceptualizer, as the nonfocal point of reference and who is part of the ground of the expression, is shown as mentally viewing a stative relationship, one that is anchored by a locative reference



Asymmetries for locating events with Cora spatial language

point ‘LR’ that is distally removed from the proximal neighborhood of the ground. Again, no element of the ground is on stage in this usage. The distal, interior relationship is confined to a physically bounded region that consists of the landmark reference point and the trajector of the locative relationship situated at some point within its search domain ‘D’. This conceptual content is being construed objectively and is tantamount to the speaker’s locus of viewing attention, i.e. the immediate scope of the expression (Langacker 1999 :6–7; 49–51; 60–61).

3.2

The topographic particles

The adverbials á and ú are the morphologically and semantically simplest of the Cora locative adverbs. The next simplest set of locative particles are the topographically oriented ones. These comprise three distinct paradigmatic sets, illustrated in Table 1. Table 1.  The topographic locative particles Slope vs. Distance here there off there

Foot of Slope Inside yuu muu uu

Outside yaa maa aa

In Slope Inside yuh muh uh

Outside yah mah ah

Head of Slope Outside yan man an

The three paradigmatic sets involve variants of a slope predication which fit into complex semantic structures along with variants of the distance and boundary predications. The presence of two predications in the meanings of a and u implies the accessing of multiple landmarks which the speaker uses simultaneously for calculating specific spatial relationships. We will observe this invoking of multiple landmarks again and again in the following discussion. The slope predication is reflected in the morphological shape of the locative particles in one of three ways: “foot of the slope” is marked by a long a or u vowel, “in the slope” is marked by a syllable-final -h, and “head of the slope” is marked by syllable-final -n. The three sets of resultant forms, corresponding to each of the regions in the slope gradient, are given above in Table 1. As usual, the locative particles are arranged vertically along the points of the distance predication. They are also set in parallel columns to display the contrastive ‘inside -outside’ categorization of the boundary predication. In order to show just how deeply and in what way the Subject of Perception is construed highly subjectively in the semantic structures associated with these morphemes, as well as illustrate the inherently asymmetric reference point nature of each, I present the following analysis of a single contrastive pair of topographic particles, áh ‘off up there at the side of the hill’ and úh ‘right straight off there up in the slope’. These





Eugene H. Casad

meanings, of course, correlate with the intrinsic axiality of a hill or mountain range. One axis is vertically oriented and the other is horizontally oriented. Before describing the composite semantic structures of this pair of slope-oriented locative particles, it will be necessary to consider the nature of the slope predication itself and its relation to the boundary predication. The abstract domain of the slope predication includes three interrelated regions, which, taken together, constitute the slope domain base. Each of these regions may serve as the reference point area for locating some entity in focus. Figure 5 presents a generalized form of the slope domain. “Head” “Face”

“Foot” V

LM

Figure 5.  Slope as Abstract Domain for Line of Sight

The slope domain in Figure 5 turns out to be the base relative to which the boundary area is defined. Within this complex base, the locative point of reference bounded area is equivalent to a speaker’s canonical line of sight, i.e., what he would see were he at the foot of the slope looking up the vertically oriented gradient. Specifically, the line of sight begins at the horizontal foot of the slope, follows the vertical axis of the slope and continues to the skyline at the head of the slope. As a natural restriction, it does not curve around to cover the head of the slope. The relationship between the generalized slope domain and the line of sight boundary area is given in Figure 5. It is important to note that the line of sight bounded area is not equivalent to the slope domain, but is rather a second landmark region with its own specialized search domain. As Figure 5 shows, one reason that the two are not equivalent is that the line of sight bounded area takes in the ‘foot’ and ‘slope’ elements of the slope domain, but excludes the area internal to the ‘head’ element. In short, the two domains are not equivalent by virtue of not being isomorphic. In addition, the speaker’s canonical position is not necessarily his actual one and the line of sight landmark region essentially takes the face of the slope as its search domain. The morphemes u ‘inside’and a ‘outside’ take on specialized meanings within this complex domain. Specifically, u ‘inside’ is broadly construed as being within the search domain of the canonical speaker’s line of sight, whereas a ‘outside’ is broadly construed as being outside that line of sight (or, directionally), at an oblique angle to it.



Asymmetries for locating events with Cora spatial language 

A consideration of the semantics of uh ‘off up there in the slope’ shows the need to characterize the medial and distal ranges in highly schematic terms and invoke a set of related variants that instantiate them in the appropriate contexts. Specifically, for the Trajector as Object of Conceptualization to be in the line of sight of the speaker, but also to be outside the medial range of the speaker, which I originally defined as the limits on his visual field (Casad 1982 :58; 88) is contradictory. However, if we construe the medial range as a highlighted focal reference point, then the various components fit. It is quite consistent for the Trajector as Object of Conceptualization to be outside of the search domain of the speaker’s proximity at the same time that he is still within the speaker’s line of sight along the slope. Figure 6 gives a diagrammatic representation of central aspects of the semantic representation of uh ‘off there straight up in the slope’. In this diagram, the viewer’s status as the non-focal point of reference is indicated by ‘V=R’. “Head” TR

“Face” “Foot”

LM V=R

‘

Figure 6.  uh ‘straight off up there in the slope’

The next locative particle I discuss is ah ‘off there out in the slope’. This particle occurs not only with spatial meanings (2a), but also with temporal ones (2b). In addition, it occurs as a bound prefixal configuration in various kinds of constructions with topographic suffixes (2c), verb and adjective stems (2)(d-e), and incorporated nominal forms (2f). (2) a. a-h tú wa-ra’a-k outside-slope we EXT-corner-leave ‘We came back around from off behind the side of the hill’. b. a-h pú nu’u i i-ra’ outside-slope SUBJ QUOT SEQ NARR-face

‘a-h-¢uná outside-slope-spring up ‘Then, so they say, she jumped right up to go meet him.’



Eugene H. Casad

c. a-h-tyi t a’-u-h-méh outside-slope-uphill SUBR DIST-ABL-slope-go.SG ‘The guy who is going uphill along the trail’ d. šá-’a-h-wii-š' muán you:PL-outside-slope-stand-DISTR:IMP you:PL ‘Get up out of bed, you guys.’ tyámwa’a m či’i e. á-h-tyee outside-slope-long really ART HOUSE ‘That building is really tall.’ f. ka=pú á-h-kwasi m pina’a NEG=SUBJ outside-slope-tail ART bird ‘That bird has no tail feathers.’ A schematic representation for ah in its topographical uses is given in Figure 7. As usual, this schema subsumes various specific configurations. The one I have in mind here locates the speaker at the canonical viewing position from the foot of the slope. This is a point of reference for defining the landmark region and the viewer is again indicated by the circle labelled V = R. As in all of our previous examples, the speaker is offstage and the Trajector may be located in any number of locations within the search domain of the landmark area. In this usage, the Trajector is located well outside of the line-of-sight landmark region LM. Also, the speaker’s vantagepoint relative to the target’s position is flexible in the distinct and felicitous usages of ah; the meaning of ah includes no specific mention of this information, and, since the speaker is not specifically mentioned, he is being construed highly subjectively. “Head” “Face”

LM TR

“Foot” V=R

Figure 7.  ah ‘Off up there at the side of the hill



Asymmetries for locating events with Cora spatial language 

3.3

The topographic adverbs

The locative particles enter into construction with members of a set of topographic suffixes to derive a set of locative adverbs which I call ‘river- oriented’ adverbs (Casad, 1977, 1982, 1988). These adverbs profile areas within the domain of the lay of the land.5They are integrated systems in their own right and are defined with respect to the specialized domains of the river and the hill areas as the array of suffixes in (3) shows. (3) a. -tye -tyapwa -ta(va)n

‘downstream’ ‘upstream’ ‘across the stream’

b. -tyi(vi) ‘uphill’ -k ‘downhill’ -ka’i(wa) ‘over the hill’ Again, in giving an account of the semantics of these suffixes, we need to invoke multiple landmarks which are simultaneously accessed by the speaker for the appropriate usages of these expressions. The base for the river domain takes the path of a river and the canonical direction of its flow as primary locative reference points and calculates certain relations in terms of them. Basically, a Trajector can be located in terms of either end of the river’s path or can be located on the opposite side of the river’s path from the position of the speaker. The speaker can also be differentially located with respect to any one of these three reference point areas. The distal points themselves are locative reference points that I label with the usual metaphors “head” and “mouth”. These correspond roughly to the source and goal, respectively, of the flow of the river. In most cases, the entire course of the river is not on the speaker’s mind, but rather an intermediate sector which is more or less bounded by his range of vision. This implies Langacker’s model of the viewing arrangement with its distinction between the maximal scope of the viewing arrangement and its immediate scope, i.e. the onstage area where events take place (Langacker 1999 :204f). In short, these two locative reference points are not the actual extremities of the river as a whole, but are rather the extremities of a highly variable neighborhood, which is frequently, though not necessarily, equivalent to what the speaker sees. Often enough it is a totally imaginary neighborhood. Figure 8 represents the conceptual base associated with the river domain. It includes three landmark regions which figure in the meanings of the topographic suffixes. All six logically possible combinations of speaker vantagepoint with respect to the locative landmark areas do occur, and each imposes a distinct configuration on the base depicted in Figure 8. The appropriate characterization of the prototypical meaning of each topographic suffix is determined by the choice of which landmark area 5. These adverbs also have additional applications to other spheres of reference such as location on a human or animal body or location within a house, but I do not discuss any of these usages here.



Eugene H. Casad

serves as the non-focal point of reference and which remaining pair is selected as simultaneous focal reference points. Therefore, neither a specific speaker’s vantagepoint nor a trajector’s position is overtly indicated in Figure 8. This diagram represents the abstract domain that is evoked by prototypical usages of each member of the set of river-oriented topographic suffixes given in (3) above. The simultaneous accessing of three locative reference points is indicated by the use of three arrows that link the viewer with each of the locative reference points. HEAD LM2

LM1

MOUTH

River

LM3

IS V=R

Figure 8.  River Domain Viewing Arrangement

For brevity’s sake, I select only a single topographic suffix, -tyapwa ‘upstream’, for discussion here. The abstract domain is that of the river. The configuration appropriate for explicating the meaning of –tyapwa ‘upstream’ is obtained by locating the target of conceptualization = trajector within the search domain of the locative landmark area labeled “head” as in Figure 9. In addition, the speaker’s vantagepoint is located within the search domain of the “mouth” landmark area along the river’s path. “Head”

“Mouth”

River TR

V

Figure 9.  -tyapwa ‘upriver’ topographic suffix

It is important to notice that these locative landmark areas are not congruent with the actual “head” and “mouth” of the riverbed, but rather are construed as being indefinitely bounded areas taking in a search domain that subtends a region at each side of the riverbed at an area sufficiently removed from the speaker’s prototypical location to



Asymmetries for locating events with Cora spatial language 

be characterized as being either downstream or upstream from him. I indicate the offstage status of the viewer in this diagram by the dotted formatting of the downriver vantagepoint area. I now turn to the distal form áhtyapwa, which I gloss as ‘off there upriver in the slope’. With respect to the various facets of the ‘river’ schema of Figure 8, then, the Trajector has, by definition, an ‘in’ relation to the relevant upstream or downstream locative landmark area, whereas he may have either an ‘in’ relation or an ‘out’ relation to the search domain of the river. Thus, in forms such as úutye ‘along the river off downstream’ the Trajector’s ‘in’ relation is to the alignment of the flow of the river. In real life, this means that the Trajector is in the search domain of a locative landmark area that the Cora speaker broadly construes as átye’e ‘along the river’. Thus it is common to hear a Cora speaker refer to an unspecified downriver location by the complex adverbial phrase úutye átye’e ‘way off yonder downstream along the bank of the river’. On the other hand, the contrastive ‘outside’ form áatye places the Trajector outside the river area designated by átye’e. This makes the phrase *aatye atye’e inappropriate. Similar reasoning presumably applies to the form áhtyapwa. The various locative landmarks evoked by áhtyapwa are depicted diagrammatically in Figure 10. Both the river and the upriver areas are landmarks that the speaker accesses simultaneously in order to characterize the Trajector’s location. The ‘onstage’ area includes the search domains of both the river and the ‘Head’ landmark areas. The search domain for the river is a region adjacent to each edge of the river channel. ‘Inside’ vs. ‘Outside’ is defined with respect to this search domain. The search domain for the ‘Head’ landmark area includes both ‘Inside’ and ‘Outside’ locations vis à vis the river channel. The offstage status of the Viewer carries over to the entire locative prefix + topographic suffix construction. “Head”

“Mouth”

River

D

V

TR

Figure 10.  áhtyapwa ‘off upriver in the slope’

4. Maximally objective object of perception Usages that exemplify a maximally objective construal of the object of perception typically entail a scene in which the speaker vantagepoint is external to the profile and the intrinsic orientation of the object of perception is irrelevant to the usage. When the



Eugene H. Casad

subject of perception is simultaneously construed as maximally subjective, the usage puts no element of the ground on stage. In other words, all the configurational properties expressed by the prefixes pertain to the internal configuration of the trajector, and can be assessed regardless of the speaker’s vantagepoint. Thus, the speaker is not involved in the guise of a spatial point of reference (Ronald W. Langacker, personal communication). I illustrate these points with examples (4) (a-d). (4) a. anšvi-ka pu yééwi w-í-h-mu’u five-ACC SUBJ QUOT in-path-slope-head ‘It has five heads’ b. w-í-h-piti in-path-slope-skinny ‘coke bottle shaped’ c. d.

an-tyí-pityíh-me top-up-skinny-COLL ‘The eating fork has sharp tines’ an-tá-tame top-across-tooth ‘The cow jaw has teeth’

The locative prefixes on the verb in (4a) designate a shape configuration that includes a stricture at some point along the primary axis of a physical object that is longer than it is wide. As one scans the profile, it narrows considerably along one relatively short section of the entity in question and then widens again. For the human or animal body, this configuration extends throughout both sides of the region from the top of the shoulders, up along both sides of the neck and ends at the top of the head. Although the body has an intrinsic axiality, this does not invoke the speaker’s position in any way: the statement that Elder Brother makes about the serpent having five heads, drawing on the locative prefix sequence u-í-h ‘inside-path- slope’ applies to both the state in which the serpent was alive, holding his neck vertically above the surface of the water in the lagoon, as well as when he was dead and stretched out horizontally on the ground.6 The examples given in (4) (b-d) reinforce this point. The coke bottle (4b) has an analogous shape with a narrowing from all sides to the mouth at its top. It makes no difference whether the coke bottle is standing on its bottom or lying on its side. Also, if it is lying on its side, it makes no difference to the speaker in which direction the narrow part of the bottle points. The same holds true for the fork, with its group of parallel, pointed tines at its business end (4c) and for the loose cow jawbone lying on the ground (4d); rightside up or upside down, the speaker would use an-tá-tame ‘it has teeth’ regardless of the way that the jawbone is positioned.

6. The QUOTATIVE clitic yéewi in (2a)is a grounding element in its own right.



Asymmetries for locating events with Cora spatial language 

The only requisite here is that the speaker must be able to see the cow teeth in actual contact with the jawbone. In summary, these examples illustrate situations that are construed as having a maximally objective object of perception and as providing a simultaneously maximally subjective construal of the subject of perception.

5. Highly objective construal of object of perception: Variations in the degree of subjectivity with which the subject of perception is construed are numerous. In many cases, the speaker vantagepoint is external, but the situation itself presents us with either a canonical viewing position important to the scene in focus or with a speaker imposed orientation important to the entity being described. The object of perception is still construed in a highly objective manner, but the viewing arrangement entails a tighter grounding of the orientation of the situation being described to the speaker or conceptualizer than is entailed by the usages described above. Examples (5) (a-c) are typical. (5) a. a-uu-tá-tapwa out-that:way-straight-broken ‘The pot is broken in two’ b. an-ká-tee top-down-long:way ‘It is a long ways down’ c. an-tí-tee top-up-long:way ‘It’s really tall’ The examples given in (5) (a-c) all reflect static situations. They also illustrate a very common Cora verb construction which is subjectless, i.e. there is no focused trajector. Similarly to the cases involving verbs with meanings such as ‘hollow’, ‘pour’ and ‘be lighted up’, the verb tee profiles a relationship that has a spatial location /distribution but is not ascribable to a specific, tangible thing in the way that a typical verbal trajector would be (Ronald W. Langacker, personal communication). In the case of the broken pot (5a), things are ambiguous: the break may be either vertical, going from top to bottom, or horizontal, going from side to side. The crucial perceptual factor is that the break leaves approximately equally sized fragments. In this case, the grounding role of the speaker is only loosely tied to the stative relationship being described, i.e. the Subject of Conceptualization is construed in a highly subjective manner. This example is essentially identical to the four examples given in (4) (a-d). Things change somewhat in (5b) and (5c). The speaker’s vantagepoint is clearly involved in the structuring of the scene presented by an-ká-tee. In this instance, the speaker, as viewer and point of reference, is positioned at top of a cliff and is looking



Eugene H. Casad

downward to the foot of that cliff. The use of the two locative prefixes in this expression is instructive. The first one in the sequence, i.e. an-, is anchored to the speaker’s position, whereas the second one ka- specifies the directionality of his/her viewing. I have represented all this diagrammatically in Figure 11. The speaker’s position in the role of a subjectively construed non-focal point of reference is indicated by the circle labeled “V = R” at the upper left of the diagram. The speaker’s line of sight is indicated by a broken arrow, which here represents visual perception. The speaker’s position is also shown to be anchored in the conceptualization of the locative prefix an- ‘at the top of ’, indicated by the label an- in this diagram. The placement of the label kaabout halfway along the path of visual perception is intended to convey the idea that the directionality of the path is more salient to this usage than is the strictly locative sense of ka-. To be sure, the path of visual perception impinges on a locative reference point at the bottom of the cliff and the location of that end point is a crucial part of the assertion associated with the use of the stative verb tee that conveys the notion of something being a long way off. I have therefore included in this diagram a salient locative reference point at the end of the path.7To reinforce the directionality of it all, I have backgrounded the salience of the speaker’s position vis à vis the destination point with a broken line style. V=R anka-

LM

Figure 11.  an-ká-tee ‘It’s a long ways down

Sentence (5c) presents us with an alternate choice of speaker vantagepoint. In this case, the speaker’s viewing position is at ground level. The use of an- in this expression does not match that of the previous example since it does not indicate the speaker’s vantagepoint. This unpredictability arises directly from the conventionalization of the usages of the Cora locative prefix sequences (Casad 1988). Often, possibly more often than not, the semantics of a locative prefix sequence is only partly determined by its 7.

For the relevance of terminal points in visual theory, see Marr (1982:71)



Asymmetries for locating events with Cora spatial language 

morphemic composition and the semantics of the overall construction reflects a variety of emergent properties. In summary, the locative prefix an- picks out the top of a slope or of an object in both (5) (b-c), but the speaker’s position can be either on top or at the opposite end of the path. The observation here is that, in both cases, the speaker is at the origin of the path coded by the directional element and therefore, the speaker’s position functions as a subjectively construed point of reference. In (5c), the overall sequence an-ti- designates a region and has itself come to take on the global meaning of ‘at the very tip top of X.’ Specific salient tip tops include those of trees, tall buildings, church towers and steeples and steep mountains such as the Matterhorn or many volcanoes. I represent the commonalities of these entities by a tall and narrow triangle in Figure 12. The tip top is a salient locative reference point and may even find its specific meaning here rooted in metonymy, i.e. it designates a salient subpart of a discrete or, at least, highly perceptible entity. The salience of the endpoint, the implicit verticality scale and the speaker’s subjectively construed vantagepoint jointly force a specific directionality on the visual path (cf. Sovorou 1993 :9). The part of the story that I have not yet told has to do with the speaker’s judgment that the degree of spatial extension associated with a region occupied by some entity or another significantly exceeds one’s normal expections about it. This precise information is conveyed by the stative verb stem tee. Again, this expression invokes no discrete entity that can function as a clausal subject. What is profiled here is both the spatial extensibility associated with a region in focus and the fact that the degree of extensibility attributed to it is judged to surpass some indefinite expected range of normal extension. I illustrate this here with the single example of an-ti-tee,. The full diagram is given in Figure 12 below. To more fully characterize the meaning of an-ti-tee, we now add a verticality scale at the left hand side of the diagram; the label ‘N’ represents the norm that one customarily expects about the size of some vertically oriented entity. That this represents an approximate, and not absolute value is indicated by the dotted oval situated approximately halfway above and below the hashmark that crosses the scale at the ‘normal’ limits. The onstage area takes in the entire range of extensibility of the vertically oriented entity the speaker has in mind. The discrepancy between the ‘normal’ mark on the verticality scale and the position of upper termination on that scale is what motivates the judgement of relative tallness that gets attributed to some unspecified entity. All of these elements are represented diagrammatically in Figure 12 below. In summary, both an-ká-tee and an-tí-tee represent situations in which the object of perception is construed in a fairly highly objective manner, situations also in which the position of the Subject of Perception is construed subjectively as a point of reference for defining the specific relationship, albeit still not brought onto the stage, i.e. the Subject of Perception is not part of the profile (Langacker 1999 :297).

Eugene H. Casad

T

LR anti-

Verticality



N

S

Figure 12.  an-tí-tee: ‘It’s a long ways up

Additional examples of such subjectless clauses as given in (5 a-c) are legion in Cora. Three more are given in (6 a-c). (6) a. an-ta-mé top-straight-go:PAST ‘He crossed the street’ b. an-tá-tee top-straight-extended ‘It is long way across to the other side’ c. á ha-úu-tee there: DIST DIST-that:way-extended ‘It is a long ways off to over there’ What is common to the examples in (6 a-c) is the directionality of movement and the anchoring of the directed movement to the speaker’s position and leading away from it. (6a) presents actual motion of someone away from the speaker’s viewing position, going across a bounded ground level surface from one side to the other. This is conveyed by the locative prefix sequence an-tá- ‘top- straight’.8The use of the verb stem me ‘go’ itself implies physical motion with a mover and a canonical directionality. The locative prefixes, then, fill in the details of the motion involved. Sentences (6b) and (6c) illustrate ‘fictive motion’, which Langacker characterizes as being both abstract and subjective (Talmy 1996; Langacker 1991 :156; 162). Both cases of ‘fictive motion’ anchor the scanning process to the speaker’s position and lead away from it to a distal locative reference point. (6c) differs from the first two examples in that its immediate scope goes from the speaker’s position to a distal point indefinitely 8. This sequence employs a third variant of an-. This particular version derives historically from the sequence a-nu-ta- (Casad, 2004:133-4).



Asymmetries for locating events with Cora spatial language 

removed from it. In summary, these three examples construe the Subject of Perception highly subjectively and present the Object of Perception highly objectively while grounding that object saliently to the position of the speaker without bringing him/ her on stage.

6. Mixed objective and subjective construals In many examples of Locative Prefix + Verb Stem constructions with mixed objective and subjective construals, we find that the meaning of at least one locative element saliently includes the speaker’s vantagepoint with a given orientation in its Image Schema. The examples that I have in mind are given in (7 a-b). The locative prefix in question is the allative ii-, which I gloss as ‘this way’. (7) a. a-ii-káa-me m avioon out-this:way-down-go:IMPERF ART airplane ‘He is coming in this way for a landing’ b. a-ii-tá-tee out-this:way-straight ‘It is a long ways across to back here’ Both of these examples illustrate a directionality of motion towards the speaker’s position. This motion may be either physical, as in (7a), or fictive, as in (7b). The prefix sequence a-ii-ká- typically designates the movement of a discrete entity along a downward path coming toward the speaker’s location, as illustrated in (7a). This sentence was spoken by a Cora sitting with me off to one side of the airport runway in Tepic, Nayarit as we were watching a DC-3 coming back from its flight into the mountains of Nayarit. The use of the deictic definite article m indicates here that the airplane was in clear sight of the speaker and also was in view by the hearer. As a grounding element, the definite article invokes both the Speaker and Hearer highly, but still does not profile them. A pictorial representation of this situation is given in Figure 13. Again, the speaker’s position, along with that of the hearer, is highly evoked in this usage, but is still ‘offstage’ in the viewing arrangement. In Figure 13, the domain is physical, three dimensional space, the perceived location of the airplane is designated by the deictic medial form of the definite article, i.e. m and the locative prefix a- ‘outside’ indicates that the scope of the entire scene selected for comment is visually accessible to the speaker. The directionality of the descending path of the airplane towards the speaker’s location is expressly indicated by the locative prefix ii-, whereas the total expanse of the downwardness of the directed path is signaled by the final locative prefix -ka ‘down’ and is calibrated against a verticality scale set to the right in Figure 13.

Eugene H. Casad 3-D Physical Space

a-

horizontal

Vertical



ii-

ka-

S-V

Figure 13.  a-ii-káa-me må avion ‘The plane is coming in for a landing’

The sequence a-plus ii-plus ta- ‘straight’/ ‘across’ refers to position at or motion from one side of a distal Locative Reference Point facing across toward the observer’s position at the other side of the boundary, as sentences (8 a-b) show. (8) a. a-ii-tá-nyeeri-’i ayún he¢e outside-this-across-visible-STAT back to way here ‘From a source on the far side of the river it is lit up coming clear across the water towards the observer’s position.’ b. a-ii-tá-tee ayún he¢e outside-this-across-long back to way here ‘It is a long ways from the far side of the river coming back across to here.’ In these two usages, a-ii-tá-tee- shows the role of ii-to mark direction toward an observer’s position in a graphic way. Several of the forms with -tee which we have seen thus far, i.e. (3)(b-c) and (4b) show that -tee marks distance from the observer’s position to a more distal point. In contrast, a-ii-tá-tee anchors the distance scale to a distal point and treats the observer’s position as the goal of the path. This is highlighted by the use of the postpositional phrase ayún he¢e ‘back on this side’. Note again that there is no subject of the clause in these sentences. Instead, we have the use of a postpositional phrase that takes a locative adverb as its object and serves to highlight the speaker’s position as a non-focal point of reference for structuring the scene he/she is describing. The various aspects of the semantics of a-ii-tá-tee are depicted diagrammatically in Figure 14. The circle with an inscribed ‘LR’ in it anchors the observer-oriented perceptual path of ii-, the goal of which is the speaker’s position, indicated by a circle labeled ‘V = R’. Again, I depict the notion of ‘a long ways’ with a distance scale set parallel to the perceptual path. This represents a crucial part of the meaning of -tee. As seen earlier in Figure 5, ‘normal’ distance is indicated by both a hashmark crossing the distance scale at roughly midpoint and an encompassing oval that represents an indefinite range of distance values that can be construed as ‘normal’, indicated in this diagram by N.



Asymmetries for locating events with Cora spatial language 

The Speaker’s position is once again strongly invoked by this usage, as evidenced by the use of the proximal form of the adverbial object of the postpositional phrase ayún he¢e ‘back here on this side’, but the Speaker himself is part of the profile of neither the verb a-ii-tá-tee nor of the postpositional phrase. LR

N T

S

Figure 14.  a-ii-tá-tee ‘Long ways back to this side’

7. Gateways for objectivizing the Subject of perception Up until now, none of the usages we have discussed have involved bringing any elements of the ground onto the stage of the viewing arrangement. This can be done, however, and quite easily. As Langacker states the case: “Facets of the ground can however be put onstage as focussed targets of conception, being profiled by expressions like I, you, here, and now. When explicitly mentioned, the speaker or hearer has a dual role as both subject and object of conceptualization (Langacker 1999 :298).

As Langacker observes, putting the Speaker onstage lessens the subject/object asymmetry (Langacker 1999 :211). The simplest way to bring the Speaker onstage in Cora is via the selection of Subject and Direct Object Prefixes. The 3rd person singular participant is the most neutral entity available for viewing situations in a detached manner. Within the conversational schema, the 3rd person entity is also the one maximally distinct from either the speaker or the hearer and is the prototype for the viewing arrangement. These points are illustrated by examples (9) (a-c). (9) a. am pú n-aa-tú’a ínyeeci DEM SUBJ me-COMPL-hit me ‘That thing right there hit me.’ b. am pú n-a-’i-tyáa-tu’a y-éh nya-ware-’e DEM SUBJ me-outside-path-middle-hit here-slope my-back-on ‘That thing hit me right here in the middle of my back.’ c. am pú na-’an-tyí-tu’a y-én nya-¢ú’u hapwa DEM SUBJ me-top-up-hit here-top my-nose on ‘That thing hit me right here on the tip of my nose.’



Eugene H. Casad

The subject of perception is brought onto the scene as the object of perception in (9a) both by means of the 1st person singular direct object prefix on the verb tú’a ‘to hit’ and by the postverbal free personal pronoun. In (9b) and (9c) the Speaker as Subject of Perception is also the Object of Perception as signaled by both the lst person singular direct object pronoun and the 1st person singular possessive prefix on the body part that serves as the object of the postposition in the post-verbal locative body part phrase. These two morphemes, taken together, suggest that the subject of perception has received a high degree of objectification. These examples also illustrate the use of the locative particles as heads of locative body part phrases, a special case of a reference point construction (Langacker 1999 :387, fn. 6). In short, the locative particle sets up a locative reference point and the postpositional phrase elaborates it. Notice that the locative particle used to designate the back of a first person singular patient is yéh ‘right here in the slope, whereas the locative particle used to designate the nose of a first person singular patient is yén ‘right here at the tip of.’ This reflects an intimate connection between the locative system and the subject and object marking system of Cora. Specifically, just as there are three distinct persons in the subject and object paradigms, the locative particle paradigm locates entities at three maximally opposed positions vis à vis the speaker along what I have called the “distance parameter” (Casad 1982 :81–2). In strictly locational. terms, the three contrastive positions of the “distance parameter” (Casad 1982 :81–2) can be labeled as ‘off there’, ‘right there’ and ‘right here’, respectively. In specific contexts, this three-fold locative distance contrast corresponds in a one to one fashion with the three-fold Speaker-Hearer-3rd Person contrast characteristic of the Conversational Image Schema (cf. Casad 1991). As seen in the following examples, the morphological chunks that reflect these three positions are Ø, m- and y-. One specific context is in the Locative Particle + Possessed Body Part Locative Phrase. The example we use to illustrate this draws on the body part mú'u ‘head’. The link between the domain of the human body and the canonical speech situation gives the distance predicates conventional variant meanings as follows: y ‘proximal’ means ‘speaker’s location’, m- ‘medial’ means ‘hearer’s location’, and Ø ‘distal’ means ‘other than speaker’s or hearer’s location’. Within the construction that draws on the pairing of these two domains, the ‘on top’ particles, for instance, instantiates the ‘inflectional’ series given in (10). It should be obvious that this particular variant of the locative phrase construction consists of paired reference point constructions. (10) a. y-é-n nya-muu-¢e-’e here-outside-top my-head-ABS-on ‘ ‘right here on the top of my head’ b. m-á-n a-muu-¢e-’e there-outside-top your-head-ABS-on ‘right there on top of your head’



Asymmetries for locating events with Cora spatial language 

c. á-n muu-¢e ‘e-n outside-top head-ABS-on-his ‘there on top of his head’ The threefold distance distinction also conventionally carries over into the sets of topographic adverbs via the locative particles which enter into construction with the topographic suffixes. The ‘inside upriver’ series is given in (11) (a-c). (11) a. yúhtyapwa b. múhtyapwa c. úhtyapwa

‘right here back in the slope upriver’ ‘right there in the slope upriver’ ‘off there in the slope upriver’

I end this paper with a discussion of the use of the proximal adverb y-ú-h-tyapwa ‘right back here in the slope upriver’. This is the most morphologically complex topographic adverb that we consider in this paper. It places the entire scene within both the domain of the river and that of the hill, as well as within the domain of the proximity of the Speaker. The upriver location, signaled by -tyapwa is reinforced and further specified by the use of the locative prefix -h “in-slope” which is conventionally understood in its most schematic sense to locate a Trajector within the domain of a locative reference point area that I conventionally call “the face of a slope”. The pairing of -h with u- ‘inside’ then tracks a path from a horizontal base straight up a slope to the top of a hill to a specific point below the top of a hill, as illustrated earlier in Figure 6. This mental tracking (cf. Matsumoto 1996 :137; Miller & Johnson-Laird 1976 :53, 244) is illustrated clearly by example (12). (12) ú-h-tyee inside-slope-long ‘It is a long ways straight up the slope’ The usage of yúhtyapwa that I have in mind here is illustrated by Sentence (13), taken from a historical text about events of the Cristero period in Nayarit, Mexico, which took place in the period from 1926–1929. (13)

áa mú véhli’i hira=’an-tathere they close circle=top-acrossk'’-kaa y-ú -h-tyapwa go around-SimMode here-inside-slope-upriver kw áaša’ata a’hna  Jesus Pineeda, San Francisco DEM ART P.N. P.N. ahtá ahna  a’at Lorenzo Estraada CNJ DEM ART person P.N. P.N. ‘That guy Jesús Pineda and that other fellow Lorenzo Estrada kept coming around, right back here upriver, near the town of San Francisco.’



Eugene H. Casad

In example (13), the locative prefix sequence an-ta- contributes schematic information about the global spatial setting within which the event occurred. Specifically, it notes that they repeatedly crossed the river into Cora country, however, it does not of itself tell the listener anything about the specific geographical region where this took place. This specific information is conveyed instead by the discontinuous locative particle + topographic adverb phrase that means “back off there upriver in the village of San Francisco.” The adverb yúhtyapwa provides us with a clear example of how one goes about heightening the role of the speaker in the grounding of the event without placing the speaker within the onstage area of the viewing arrangement. It also illustrates the joint operation of multiple reference points to establish the location of the Trajector event. The next step in accounting for the way that these multiple points are simultaneously employed in a single viewing arrangement appropriate to the distal topographic adverb úhtyapwa is represented diagrammatically in Figure 15. “Head” L: RP-B

D

L: RP UP

“Mouth”

L-RP: River

TR

V=S

Figure 15.  uh-tyapwa ‘inside upriver’ topographic adverb

The three locative landmark regions are the river (LM 1:River), the boundary region adjacent to the river, on both sides, that separates entities from either being in that region’s search domain or outside of it (LM 2) and the the “head” of the river region (LM 3:up). The Speaker (= Viewer) is located in the “mouth” region downstream from the Trajector. The speaker establishes mental contact with the Trajector event by accessing all of the locative landmark regions simultaneously. A natural way to model this in procedural or computational terms would begin with placing one’s focus of attention on the part of the scene subtending the locative landmark region with the widest scope, continuing on to the one with the next widest scope and finishing with the locative landmark region with the most restricted scope, analogously to the account that Langacker gives for the nested prepositional phrases of English (Langacker 1999 :6, 49ff.; 60–1). 1 simply use a single broken arrow to indicate this process in order to keep the diagram simple and readable. I have also left the distance parameter implicit in this diagram. The distal viewing position is the most neutral one for the locative adverbs and for this configuration we can as a convention



Asymmetries for locating events with Cora spatial language 

simply say that the boundary of the region in which the Speaker = Viewer is located separates ‘distal’ locations from ‘proximal’ ones. The ‘proximal’ viewing arrangement builds on all the complexity shown in Figure  15, but highlights the Speaker’s role as ground of the utterance, extending his ‘proximal’ point of reference area to subtend also the location of his upriver target. This is diagrammed in Figure 16. To convey the evoking of the Speaker’s position as a subjectively construed point of reference, I draw on Langacker’s notation for the ground of an utterance: Both the Speaker and Hearer are overtly indicated as being in a reciprocal communicative relationship. The Speaker is indicated by a circle labeled ‘S’, whereas the Hearer is indicated by a circle labeled ‘H’ and their communicative interaction is indicated by the short double headed arrow placed between the two circles. The bolding of the circle corresponding to the Speaker indicates that the Speaker is selecting his/her own position as a special case: i.e. the Trajector event is distally located vis à vis the speaker in objective terms, but the Speaker is anchoring that distal area to his own location at the time of speaking. Another way of stating this is to say that the Speaker establishes mental contact with another entity in a different location by mentally extending his own position outward so that it comes to include that distal location. I have indicated this by a large oval that includes both the Speaker and his Target. That oval is labeled ‘Proximal’. This is also categorized as Locative Landmark 3 in this diagram. The other landmark regions are (a) the River, (b) the Head area upriver from the Speaker and (c) the Boundary area that separates ‘along the river’ locations from ‘away from the river locations’. L-RP4: Boundary

D

L-RP2: L-RP1: River Head D TR

L-RP3: Proximal

Mouth

S

H

Figure 16.  yúhtyapwa: ‘Right back here upriver’

8. Conclusion In this paper we have discussed several particularly salient asymmetries and have explored their roles in motivating usages of Cora spatial language to locate events, situations, conditions, relations, people and things. The first of these asymmetries is that of the speaker’s role as observer: in framing his conceptualizations for linguistic expression, he selects a particular viewing arrangement with a specific focus. The second asymmetry is that of locating entities within some field of spatial extension: spe-



Eugene H. Casad

cific entities are located spatially with respect to some locative reference point. As we have seen, many of the Cora spatial expressions are conceptually and morphemically complex and invoke multiple locative reference points which the speaker accesses simultaneously in order to locate the entity that he has in mind. Finally, the most high level asymmetry, i.e. the Subject of Perception vs. the Object of Perception contrast relates to these other asymmetries in subtle ways. Certain grammatical correlates emerge from this study that explain the consistency for the adverbial sets of spatials to construe the Subject of Perception in a highly subjective manner. They often occur sentence initially and may be the base for attaching a subject clitic. They also incorporate a series of morphemes that calibrate the position of the entity in mention or the event being described vís à vis the speaker in terms of degrees of distance without placing the Speaker on stage. Finally, they are relatively free in word order occurrence in a sentence and often function as antecedents to more complex spatial expressions that elaborate them in greater detail. The three distance morphemes are inherently deictic and evoke elements of the ground in three different ways, without profiling any one of them. In contrast, several grammatical correlates suggest why the locative verbal prefixes construe the Object of perception in a highly objective way in their usages. For one, they provide highly schematic settings for the scenes designated by the nouns, verbs and adjectives they combine with for describing both static situations and dynamic interactions between discrete entities, as well as the manner and direction of motion. Second, they have grammaticalized in ways that one would not expect highly subjective elements to do. Thus, certain combinations are possessive statements, whereas others correlate with specific tense-aspect usages. Finally, the distance parameter, so relevant to the adverbs, demonstrative pronouns and definite articles of Cora, occurs in a much more degraded form in the locative prefix system. The distal locative prefix á'- is very common and signals that the event being related by the speaker is outside the speaker’s immediate location at the time of speaking. The proximal and medial distance morphemes seldom occur tightly bound to the verb word.9 In conclusion, the intrinsically asymmetric subjective-objective continuum plays a highly significant role in the explanation of the overall functioning of the Cora system of spatial language. In particular, the difference in profiling between the locative particles and topographic adverbs and the locative prefix + verb stem constructions explains the differential roles, i.e. the locative particles and the topographic adverbs profile regions, whereas the verbal constructions profile processes, states and things.

9. El Nayar Cora has a rule of glottal deletion when followed by a high-toned vowel (Casad 1984:159). This appears to be the primary context in which y- and m- function as locative prefixes. It is interesting that neither Mesa del Nayar Cora nor Santa Teresa Cora have this rule of glottal deletion, but recent work parsing and glossing El Nayar Cora text materials leaves little doubt, to me at least, about the status of y- and m- as bonafide locative verbal prefixes.



Asymmetries for locating events with Cora spatial language

References Casad, E. H. 1977 Location and Direction in Cora Discourse. Anthropological Linguistics, Vol 19, No. 5:216–41. Casad, E. H. 1982 Cora Locationals and Structured Imagery. La Jolla, CA: UCSD Doctoral Dissertation. Casad, E. H. 1988 Conventionalization of Cora Locationals. In: Brygida Rudzka-Ostyn, ed. Topics in Cognitive Linguistics:345–78. John Benjamins Series Mayor on Theoretical Issues in Contemporary Linguistics. Amsterdam/ Philadelphia: John Benjamins. Casad, E. H. 1993 On the notions ‘Locative,’ ‘Path,’ ‘Source’ and ‘Goal.’ In: Richard A. Geiger and Brygida Rudzka-Ostyn, eds. Conceptualizations in Natural Language Processing:603–54. Berlin/ New York/ Amsterdam: Mouton de Gruyter. Casad, E. H. 1996 What good are locationals. In: Martin Pütz and René Dirven, eds. Language and the Construal of Space. Cognitive Linguistics Research 8. Berlin/ New York/ Amsterdam: Mouton de Gruyter. Casad, E. H. 1998 Lots of ways to GIVE in Cora. In: John Newman, ed. The Linguistics of Giving:135–174. Amsterdam/ Philadelphia: John Benjamins. Casad, E. H. 2001 Where do the senses of Cora va’a- come from? In: Hubert Cuyckens and Britte Zawate, eds. Polysemy in Cognitive linguistics: Selected papers from the Fifth International Conference of the ICLA, Amsterdam, 1997:83–114. Current Issues in Linguistic Theory 177. Amsterdam/ Philadelphia: John Benjamins. Casad, E. H 2004 Imagery through the Ages. In: Eds. Barbara Lewandowska-Tomaszcyk and Alina Kwiatkowska, eds. Imagery in Language. Festscrift in Honour of Professor Ronald W. Langacker:115-158. Frankfurt am Main-Berlin-Bern: Peter Lang: Europäischer Verlag der Wissenschaften. Casad, E. H 1985 ––– and R. W. Langacker. ‘Inside’ and ‘Outside’ in Cora Grammar. IJAL 51:247–281. Harder, P. 1996. Functional Semantics: A Theory of Meaning, Structure and Tense in English. (Trends in Linguistics Studies and Monographs 87). Berlin/New York: Mouton de Gruyter. Hawkins, B. W. 1984. The Semantics of English Spatial Prepositions. [Unpublished Ph.D. Dissertation, University of California, San Diego]. Langacker, R. W. 1990. Subjectification. Cognitive Linguistics 1.5–38. Langacker, R. W. 1991 Concept, Image, and Symbol: the Cognitive Basis of Grammar. Cognitive Linguistics Research 1. Berlin and New York: Mouton de Gruyter. Langacker, R. W. 1999 Grammar and Conceptualization. Cognitive Linguistics Research 14. Berlin and New York: Mouton de Gruyter. Miller, G. A. and P. N. Johnson-Laird. 1976. Language and Perception. Cambridge, MA: Harvard University: Belknap Books. Marr, D. N. 1982. Vision. New York: Freeman and Co. Matsumoto, Y. 1996. Subjective Motion and English and Japanese Verbs. Cognitive Linguistics 7:124–156 Pike, Kenneth L. 1960 Language in Relation to a Unified Theory of the Structure of Human Behavior. Part III. Preliminary Edition. Glendale, CA: Summer Institute of Linguistics. Svorou, S. 1993. The Grammar of Space. Typological Studies in Language 25. Amsterdam/ Philadelphia: John Benjamins.





Eugene H. Casad Talmy, L. 1996. Fictive motion in language and ‘ception’. In: Paul Bloom, Mary A. Peterson, Lynn Nadel and Merril F. Garrett, eds. Language and Space: 211-276. Cambridge, MA / London: MIT Press / Bradford. Verhagen, A. 1995. Subjectification, syntax and communication. In: Dieter Stein and Susan Wright, eds. Subjectivity and Subjectivisation: Linguistic Perspectives: 103-128. Cambridge: Cambridge University Press.

Spanish (de)queísmo Part/whole alternation and viewing arrangement Nicole Delbecque University of Leuven

The introduction situates the approach taken and presents the structure of the paper (Section 1). The classical definition of (de)queísmo serves as starting point (Section 2). The analysis is done in terms of viewing arrangement (Section 3) and of part/ whole alternation (Section 4). By considering the head elements one by one, i.e., adjectives and nouns as well as verbs, it is possible to arrive at a global view on the variation (Section 5). The distribution of the verbs in frequency classes according to the Ø/de alternation confirms the predominance of the internal perspective (Ø) and suggests that the minority choice (de) of the external perspective is also systematic and structurally motivated (Section 6). The conceptual difference is further corroborated in the translation (Section 7). Finally, the semantic import of the Ø/de alternation should not be confounded with the inferences it contributes to canalize regarding the attitude of the speaker (Section 8).

Keywords: viewing arrangement, part/whole alternation, Spanish clause complementation, syntactic variation, sociolinguistic variable, pragmatic, evidentiality, subjectification, grammaticalization, speech or thought representation, relational profile, translation, inferences, construal, subject conceptualizing constructions, deictic center, ground, re-enactment, conceptualizer, subjective, Genitive, Ablative, figure-ground, mental space, clause complementation, saliency effects, vantage point, conceptualizing subject, unmarkedness, markedness, onstage conceptualizer, face preserving strategy, partitioning, paradigmatic, frequency, actual ground, stage-management

1. Introduction The aim is to present a global account of a double phenomenon of syntactic variation in Spanish clause complementation, viz., the alternate use of de que ‘of that’ for the simple que ‘that’-complementizer (dequeísmo), and the reverse, que for de que (queísmo). Although both phenomena are documented since the first stages of the language (Kany 1945, Cano Aguilar 1985, Boretti de Macchia 1989, Serradilla Castaño 2000, del



Nicole Delbecque

Moral 2004), they are traditionally set apart as ‘non-standard’ variation and condemned as ‘deviant’ by prescriptive grammar. Century-long stigmatization, especially of dequeísmo, has not prevented (de) queísmo to persist, however, and quantitative studies made clear that it is not a sociolinguistic variable as defined by the variationist paradigm, since it does not fully display the expected stratification in function of social class, age, sex, speech style, etc. (e.g., Bentivoglio and Galué 1998). Modern descriptive studies have therefore attempted to cast the variation in pragmatic terms, sometimes in combination with a mitigated sociolinguistic approach in terms of hypercorrection (Bentivoglio 1976, 1980; Arjona 1978). Dequeísmo has been suggested to have an attenuating effect on the assertion (Bentivoglio and D’Introno 1977; Bentivoglio 1980). This line of thought, first pursued by García (1986), has nowadays fashionably been reformulated in terms of evidentiality (Schwenter 1999) and subjectification (del Moral 2004). However, invoking extra-linguistic forces, be they social or attitudinal, fails to specify the proper grammatical meaning of the alternation. While sociolinguistic approaches direct the attention to the social profile of the speakers and the type of setting, pragmatic explanations put the burden of proof on the speech participants and what happens in the communicative interaction, without ascertaining that the invoked inferences are not triggered by elements of the context, i.e., without demonstrating that they really correspond to the que / de que alternation as such. From the data available, the variation still seems to be in an ongoing process of increase. As it results from a grammaticalization process, there are reasons enough to search for the grammatical meaning conveyed by the Ø/de alternation. The paper is organized as follows. It starts off from the traditional definition of queísmo and dequeísmo (Section 2), before proposing an account that connects the alternation with a basic meaning distinction linked to the use of the preposition de, and analyzes it as the manifestation of a more encompassing, cognitively motivated choice in viewing arrangement (Section 3). I will argue that the que-construal presents the speech or thought representation from the vantage point of the main clause subject, in terms of reenactment, while the de que-construal represents the intensional perspective of the speaker, who selects the speech or thought representation for current expressive purposes. As a corollary of the higher-order role of the speaker as conceptualizer, a derivative dimension comes in that downgrades the ‘original’ sayer-cognizer’s expressivity. The notions of selection and derivation are consistent with the overall schematic meaning of the preposition de and capture the difference in relational profile among the two construals, as well as the saliency effects each is susceptible of producing (Section 4). Rather than being a disturbing factor, (de)queísmo provides any kind of head with two complementation types: one with an exclusively internal viewing arrangement and one in which the speaker stage-manages the relational profile of the speech or thought representation. Even if the partitioning view (de) is typical of nominal and adjectival heads, the speaker-conceptualizer (C0) can choose to remain backstage (Ø)

Spanish (de)queísmo 



instead of acknowledging current access to the mental domain instantiated by the complement clause (Section 5). The on-line pan-Hispanic reference corpus of actual Spanish (CREA) has been systematically scrutinized in order to measure the extent of the variation with verbal heads. The skewing of the distribution at type, token and frequency level supports the idea that the Ø-construal is the unmarked option both in probabilistic and in cognitive-functional terms (Section 6). Another heuristic consists in turning to translation, since the existence of different construals may be expected to be reflected in the use of distinct verbal predicates in other languages, e.g. in English (Section 7). Additional inferences regarding the speaker’s position, e.g. in terms of assertion/ attenuation, certainty/doubt, direct/indirect evidence, have to be separated from the very meaning structure of the complementation type (Section 8).

2. Deviant uses? The labels queísmo and dequeísmo go back to the seventies. The coinage, which seems to be due to A. Rabanales (1974), is rooted in a norm-based approach. Queísmo is indeed defined as the “omission” of the preposition de ‘of ’ before the que ‘that’- complementizer where it is normally required by the head predicate, generally a verb, more rarely a noun or adjective (Gómez Torrego 1999: 2133). In (1) and (2), e.g., me alegro que ‘I am happy that’ and me acuerdo que ‘I remember that’ are seen as “deviations” from the normative me alegro de que ‘I am happy of-that’ and me acuerdo de que ‘I remember of-that’, respectively. (1) – – ‘–

Estaba pensando que es la habitación más bonita que he visto en mi vida. Gloria sonrió visiblemente halagada. Vaya, me alegro que te guste; la he puesto yo. I was thinking that it is the most beautiful room (that) I’ve ever seen in my life. Gloria smiled visibly flattered. – Come on, I am happy that you like it (lit.: “it pleases you”). I was the one who arranged it.’ (C. Martín Gaite, 1976 (novel, Spain)) (2) si le veía era a las nueve de la noche, una vez al día le veía y qué sé yo y en ese sentido sí, perdí mucho contacto con él. Pero después cuando dejó la política me acuerdo que nos íbamos todos los fines de semana a la estancia y así (Paraguay, Encuesta 120) ‘if I saw him it was at nine p.m., once a day I saw him and what do I know and in that sense yes, I lost a lot of contact with him. But later, when he quitted politics, I remember that we went every weekend to the country house and so.’



Nicole Delbecque

Dequeísmo designates the reverse phenomenon, viz., so-called “abusive insertion” of the preposition de before the finite complement clause where it is deemed superfluous (3). Submitted to strong censorship, it is hardly ever found in public writings nowadays. Most grammarians consider it to belong to low, informal registers of oral speech (Gómez Torrego 1999: 2107 ff.). Yet, a quick exploration reveals its presence even in quite formal settings. In (4) and (5), for instance, drawn from the “macrocorpus of educated speech” (Samper Padilla et al. 1998), indicar ‘indicate’ and sabemos ‘we-know’ are followed by de que ‘of that’ instead of the habitual que ‘that’. (3) “Tendencia a anteponer la preposición de al que preferentemente gramemático, cuando la norma “oficial” no hace esperar su presencia” (Rabanales 1974: 415) ‘Tendency to introduce the preposition de before the preferably grammemic que, when the “official” norm does not make it expected’

(4) Esto nos puede a nosotros indicar de que no todas las cosas que él estaba entonces haciendo, las consideraba él mismo perfectas, sino que podían salir no buenas. (HC-ME7)1 ‘This can show us (of) that not all things he was doing then, he himself considered perfect, but rather that they could end up being not good.’

(5) Nada más que sabemos de que Dios se ha dirigido al hombre y todo lo que le ha dado al hombre es para el hombre. (HC-ME7) ‘Nothing else than that we know (of) that God has addressed mankind and everything he has given to mankind is for mankind.’

Meta-linguistic comments found in America (6a) and in Spain (6b) testify to the ongoing stigmatization of dequeísmo. The negative attitude generated by the prescriptive tradition on both sides of the Ocean may well put a brake on usage, and perhaps explains the relative lack of interest for the functionality of the alternation. (6a) (...) siempre la misma pizza, siempre la misma amiga, no importa que fueran diferentes muchachas, porque insistían en comentar la película con un mismo fervor por los “pienso de que”, hasta hacerme sentir que la pizza y el dequeísmo eran un anticipo generoso del infierno. (V. Kociancich, 1982 (novel, Argentina)) ‘Always the same pizza, always the same friend, it’s not important that it were different girls, because they insisted on commenting the movie with a similar fervor with the “I think of that”, to the extent of giving me the feeling that the pizza and the dequeísmo where a generous anticipation of hell.’ (6b) Si se cumple este pronóstico y la pequeña pantalla arroja sobre nosotros algo de luz, juro que no protestaré cuando se utilice para decir “pienso “de” que estamos viviendo la cultura “de”...” Al fin y al cabo, es igual un “de” más que un “de”

1. HC stands for “Habla culta” ‘educated speech’, followed by the first letters of the country (ME: Mexico), and completed by the number identifying the interview.

Spanish (de)queísmo 



menos. Por lo que a mí respeta, me limitaré a decir amén. (El Mundo 13/04/1996, column by L. Rico, producer) ‘If this prognosis is fulfilled and the small screen sheds some light on us, I swear that I will not protest when someone uses to say “I think ‘of ’ that we are living the culture ‘of ’…” In the end, it is the same an “of ” more or an “of ” less. As far as I am concerned, I will limit myself to saying amen.’ According to the main stream view, the presence or absence of de ‘of ’ before que ‘that’ does not in any way affect the grammatical relation between the head and the complement. As verbs are supposed to arbitrarily take one complementizer or the other, so called deviant uses of de are thought of as a matter of confusion between verb complementation and noun complementation. Spanish nominal heads are indeed canonically followed by de que (7). The idea is that, by analogy with complex predicates of the form [verb + nominal group] followed by a complement clause (CL) introduced by de que (8), speakers are prone to put de que after the corresponding simple verbs – esperar ‘hope’, temer ‘fear, and the like (9). A similar reasoning is adduced for the reverse situation: insofar as the use of de que ‘of that’ is not perceived as being motivated, it easily reduces to que ‘that’. This would not only hold for complex predicates, such as darse cuenta ‘giveoneself account, realize’ (10), but also for simple ones, e.g. quejarse ‘complain’ (11). (7a) canonical coding with nominal head: de que CL (7b) la esperanza de que ‘the hope of that…’, la prueba ‘the proof of that…’, el temor de que ‘the fear of that…’, etc. (7c) (...) pero lo cierto es que las esperanzas de que se produzca quedaron en alto. (La Vanguardia, 30/11/1995 (Spain)) ‘but what is certain is that the expectations (of) that it would happen remained high‘ (8a) canonical coding with complex [V-N] predicate: de que CL] (8b) Aun así, Alborch tiene la esperanza de que en los próximos tres meses se llegará a materializar su petición de amnistía. (La Vanguardia, 02/10/1995) ‘Even so, A. has the hope (of) that in the next three months his demand of amnesty will come to materialize.’ (9a) canonical coding with verbal head: que CL → non-canonical coding: de que CL = “dequeísmo” (9b) Vivian, de 85 años, no espera que su nombre figure en el testamento de su marido (La Voz de Galicia, 1991) ‘V., of 85 years old, does not expect that her name figures in the testament of her husband’ (9c) “Me he quedado bastante sorprendido. Hay algunos pasos que todos hemos estado esperando, y, francamente, seguimos esperando de que lo tome antes de julio”. (Diario de las Américas, 10/02/1997)



Nicole Delbecque

‘It came rather as a surprise to me. There are some steps that we all have been expecting, and, frankly, we keep expecting him to take before July.’ (10a) canonical coding with complex predicate: → non-canonical coding:

de que CL que CL = “queísmo”

(10b) “Y qué quiere que le cuente yo de mi sobrino? ¿No se da cuenta de que queda muy ridículo que yo diga que es muy buen chico?” (El Mundo, 27/11/1994) ‘And what do you want me to tell you of my nephew? Don’t you realize (of) that it’s ridiculous for me to say that he’s a good guy?’ (10c) Bueno, ¿pero no se da cuenta que su actitud, medio como suplicante, la alejaba de usted todavía más? (Radio Continental 11/12/98 (Spain)) ‘Okay, but don’t you realize that your attitude, half kind of supplicating, was driving her away from you even more?’ (11a) if [V de que CL ] = canonical coding → then [V que CL] = non-canonical coding = “queísmo” (11b) Ferrer se quejó de que la dirigencia eduardista está entrando en un proceso polémico irritante. (El Universal, 21/04/1993, Venezuela) ‘F. complained (of) that E. leadership is entering an irritating polemic process’ (11c) En Rusia se quejan que los expertos occidentales ganan en un día lo que un profesor en un año (El Universal, 21/04/1993) ‘In Russia they complain that the occidental experts earn in one day what a professor (earns) in one year’ The rationale of the structuralist approach is summarized in Figure 1. Nominal heads axiomatically require the de que ‘of that’ complementizer. Verbal heads, for their part, split into two lexical classes: on the one hand, verbs that take que ‘that’ and, on the other, those governing de que ‘of that’. Insofar as this distribution is taken for granted, uses that depart from it are considered to be performance errors that arise under analogical pressure in one sense or the other. From a normative point of view, the use of such deviating blends is said to be typical of uneducated speakers and, hence, symptomatic for the existence of lower social class speech varieties.

canonical  noncanonical

[verbal head]

[nominal head] complex V-N head

de que CL

de que CL

que CL queísmo

que CL queísmo

Figure 1.  Structuralist approach of queísmo and dequeísmo

[verbal head] que CL de que Cl dequeísmo

Spanish (de)queísmo 



Analogical reasoning constitutes a first step towards a global account, since it is necessary to identify the head in order to determine the expected form of the complementizer and correctly situate the minority forms. However, this should be done without prescriptive bias. Queísmo and dequeísmo cannot be simply done away with as arbitrary, unmotivated deviations.

3. The {Ø/de} alternation: a matter of viewing arrangement (De)queísmo typically occurs in complex clauses with animate subject and sentential complement. According to recent counts (De Mello 1995: 121) more than 80% of the cases are found in this environment. This explains why this context has received most attention. In line with one of the basic tenets of cognitive grammar, the Ø/de alternation is to be seen as a matter of construal. Far from being meaningless, it can be assumed to have a crucial bearing on the relational profile. The context in which the alternation occurs, corresponds to what Achard (1998: 46), in his analysis of complex clauses in French, has coined “subject conceptualizing constructions” (CSC): “CSC verbs present the sort of mental operation the main clause subject performs on the complement content. In the CSC cases, the event in the complement is viewed as dependent on its perception, report, desire, etc.” The content of the complement clause is hooked on the main, reporting clause in two alternate ways. While queísmo is a way of transitivizing an otherwise intransitive construal, dequeísmo is a way of intransitivizing an otherwise transitive construal. The Ø/de alternation thus hinges on how the dependence relationship between the reporting and the reported clause is conceived of. Especially crucial is the way of understanding the conceptualizer role of the subject entity, the ‘original’ sayer-cognizer. In both cases, the main, reporting clause is conceptually dependent on the reported clause for its semantic completion, while at the same time it is the profile determinant of the composite construction: e.g., (12) designates the longing of the interlocutor, not the arrival of the holidays. Yet, in the de-construal the conceptualizing subject is not held speech functionally responsible for the reported clause. The link between the current – reporting – and the represented conceptualizing situation (estabas deseando ‘you were longing’) is of a weaker and more inferential nature. The echoed speech or thought is not primarily construed from the intensional perspective of the conceptualizing subject of the reporting clause, but from the spatiotemporal deictic center, i.e., the actual ground. The speaker is selecting the representation for current expressive purposes, thus downgrading the ‘original’ sayer-cognizer’s expressivity. The preposition de introduces a mental space that is conceptually independent from the reporting clause: it signals appropriation by the current speaker.

 Nicole Delbecque

(12) Estabas deseando de que llegaran las ferias para para estar de fiesta, ¿no? (Conversation, Alcalá de Henares, 1992) (dequeísmo) ‘You were longing for the holidays to arrive to make merry, no?’ (13) Porque no nos olvidemos que nosotros somos hijos de tres que siempre yo destaco, los esos tres caraqueños notables: Andrés, Francisco y Simón: Andrés Bello, Francisco de Miranda y Simón Bolívar (Venezuela, Encuesta 157, 1977) (queísmo) ‘Because we should not forget that we are children of three that I always mention, those three remarkable inhabitants of Caracas’ In the transitive construal, on the contrary, e.g. (13), the speech or thought representation is understood with respect to the original sayer-cognizer (which may be the speaker in referential terms). It is presented in terms of reenactment. In this type of speech or thought representation, the conceptualizing subject’s deictic center is fully operative. The illusion of mimesis is maintained, since nothing (Ø) interferes. Without preposition, the complement clause (CL2) is construed with maximum objectivity from the vantage point of the main clause subject (C1). By using the minimal subordination marking – the simple que-complementizer -, the speaker makes the structure analogous to the unmarked accusative with nominal complements (cf. Delbecque 2000). Although the whole sentence of course represents the conceptualization of the speaker, the speaker’s role remains external to the scene, so that the main subject (C1) is perceived as the sole conceptualizer of the content of the complement clause (CL2). In contrast, the oblique representation testifies to the speaker’s intervention: rather than simply echoing or reenacting, (s)he selects a thought or discourse representation. De marks the speaker’s intervention beyond the default conceptualization of the reporting clause (CL1). It is a device for introducing a subjective speaker oriented viewing arrangement in an otherwise objective construal of the complement scene (CL2).2 By gaining independent access to the complement process, the speaker takes on a higher-order role as conceptualizer. This global account of the impact of the Ø/de alternation on the viewing arrangement of conceptualizing subject (CSC) constructions is summarized in Figure 2. Figure 2.  The impact of the Ø/de alternation on the viewing arrangement of the conceptualizing subject constructions Symbols: C0 stands for the speaker as conceptualizer, C1 for the main clause subject, CL2 for the complement clause viewing arrangement Ø (Accusative) de (Oblique)

C1 C0

conception of CL2 solely of C1 not solely of C1

2. It would be a step too far, however, to consider it an instance of subjectification in the sense of an “expression of the speaker’s belief or speaker’s attitude towards what is said” (Traugott 1995: 31). See section 8.

Spanish (de)queísmo 



4. The meaning import of the relational marker de The choice of the relational marker is not arbitrary. The use of the preposition goes back to the Genitive and Ablative cases in Latin. It has an essentially partitive and extractive meaning. It thus conveys a ‘derivative’ conceptualization of the main subject’s participation in the conceptualization of the scene evoked by the complement clause (Figure 3). de Genitive

partitive

Ablative

extractive

derivative

Figure 3.  The schematic meaning of the preposition de ‘of ’

Any preposition instantiates an idealized geometric model (cf. Herskovits 1986). Figure 4 sketches the relational profile brought in by the Spanish preposition de. This representation draws on the relational semantics of the preposition as sketched by Pottier (1962) and on Kemmer and Shyldkrot (1996)’s account of the use of the prepositions à and de in French infinitive clauses. In terms of figure-ground alignment, the verbal predicate functions as trajector or ‘prepositional subject’, and the clausal complement as landmark or ‘prepositional object’. The trajector (figure) is labeled B and the landmark (ground) A so as to conventionally symbolize the conceptual precedence of the reported propositional content on the reporting activity. The vertical line stands for the speech and thought representation predicate, and the circle for the reported content. The dashed line evokes the force dynamics, and the arrow reflects its direction. The landmark is set off against a base that is made up of a whole set of propositional contents that could alternately be envisaged by the speaker, next to or instead of the chosen one, and ascribed to the primary trajector, viz., the conceptualizing subject of the reporting predicate. The elaboration site thus becomes fore-grounded with respect to an implicit background of composite structure, which is best characterized by its open character. The message is that more could be unveiled than the selected portion, which stands out in the scene as being most nearly adjacent to the vantage point of the conceptualizing subject. The preposition signals that the conceptualizer role of the main subject (C1) is a derivative one. This means that the subject entity C1 draws his/her participation from something established independently from him/her. Put otherwise: the conceptualization of CL2 is presented as inherited from another mental space, i.e. outside the realm of C1. This contrasts with the non-derivative profile that is associated with the non-



Nicole Delbecque

prepositional clause complementation, the conceptualization of which is by default taken to originate inside the realm of C1. B tr de

A lm

Figure 4.  The figure (B = trajector tr) / ground (A = landmark lm) alignment expressed by de

In each construal various dimensions regarding the relation between the conceptualizing subject C1 and the content of the complement clause CL2 can become more or less salient according to the context and depending on the semantics of the verb. The saliency effects, listed in Figure 5, can, in turn, induce certain inferences concerning C0’s communicative intentions and the attitude of both C0 and C1. These discourse effects will be taken up in Section 8. Figure 5. Relational profile and saliency effects of de vs. Ø de

Ø

initiation of conceptualization of CL2 lies

outside the realm of C1

inside the realm of C1

relational profile

derivative

non-derivative

potentially salient dimensions

part-whole continuity selective

one-to-one matching punctual emergence global coverage

The derivative / non-derivative contrast applies regardless of the person C1 refers to. With first person, the presence of de enhances the dual role of the speaker: endorsing the conceptualizer role of CL2 is not incompatible with the ability to envisage things transcending the here-and-now of the speech event. In (14), the reason the speaker invokes for rejoicing (de que te vayas ‘(of-)that you leave’) is not one she comes up with while speaking. Compare with (1), where the joy (que te guste ‘that you like it’) emerges from the very interaction. Just like there are at least two ways of rejoicing, there is



Spanish (de)queísmo 

also more than one kind of remembering.3 In (15), the speaker is retrieving from her remembrances of infancy what she considers eligible in answer to the interviewer’s question: in this context it is plainly justified to use me acuerdo de que since the presence of the preposition precisely conveys the image of things being picked up from an external vantage point, out of a stock of available memories.4 In (2), in contrast, using the preposition would be at odds with the absence of external vantage point: here, the speaker finds herself submerged in the atmosphere of the past she is evoking. Not using de is consistent with the impression that she is reenacting, re-experiencing, going through it all again as if her mind were still there. (14) – Vaya por Dios, chica, cuánto lo siento -dijo Isabel-, yo también te había tomado mucho cariño. Pero por ti me alegro de que te vayas. En esta casa no hay quien pare. Ya te lo dije ayer. (C. Martín Gaite, 1976 (novel, Spain)) – ‘Well by God, girl, how much I regret it – said Isabel – I had also got to like you a lot. But for you I’m happy (of) that you’re leaving. In this house no one remains. I already told you yesterday.’ (15) – Yo me acuerdo, pero te estoy hablando de cuando yo era así como tenía como tres o cuatro años porque yo me acuerdo de eso muchísimo. – ¿Y cómo era la vida que llevaban allí en esa casa? – Bueno, en esa casa yo me acuerdo de que mi papá iba a trabajar y mi abuelo también, entonces mis tías y mi mamá se quedaban en la casa haciendo los oficios del hogar (Venezuela, interview 66, 1987) – ‘I remember, but I’m talking to you about when I was about three or four years old, because I remember a lot about that. – And how was the life you were leading there in that house? – Well, in that house I remember (of) that my dad went to work and my grandfather also, so my aunts and my mummy stayed at the house, doing the household’ The functionality of transitively construed reflexive predicates such as alegrar-se que ‘rejoice-oneself that’ and acordar-se que ‘remember-oneself that’ is thus to privilege the inside-view of the conceptualizing subject (C1), in contrast with the oblique construal (de que ‘of that’) that superimposes the outside-view of the speaker-enunciator (C0). In

3. In fact, alegrar is most commonly construed as psych-verb, i.e., with dative and subjectclause (me alegra que... ‘to-me enjoys that…’), which is still another way of conceptualizing the predication relation. 4. The same derivative, partitioning relational profile applies to the verbs that are canonically construed with de que-clause. About twenty enter the series of venir ‘come’, dudar ‘doubt’, hablar ‘talk’, another eighty are reflexive ones that express a mental state, e.g. admiration (admirarse), joy (alegrarse), repent (arrepentirse), amazement (asombrarse), shame (avergonzarse), in reaction to a specific source-event, selected among other possible ones. See Section 6, Figure 7.



Nicole Delbecque

view of the difference in viewing arrangement, examples such as (1) and (2) cease to appear as manifestations of deficient linguistic competence. The same holds in the opposite direction. The usual way of construing communication and propositional attitude predicates is illustrated in (16) and (17). Whatever the type of speaker or situation, the transitive construals yield an unquestioned matching between the scope of the speech or thought representation and the vantage point of C1. The subordinate CL2 exhausts what is relevant to the conceptualizing subject from within its discourse space, i.e. without extra projection from the speaker’s actual deictic center. (16) El señor senador Ulloa me indica que desea abstenerse de votar. (Argentine Senate, 24/02/1999) ‘Mr. Senator U. signals me that he wishes to refrain from voting.’ (17) Pero no me basta porque ya sabemos que estamos por el buen camino. (TVE 1, 05/03/87 (Spanish Television)) ‘But to me it’s not sufficient since we already know that we’re on the right track.’ With these verbs, dequeísmo only occurs very sporadically. Yet, examples (4) and (5) were produced in a rather formal speech situation. Here too, my claim is that the presence of the preposition is not meaningless. In (4) CL1 is construed so as to downgrade the conceptualizer’s initiative: the individual is absorbed in the plurality of ‘us’ (nosotros) and instead of occupying subject position it takes on the dative role, i.e. that of the affected (potentially agentive) participant, in internal focus position (a nosotros ‘to us’) right after the epistemic modal auxiliary (puede ‘may’). As a result, indicar ‘indicate’ depicts a trace keeping attributive relation, not one of effective communication. It can be paraphrased as: ‘we find in this an indication, a symptom of the state of affairs evoked in CL2’. In (5) the receptive attitude holds vis-à-vis parcels of a religious belief system that has come down from ages: ‘to know’ (saber) takes on the sense of ‘having knowledge of ’. At inferential level, the utterance perspires the humility of laying one’s destiny in the hands of a divinity, the plans of which are acknowledged to be beyond grasp. Cognitively speaking, there is nothing more deviant in examples (1) and (2) than in (16) and (17), and nothing more irregular in (4) and (5) than in (14) and (15). The {Ø/de} alternation signals whether the initiation of the conceptualization of CL2 lies inside (Ø) or outside (de) the realm of C1. While without preposition the interclausal relation can be called non-derivative, the preposition makes it derivative.

5. The nominal-verbal head analogy revisited The meaning structure spelled out in Sections 3 and 4 applies to nominal and adjectival heads as well as to verbal ones. The global picture given in Figure 6 integrates the

Spanish (de)queísmo 



minority uses in the overall system of the que and de que complementation. The number of pluses gives an approximate idea of the extent of their usage according to the type of head. The brackets signal the constructions that fall under the traditional definition of (de)queísmo. This integrative representation replaces the partial and prescriptive one given in Figure 1 (Section 2). Whereas in the traditional view, (de)queísmo disturbs an otherwise quasi complementary distribution between verbal and nominal complementation, in the present approach it testifies to the availability of two complementation types for any kind of head: one with an exclusively internal viewing arrangement (que) and one in which the speaker stage-manages the relational profile of the speech or thought representation (de que). Figure 6.  Integrative approach of the que and de que-complementizers head

que-CL

de que-CL

Verb Complex predicate

++++++++++++++++[++] [+++]

++[+] ++++++

Adjective

[++]

++++++++

Noun

[+]

+++++++++++

In terms of usage tendencies, the oblique construal prototypically characterizes nominal and adjectival heads. Yet, the use of the non-oblique que has never been censured as strongly as the oblique construal with usually transitive verbs. Turning to simple que may even be perceived as more appropriate for everyday speech since, stylistically speaking, nouns and adjectives construed with a complement clause rather belong to elaborate registers. The true reason, however, for the general acceptance of queísmo with nominal and adjectival heads is a principled one: its range is, by definition, highly restricted. As predicted by the present account, it requires the head to be compatible with an exclusively internal viewing arrangement. Relatively few nouns and adjectives fulfill this condition. The phenomenon can therefore easily pass unnoticed. The relative unmarkedness of queísmo further contrasts with the striking markedness of dequeísmo, which relies on a morphologically complex form, viz., preposition plus conjunction. Unlike queísmo, dequeísmo can only pop up after a verbal head and yields an oblique construal that fits into the niche of the regular de que-complementation. This niche, however, is doubly marked: quantitatively by being very small, and qualitatively by comprising a significantly high rate of reflexive-marked predicates, oftentimes referred to as “pronominal verbs” in the Romance tradition. This is only part of the story, however. Outside the domain of verb complementation, the reverse image obtains, since with nominal and adjectival predicates the preposition de does not stand for the marked but for the unmarked complementation. The



Nicole Delbecque

oblique construal thus brings the complementation in line with that of nominal and adjectival heads. This paradigmatic entrenchment somehow retrieves the situation and rescues dequeísmo. It persists, no matter how strong the stigmatization, since it also corresponds to a regular pattern of the language and displays a function that is complementary to that of queísmo. The remainder of this section surveys the functionality of the two construals with the minority head types.

5.1

Predicative adjectives

Predicative adjectives can only participate in the Ø/de alternation when they profile a conceptualizer-conceptualization relationship. Otherwise, there is no point in distinguishing between an on-stage and an off-stage viewing arrangement. Therefore, adjectives such as acertado ‘to-the-point’, adecuado ‘adequate’, cierto ‘true’, curioso ‘strange’, dudoso ‘doubtful’, gratificante ‘gratifying’, posible ‘possible’ automatically take a que subject-clause. With mental state adjectives, on the contrary, e.g. cansado ‘tired’, consciente ‘conscious’, contento ‘glad’, harto ‘fed up’, satisfecho ‘happy’, preocupado ‘preoccupied’, triste ‘sad’, the enactment of the stance of C1 can be presented either with or without overt stage-managing by the speaker-conceptualizer C0, e.g. (18) vs. (19). In (18) the speaker suggests that from his/her actual ground (C0) the state of mind (s)he attributes him/herself as experiencer (C1) is to be envisaged in derivative terms. The oblique construal reflects the ability to come up with other reactions than the one acknowledged, and also to develop the same kind of reaction (satisfaction (18a), sadness (18b), confidence (18c)) with respect to other states of affairs. The direct construal, in contrast, does not relate the experiential C1-CL2 relation to the higher-order processing at C0 level. The experience is merely envisaged from within the onstage domain (C1), so that the attitude towards the content of CL2 is taken at face value as the one matching the mental space of C1. Since the Ø-construal is opaque as to the position of C0, it is easily found in the reporting discourse of journalists, e.g. (19), (20b), (21b). (18a) Como presidente que fui de la Convención Constituyente, y teniendo en cuenta que en aquel entonces este tema fue aprobado por unanimidad, me siento muy satisfecho de que hoy estemos tratando este asunto. (Argentine parliament, July 1 1998) ‘As president of the Constitution Convention, and taking in account that at that time this theme had been approved by unanimity, I feel very satisfied (of) that now we are treating this issue.’ (18b) – Perdí muchos amigos y algunos familiares. -Mairena hundió la mirada en la grama-. Y no estoy resentido, sólo un poco triste de que en la repartición de colores a mí me haya tocado éste. (R. Quesada, 2000 (novel, Honduras)) ‘– I lost many friends and some relatives. –Mairena looked away into the greensward-. And I am not resentful, just a little sad (of) that in the partition of colors I got this one.’



Spanish (de)queísmo 

(19a) Callejas dijo estar satisfecho que la Policía Internacional (INTERPOL), luego de realizar una serie de investigaciones, no haya encontrado los pasaportes falsos de los que tanto han hablado muchos de sus opositores políticos para dañar su imagen. (La Tribuna, 02/07/1997 (Honduras)) ‘C. said to be satisfied that Interpol, after realizing a series of investigations, had not found the false passports that were so much talked about by his political opponents to damage his image.’ (19b) San Pablo decía: “¿Quién de vosotros está triste que no esté triste yo?”; es una forma de patentizar la solidaridad humana. (Proceso, 21/07/1996 (Mexico)) ‘Saint Paul said: “Who of you is sad that I am not sad?”; it is a way of demonstrating human solidarity.’ With mental state adjectives the construal alternation thus follows the same rationale as with their verbal counterparts. Leaving aside aspectual differences, contento ‘glad’ (20) and consciente ‘conscious’ (21) present the same alternation as the psych-verb alegrarse ‘rejoice’ (cf. (14), (1)) and the epistemic saber ‘know’ (cf. (5), (17)). In (20a), as in (14), the onstage information has presuppositional status (viz., ‘Ileana is to marry Diego’) and is retrieved by the ‘actual ego’ as a motive (source) of satisfaction for the discourse protagonist, in this case the ‘narrated ego’; mentioning one of the consequences (‘she will get to know the land of her ancestors’) further enhances the derivative nature of the evocation. In (20b), on the contrary, as in (1), things are envisaged from the experiential ground of C1: the Ø construal profiles a maximally cohesive relation, i.e., it does not dissociate the affected participant C1 from the experience evoked by CL2. By the same token, the Ø construal does not overtly ground the relation in the speaker’s deictic center (C0). (20a) Estoy muy contento de que te cases con Diego, Ileana. Y de que conozcas la tierra de tus ancestros. (M. Vargas Llosa, 1993) ‘I am very happy (of) that you marry (with) Diego, Ileana. And (of) that you will know the land of your ancestors.’ (20b) El presidente del Banco Wiese, Guillermo Wiese de Osma, al ser entrevistado en su domicilio, dijo: “Estoy muy contento que hay justicia en el Perú”. (Expreso, 01/08/1987 (Peru)) ‘The president of the Wiese Bank, G.Wiese de Osma, when interviewed at his home, said: “I am very happy that there is justice in Peru”.’ Likewise, with the obliquely construed consciente ‘conscious’, the state of consciousness is acknowledged to fall within the reach of a wider belief system (21a), i.e., one also accessible from the actual mental space of the speaker, cf. the transcending nature of the receptive attitude depicted in (5). The Ø construal, on the contrary, lacks this transparency: the retrieval of what occupies C1’s mind while facing what awaits her,



Nicole Delbecque

passes through identification with her experiential base; this is corroborated by the comment at the end of fragment (21b) (‘that lesson she will never forget’).5 (21a) Salir cuanto antes para que la pequeña no se críe en la cárcel. Es consciente de que los delitos hay que pagarlos, pero prefiere mirar hacia delante. (16/11/96, TVE 1 (Spanish television)) ‘Get out as soon as possible in order for the girl not to be raised in prison. She is aware (of) that the delicts have to be paid for, but she prefers to look ahead.’ (21b) No quiso hablar de la final, porque explicó que primero piensa concentrarse en el partido contra Conchita. Arantxa es consciente que el año pasado, cuando todo el mundo, incluso ella misma, se veía en la final, se le cruzó Mary Joe Fer­ nández. Y esa lección no la olvidará nunca. (La Vanguardia, 01/06/1994) ‘She did not want to talk about the final, because she explained that first she intends to concentrate on the match against Conchita. Arantxa is conscious that last year, when everybody, herself included, saw her in the final, Mary Joe Fernández came across. And that lesson she will never forget.’

5.2

Predicative nouns

Nominal heads are made up of an abstract noun in singular, preceded by a determiner. The construal yields a meta-linguistic discourse move that deserves to be analyzed in its own right (Delbecque: in prep.). May it suffice here to signal that, as predicted by the present account, the oblique construal is the expected pattern, since the noun generally reflects the speaker’s way of categorizing the content of the clausal complement. This is also in line with nominal complementation in general.6 Bridging the equivalence relation between the nominal category and the propositional content by means of de shows that ascription of class membership to the propositional content proceeds from a choice: it is done on the background of other contents also susceptible of fitting the same abstract category. Although this pattern yields the same schematic meaning as the oblique construal with verbal head, it only exceptionally alternates with the bare construal. Nominal heads rarely qualify for construal alternation because they mostly convey an unequivocally C0-based conceptualization, e.g. 5. Unlike the verbal counterpart creer ‘believe’, the adjective creído ‘trustful’ fits the oblique construal: (i) Me acuerdo que Milans, como nieto e hijo de generales, era un chico creído de que podría llegar hasta donde fuera. (El País, 02/08/1986) ‘I remember that M., as grandson and son of generals, was a boy confident (of) that he could get anywhere.’

6. The noun-noun combination, e.g. niño probeta ‘test-tube baby’, is extremely rare in Spanish.



Spanish (de)queísmo 

declaración ‘declaration’, hipótesis ‘hypothesis’, historia ‘history’, percepción ‘perception’, prueba ‘proof ’. So there is no need for restricting the scope to the viewing arrangement of the onstage conceptualizer C1, which is the raison d’être of the Ø-construal.7 (22a) La hipótesis de que nada existiera no encierra, en sí misma, ninguna contra­ dicción. (Filosofía (handbook), Madrid 1998) ‘The hypothesis (of) that nothing would exist does not include, in itself, any contradiction.’ (22b) Delacroix (...) escribió esa famosa declaración de que también era capaz de hacer cosas en arquitectura, escultura y pintura como el mejor. (F. Calvo Serraller, Historia del Arte, 1997) ‘Delacroix (…) wrote that famous declaration (of) that he was also capable of doing things in architecture, sculpture and painting like the best.’ Yet, it is sometimes appropriate to turn to the Ø-construal, especially in reporting contexts. It allows the journalist to echo a speech or thought representation while avoiding to be held speech-functionally responsible for attributing it to the onstage conceptualizer C1. In (23a), e.g., the ‘conviction’ goes on the account of the Israelis, in (23b) on that of the politician who makes the statement. (23a) Los “israelíes tienen la convicción que los iranios están tan desesperados por obtener material militar” que la entrega de armas a Teherán tendrá por resultado “cambios favorables, a largo plazo, en las posiciones del gobierno iranio”, indica el documento. (El Tiempo, 11/01/1987 (Colombia)) ‘The “Israelis have the conviction that the Iranians are so desperate to obtain military material” that the delivery of arms to Teheran will have as result “favorable changes, at long term, in the positions of the Iranian government”, indicates the document.’ (23b) No todos pertenecemos a un mismo partido o apoyamos a un mismo candidato, pero todos somos panameños con el deseo que la Nación sea bien dirigida. (La Estrella de Panamá, 09/11/2003) ‘We do not all belong to the same party nor do we support the same candidate, but we are all Panamanians with the desire that the Nation be well led.’ Generally speaking, the speaker’s face preserving strategy may well be the driving force behind the construal alternation. In (24a), the oblique construal signals that the narrating self (C0) retrospectively widens the scope of the pretexts available to the narrated self (C1), hereby inducing the inference that the choice was an adequate one. In (24b), however, the speaker has no reason to stage-manage a disenchanting excuse, so the speech representation remains confined to the C1-domain (Castellanos). 7. The wide-scope representation introduced by these nouns contrasts with the onstage scope conveyed by the verbal counterparts declarar ‘declare’, percibir ‘perceive’, probar ‘prove’, which systematically enter the Ø-construal.



Nicole Delbecque

(24a) Entonces yo entraba con la excusa de que yo iba a buscar a mi papá, que era funcionario del Concejo Municipal. (informal interview, Venezuela, 1987) ‘Then I entered with the excuse (of) that I was looking for my dad, who was a City Council administrator.’ (24b) Como el único lugar que tienen para “colarse” es la Alcaldía Municipal, el propio “Gordito” Castellanos estuvo escuchando sus peticiones y hoy continuará atendiéndolos. Unos salen alegres y otros con los rostros desencajados, porque después que tanto le metieron el hombro, los ha desinflado con la excusa que no hay cama para tanta gente. (La Tribuna, 08/01/1998 (Honduras)) ‘As the only place they have to “gather” is the City Hall, the proper “fatty” Castellanos was listening to their petitions and will keep attending them today. Some leave good-humored and others with drawn faces, because after having shouldered him so much, he put them off with the excuse that there is no room for so many people.’ In (25a), the control exerted by the narrator (C0) is in contrast with the limited insights of the character within the narrated universe (‘the suspicion does not cross the mind’ of C1). In (25b), the hypothesis is exclusively linked to the domain where it has originated (‘according to the Chamber’). (25a) Si hay algo, tiene que estar en el escritorio: fotos, copias de cartas, recortes de revistas, allí guarda todo lo que podría delatarla. Además, no se le cruza por la cabeza la sospecha de que estén espiándola. Se siente a salvo. (Martínez, 2002 (Argentine, novel)) ‘If there is something, it has to be in the writing-desk: photos, copies of letters, press-cuttings, there she guards anything that could betray her. Moreover, the suspicion (of) that they could be spying on her does not (even) cross her mind. She feels save.’ (25b) Según la Cámara, ha surgido la sospecha que ese dinero, aparentemente depositado en Cuba, se estaría invirtiendo actualmente en la compra de bienes inmuebles en la Argentina. (Diario La Prensa, 06/05/1992 (Argentina)) ‘According to the Chamber, the suspicion has arisen that this money, apparently deposited in Cuba, would be actually invested in the acquisition of real estate in Argentina.’ The construal alternation is not limited to nominals such as convicción ‘conviction’, deseo ‘desire’, excusa ‘excuse’, sospecha ‘suspicion’, tesis ‘thesis’, e.g. (23)-(25), it also occurs with conciencia ’conscience’, miedo ‘fear’, pretexto ‘pretext’, tesis ‘thesis’, and even with the most general label idea ‘idea’ In (26a) the speaker (C0) retrieves a prepackaged view that circulates in the speech community, as a collective property so to say. In (26b) the reporter (C0) refrains from acknowledging access to the idea put forward in the Collection (C1).

Spanish (de)queísmo 



(26a) – Me deprime si no sé qué pasa ahí dentro. – Pues hazte a la idea de que nunca sabrás cómo son. (14/03/89, TVE 1 (Spain)) ‘– It depresses me not to know what happens inside there. – Well get used to the idea (of) that you will never know how they are.’ (26b) Las barreras lingüísticas entre el español, el portugués, el inglés y el francés han caído, de hecho, y la Colección comprueba la idea que nuestra diversidad es una riqueza y no solamente un obstáculo. (ABC, 24/05/1989 (Spain)) ‘The linguistic barriers between Spanish, Portuguese, English and French came down, in fact, and the Collection validates the idea that our diversity is a luxury and not only an obstacle.’

5.3

Complex predicates

The construal alternation with complex predicates is based on the same distinction in viewing arrangement. It does not mechanically follow from formal attraction towards the canonical clause complementation with transitive verbs (cf. Section 2, Figure 1). The nominal predicate indeed combines with ‘light’ verbs such as dar ‘give’, tener ‘have’, tomar ‘take’ that are incompatible with clause complementation when they do not form part of a complex predicate. The most cited case is that of darse cuenta ‘give-oneself account, realize’ (27). The pragmatic notions of “closeness to the speaker” (García 1986) and “direct evidence” (Schwenter 1999) have been adduced to explain the high frequency of queísmo especially with first person subject.8 Under the present account, however, these are but possible inferences, drawn from the tendency to identify the mental space of the onstage self (C1) with the knowledge domain of the speaking ego (C0), taking referential co-extensiveness for granted. In fact, the semantic mechanism is the same as with other epistemic predicates, e.g. saber ‘know’ (17) and consciente ‘aware’ (21b), to which the invoked pragmatic notions hardly apply, even with first person conceptualizer. Insofar as the reflexive dative in darse cuenta enhances the referential identity with C0, it may favor the assimilation of C0‘s viewing perspective to that of C1. Yet, it does not determine it. In (27a), e.g., what enters the conscience of C1 extends to a knowledge domain that is also accessible to C0; the oblique is consistent with the control exerted by C0 on the role played onstage as C1 (“feeling insignificant”). The Ø-construal, in contrast, is opaque in this respect, even with other than first person subjects: in (27b), e.g., only the awareness attributed to C1 comes into the picture, independently of C0, the speaker-conceptualizer at the moment of speech.

8. Cf. figures in Bentivoglio & D’Introno (1977), and Bentivoglio & Galué (1998).



Nicole Delbecque

(27a) También me siento insignificante cuando me doy cuenta de que en este país, lo realmente importante, para toda la sociedad y también para los políticos, es el fútbol. (ABC Cultural, 22/11/1996 (Spain)) ‘I also feel insignificant when I realize (of) that in this country, what really matters, for the whole society and also for the politicians, is football.’ (27b) Smith, que procedía del campo fotográfico, era partidario de los planos de gran tamaño, más tarde conocidos como primeros planos, pero pronto se da cuenta que en ellos no puede mostrar más que una parte de la escena, lo que le lleva a la alternancia de los grandes planos con otros de campo más amplio. (P. Del Rey del Val, 2002 (essay, Spain)) ‘Smith, who came from the domain of photography, was in favor of planes of great shape, later known as first planes, but he soon realizes that in them he can show only one part of the scene, what leads him to the alternation between the big planes with others of wider scope.’ Examples such as (27) suggest that the predominance of the Ø-construal has more to do with C1’s prominence and agentivity than with the person as such. The Ø-construal is indeed also typical of epistemic predicates that are not reflexive but point to an active cognizer, viz., tener/tomar en cuenta ‘have/take into account’. The voluntary intake depicted in (28a) yields an unquestioned matching between the scope of the thought representation and the vantage point of C1. In contrast, with the experiential caer en (la) cuenta ‘fall into (the) account’, the subject rather appears as the affected participant. Hence, the oblique construal prevails, regardless of the person of the subject. The context of (28b) explicitly refers to the doubling of the ego (“I observe myself ”, “have operated in me”). (28a) Al recordar sus años de exilio tienen en cuenta que sus hijos han pasado parte de la infancia en Galicia o en Cataluña y que España ha dejado una profunda huella en su corazón. (web 1999, Spain) ‘When remembering their years of exile they take into account that their children have lived part of their infancy in Galicia or in Cataluña and that Spain has left a profound trace in their heart.’ (28b) En cuanto me observo un poco -hay que partir del cuerpo, más que del alma, para reflexionar, aconsejaba Nietzsche-, caigo en la cuenta de que unos albañiles, fresadores, mecánicos, pintores, electricistas y carpinteros han trabajado en mí. (F. Umbral, 1975 (novel, Spain)) ‘When I observe myself a little -one has to start from the body, rather than from the soul, to reflect, recommended Nietzsche-, I realize (of) that bricklayers, milling machine operators, painters, electricians and carpenters have operated in me.’ In (29a), the oblique construal patterns with the same kind of self-doubling in CL2 (“I am born again”). Compare with the Ø-construal in (29b): here, CL2 exhausts what is



Spanish (de)queísmo 

relevant to the C1, from within its discourse space, i.e. the actor as conceptualizing subject within the narrated scene, without extra projection from his actual deictic center as C0. (29a) Yo no he percibido nunca eso. Que abran los brazos y nada más. No lo sé, pero a mí me da la sensación de que estoy volviendo a nacer.(Madrid, 28/01/93, TVE 1) ‘I have never perceived this. That they open the arms and nothing more. I don’t know, but to me it-gives the sensation (of) that I am born again.’ (29b) (...) me cortaron el pelo a tirones porque no sabían, y yo tenía el pelo muy muy lleno de grasa del teatro. A ver quién le da la primera patada a este maricón. Yo mismo. Me golpeaban con los puños de las pistolas, tan fuerte que a mí me daba la sensación que me pegaban tiros. (Madrid, 06/03/93, TVE 1) ‘(…) they cut my hair snatching because they didn’t know, and I had my hair very very full of grease from the theater. Let’s see who gives the first kick to this queer. Me precisely. They beat me with the handles of the pistols, so hard that it gave me the sensation that they were shooting me.’

6. Construal alternation with verbal head: degrees of markedness Since the verbal heads form the vast majority, irrespective of the type of speech, and the Ø-construal constitutes the predominant pattern, it easily passes unnoticed whether it is an instantiation of the regular pattern or represents the minority use with a given verb. For a few entries, e.g., avisar ‘advise’, dudar ‘doubt’, informar ‘inform’, sospechar ‘suspect’, most dictionaries register both complementation types. Example (30) illustrates the effect of switching, at a few lines interval, from the internal to the external, derivational, viewing arrangement with informar. (30a) El ministro informó que entre los dos ministerios que él dirige se han gastado algo menos de 700 millones de pesetas de la partida presupuestaria de los fondos reservados en el último año. (El Mundo, 01/06/1995) ‘The minister informed that among the two ministries that he leads they spent a bit less than 700 million pesetas of special budget of the reserved funds in the last year.’ (30b) Según las distintas fuentes consultadas por este diario, Belloch también informó de que en ambos ministerios se han tomado las medidas oportunas para establecer un cierto control sobre estos gastos. (El Mundo, 01/06/1995) (from the same article as (30a)) ‘According to the different sources consulted by this newspaper, Belloch also specified (of) that in both ministries the appropriate measures have been taken to establish a certain control on these expenses.’



Nicole Delbecque

While (30a) represents simple reenactment of the speech of C1 (‘the minister’), (30b) views it from the here-and-now of the journalist, who has been gathering information from different sources (según las distintas fuentes consultadas). By using de he signals that he is picking out a relevant point, suggesting by the same token that it is only part of the story. Can the (30b) construal then be termed more “marked” than the other? Yes, if by markedness is meant that (i) the relation to the speaker’s ground makes for a supplementary meaning layer, or (ii) the oblique construal is far less likely to occur although the number of potential candidates (transitive verbs) is considerably higher than for queísmo. No, if by “marked” were meant “deviant”. There is nothing more deviant in using de with informar (30b) than in the Ø-construal with darse cuenta ‘become aware’ (27b), for instance. Yet, as reported in Gómez Torrego (1999: 2118), informar is one of the verbs some authors associate with dequeísmo and others with queísmo. Although the discussion becomes pointless under the present account, it is revealing for the existence of diverging preferences in the speech community. It can also be assumed to reflect a far greater deal of variation within the speech of particular individuals than has generally been taken note of. The traditional premise that the complement choice is fully governed by the verb rests on a categorical conception of verbal meaning. So, even if the constructional variation is recognized in terms of valence, it is usually not considered to affect the verbal meaning proper. Classically recognized alternations are those between unitransitivity and ditransitivity, or between animate and inanimate direct object, e.g., informar a alguien de algo ‘inform (to) someone of something’ vs. informar algo a alguien ‘inform something to somebody’ (Cano Aguilar 1981: 340). However, for some verbal lexemes that are alternately construed transitively and intransitively with de-oblique (not necessarily a clause), it is common practice to provide two separate, so called “homonymic” entries, e.g. responder ‘respond’/ responder de ‘assume responsibility for’; maldecir ‘detest’/ maldecir de ‘slander’, presumir ‘presume’/ presumir de ‘take pride in’. Compare the use of responder in (31a) vs. (31b).9 (31a) A lo que le respondieron todos: “¡Los cargos, don Natalio, los cargos!” (ABC, 11/10/1982 (Spain)) ‘To which all answered him: “The charges, don Natalio, the charges!”’ (31b) Manuel Becerra dijo que “los carniceros respondemos de la carne” y explicó que trescientos profesionales reunidos el sábado constataron la recuperación de ventas al 80%. (La Voz de Galicia, 18/12/2000) ‘M.B. said that “the butchers we stand surety for (lit.: of) the meat” and he explained that three hundred professionals gathered on Saturday noticed the recovery of the sales to 80%.’

9. For an in-depth analysis of responder, see Delbecque (2004).



Spanish (de)queísmo 

My claim is that this split is unwarranted. These verbs in fact exploit the same schematic distinction between a globalizing vs. partitioning view, be it in a perceptually more salient fashion. On the basis of the organization of the verbal lexicon as such, no principled dividing line can be drawn, since the Ø/de alternation is a paradigmatic operator that doubles the semantic potential of the lexicon and of the constructions. It is a priori available for any kind of speech or thought representation, provided the verb’s meaning structure fits the corresponding relational profiles. This is not to say, of course, that the two construals will be equally attested in usage. The availability of two constructional schemas – each yielding a proper meaning structure – is not to be confused with the relative ease and frequency with which these schemas apply with individual verbs. The vast majority, over five hundred fifty, take a Direct Object clause (Levy Podolsky (1983), Subirats-Rüggeberg (1987), Delbecque and Lamiroy (1999)). The number of verbs that take a de que-clause is nearly five times lower. Adding those construed with que-subject clause, another three hundred, makes the disproportion even more striking. The significant difference at the level of the available lexemes (types) is further strengthened at the level of the frequency of use (tokens). While quite a few transitive verbs construed with que-clause rank high on the frequency lists, the intransitive ones construed with de que-clause occur only marginally, and are more likely to be found in written language, especially in fiction prose (cf. Juilland 1964). The contrast thus operates both at type and at token level. Likewise, all counts confirm that queísmo is quite common, dequeísmo rather marginal. The preference for the Ø-construal thus seems to hold at all levels. The frequency differences are statistically so robust that they point to an underlying cognitive-functional regularity that can be formulated as follows: the minimal separation between head and complement iconically reflects that CL2 remains within the scope of the conceptualizing subject C1. As transitivizing device, queísmo brings the relational profile in conformity with the majority use. Put otherwise: it yields a transposition from a marked to an unmarked paradigm. The constancy and relative success of the oblique construal, on the other hand, is not unmotivated, either. It is revealing for the possibility to transcend the C1 initiated conceptualization, inserting it in an ampler epistemic base, beyond the realm of C1’s conscience, memory storage, creativity, perception. The alternative thus also obeys a proper language logic, the consistency of which has to do with overall functionality, economy and transparency. The balancing of markedness strategies is an element hereof. Prior studies provide interesting information on the relative frequency of the complementation types per verb.10 The figures should be handled with care, however, since most of them are drawn from too small an empirical basis. Systematic searches in the on-line thesaurus of the Spanish Academy (CREA) contribute to enhance the reliability 10. Cf. Bentivoglio & D’Introno (1977), Boretti de Macchia (1989), DeMello (1995), Bentivoglio & Galué (1998), Schwenter (1999), del Moral (2004).



Nicole Delbecque

of the estimations and permit tentative extrapolations, disregarding possible peaks in one direction or the other due to idiolectal, dialectal, sociolectal or stylistic factors. Markedness in probabilistic terms can be measured by ordering the data of the skewed distribution along a cline going from the (nearly) categorical use of one construal to that of the other. It may presently suffice to distinguish a gradient of six separate clusters. Figure 7 conjointly reflects decrease of likelihood of occurrence of the Ø-construal and increase of the oblique one. A closer look at the groupings suggests that frequency of use operates as a driving force behind prescriptive and proscriptive statements. Using the preposition is doomed “dequeísta” for rank 1 and 2-verbs (32), and leaving it out is doomed “queísta” with rank 5 and 6-verbs (33).11 It is no coincidence that the norm most strongly rejects the alternation for verbs that belong to the ranks 2 and 5. The verbs of the outer classes escape attention because they hardly display any variation. Rank 6 verbs typically convey partial or derivative participation in a reactive process. (32) Tengo entendido de que es usted el autor del libro titulado, ‘Cuidado con las señoras’. (Tele 5, 21/10/96 (Spain)) ‘I have the understanding (of) that you are the author of the book titled “Watch our for the ladies”.’ (33a) Nadie disfruta que se le diga que los consumidores preferirían que él o ella se retirara de los negocios o encontrara un trabajo diferente. (La Hora, 08/04/1997, Guatemala) ‘Nobody likes being told that the consumers prefer him or her to leave business or find another job.’ (33b) En una reunión que hubo se veía muy serio y después me enteré que ese día estaba enfermo. (Vea on line, 11/05/2003, Puerto Rico) ‘In a reunion that took place, he seemed very serious and later on, I found out that that day, he was ill.’ 11. Compare with the so-called canonical obliques: (i) O, por lo menos, quería vivir más que lo que quería morirse. Porque a pesar de todo lo dicho, este hombre disfrutaba viviendo, y quienes lo rodeaban, él así lo creía, disfrutaban de que existiera. Pero atención: aquel goce mutuo debió mantenerse siempre “a distancia”. (J. Bucay, short story, 2002, Argentine) ‘At least, he wanted to live more than that he wanted to die. Because in spite of everything that has been said, this man enjoyed living and those who surrounded him, he believed so, enjoyed (of) that he existed. But pay attention, this mutual enjoyment always had to keep itself at a distance’.

(ii) “Yo venía de hacer un documental en San Nicolás sobre Somisa (Sociedad Mixta Siderurgia Argentina), en 1992, y ahí me enteré de que existía Sierra Grande. (El Clarín, 28/02/1997 (Argentine newspaper)) ‘I had just finished a documentary in San Nicolás about Somisa, in 1992, and there I found out about the existence of Sierra Grande (lit.: of that S. G. existed).’

Spanish (de)queísmo 



Figure 7.  Alternation cline: from (quasi) categorical Ø to (quasi) categorical de 1. Ø

>500 verbs = ± 70% Ø: ±100 %

adivinar ‘guess’, admirar ‘admire’, adorar ‘adore’, asumir ‘assume’, constatar ‘observe’, considerar ‘consider’, descartar ‘dismiss’, enunciar ‘enunciate’, ignorar ‘ignore’, imaginar(se) ‘imagine’, querer ‘want’, reconocer ‘recognize’ (± 300 unitransitives); aceptar ‘accept’, aconsejar ‘advice’, anunciar ‘announce’, confirmar ‘confirm’, explicar ‘explain’, mencionar ‘mention’, proponer ‘propose’, repetir ‘repeat’, replicar ‘reply’ (± 200 ditransitives)

2. Ø

± 50 verbs = ± 7% Ø: 99 %

afirmar ‘claim’, agradecer ‘thank’, comentar ‘comment’, contar ‘tell’, creer ‘believe’, decir ‘say’, entender ‘understand’, oír ‘hear’, olvidar ‘forget’, opinar ‘hold an opinion’, pedir ‘ask’, pensar ‘think’, recordar ‘recall’, saber ‘know’, sentir ‘feel’, suponer ‘suppose’

3. Ø

± 10 verbs >60% = Ø

cuidar ‘care’, informar ‘inform’, maldecir ‘curse’, necesitar ‘necessitate’, presumir ‘presume’, responder ‘respond’, sospechar ‘suspect’

4. de

± 10 verbs de: >60%

acordarse ‘remember’, advertir ‘notify’, avisar ‘advise’, dudar ‘doubt’, precisar ‘need’, quejarse ‘complain’, sufrir ‘suffer’

5. de

± 20 verbs= ±3% de: >95%

alegrarse ‘rejoice’, asegurarse ‘make sure’, asombrarse ‘wonder’, congratularse ‘congratulate o.s.’, disfrutar ‘enjoy’, enterarse ‘learn’, hablar ‘talk’, jactarse ‘boast’, lamentarse ‘regret’, olvidarse ‘forget’

6. de

±100 verbs = ±15%

admirarse ‘admire’, arrepentirse ‘repent’, avergonzarse ‘shame o.s.’, beneficiarse ‘benefit’, burlarse ‘laugh at’, deleitarse ‘be delighted about’, dolerse ‘mourn over’, percatarse ‘notice’, (pre)ocuparse ‘take care’, vengarse ‘revenge o.s.’(± 80 reflexives) gustar ‘get a take to’, partir ‘start’, provenir ‘proceed’, tratar ‘try’, venir ‘come’ (± 20 non refl.)

de: ±100%

In the central categories, the variation is so omnipresent that it transcends the (de) queísmo anathema. At rank 3 usage oscillates at the advantage of the Ø-construal ((30), (31), (34)), at rank 4 at that of the oblique ((2)-(15), (35), (36)). Since with these verbs the two construals are more or less in balance, neither one can reasonably be considered “deviant”. Hence, both tend to count as canonical realizations, i.e. they are generally considered to be part of the individual verb’s valence, in contrast with the minority uses registered at the other ranks. (34a) “Oficiamos de auditor, cuidando que se respeten los derechos humanos de aque­ llos que no se encuentran en condiciones de hacerlos valer por sí mismos”, concluyó Sigón. (La Nueva Provincia, 15/03/1997 (Argentine)) ‘“We work as auditor, checking (‘making sure’) that the human rights are respected, of those that are not in the condition of asserting them by themselves”, concluded Sigón.’



Nicole Delbecque

(34b) Informar con exactitud y con verdad; no omitir nada de lo que el público tenga derecho a conocer; (...); cuidar de que en las informaciones no se deslice la intención personal del que la redacta (La Prensa, 21/05/1992, (Argentine newspaper; quoting the ethical code for journalists)) ‘Inform with exactitude and with truth; not omit anything that the public has the right to know; (…); make a point of not slipping in the information the personal intention of the one who writes it up’ (35a) Carlos Cariola les mostró el camino. No dudo que, inspirados en su fecunda vida y obra, todo lo pueden conseguir. He dicho. (Chilian parliament, Session 21, 1995) ‘Carlos Cariola showed them the way. I don’t doubt that, inspired by his fruitful life and work, you are able to reach it all (“all it you (guys) can obtain”). I have spoken.’ (35b) Yo desconfío de la implementación, en el sentido de que es una cosa muy rápida, y sobre todo dudo de que el maestro pueda captar así, el maestro paraguayo, así, en general, pueda captar fácilmente los fundamentos teóricos que apoyan a la reforma educativa. (Paraguay, Encuesta 103, 1993) ‘I distrust the implementation, in the sense that it is a quick thing, and above all, I doubt (of) that the teacher can understand so, the Paraguayan teacher in general, can easily understand the theoretical foundations that support the educational reform. (36a) Si me lo permite vuestra majestad, quisiera ser mi humilde persona la primera en intervenir para manifestar (se acalora) que no puedo entender, mi señor, cómo haya hijos de mala hembra que deseen veros derrocado, mi rey, con lo que vos habéis hecho y aún hacéis de bien (...) (A punto de llorar.) Yo, majestad, no sufro que haya ingratos que tan mal se porten con su rey, con el único que ha sabido alejar a esta nación de la secular pobreza y de la desigualdad entre los suyos (J. Márquez, 1990 (play, Spain)) ‘If your majesty allows me, my humble self would like to be the first in intervening to say (gets excited) that I can’t understand, milord, how there are children of a bad woman that wish to see you deposed, my king, since you have done and still do good (…) (About to cry). I, your majesty, do not tolerate (lit.: suffer) that there are ungrateful people that behave so badly towards their king, the only one who has been able to move this nation away from the secular poverty and the inequality amongst his people. (36b) Y yo sufrí de que una mujer así no fuera la mía, y yo tuviera que volver a casa y cargar contigo, que eres un retal. (M.Hidalgo, 1998) ‘And I suffered (of) that such a woman wasn’t mine, and I had to go home and put up with you, piece of dirt.’ (novel, Spain)

Spanish (de)queísmo 



While the Ø-score amounts to some five per cent with rank 5 verbs, the chance for the oblique to occur with rank 2 verbs is less than one per thousand. The difference in likelihood is probably revealing for a difference in cognitive complexity. However, frequency does not measure the nature of things: it is functionally as sound to turn to the oblique with highly frequent verbs of communication (decir ‘say’) and cognition (creer ‘believe’, pensar ‘think’) as it is with verbs of the next ranks.12 Even if the wide range in degree of variation admittedly remains a puzzle at the level of the individual verbs, the fact that most of them admit the alternation testifies to the interaction between constructional meaning and lexical semantics.

7. Translation: indirect evidence for a difference in relational profile Turning to translation can be considered a valuable heuristics in gathering external evidence for the existence of alternating construals. In languages that lack this kind of alternation, the difference in meaning pattern has to be expressed by other means. It may, for instance, show up in the use of distinct verbal predicates. If we look at English, quite a few Spanish verbs receive a different translation in function of the construal used. The following list is illustrative of this semantic split. afirmar agradecer acordarse alegrarse comentar comprender cuidar darse cuenta enterarse decir disfrutar dudar entender esperar gozar hablar informar

Ø que – CL2 claim thank recall be glad comment understand watch, see, make realize find out say enjoy doubt understand hope enjoy envisage inform, let know

de que – CL2 be positive about be grateful have recollection of rejoice at comment upon have some understanding of take care of, look after become aware receive the information tell make the most of have doubts about have some understanding of have the expectation take advantage of talk about inform about, give notice of

12. According to del Moral (2004: 201), over the last five hundred years dequeísmo has been increasing with communication and cognition verbs, while it has been decreasing with emotion verbs (ranks 5 & 6).



Nicole Delbecque



lamentarse leer maldecir necesitar oír olvidarse percatarse preocuparse presumir quejarse reconocer recordar saber sospechar sufrir

regret read detest need, necessitate hear forget realize worry presume complain, wail recognize recall know suspect suffer, endure

mourn over do some reading on slander be in need of, lack catch sth on lose sight of become aware, notice take care take pride in complain about acknowledge have recollection of have (some) knowledge of be suspicious of suffer from

Many more verbs have a meaning potential that allows them to profile either type of relationship. Since this kind of polysemy can also apply to the target language, the difference does not necessarily show up in the translation. As a consequence, the relational profile is not always readily detectable. It is my contention, however, that its veiled nature does not hamper its effectiveness at the level of grammatical organization. On the contrary, it makes the Ø/de variation all the more functional, since it overtly codes a category that can remain covert in other languages.

8. Possible inferences At discourse level, the viewing arrangement can orient towards certain inferences regarding C0’s communicative intentions and the attitude of both C0 and C1 towards the propositional content of CL2. Grossly speaking, the Ø-construal suggests that all the relevant ‘information’ comes into the picture, untouched, reproducing the authentic form it originated in. As the oblique construal, on the other hand, stands for selective speech or thought representation, it points to C0’s ongoing processing activity, thus drawing attention to the speaker’s attitude and intentions. By inference, the oblique can then be considered to suggest that the representation is susceptible of departing from genuine reenactment or reproduction in a mitigating or derogating sense. The propositional attitude conveyed depends on the verbal head (V1). Communication predicates naturally relate to illocution, and epistemic predicates bring in an evidential dimension. With first person subject (C1), the head is by default taken to mark off the speaker’s stance (C0). While the Ø-construal unquestionably instantiates a one-toone matching between V1 and CL2, the oblique recovers a reasoned association, thus yielding a less absolute, looser, and perhaps questionable identification of C1 with C0.

Spanish (de)queísmo 



The proposal to analyze iconic proximity (Ø) vs. distance (de) between head and complement in terms of speaker attitude was made by Bentivoglio and D’Introno (1977). It gave rise to the reformulations summarized in Figure 8. Beyond terminological differences, the basic postulate remains the same: Ø is the pragmatically strong variant, de the weak one. Though suggestive, the assumptions evoked by these labels remain quite elusive. They rest upon conjectures that overestimate the communicative impact of (de)queísmo. In drawing inferences regarding the stance of the speaker (C0), the complex interplay between three components is to be reckoned with, viz. C1, V1 and CL2. High/low speaker “commitment” can relate to (i) the appropriateness of V1 to qualify C1’s speech or thought representation, (ii) the conformity of V1 with the way C1 would qualify the represented speech or thought, (iii) the exactness, reliability, truthfulness of the representational content (CL2), (iv) the adoption of the C1 stance by C0, and (v) actualization of the CL2 content for current communicative purposes. Figure 8.  Functional-pragmatic accounts of (de)queísmo Ø

de

Bentivoglio and D’Introno (1977: 73)

iconic proximity

iconic distance

strong assertion / presuppostional content

attenuation

García (1986: 52)

endorsement/certainty relevant / truthful

distance irrelevant / hypothetic

Schwenter (1999: 75 )

direct evidence truth commitment / certainty

indirect evidence hearsay / inference

del Moral (2004: 315)

objective high speaker commitment

subjective low commitment

If any, it is perhaps dimension (v) the oblique most easily departs from, by virtue of the derivative connection it introduces (cf. Figure 3). With first person subject, the iconic distance can make the (re)enactment appear as dissociated from the speaker’s actual ground and, hence, be perceived as conveying ‘lack of commitment’. In the Ø-construal, in contrast, the dual role of the speaker goes unnoticed: since C1’s vantage point is by default assimilated to that of C0, the speaker is perceived as committed to CL2.13 Some of the most frequent transitive verbs, e.g. creer ‘believe’, decir ‘say’, pensar ‘think’, saber ‘know’, suponer ‘suppose’ (Figure 7, rank 2) present parenthetical and semi-auxiliary uses that display their ability to express evidential and modal epistemic values. When functioning as V1 they may therefore also be assumed to have a balancing 13. The C1 / C0 duality becomes more readily apparent when V1 does not come in the present tense.



Nicole Delbecque

function: in (37), e.g., pienso ‘I-think’ modulates affirmative answers. It is a commonly used hedge in rather formal interviews. (37a) – Yo la [= tu mamá] he visto muy orgullosa [...] además ya le darás otras satisfacciones; ahora estás en el camino de darle satisfacciones. – Sí, pienso de que sí, ¿no?, no sé. (Buenos Aires 77/11: 1–9) ‘– I’ve seen her [= your mom] very proud [...] moreover you’ll give her other satisfactions; how you’re on the way of giving her satisfactions. – Yes, I think (of that) yes, no?, I don’t know.’ (37b) – Y vos, ¿qué decís de vos mismo? vos ¿sos un buen amigo? – Sí, yo pienso que sí, este, en, la, la amistad es una de las cosas que más o menos ando bien (Buenos Aires 77/13: 26–34) ‘– And you, what do you say about yourself? are you a good friend? – Yes, I think (that) yes, eh, in, the, (the) friendship is one of the things that more or less I am good in.’ If (de)queísmo would induce a “doubt” (37a) vs. “certainty” (37b) reading, as claimed by García (1986: 57), inverting the type of complementation could be expected to induce the opposite outcome. Yet, the interpretation remains unaltered. This suggests that it does not hinge on the construal alternation but is licensed by independent contextual clues, viz., the double hesitation tag (“isn’t it?, I don’t know”) in (37a), and personal pronoun expression (“yo”)14 and subsequent self-appraisal in (37b). In answers, the choice follows the rationale of the question, it does not appear out of the blue. In (37b), only the interviewee’s experiential basis is put into play; in (37a), however, the question involves a third party: the evaluation does not only depend on her own standards but also on her mother’s. The oblique reflects stage monitoring: the higher processing complexity is motivated by the acknowledgment of not being in a position to fully ascertain the course of things. Yet, (partial) lack of control is not to be equated with lack of self-confidence. On the contrary, the construal counts as a face preserving strategy: it makes for a safe discourse move. The notions of evidential and epistemic modality cannot fully capture the discourse effects triggered by the construal alternation, since the way a piece of information is accessed does not necessarily shed light on the degree of certainty regarding its validity or truthfulness. In (38a), e.g., the oblique does not merely echo the parenthetical forms pienso yo ‘I think’ and yo creo ‘I believe’, it makes the thus introduced view transcend the individual approach. The Ø-construal, in contrast, yields an immanent representation, without additional conceptual layer: in (38b), the forecast fully emerges from the momentary situation the TV-presenter is in.

14. Since person marking is morphologically incorporated in the verbal ending, subject expression is discursively marked (cf. Delbecque 1992). Using the personal pronoun subject remains the minority option even in regions where it seems to be less marginal than in peninsular Spanish.



Spanish (de)queísmo 

(38a) De ahí de ahí, pienso yo, que los así como la poesía es un arte intuitivo, y las mejores poesías yo creo que se escriben en la juventud, a mí me da la impresión de que la novela es un arte más deductivo y que las mejores novelas se escriben pasada la madurez. (Madrid, 02/11/96, Cadena SER) ‘Therefore therefore, I think, just like poetry is an intuitive art, and the best poems I believe that they are written in youth, to me [it] gives me the impression (of) that the novel is a more deductive art and that the best novels are written after maturity.’ (38b) Bueno, es que yo creo que hablar de triunfo del pe-pe me parece una hipótesis en estos momentos de ciencia ficción. Me da la impresión que a lo largo de esta lucha electoral los resultados van a ir variando de hora en hora, al alza o a la baja, hay que esperar a los resultados oficiales. (Madrid, 26/05/91, TVE) ‘Well, it is that I believe that talk of triumph of the P.P. appears to me as a hypothesis in these moments of science fiction. [It] gives me the impression that all along this electoral battle the results will be varying from hour in hour, upwards or downwards, we have to wait for the official results.’ The predicates beyond rank 2 (Figure 7) do not fit the inferential values listed in Figure 8. The construal alternation does not affect the hearsay dimension inherent to the meaning structure of, e.g., enterarse ‘learn, be told, get to know’, a rank 5 predicate. However, since the choice reflects opacity (Ø) vs. transparency (de) as to C0’s position, it indirectly enhances vs. downgrades C1’s protagonist status: in (39a), the viewing is done from within the discourse space of C1, the narrated self (yo, who pulled down “that peg”); in (39b), the evocation is filtered by the narrating ego (C0), who abstracts away from concrete anchoring (alguien ‘somebody’, en alguna parte ‘somewhere’). At inferential level, the CL2 content of (39a) appears as primarily relevant within the scope of the narrated universe, while that of (39b) extends to the speaker’s current discourse space. (39a) yo ni me enteré que me había llevado el cono ese... (Bentivoglio and Galué 1998: 144) ‘I wasn’t even aware that I had taken that peg along with me’ (39b) me enteraba de que alguien cantaba en alguna parte y volaba... (Bentivoglio and Galué 1998: 145) ‘I heard (of) that somebody sang somewhere and flew…’ In terms of internal cohesion, the Ø-construal ties CL2 to V1, thus setting off C1. The oblique, on the other hand, strengthens the bond between C1 and V1, at the same time inducing a derivative relation with CL2. Figure 9 shows how the image schematic difference pervades various levels of conceptualization. While the Ø-construal appears as a concrete snapshot, centered on C1, the oblique rather foregrounds V1, thus depicting an attributive relation. In line with the immanent vs. transcendent viewing arrangement (Cf. Section 3, Figure 2), the representation is either perceived as mimetic reporting by a genuine recorder-displayer (Ø) or as a selective arrangement put in a



Nicole Delbecque

reasoned perspective (de) by the speech-responsible stage-managing speaker-enunciator C0. While the Ø-construal instantiates a seemingly unimpaired content, as an on the spot picture, the oblique adjusts the representation to actual communicative purposes and face needs. Figure 9.  Deconstruction of the meaning layers involved in the two construals

foregrounding image schema viewing arrangement representation speaker (C0) access speaker role CL2 content inference

Ø

de

[C1] – [V1 que CL2] conceptualizer C1 concrete snapshot immanent mimetic opaque record & display unimpaired on the spot picture

[C1-V1] de [que CL2] conceptualization V1 attributive relation transcendent stage-managed transparent select & arrange in perspective adjusted retrieval

Further inferences regarding modal values, e.g. reliability and commitment, hinge on the meaning of the head predicate, and can somewhat vary according to the context and the receiver’s background knowledge and expectations.

9. Conclusion The Ø/de alternation is not a matter of free variation. It yields a semantic conditioning at the level of the head-complement relation: it depicts a part vs. whole relation that functions as a device for modulating the relational profile of the complex clause in terms of viewing arrangement, and thus reflects one’s choice to remain backstage as enunciator (Ø) or not (de). In the Ø construal, the speech and thought representation is (re)enacted solely from the perspective of the conceptualizing subject (C1). Hence the impression of facing the full picture of what is relevant for C1 at the moment of conceptualization. In the oblique construal, on the contrary, the speaker (C0) intervenes from the here-andnow: the preposition signals the derivative nature of the speech and thought representation that is put onstage. This yields as default reading that the representation is selective, that only part of the story comes into picture, viz., what the speaker-enunciator (C0) considers relevant. The persistence of the alternation stems from its overall functionality. The analysis in terms of covert vs. overt stage-management by the speaker testifies to the conceptual latitude of the head, and holds independently of the head’s category. While the

Spanish (de)queísmo 



Ø-construal is the marked option with nominal and adjectival head, it is the unmarked one with verbal head. With the latter, informal usage extends the alternation far beyond the lexicalized cases. Construing a verbal predicate with de-oblique manifests its ability to integrate a speaker-grounded viewing arrangement. Rather than turning to less frequent, high register verbs, the speaker obtains a similar effect simply by varying the construal of frequently used ones. Discursive and argumentative values operate at inferential level and do not as such belong to the meaning structure of the complementation type. There is no one-to-one relation between construal choice and illocutionary strength. Neither construal stringently orients toward a particular evidential or epistemic modal reading; the latter first and foremost rely on verbal tense, aspect and person, as well as on the verb’s lexical semantics, of course. The felicitousness of the construal choice is a discourse functional matter related to register, genre requirements and face preserving strategies.

References Achard, M. 1998. Representation of Cognitive Structure. Syntax and Semantics of French Sentential Complements. Berlin/New York: Mouton de Gruyter. Arjona, M. 1978. Anomalías en el uso de la preposición de en el español de México. Anuario de Letras 16: 67–90. Bentivoglio, P. 1976. Queísmo y dequeísmo en el habla culta de Caracas. In Colloquium on Hispanic linguistics, F. Aid, M. Resnick and B. Saciuk (eds), 1–18. Washington: Georgetown University Press. Bentivoglio, P. 1980. El dequeísmo en Venezuela: ¿Un caso de ultracorrección? Boletín de Filología de la Universidad de Santiago de Chile 31: 705–719. Bentivoglio, P. and D’Introno, F. 1977. Análisis sociolingüístico del dequeísmo en el habla culta de Caracas. Boletín de la Academia Puertorriqueña de la Lengua Española 6: 58–82. Bentivoglio, P. and Galué, D. 1998. Ausencia y presencia de la preposición de ante cláusulas encabezadas por que en el español de Caracas: Un análisis variacionista. Boletín de Filología de la Universidad de Santiago de Chile 37: 139–59. Boretti de Macchia, S. 1989. (De)queísmo en el habla culta de Rosario. Anuario de Lingüística hispánica 5: 27–49. Cano Aguilar, R. 1981. Estructuras sintácticas transitivas en el español actual. Madrid: Gredos. Cano Aguilar, R. 1985. El régimen de las oraciones completivas en el español clásico. In Philologica hispaniensia II in honorem Manuel Alvar, 81–93. Madrid: Gredos. Delbecque, N. 1992. Por qué y cómo integrar la variación en la descripción gramatical. Lingüística Española Actual 14: 5–68. Delbecque, N. 2002. A construction grammar approach to transitivity in Spanish. In The Nominative/Accusative. Case and Grammatical Relations across Language Boundaries, K. Davidse and B. Lamiroy (eds), 81–130. Amsterdam: J. Benjamins. Delbecque, N. in prep. Usage and function of Spanish propositional attitude nominals.



Nicole Delbecque Delbecque, N. and Lamiroy, B. 1999. La subordinación sustantiva: Las subordinadas enunciativas en los complementos verbales. In Gramática Descriptiva de la Lengua Española, I. Bosque and V. Demonte (eds), Chapter 32: 1965–2081. Madrid: Espasa 1999. del Moral, G. 2004. Grammaticalization of Spanish de: reanalysis of Spanish (de)queísmo in Southern Cone Dialects. Ph. D. diss., Department of Linguistics, Illinois University. DeMello, G. 1995. El dequeísmo en el español hablado contemporáneo: un caso de independencia semántica. Hispanic Linguistics 6/7: 117–152. García, E. 1986. El fenómeno (de)queísmo desde una perspectiva dinámica del uso comunicativo de la lengua. In Actas del II Congreso Internacional sobre el Español de América, J. Moreno de Alba (ed.), 48–65. México: UNAM. Gómez Torrego, L. 1999. La variación en las subordinadas sustantivas: Dequeísmo y Queísmo. In Gramática Descriptiva de la Lengua Española, I. Bosque and V. Demonte (eds), Chapter 34: 2105–2148. Madrid: Espasa. Herskovits, A. 1986. Language and Spatial Cognition. Cambridge: Cambridge University Press. Juilland, A. and Chang-Rodríguez, E. 1964. Frequency Dictionary of Spanish Words. London/The Hague: Mouton. Kany, Ch. 1945. American-Spanish Syntax. Chicago: University of Chicago Press. Kemmer, S. and Bat-Zeev-Shyldkrot, H. 1996. The semantics of empty prepositions in French. In Cognitive linguistics in the redwoods: the expansion of a new paradigm in linguistics, E. Casad (ed.), 347–388. Berlin/New York: Mouton de Gruyter. Levy Podolsky, P. 1983. Las completivas objeto en español. México (DF): El Colegio de México. Pottier, B. 1962. Systématique des éléments de relation. Étude de morphosyntaxe structurale romane. Paris: Klincksieck. Rabanales, A. 1974. Queísmo y dequeísmo en el español de Chile. In Estudios filológicos y lingüísticos. Homenaje a Angel Rosenblat en sus 70 años, 413–44. Caracas: Instituto Pedagógico. Real Academia Española, on-line. Banco de datos (CREA), Corpus de referencia del español actual.  [searches year 2005] Samper Padilla, J. A., Hernández Cabrera, E., C. and Troya Déniz, M. (eds) 1998. Macrocorpus de la norma lingüística culta de las principales ciudades del mundo hispánico. CD-ROM: Universidad de Las Palmas de Gran Canaria. Schwenter, S. A. 1999. Evidentiality in Spanish Morphosyntax: a Reanalysis of (de)queísmo. In Estudios de variación sintáctica, M.ª-J. Serrano (ed.), 65–87. Madrid: Vervuert. Serradilla Castaño, A. M. 2000. Acerca de las construcciones directas y preposicionales ante que completivo. La situación del español preclásico y del español de América. In Estudios sobre el español de América. Actas del Vº Congreso Internacional de “El Español de América” (Burgos, 6–10 de noviembre de 1995), H. Perdiguero and A. Álvarez (eds), 1301–1310. CDROM: Universidad de Burgos. Subirats-Rüggeberg, C. 1987. Sentential Complementation in Spanish. Amsterdam/Philadelphia: John Benjamins. Traugott, E. C. 1995. Subjectification in grammaticalization. In Language, Subjectivity and Subjectivisation, D. Stein and S. Wright (eds), 31–54. Cambridge: Cambridge University Press.

What does coordination look like in a head-final language?*, ** Nayoung Kwon and Maria Polinsky University of California, San Diego

Theories of clause linking treat coordination and subordination as mutually exclusive. In this paper, we examine a case where the surface distinction between coordination and subordination is obscured. Concentrating on the -ko construction in Korean, we show that a clause chain can be structurally ambiguous – it can either have all the properties of a coordinate structure or all the properties of a subordinate structure. The choice between the two types is determined by the construal of the events in question as parallel (coordination) vs. causal/sequential (subordination). We present diagnostics for determining subordination vs. coordination and show a correlation between syntactic and semantic properties involved in such structures.

Keywords: coordination, subordination, clause chain, Coordinate structure constraint (CSC), Korean, -ko construction, causality, sequentiality, coherence, clause linkage

1. Introduction Coordination is a structure in which two or more elements are joined in such a way that each of them could possibly be the head of that structure, and neither dominates the other one. Subordination is a structure in which two elements are joined in such a way that one of them is dominated by the other. This difference in terms of structural symmetry/asymmetry translates into a number of criterial distinctions, for example, the requirement that the conjoined phrases match in category, or the Coordinate Structure Constraint (CSC), well-known since Ross’s seminal work (1967). However, * We would like to thank Grant Goodall, Beth Levin, Barbara Lewandowska, Jakov Testelets, audiences at the 2003 BLS meeting, Stockholm University, and ICLC 9 (Seoul, 2005) for helpful comments on this paper. All errors are our responsibility. ** The following abbreviations are used: acc–accusative, dat–dative, decl–declarative, fut– future, gen–genitive, nom–nominative, past–past tense, pl–plural, prs–present, rel–relative marker, top–topic.



Nayoung Kwon and Maria Polinsky

it has long been noted that a superficially coordinate structure can either have all properties of coordination or all properties of subordination. For example, coordination is known to tolerate category mismatches (Culicover and Jackendoff 1997), cf. (1b), and violations of CSC do not always lead to ungrammaticality (Ross 1967; Schmerling 1975; Goldsmith 1985; Lakoff 1986; Kehler 2002), as shown in (2b). (1) a. b.

John drank another can of beer and we left. Another can of beer and I’m leaving.

(2) a. b.

What did the chair decide __and the committee vote on __? What did the committee meet and the chair decide __?

Researchers have tried to account for such unexpected phenomena in terms of structural ambiguity (Goodall 1987), mismatches between semantic and syntactic clause linkage types (Culicover and Jackendoff 1997), continuum between coordination and subordination (Foley and Van Valin 1984), or semantically determined selection of coordination vs. subordination (Na and Huck 1992, Kehler 2002). With few exceptions (Yuasa and Sadock 2002; Haspelmath 1995; Cristofaro 2003; Wälchli 2005), most of the work on the distinctions between coordination and subordination has been carried out using data from English or structurally similar languages, where the initial cut between coordination and subordination seems quite straightforward. It is much less clear what happens in a language that does not have such a straightforward contrast, and the goal of this paper is to examine one such language, Korean. Korean differs from English in that it has very little true coordination. According to some descriptions of Korean grammar, “[t]he distinction between coordination and subordination is not clear-cut and is a matter of degree” (Sohn 2001: 304). The criteria used to distinguish coordination and subordination seem primarily semantic, an issue to which we will turn below. Abstracting from semantic criteria for a while and judging by morphosyntactic encoding, the majority of clause linkage is done using a single finite predication, with a number of non-finite clauses that either linearly precede it or appear center embedded in it. Structures of this kind, illustrated in (3), are often referred to as clause chains. In (3), there is a single finite predicate (teylyewassta ‘brought’, underlined) and connected to its clause, are five apparently non-finite clauses with predicates linked to the rest of the sentence by the element –ko ‘and’ (in boldface below). (3)

(na-nun) yangka-eyse patao-n ket-tul nayngcangko-lang (I-top) two.family-from receive-rel thing-pl refrigerator-and kimchi nayngcangko-ey katuk chaywu-ko ppalay tolli-ko kimchi refrigerator-at fully fill-and laundry run-and naccam ca-ko cemsim mek-ko khemphyuthe com ha-ko nap sleep-and lunch eat-and computer little do-and Kkomayngi aykyen hothel-eys teylyewa-ss-ta. Komayngi pet.dog hotel-from bring-past-decl



What does coordination look like in a head-final language 

‘I filled up the regular refrigerator and kimchi refrigerator with things that I received from my relatives, did laundry, took a nap, had lunch, worked a little bit on the computer, and brought Komayngi back from the pet hotel.’ (Internet) Loosely defined, a clause chain is a sequence of two or more clauses; in that sequence, the predicate of one clause has a distinctive structure that occurs only once in the entire chain, while all the other clauses have predicates of a different structure chaining (Haiman 1985; Longacre 1985). More specifically, one predicate is finite and all the other predicates in the chain seem non-finite, hence dependent on the finite predication for the well-formedness of the entire chain.1 Clause chaining is very common cross-linguistically, and although it is more prominent in head-final languages, it is also found in English:

(4) Having heard how the great Mr. Brandon was to dine with them upon that day, the simple girl had been showing her respect for him… (Thackeray)

A structure with a single finite verb and a series of non-finite verbs is expected to show properties of subordination. To anticipate our conclusions, we show that clause chains are structurally ambiguous, showing either all properties of coordinate structures or all properties of subordinate structures. Crucially, there are no cases where the behavior of a clause chain is intermediary between coordination and subordination. Thus, the actual morphosyntactic appearance of the sentence may not be sufficient to identify it as a subordinate structure. We will show that the structural ambiguity of Korean clause chains can be resolved in a principled way. In what follows we will limit our discussion to clause chains with the complementizer –ko, which can appear on all clauses, with the exception of the finite one.2 In the ko-construction, only the finite predicate can bear the polarity marker, which suggests that this construction instantiates subordination, at least from the morphosyntactic standpoint. Ko-constructions are treated as subordinate structures by some researchers (Yi 1997 and references therein; Rudnitskaya 1998); however, Yoon (1994) and Sohn (2001: 203) identify them as coordinate. This difference of opinion already indicates that their status is not immediately obvious. The rest of the paper is structured as follows. Section 2 introduces general properties of the ko-construction, and sections 3 and 4 show that the construction can behave either as coordinate or as subordinate, respectively. In section 5 we address the issue of 1. While some researchers insist on distinguishing clause chains from standard subordination (Longacre 1985; Good 2003), the motivation for such separation is unclear. On the surface, chains seem to instantiate a particular case of subordination, and the only difference between chains and subordination in a more traditional sense seems to be in the number of non-finite clauses attached to the finite predication—hardly a reason for introducing a qualitative distinction. 2. It is generally assumed (often without much discussion) that –ko is some kind of a conjunction. Chung (2005), however, analyzes it as a mood marker.

 Nayoung Kwon and Maria Polinsky

this contradictory behavior and argue that the coordinate vs. subordinate behavior of the construction can be predicted on the basis of event structure. Section 6 summarizes the results and discusses outstanding questions.

2. General properties of the ko-construction Korean is a strictly head-final language with the predominantly SOV word order. Extraposition is extremely rare (Sohn 2001: 267), and under clause linkage, the finite clause either follows all the non-finite clauses, or has them center embedded—in any event, this means that the finite clause predicate always appears last, which provides us with a simple linear means of identifying it. In what follows, we will be distinguishing between the non-final clause (NFC) and final clause (FC). As mentioned above, tense marking in the NFC of the ko-construction is optional and polarity marking in the NFC is impossible; compare the tense and polarity of the NFC in (5): (5)

John-i chayk-ul ilk-(ess)-(*ta)-ko John-nom book-acc read-(past)-decl-and Mary-ka tibi-lul po-ass-ta. Mary-nom TV-acc see-past-decl ‘John read a book and Mary watched TV.’

At first blush, only the final clause (FC) is finite, while NFC is nonfinite. This structural asymmetry suggests that in the ko-construction, one clause syntactically dominates the others. However, the situation is more complex. As we will show in the next two sections, the ko-construction sometimes behaves as a coordinate structure, and sometimes, as a subordinate one.

3. Ko-construction as a coordinate structure Despite the apparent structural asymmetry, the ko-construction can behave as a standard coordinate structure. Evidence for its coordination behavior comes from several quarters: permutation without meaning change, backward pronominalization, sensitivity to the CSC, and center embedding. In standard coordination, permutation does not lead to meaning change (6). Similarly, the ko-construction does not show meaning change after permutation shown in (7b)—modulo the requirement that the final clause bear the polarity marker. (6) a. John likes Jane and loves Mary. b. John loves Mary and likes Jane.



What does coordination look like in a head-final language 

(7) a. John-i Jane-ul cohaha-ko Mary-lul salangha-ess-ta. John-nom Jane-acc like-and Mary-acc love-past-decl ‘John likes Jane and loves Mary.’ b. John-i Mary-lul salangha-ko Jane-ul choaha-ess-ta. John-nom Mary-acc love-and Jane- acc like-past-decl ‘John loves Mary and likes Jane.’ Backward pronominalization is impossible in coordinate structures (8), and possible in subordinate structures only (9), cf. (Culicover and Jackendoff 1997).

(8) *Another picture of himselfi has appeared in the newspaper, and Susan thinks Johni will definitely go out and get a lawyer.



(9) If another picture of himselfi appears in the newspaper, Susan thinks Johni will definitely go out and get a lawyer.

As befits a coordinate structure, the ko-construction does not allow backward pronominalization, which is illustrated in (10). (10)

*cakii-ka Sue-lul cohaha-ko Tomi-i John-ul silhehay-ss-ta. self- nom Sue-acc like-and Tom-nom John-acc hate-past-decl (‘Tom liked Sue and disliked John.’) lit.: ‘Hei liked Sue and Tomi disliked John.’

Next, the well-known Coordinate Structure Constraint states that in a coordinate structure, no conjunct may be moved, nor may any element contained in a conjunct be moved out of that conjunct (Ross 1967). (11) a. Time will bring pain and take my love away. b. *Whati will Time bring pain and take away __i? Topicalization in one clause of the ko-construction leads to ungrammaticality, which shows that the CSC is operational.  In Korean, topicalization is achieved by using a dedicated topic marker –(n)un ‘as for, concerning’ (Sohn 2001). Example (12) shows that ko-construction shows ungrammaticality after topicalization in one clause, just as in its English equivalent, while sentence (13) is grammatical since topicalization has applied across the board. (12)

*Maryi-nun John-i Jane-ul cohaha-ko Tom-i __i cohaha-n-ta. Mary-top John-nom Jane-acc like-and Tom-nom like-prs-decl (‘Mary, John likes Jane and Tom likes her.’) lit.: ‘Mary, John likes Jane and Tom likes.’

(13) Janei-un John-i __i cohaha-ko Maryj-nun Tom-i __j cohaha-n-ta. Jane-top John-nom like-and Mary-top Tom-nom love-prs-decl ‘Janei, John likes (heri), and Maryj, Tom likes (herj).’



Nayoung Kwon and Maria Polinsky

Relativization out of a single conjunct also leads to ungrammaticality in a coordinate structure. The ko-construction shows ungrammaticality under relativization: (14)

*John-i Jane-ul cohaha-ko Tom-i ___i cohaha-n Maryi John-nom Jane-acc like-and Tom-nom like-rel Mary (‘Mary who is such that John likes Jane and Tom likes her’) lit.: ‘Maryi who John likes Jane and Tom likes __i.’

In short, the ko-construction seems to obey the CSC. Next, center embedding is disallowed in coordinate structures but possible under subordination, as shown in the following English examples: (15) a. *John, and everyone was talking about center embedding, fell asleep b. John, while everyone was talking about center embedding, fell asleep As in English coordination, the ko-construction does not seem to allow center embedding. (16)

*Mary-ka [John-i yakwu-lul cohaha-ko] Mary-nom John-nom baseball-acc like-and nongkwu-lul silheha-ess-ta. baseball-acc hate-past-decl (‘John liked baseball and Mary disliked basketball.’)

Thus far the Korean ko-construction shows all relevant properties of coordination. However, these properties do not seem to be present in all ko-constructions.

4. Ko-construction as a subordinate structure Thus far we have shown that the ko-construction exhibits properties typically found in coordinate structures. In this section, we examine other instances of the ko-construction and show that they behave as subordinate structures. In the preceding section, we showed that permutation in the ko-construction does not change meaning, consistent with what normally happens under coordination. However, the following example illustrates an occasion of permutation in the ko-construction leading to meaning change: the meaning of (17a) is different from the meaning of (17b). This shows that the ko-construction sometimes behaves as a subordinate structure—a surprising result given that we have just observed its “coordinate behavior”. (17) a. Tom-i cip-ey o-ko Mary-ka tochakha-ess-ta. Tom-nom home-to come-and Mary-nom arrive-past-decl ‘After Tom came home, Mary arrived.’ ≠ b. Mary-ka tochakha-ko Tom-i cip-ey o-ass-ta. Mary-nom arrive-and Tom-nom home-to come-past-decl ‘After Mary arrived, John got home.’



What does coordination look like in a head-final language 

Contrary to the ban on backward pronominalization, illustrated in (10) above, some instances of the ko-construction allow backward pronominalization as well (18), which again attests to subordinate structure behavior. (18)

cakii-ka silswu-lul ha-ko Tomi-i na-eykey hwa-lul nay-ss-ta. self-nom error-acc do-and Tom-nom I-dat anger-acc give-past-decl ‘Tom got mad at me after he made an error.’ (lit.: ‘Hei made a mistake, and Tomi got mad at me.’)

Turning now to the CSC, recall that its violation should not result in ungrammaticality in a subordinate structure. Examples (19) and (20) illustrate topicalization and relativization, respectively, in just one clause of the ko-construction. Despite the asymmetrical topicalization and relativization, the sentences are grammatical. (19)

Tayceni-ulo-nun, John-i hankwuk-ey ipkwukha-ko(se) Taycen-to-top, John-nom Korea-to enter-and, Tom-i ___i isaha-ess-ta. Tom-nom move-past-decl ‘As for Taycen, after John entered Korea, Tom moved (to it).’

(20)

[Mina-ka pheynci-lul ssu-ko(se) ___i ka-n] hakkyoi Mina-nom letter-acc write-and go-rel school ‘the school that Mina went to after she wrote a letter’ (lit.: ‘the school that Mina wrote a letter and went to’)

Lastly, a subset of ko-constructions allows center embedding, contrary to the data in (16) above but in keeping with standard subordination (van Oirsouw 1987). (21) a.

John-i hakkyo-ey ka-ko Mary-ka John-uy pang-ey John-nom school-to go-and Mary-nom John- gen room-to mollay tule ka-ess-ta. sneak enter go-past-decl ‘John went to school and Mary sneaked into John’s house.’

b.

Mary-ka [John-i hakkyo-ey ka-ko] John-uy pang-ey Mary-nom John-nom school-to go-and John-gen room-to mollay tule ka-ess-ta. sneak enter go-past-decl ‘Mary, after John went to school, sneaked into John’s house.’

All the properties of the Korean ko-construction discussed in this section indicate that it involves subordination. In sum, a strange ambiguity arises: in some cases, the koconstruction behaves as a coordinate structure; in other cases it shows all the properties of a subordinate structure, as shown in Table 1 below. Crucially, there are no cases where one and the same instance of this construction shows some properties of a coordinate structure and some properties of a subordinate structure.



Nayoung Kwon and Maria Polinsky

Table 1.  Coordinate vs. subordinate properties of the ko-construction Coordinate structure

Subordinate structure

    

    

Permutation without meaning change Backward pronominalization Topicalization in only one clause Relativization out of only one clause Center embedding

In the next section, we attempt to reconcile this contradictory behavior of different instances of the ko-construction.

5. Reconciling parallelism and asymmetry In the preceding sections we showed that the ko-construction either demonstrates all properties associated with coordinate structures or all properties associated with subordinate structures. It should be noted that the ko-construction does not represent instances of co-subordination, under which coordinate (parallel) and subordinate (asymmetrical) properties co-occur within one and the same sentence (Foley and Van Valin 1984; Haspelmath 1995). On the contrary, the phenomenon observed here is ‘all or nothing’, either all coordination or all subordination. This suggests that the construction may be structurally ambiguous. If so, one should be able find a principled distribution of its coordinate vs. subordinate properties.

5.1

Structural ambiguity

The crucial descriptive generalization with respect to the ko-construction is that all the sentences showing coordinate properties are possible with overtly marked tense on the NFC predicate, while sentences showing subordinate structures do not permit such tense marking.3 The possibility or impossibility of tense marking on the NFC predicate also correlates with differences in interpretation discussed in section 5.2 below. To illustrate, consider the following example:

3. Of the Korean tenses, past is the clearest one, both in terms of its identification as tense (and not aspect or modality) and in terms of its overt marking. In the examples below, we will be mostly using past tense for expository purposes. See Sohn (2001: 358, 361) for the arguments that the forms identified as present tense (-nun) and future (marked by –(u)l- or –keyss) are actually mood.



What does coordination look like in a head-final language 

(22) a.

palam-i pwul-ko pi-ka wa-ss-ta. wind-nom blow-and rain-nom come-past-decl ‘The wind was blowing and it was raining.’ ‘The wind blew and then it rained.’

b.

pi-ka [palam-i pwul-ko] wa-ss-ta. rain-nom wind-nom blew-and come-past-decl ? ‘The wind was blowing and it was raining.’ ‘It rained after the wind blew.’

In (22a), overt tense marking on the NFC predicate is possible, but in (22b) it leads to ungrammaticality: (23) a.

palam-i pwul-ess-ko pi-ka wa-ss-ta. wind-nom blow-past-and rain-nom come-past-decl ‘The wind was blowing and it was raining.’ *‘The wind blew and then it rained.’

b. *pi-ka [palam-i pwul-ess-ko] wa-ss-ta. rain-nom wind-nom blow-past-and come-past-decl (‘The wind blew and it rained.’) If tense is overtly marked on the predicate of the ko-clause, the whole construction shows all the properties associated with coordination. This suggests that the construction without overt tense marking in the NFC is ambiguous between tensed NFC and non-tensed NFC. The former instantiates a coordinate structure, the latter, a subordinate one. Some additional examples: (24)

*[Mina-ka __i hapkyekhay-ess-ko] [emeni-ka kippum-uy Mina-nom pass-past-and mother-nom joy-gen nwnmwul-ul hulli-n] tayhaki tear-acc shed-rel college (‘the college that Mina got into and her mother shed tears of joy’)

(25) [sonyen-i ___i namki-ko hakkyo-lo ttena-n] pheyncii boy-nom leave-and school-to depart-rel letter ‘a letter that the boy left and went to school’ In (24), the extraction of a DP out of the ko-construction with optional tense in the NFC is impossible—this is a true coordinate structure, and the desired relativization clearly violates the CSC. In (25), however, tense marking is impossible, and relativization is legitimate. Another piece of evidence in support of structural ambiguity comes from scrambling. Although Korean allows cross-clausal scrambling (Sohn 2001 :295), scrambling out of some ko-constructions is impossible (26b), while other ko-constructions allow



Nayoung Kwon and Maria Polinsky

it (26c). The difference again correlates with the availability of overt tense marking – it is possible in (26b) and impossible in (26c), presumably under subordination. (26) a. Inho-nun [olaysstongan TV-lul po-ko] Mina-eykey malhay-ss-ta. Inho-top long TV-acc watch-and Mina-dat talk-past-decl ‘Inho watched TV and talked to Mina for a while.’ b. *Inho-nun Mina-eykeyi [ olaysstongan TV-lul po- ass- ko] __i Inho-top Mina-dat long TV-acc watch- past- and

malhay-ss-ta. talk-past-decl ‘Inho for a while watched TV and talked to Mina.’

c. Inho-nun Mina-eykeyi [olaysstongan TV-lul po- ko] __i malhay-ss-ta. Inho-top Mina-dat long TV-acc watch- and talk-past-decl ‘Inho watched TV for a while and then talked to Mina.’ In sum, there is a clear correlation between the availability of overt tense marking and the coordinate interpretation. The ko-construction is therefore ambiguous between the interpretation with the NFC tensed and the one with the NFC untensed. The use of a tensed NFC (including those clauses where tense is morphologically null) gives rise to the coordinate structure. The untensed NFC—where tense is impossible in principle, not just unexpressed—represents a syntactic structure with less functional architecture than found in a free-standing clause.4 Such a structure is embedded under the finite predicate, giving rise to subordination effects.

5.2

Semantic correlates of structural ambiguity

Tense specification in NFC also has discernible semantic correlates. In order to present these correlates, we will need to introduce some terminology here. Assuming that the entire ko­-construction describes a particular event, the states of affairs expressed by separate clauses in that construction are subevents. For the sake of simplicity, we will limit the discussion here to two subevents, one expressed by the NFC (ko-clause), the other expressed by the FC. In the absence of contextual cues, structurally ambiguous ko-construction is also ambiguous semantically, allowing a reading in which the subevents are simultaneous and a reading in which the subevents are sequential:

4. See Chung 2004, Bhatt 2005 (among others) on the category of the embedded clause (or clauselet, in Chung’s terms). For our purposes, it is crucial that the embedded ko­-clause is structurally smaller (less complex) than the coordinated ko-clause. Nothing hinges on the actual category of the embedded clause, and we leave this issue for future discussion.



What does coordination look like in a head-final language 

(27)

Mina-ka pap-ul ha-ko Chelswu-ka kwuk-ul kkuly-ess-ta. Mina-nom rice-acc do-and Chelswu-nom soup-acc boil-past-decl ‘Mina cooked rice and Chelswu cooked soup (at the same time).’ ‘Mina cooked rice and then Chelswu cooked soup.’

If the NFC has overt tense marking, the whole construction can have a range of interpretations, as long as the respective subevents are not construed as sequential.  The sequential reading is avoided. This is particularly apparent in ko­-constructions with a tensed NFC, where each clause has a different main participant—for such subevents, it is easier to construe each of them as developing in parallel, temporally co-extensive (Givón 1981). Relevant examples are presented in (28) and (29). (28)

cencayng-i theci-ess-ko manhun salam-tul-i cwuk-ess-ta. war-nom happen-and many person-pl-nom die-past-decl ‘The war broke out, and many people died/were dying.’ *‘The war broke out and then/as a result many people died.’

(29)

Yenghi-ka swukcey-lul hay-ss-ko, Yenghi-nom homework-acc do-past-and Mina-ka phiano-lul chy-ess-ta. Mina-nom piano-acc play-past-decl ‘Yenghi did her homework, and Mina was playing the piano.’ *‘Yenghi did her homework and then Mina played/was playing the piano.’

The absence of sequential interpretation is also clear from the following examples, where the continuation that presents the two subevents as sequential results in incongruity. (30)

palam-i pwul-ess-ko pi-ka wa-ss-ta. wind-nom blow-past-and rain-nom come-past-dec ??Kulentey pi-ka mence wa-ss-ta. but rain-nom first come-past-decl ‘The wind was blowing and it was raining. But it rained first.’

As long as the sequential interpretation, associated with subordination, is avoided, the actual interpretation of a particular coordinate sentence may vary, from strict simultaneity of the relevant subevents (30), to their co-extensivity in some general temporal frame reference such as a day (31), to the construal of the subevents as independent of each other (32). All these more specific readings seem to be implicatures stemming from particular event descriptions. As implicatures, they can be canceled. (31)

John-i 2 si-ey hakkyo-eyse nolay-lul ha-*(ess)-ko John-nom 2 hour-at school-at song-acc do-past-and Tom-i 4 si-ey cip-eyse siksa-lul ha-ess-ta. Tom-nom 4 hour-at home-at meal-acc do-past-decl ‘John sang a song at 2 at school, and Tom had a meal at 4 at home.’



Nayoung Kwon and Maria Polinsky

(32)

Ilun achim chengsopwu-ka kil-lul chengsoha-ko congil early morning janitor-nom street-acc clean-and all day acwumeni-tul-i hankaloi keli-lul kenil-ko ttayttaylo ladies-pl-nom leisurely street-acc walk-and sometimes ai-tul-i sikkulepkey ttuyeno-n-ta. child-pl-nom loudly run.play-prs-decl ‘In the morning, janitors clean the street, all day, ladies stroll around the street, and sometimes, children run around noisily.’

If the subevents within an event involve the same main participant the temporal coextensivity is not as apparent; after all, one and the same main participant often performs different actions sequentially. With a single main participant involved in both subevents, the construal is that the subevents are independent of each other and do not enter into a cause-effect relation, for example: (33)

nayil John-i chengso-lul ha-ko wumak-ul tul-ul tomorrow John-nom cleaning-acc do-and music-acc listen-fut kes-i-ta. ama kupakkey talun il-to ha-l kes-i-ta. thing-be-decl. probably else different job-also do-fut thing-be-decl ‘Tomorrow, John will clean and (then) listen to music. Probably he will do some other things as well.’

Ko-constructions that involve embedding and do not allow for an overt tense marking in the NFC are interpreted as sequential. Compare (28) above with (34), where the NFC is center-embedded and cannot have overt tense marking: (34) manhun salam-tul-i [cencayng-i theci-(*ess-)ko] cwuk-ess-ta. many person-pl-nom war-nom happen-past-and die-past-decl ‘Many people died after/since war broke out.’ The sequential interpretation cannot be canceled, which suggests that it is built into the meaning of the subordinate ko-construction: (35) a.

pi-ka [palam-i pwul-ko] wa-ss-ta. rain-nom wind-nom blow-and come-past-decl # Kulentey pi-ka mence wa-ss-ta. but rain-nom first come-past-decl ‘The wind blew and it rained. #But it rained first.’

b.

pi-ka [palam-i pwul-ko] wa-ss-ta. #Kulentey rain-nom wind-nom blow-and come-past-decl but pi-ka palam-i pwu-l-ttay tongsiey wa-ss-ta. rain-nom wind-nom blow-rel-time simultaneously come-past-decl ‘The wind was blowing and it was raining. #But it rained at the same time as the wind was blowing.’



What does coordination look like in a head-final language 

The strong sequential interpretation found under subordination in turn gives rise to the cause-effect interpretation (cf. also Rudnitskaya 1998). This can be seen in (34) above, which can be interpreted as a description of the cause (the war) and the effect (people dying), and in the following example: (36) nwun-i o-ko kil-i mikkulewue-ci-ess-ta. snow-nom come-and road-nom slippery-become-past-decl ‘It snowed and then the roads got slippery.’/ ‘The roads got slippery because it had snowed.’ However, the result interpretation can easily be canceled, which suggests that it is an implicature of the asymmetrical ko-construction. Example (34) above is compatible with the following continuation, which cancels the cause-effect relationship between the war and the deaths: (37)

manhun salam-tul-i cencayng-i theci-ko, many person-pl-nom, war-nom break out-and cwuk-ess-ta. Kulentey cenyempyeng-ulo cwuk-ess-ta. die-past-decl but plague-with die-past-decl ‘Many people died after/since the war broke out. But they died of plague.’

Likewise, in the next example, the sale of the house and the move are sequential but not causally related: (38)

isa-nun Inho-ka [ku cip-ul pal-ko] moving-top Inho-nom that house-acc sell-and hay-ss-ta. Mina-lul manna-se isa-lul hay-ss-ta. do-past-decl Mina-acc meet-because moving-acc do-past-decl ‘Inho moved after he sold the house. He moved because he met Mina.’

In sum, the structural ambiguity of the ko-construction finds correspondence in semantic ambiguity. The main distinction is between the sequential and non-sequential interpretation of the ko-construction. The sequential interpretation correlates with true subordination, while the coordinate ko-construction normally leads to the construal of events as temporally co-extensive, albeit in a very vague way. The presence of the causal interpretation is often implicated in the loss of coordinate properties (cf. the English example above, (1b), which has a clear cause-effect reading), but as the Korean data show this interpretation is not built into the inherent meaning of the subordinate construction. Rather, it is an implicature that arises due to the sequential reading.

 Nayoung Kwon and Maria Polinsky

6. Conclusions This paper examined a particular instance of clause chains in Korean, ones in which the non-final construction is headed by the complementizer -ko. We found that these clause chains, which morphologically appear to be subordinate structures, show exactly the same contrast between coordinate and asymmetric clause linkage as the purportedly coordinate structures in English, which sometimes exhibit subordination properties. Clause chains are thus ambiguous between coordinate structures, where the NFC can have tense (but not mood/polarity), and subordinate structures where the NFC cannot have tense. The contrast between coordination and subordination is therefore due to the independent tense on the combined clauses. To put it somewhat differently, the ko-construction is ambiguous between “masked” coordination and “masked” subordination—since the morphological marking is not always overt, extra work is needed to determine if a particular clause chain exhibits properties associated with coordination or subordination. Crucially, there are no situations when the ko-constructions discussed above have some coordinate and some subordinate properties. While we cannot speak for all other languages that have been claimed to have such “mixed” constructions, we would like to suggest that at least some of the mixed constructions may exhibit structural ambiguity similar to that found in Korean. The observed structural ambiguity in Korean clause chains also has regular semantic correlates: The coordinate ko­-construction favors a temporally co-extensive interpretation of the events represented in the sentence, while the subordinate construction is compatible with the sequential interpretation of events. The sequential interpretation often receives the additional cause-effect reading, which we suggest to be inferential, not built into the meaning of the construction. Further research is needed to determine if this conclusion is on the right track. Unlike the situation in more familiar languages such as English, tense in Korean is not the defining feature of finiteness. In Korean, tense specification is separate from finiteness—the latter is associated with polarity or mood. In the ko-construction, only the rightward predicate can be marked for polarity: (39)

John-i chayk-ul ilk-(ess)-(*ta)-ko, John-nom book-acc read-(past)-decl-and Mary-ka tibi-lul po-ass-ta. Mary-nom TV-acc see-past-decl ‘John read a book and Mary watched TV.’ (=(5) above)

The idea that finiteness is not defined by tense is not new; it is intriguing however that the more compelling arguments for the dissociation of these categories come from head-final languages (Whitman 1989; Shim 1995; Zanuttini and Portner 2003; Aygen 2004, and others), where tense marking and finiteness marking (expressed by mood or polarity) are often morphologically distinct.



What does coordination look like in a head-final language 

This points to a principled way in which languages may vary. In English, tense and finiteness are hard to separate, which is why clauses that appear to be non-finite are subordinate structures (but finite ones are not always coordinate ones, as illustrated by (1b) and (2b) above). In head-final languages like Korean, where mood/polarity and tense are expressed by distinct morphosyntactic means, the familiar conditions on coordination appear in a very different guise. Thus, English seems to have superficial coordination which masks subordination. Korean on the other hand, has superficial subordination which masks coordination. Hence, the morphosyntactic boundaries between the two phenomena are determined by language-particular constraints, such as the morphosyntactic locus of finiteness. However, the actual contrast between coordination and subordination follows principled rules familiar from more commonly studied languages.

References Aygen, Gulsat. 2004. Finiteness, case, and clausal architecture. MIT Occasional Papers in Linguistics 23. Cambridge, MA: MIT Department of Linguistics and Philosophy. Bhatt, Rajesh. 2005. Long distance agreement in Hindi-Urdu. Natural Language and Linguistic Theory 23: 757–807. Chung, Sandra. 2004. Restructuring and verb-initial order in Chamorro. Syntax 7: 221–255. Chung, Daeho. 2005. What does bare -ko coordination say about post-verbal morphology in Korean? Lingua 115: 549–568. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Culicover, Peter and Jackendoff, Ray. 1997. Semantic subordination despite syntactic coordination. Linguistic Inquiry 28: 195–217. Foley, William and Van Valin, Robert. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press. Givón, Talmy. 1981. Topic continuity in discourse: The functional domain of switch-reference. In Switch reference and Universal Grammar, John Haiman and Pamela Munro (eds), 51–82. Amsterdam: John Benjamins. Goldsmith, John. 1985. A principled exception to the Coordinate Structure Constraint. Proceedings of Chicago Linguistics Society 21: 133–143. Good, Jeff. 2003. Clause combining in Chechen. Studies in Language 27: 113–170. Goodall, Grant. 1987. Parallel structures in syntax: Coordination, causatives, and restructuring. Cambridge: Cambridge University Press. Haiman, John. 1985. Symmetry. In Iconicity in syntax, John Haiman (ed), 73–95. Amsterdam: John Benjamins. Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Converbs in cross-linguistic perspective, Martin Hasplemath and Ekkehard König (eds), 1–55. Berlin: Mouton de Gruyter. Kehler, Andrew. 2002. Coherence, reference, and the theory of grammar. Stanford, CA: CSLI. Lakoff, George. 1986. Frame semantic control of the Coordinate Structure Constraint. Proceedings of Chicago Linguistics Society 22(2): 152–167.

 Nayoung Kwon and Maria Polinsky Longacre, Robert. 1985. Sentences as combinations of clauses. In Language typology and syntactic description, Timothy Shopen (ed), 235–286. Cambridge: Cambridge University Press. Na, Younghee and Huck, Geoffrey. 1992. On extracting from asymmetrical structures. In The joy of grammar, Diane Brentari et al. (eds), 251–274. Amsterdam: John Benjamins. Oirsouw, Robert R. van. 1987. The syntax of coordination. London—New York: Croom Helm. Ross, John Robert. 1967. Constraints on variables in syntax. Ph. D. Dissertation, MIT. Rudnitskaya, Elena. 1998. Syntactic properties of the Altaic coordination construction in Korean. Sprachtypologie und Universalienforschung 51: 179–198. Schmerling, Susan. 1975. Asymmetric conjunction and rules of conversation. In Syntax and Semantics 3: Speech acts, Peter Cole and Jerrold Sadock (eds), 211–231. New York: Academic Press. Shim, Sang-Wan. 1995. A study on the Korean subjunctive complementizer –ki. Language Research 31: 55–85. Sohn, Ho-Min. 2001. The Korean language. Cambridge: Cambridge University Press. Wälchli, Bernhard. 2005. Co-compounds and natural coordination. Oxford: Oxford University Press. Whitman, John. 1989. Topic, modality and IP structure. In Harvard International Symposium on Korean Linguistics III, Susumu Kuno et al. (eds), 341–356. Seoul: Hanshin. Yi, Eun-Young. 1998. Adjunction, coordination and their theoretical consequences. In Cornell Working Papers in Linguistics, Niken Adisasmito-Smith and Tobey Doeleman (eds), 188–252. Ithaca, NY: Cornell University. Yuasa, Etsuyo and Jerry M. Sadock. 2002. Pseudo-subordination: A mismatch between syntax and semantics. Journal of Linguistics 38: 87–111. Yoon, James. 1994. Korean verbal inflection and checking theory. MIT Working Papers in Linguistics 22: 251–270. Yoon, James. 1997. Coordination (a)symmetries. Harvard Studies in Korean Linguistics VII, 3–30. Zanuttini, Raffaella and Portner, Paul. 2003 Exclamative clauses: At the syntax-semantics interface. Language 79: 39–81.

Verb serialization as a means of expressing complex events in Thai Kingkarn Thepkanjana

Chulalongkorn University, Bangkok

Verb serialization is generally defined as a linguistic device by means of which two or more verbs or verb phrases are juxtaposed without any linker. The output of verb serialization is called “serial verb construction”, henceforth SVC. This paper aims to (1) examine the characteristics of events expressed by SVCs, (2) analyze the relationships between serial verbs, and (3) investigate constraints on the serializability of verbs. This paper limits its scope to basic nongrammaticalized SVCs in Thai, which consist of two verb phrases only. It is argued that (i) each SVC expresses a single whole, complex event, and (ii) verbs do not serialize freely. This paper accounts for the serializability of verbs by analyzing semantic relationships between verbs in SVCs.

Keywords: verb serialization, serial verb construction, event, complex event, serializability, Cognitive Grammar, valence relations, correspondences

1. Introduction * Verb serialization is generally defined as a linguistic device by means of which two or more verbs or verb phrases with shared nomimal arguments are put into juxtaposition without any linker. The output of verb serialization is called “serial verb construction”, henceforth SVC. However, cases that are called SVCs in linguistic literature are found not to constitute the same phenomenon. There are cases that all linguists working on this topic categorize as SVCs and there are some other cases where they have conflicting opinions. SVCs are prevalent in Chinese, in the languages of West Africa, Southeast Asia, Papua New Guinea, Oceania, pidgins and creoles. Previous research works on SVCs have mostly dealt with such issues as the syntactic structures of SVCs, the differences between SVCs and related constructions such as coordinate and subordinate constructions, the types of event expressed by SVCs, and the grammaticalization * This research is supported by a Basic Research Grant from the Thailand Research Fund (No. BRG 4780019).

 Kingkarn Thepkanjana

of serial verbs into grammatical markers. Many linguists working on SVCs, such as Bruce (1988), Durie (1997), Foley and Olson (1985) describe them as expressing a “single event”. This paper aims to make an in-depth investigation of the properties of the so-called “single event” expressed by SVCs in Thai. However, the notion of an event is elusive. Different researchers may give different definitions of this term. This paper discusses the different meanings of the term “event” in section 2. The other issue that this paper aims to investigate is the serializability of verbs. The term “serializability” is defined as the potentiality of verbs to be serialized with one another. Verbs in a serializing language do not serialize with each other freely. The serializability is subject to certain constraints. This paper also aims to investigate the constraints on the ability of verbs to co-occur with each other in SVCs. In accounting for the serializability of verbs, this study draws on Langacker’s notions of valence relations and correspondences in Cognitive Grammar (Langacker 1987, 1999). The two theoretical notions will be discussed in section 4. SVCs in Thai consist of two or more verbs or verb phrases. SVCs which consist of only two verbs or verb phrases are called “basic SVCs” and ones which consist of more than two verbs or verb phrases are called “complex SVCs” (Chuwicha 1993). Chuwicha (1993) classifies basic SVCs into twelve patterns each of which is described by means of two syntactico-semantic types of verb co-occurring in a fixed linear order. According to Chuwicha (1993), complex SVCs consist of more than two basic SVCs embedded within one another. This paper limits its scope of study to only four patterns of basic SVCs in Chuwicha’s terms. The four patterns under investigation have many properties in common. Firstly, they are lexical verbs, not grammaticalized serial verbs which are prevalent in a large number of SVCs in Thai. Therefore, SVCs with grammaticalized verbs such as in (1) and (2), which are derived from the lexical verbs yùu ‘live’ and wâa ‘say’, respectively, are excluded from this study. Secondly, each verb in the SVCs being examined is semantically complete; it does not require another verb to complete its sense. SVCs with complement-taking verbs such as yàak ‘want’ in (3) and waˇŋ ‘hope’ in (4) are thus excluded. Thirdly, by virtue of being a lexical verb, verbs in each SVC in the four patterns under investigation have subject arguments which can be identical or different. In the case of the latter, the subject argument of the second verb is identical with the direct object argument of the first. The SVCs containing one of the verbs with the modifying function which does not have any subject argument such as in (5) and (6) are excluded from this study. The four patterns of SVCs under investigation are described in section 2. (1) kha&w kin he eat ‘He is eating.’

khâaw yùu rice PROG

(2) kha&w khít wâa kha&w tham thùuk he think COMP he do right ‘He thought he had done the right thing.’



Verb serialization as a means of expressing complex events in Thai 

(3) kha&w yàak pay tàaŋpràthêet he want go abroad ‘He wants to go abroad.’ (4) kha&w wâŋ tæŋŋaan kàp thәә he hope marry with you ‘He hopes to marry you.’ (5) kha&w phimdìit khlfŋ he type skilful ‘He types skilfully.’ (6) kha&w tffp thùuk he answer correct ‘He answered correctly.’

2. Relations between events expressed by verbs in basic SVCs In order to understand the characteristics of events expressed by verbs in basic SVCs, we need to analyze the relationships between those events. As mentioned above, this study limits its scope to only four patterns of SVC. Before I discuss the four patterns of SVC and investigate the properties of the events expressed by them, it is necessary to examine the possible meanings that the term “event” can express. Pawley (1987) is aware of the difficulty in characterizing events in a precise way. The difficulty arises from the lack of a well-developed metalanguage for talking about conventions for encoding events in linguistics. Therefore, he postulates a number of technical terms related to “event” and gives definitions to them in order to provide a “rough working platform” (Pawley 1987 :335) for accounting for how events are encoded in Kalam and English. The terms which are relevant to this study are as below (Pawley 1987: 335–336). 1. Event classifier: A verb stem denoting a kind of event, which can be conceptually simple or complex, such as hit, kill, eat, lie. 2. Conceptual event: The meaning of a clause containing a single event classifier, and denoting a bounded activity. It is usually placed in a particular time and place. 3. Event sequence: Two or more conceptual events, each of which is expressed by a separate clause. 4. Episodic event: A conceptual event which denotes a sequence of more or less separate acts or events. All of the events expressed by (7) constitute an episode or an episodic event.

(7) Mary painted a landscape; Bill built his own house; Lisa bludgeoned her father to death with forty whacks of an axe; then she gave her mother forty-one. (Pawley 1987: 336)

 Kingkarn Thepkanjana

5. Simple event: A conceptual event which comprises a single action, unrepeated. It is typically an event that begins and ends within the space of a few seconds or less. The event expressed by each of the sentences in (8)-(10) exemplifies the simple event.

(8) Bill released the rope.



(9) John winked.

(10) Mary gave me a slap on the face. (Pawley 1987: 336) 6. Complex event: Any conceptual event that is not a simple action but is not necessarily episodic. There are a large number of conceptual events occupying various points on the scale of complexity between simple and episodic events. Each of the sentences in (11)-(13) expresses a complex event. (11) Bill hit a ball through the window. (12) Fido is fetching his stick. (13) Mary’s words made me angry. (Pawley 1987: 336) 7. Objective event: The reality which event expressions refer to or which language users imagine they refer to. It is noted that the term “event” according to Pawley refers to three things, i.e. (1) the conceptual event, (2) the objective event, and (3) the linguistic expression which expresses a conceptual event. The conceptual event is the meaning of a verb stem and of a single clause. The conceptual event can be simple in that it comprises a single and unrepeated action. A number of conceptual events can occur in sequence and constitute an episodic event if they are expressed by a number of single clauses occurring consecutively. The conceptual event can be complex in that it consists of more than one simple action. However, the actions it comprises are more closely related to one another in time and space than those in an episodic event. In short, a conceptual event is an event in the mind of the speaker and hearer which is designated by a linguistic expression. On the other hand, an objective event is viewed as a “real” event in the world, not an event in the speaker’s or hearer’s mind. The last thing the term “event” refers to is a linguistic element which expresses an event, namely, a verb stem. The label used for this kind of event is “event classifier”. In each basic SVC, which comprises two verbs or verb phrases, each verb is an event classifier which denotes a conceptual event. Therefore, each basic SVC consists of two event classifiers. I will analyze the conceptual event and the objective event each pattern of SVC denotes below. The four patterns of SVCs are described in terms of syntactico-semantic verb types followed by some examples as follows.



Verb serialization as a means of expressing complex events in Thai 

1. Primary action verbs + Non-primary action verbs (14) kha&w takoon tffp he shout answer ‘He answered by shouting.’ (15) kha&w bòok mшш laa he wave hand bid farewell ‘He bade farewell by waving his hand.’ (16) kha&w kwàk mшш rîak cha&n he wave hand call I ‘He called me by waving his hand.’ (17) kha&w phayáknâa he&ndûay he nod agree ‘He agreed by nodding.’ (18) kha&w wây khffpkhun he raise the hands pressed together thank you ‘He thanked you by raising his hands to his chest and pressing.’ (19) kha&w khàp rót fàa faydææŋ he drive car violate red traffic light ‘He violated the red traffic light by driving.’ According to Chuwicha (1993), primary action verbs denote events which are carried out in an exact physical means and manner. In other words, primary action verbs are not vague as to the physical means and manner of carrying out the actions indicated by them. Primary action verbs are classified by Chuwicha (1993) into five subclasses: (1) physical displacement verbs, such as dәәn ‘walk’, wîŋ ‘run’, bin ‘fly’, khlaan ‘crawl’, (2) body part motion verbs, such as phaŋòk (hu&a) ‘lower (head)’, pròp (mшш) ‘clap (hands)’, phayák (nâa) ‘nod (face)’, ha&n (nâa) ‘turn (face)’, (3) posture verbs, such as nâŋ ‘sit’, yшшn ‘stand’, nffn ‘sleep’, (4) auditorily and visually physical action verbs, such as faŋ ‘listen’, mffŋ ‘look at’,?àan ‘read’, phêŋ ‘gaze’, and (5) other action verbs which express physical actions carried out by means of more than one body part, such as kin ‘eat’, phûut ‘say’, khayàw ‘shake’, chìik ‘tear’, takoon ‘shout’, láaŋ ‘wash’, pîŋ ‘grill’. In contrast, nonprimary action verbs are vague as to the means and manner of carrying out the actions expressed by them. In other words, it is not known how the actions expressed by nonprimary action verbs are exactly carried out. Nonprimary action verbs can be subclassified into physical action verbs such as pay ‘go’, maa ‘come’, khâw ‘enter’, ffk ‘exit’, rîip ‘hurry’, fшk ‘practise’, klææŋ ‘pretend’, laa ‘bid farewell’, tffp ‘answer’, and mental action verbs such as khít ‘think’, wikhrf? ‘analyze’. The SVCs having this pattern refer to one objective physical event. The first verb in each SVC expresses the exact physical means and manner of carrying out the action expressed by the second verb. The actions indicated by the two verbs in each SVC were

 Kingkarn Thepkanjana

performed by the same agent, and took place at the same time and place but refer to different aspects of the same objective event. For example, in (14) the agent answered a question by shouting. The action performed by the agent can be named shouting or answering at the same time. In (15), the agent bade farewell by waving his/her hand. The linear order of verbs in this type of SVC is not iconic since the two actions expressed by the two verbs in the SVC are cotemporal. This kind of event is conceptually complex in a different way from that described by Pawley (1987). It does not comprise “subevents” which occur in close sequence. Rather, it comprises multiple “event-facets” (Enfield 2002) which are not separable from one another. They are overlaid to form a detailed description of a single objective event. 2. Primary action verbs + Primary action verbs (20) kha&w thfft plaa kin he fry fish eat ‘He fried fish to eat.’ (21) kha&w cùt bùrìi sùup he lit cigarette smoke ‘He lit a cigarette to smoke.’ (22) kha&w yók mшш wây he raise hands pay respect by pressing the hands together ‘He raised his hands to pay respect.’ (23) kha&w rîit sш a sày he iron shirt put on ‘He ironed the shirt to wear.’ (24) kha&w yìp khano&mpaŋ kin he grab, pick up bread eat ‘He picked up a piece of bread to eat.’ (25) kha&w ŋæ&æn mffŋ thffŋfáa he raise one’s head see sky ‘He raised his head to see the sky.’ (26) kha&w ha&n la&ŋ phiŋ phanák kâw?`ii he turn back lean against backrest chair ‘He turned around to lean against the back of a chair.’ The SVCs in this pattern express two physical actions which are performed by the same agent but which occur in close sequence without a noticeable time span in between. The two verbs in this type of SVC are not vague as to the means and manner of carrying out the actions indicated by the verbs. The order of the verbs in this type of SVC is obviously iconic. The second actions are typically interpreted as the purpose of carrying out the first actions. Therefore, it is not necessary that the second events denoted by the second verbs, which are the purposes of performing the first actions,



Verb serialization as a means of expressing complex events in Thai 

really took place. Since the two actions expressed by each SVC in this pattern are performed by the same agent, are interpreted as an action-purpose sequence of events, and are conceptualized as occurring at more or less the same time and place if the purpose event does take place, they can be easily thought of as constituting a single objective event. However, they constitute a complex conceptual event because they comprise two conceptual actions expressed by two verbs or two event classifiers which are closely related to each other. 3. Primary action verbs Physical action verbs

+ Physical process (inchoative) verbs

(27) kha&w paa kææw tææk he throw glass broken ‘He threw a glass and it was broken.’ (28) kha&w yìap klfŋ bææn he step on box flat ‘He stepped on a box and it became flat.’ (29) kha&w thúp kææw tææk he hit, pound glass broken ‘He hit a glass and it was broken.’ (30) kha&w khâa phûuráay taay he kill criminal dead ‘He killed the criminal.’ (31) kha&w sák sш a sà?àat he wash shirt clean ‘He washed a shirt and it became clean.’ (32) kha&w chìik phâa khàat he tear cloth torn ‘He tore a piece of cloth.’ (33) kha&w rîit sш a rîap he iron shirt smooth ‘He ironed the shirt and it became smooth.’ The two verbs in each SVC in the third pattern express two events occurring in close sequence. Like the second pattern of SVC, there is hardly any noticeable time span between the two events. The first event is an action whereas the second one can be interpreted as either a process or the resulting state of an entity indicated by the direct object argument of the first verb. It is noted that stative verbs in Thai are ambiguous in that they can express either processes or states. In short, the two verbs in this type of SVC express a physical action carried out on a patient entity and a physical resulting state occurring to the patient entity, respectively. Like the second pattern of SVC, the order of this pattern of SVC is iconic. Since this type of SVC is interpreted as a cause-

 Kingkarn Thepkanjana

result sequence of events occurring right after each other at the same place, they can readily be considered as constituting a single objective event like the second pattern of SVC. This pattern of SVC expresses a complex conceptual event in that it comprises two conceptual events corresponding to the cause and result which occur at the same place without any noticeable time span. Notice that this pattern of SVC is complex in a different way from the second one. 4. Posture verbs + Action verbs (34) kha&w nâŋ àan nâŋsш ш he sit read book ‘He sat reading a book.’ (35) kha&w yшшn rffŋphleeŋ he stand sing ‘He stood singing.’ (36) kha&w nffn rffŋhây he lie cry ‘He lay crying.’ This type of SVC expresses the meaning that the subject entity performs an action denoted by the second verb while being in a certain posture denoted by the first verb. The two verbs in this pattern of SVC denote one single objective event because they express an action performed by the same agent at the same time and place. Like the first pattern, this pattern of SVC expresses multiple event-facets which provide a detailed description of an objective event. Therefore, it expresses a complex conceptual event like the first pattern of SVC. One might wonder whether we can analyze the verb types of each pattern of SVC above in terms of their lexical aspect, which is also called Aktionsart, or not. Lexical aspect is the inherent temporal structure of a situation. Most semantic analyses of the lexical aspect are presented in terms of Vendler’s event types (Vendler 1967). Vendler distinguishes four types of event based on three semantic features, namely, stative/dynamic, durative/punctual, and bounded/unbounded (or telic/atelic). The four event types are states, activities, achievements and accomplishments. States describe situations that do not change over time, that are extended in time and that do not have an inherent endpoint, such as love and be beautiful. Activities describe situations that involve change over time, but are unbounded and durative, such as run and eat. Achievements also describe dynamic situations but involve a change of state that is instantaneous or punctual. The punctual change of state ends in a resulting state. Two examples of achievements are reach the summit and shatter. Accomplishments are processes that are durative and lead to a natural endpoint, such as kill and read the book. Using Vendler’s terminology, the first pattern of SVC consists of two activity verbs; the second one also consists of two activity verbs, the third one consists of an activity verb followed by an accomplishment or a state verb; the fourth one consists of two activity verbs. It is



Verb serialization as a means of expressing complex events in Thai 

obvious that Vendler’s classification of event types is too coarse for our purpose. We need a classification of event types that is more fine-grained than that of Vendler. To summarize, the four patterns of SVCs above can be classified into two types according to the temporal relationship between the events expressed by the verbs in the SVCs. The first and the fourth patterns fall into the first type in which the two events expressed by the two verbs are performed by the same agent and occur simultaneously. The second and the third patterns fall into the second type in which the two events expressed by the two verbs have different subject arguments and occur in close sequence without a noticeable time span. However, it is argued that each of the four patterns of SVCs expresses one complex conceptual event.

3. Constraints on the serializability of verbs In this section, I will examine the factors which constrain the ability of verbs to cooccur in such a way that they result in well-formed SVCs. It is argued in this paper that there are two constraints on serializability, namely, the schematic constraint and the lexical constraint. The schematic constraint involves the syntactico-semantic verb types which co-occur as the first and the second verbs in basic SVCs. This constraint is therefore realized in terms of the two specific verb types which co-occur with each other in the patterns of SVCs described in section 2. This constraint operates at the schematic level in a taxonomic hierarchy in that the verb types in question are superordinate structures of lexical verbs. Such superordinate structures are termed “schemas” in Cognitive Grammar (Langacker 1987: 68). Thus, this constraint is called the schematic constraint. To recapitulate, there are four patterns of SVCs under investigation as below. These patterns can be viewed as the schematic constraint on the serializability or co-occurrence of verb types in SVCs within the scope of study. 1. Primary action verbs + Non-primary action verbs 2. Primary action verbs + Primary action verbs 3. Primary action verbs + Physical process (inchoative) verbs Physical action verbs 4. Posture verbs + Action verbs The schematic constraint is not sufficient for the well-formedness of SVCs in Thai. It is not the case that any verbs which instantiate the schemas appearing in the four patterns of SVCs above can co-occur with each other in SVCs. It is found that only certain verbs which instantiate their respective schemas can co-occur. The examples below illustrate this point. 1. Primary action verbs + Non-primary action verbs (37) * kha&w yók mшш khffpkhun he raise hands thank

 Kingkarn Thepkanjana

(38) * kha&w kwàk mшш fàa faydææŋ he wave hands violate red light (39) * kha&w khàp he drive

rót car

rîak call

(40) * kha&w phayáknâa laa he nod bid farewell 2. Primary action verbs + Primary action verbs (41) * kha&w pîŋ plaa klшшn he grill fish swallow (42) * kha&w sák phâa rîit he wash clothes iron (43) * kha&w yép sш a hâyyшшm he sew shirt lend (44) * kha&w cùt bùrìi thíŋ he light cigarette throw away 3. Primary action verbs + physical process (inchoative) verbs Physical action verbs (45) * kha&w hu&ŋ khâaw dш at he cook rice boil (46) * kha&w chìik kràdàat yáp he tear paper wrinkled (47) * kha&w thúp kæ æw bææn he hit strongly glass flat (48) * kha&w khâa phûuráay cèp he kill thief hurt 4 Posture verbs + action verbs (49) * kha&w dәәn làp he walk fall asleep (50) * kha&w nffn khii càkkayaan he sleep ride bicycle (51) * kha&w nffn ffkkamlaŋkaay he sleep exercise (52) * kha&w yшшn khàp rót he stand drive car



Verb serialization as a means of expressing complex events in Thai 

The SVCs in sentences (37)-(52) obey the schematic constraint as the two verbs in each sentence fall into the two verb classes in each pattern of SVC. However, these sentences are not well-formed. This fact shows that the schematic constraint is not sufficient to produce well-formed SVCs. I argue that SVCs must also obey another constraint, namely, the lexical constraint. The lexical constraint requires that verbs which have a potential to be serialized express commonly associated events. The lexical constraint can be viewed as a pragmatic constraint since it requires real-world knowledge to determine whether two events are associated in the real world or not. Sentences (37)-(40) are not acceptable because of two reasons. In (37) and (40), the first verbs in these sentences do not express conventional means and manner in performing the actions indicated by the second verbs. In Thai culture, it is not customary to express appreciation by raising one’s hands as expressed in (37) and to bid farewell by nodding as expressed in (40). On the other hand, in (38)-(39), it is not conceivable that one can perform the actions indicated by the first verbs while performing those indicated by the second verbs. In other words, it is practically not possible to perform the actions expressed by the first verbs and the second verbs simultaneously. Sentences (41)-(44) are not well-formed for two reasons. Sentences (41)-(42) express plausible sequences of events; however, they are ill-formed because our realworld knowledge suggests that such sequences of events do not occur right after each other. There is a span of time between the two events in each sequence of events. As for sentences (43)-(44), they do not express conventionalized scenes or prototypical sequences of events in the real world. We can conclude that, in order for this type of SVC to be well-formed, it must express a conventionalized sequence of actions occurring right after each other without a time span. Sentences (45)-(48) are ill-formed because they do not express conventionalized sequences of causing and resulting events. All sequences of events expressed in (45)-(48) are practically impossible. In order for this type of SVC to be well-formed, it must express a conventionalized sequence of causing and resulting events occurring without an intervening time span such as below. (53) kha&w hu&ŋ khâaw sùk he cook rice cooked ‘He cooked rice (and it was cooked.)’ (54) kha&w chìik kràdàat khàat he tear paper torn ‘He tore paper (and it became torn).’ (55) kha&w thúp kææw tææk he hit strongly glass broken ‘He hit a glass and it broke.’ (56) kha&w khâa phûuráay taay he kill thief dead ‘He killed the thief (and he died.)’

 Kingkarn Thepkanjana

Sentences (49)-(52) are ill-formed because the sequences of events expressed by them are practically impossible for the same persons to perform simultaneously. In other words, their ill-formedness arises from the impossibility of the two events denoted by the first and the second verbs to take place simultaneously in the real world. In summary, the lexical constraint requires that SVCs consist of verbs which are semantically and pragmatically associated in the following ways. Firstly, they must express scenes which have the possibility of taking place in the real world. Secondly, the possible scenes must also be prototypical or conventionalized in the real world. There are two types of conventionalized scenes exhibited by SVCs, namely, (i) the scenes in which actions are performed by conventionalized means and in conventionalized manners, and (ii) the scenes in which conventionalized sequences of actions take place without an intervening time span. It is noted that these SVCs do not have a high degree of productivity. Rather, they exhibit word-like features in that they express scenes which are highly conventionalized. This corresponds with Bruce (1988)’s claim that SVCs in Alamblak exhibit a high degree of lexicalization and therefore form part of a structural continuum between syntactic units and lexical items. Notice that the four patterns of SVCs under investigation have some properties in common with the coordinate construction. Stassen (1985) classifies conjunct clauses in the coordinate constructions into two types based on the temporal relationship between them, namely, “consecutive chains” or “C-chains” and “simultaneous chains” or “S-chains”. The C-chains are clauses which express events which take place in temporal sequence. The clauses which are C-chains are iconic in that the linear order of the conjoined clauses reflects the temporal order of the events denoted by the clauses. The S-chains are clauses which represent simultaneous events. Since the two events expressed by the two verbs in the SVCs being investigated can occur sequentially and simultaneously, Stassen’s notions of C-chains and S-chains can be applied to the SVCs too. This is the first similarity between SVCs and coordinate constructions. Wierzbicka (1980) argues that joining two sentences with a conjunction is possible only when a speaker can conceive of the two events as a single whole. The single whole corresponds to the Gestalt notion of a single, unified figure. According to Wierzbicka (1980), the conceptualization of two events as a single whole requires finding a common denominator between the conjoined elements. In other words, the speaker must conceptualize the two events as a single whole having something in common. According to Croft (2001), the requirement to find a common denominator between conjoined elements explains why some coordinate constructions are acceptable such as (57) whereas some others sound odd such as (58). (57) The sun was shining and the birds were singing. (58) ?? John kissed Mary on the nose and kangaroos are mammals. (Croft 2001: 336)



Verb serialization as a means of expressing complex events in Thai 

The requirement for SVCs to obey the lexical constraint, which depends to a large extent on real-world knowledge, suggests that SVCs must express a single, unified figure as does the coordinate construction. In order for a basic SVC to be well-formed, the speaker and hearer must be able to conceptualize of two events expressed by the two verbs in the basic SVC as a single, complex conceptual event which is a single, unified but complex figure in Gestalt terms. In coordinate constructions, not everything can be naturally conjoined. In the same vein, not everything can be naturally serialized in SVCs.

4. An account of the constraints on serializability in terms of valence relations The term “valence”, which is better known as “valency”, is a theoretical construct used in various syntactic theories such as valency grammar, dependency theory and some kinds of functional grammar. It refers to a kind of dependency property of lexical items especially verbs. It is seen as the capacity a lexical item has for combining with other sentence constituents, in a similar way to that of the valency of a chemical element which refers to its capacity for combining with a fixed number of atoms of another element. The term “valency” is typically used to describe the capacity of verbs to combine with other constituents which are typically nouns.

4.1

Valence relations in Cognitive Grammar

Langacker (1987) postulates the term “valence relations” in his Cognitive Grammar. Valence relations exist between two component structures which combine to form a composite structure, which is a more elaborate expression. Valence relations are based on the sharing of elements between component structures in the same way as those between atoms which are based on the sharing of electrons. “It is only by virtue of having certain substructures in common that two component expressions can be integrated to form a coherent composite expression (Langacker 1987: 278).” In other words, two symbolic structures are combined by virtue of a set of “correspondences” which link shared substructures within the two structures (Langacker 1999: 66). Correspondences can hold between facets of two component structures, which are called “horizontal correspondences”, and can hold between facets of the component and composite structures, which are called “vertical correspondences.” Horizontal correspondences constitute the one invariant feature of valence relations. Moreover, there are both semantic as well as phonological correspondences. The type of correspondence which is relevant to this study is the semantic one. For example, the two semantic components [UNDER] and [THE TABLE] are integrated into [UNDER THE TABLE] (Langacker 1987 :279). The predication [UNDER] profiles a stative spatial relationship whereas [UNDER THE TABLE] profiles an entity which has a shape specification in three-dimensional space. The integration of the two predications above depends on a

 Kingkarn Thepkanjana

correspondence between the landmark of [UNDER] and the profile of [THE-TABLE]. These are substructures which are construed as identical.  A composite structure is formed by superimposing corresponding entities and merging their specifications. The component predications are integrated by virtue of their overlapping substructures. Identifying correspondences is therefore crucial to the full description of a construction. Figure 1 illustrates the horizontal and vertical valence relations of the composite structure [UNDER THE TABLE]. The correspondences are represented by dotted lines.

Figure 1.  Valence relations of [UNDER THE TABLE] (Langacker 1987: 280)

Linguists who work on valency or valence relations have focused their attention on verbs and their capacity to take nominal arguments. Consequently, there is a tendency to assume that only relational predications especially verbs are capable of having valence relations. According to Langacker (1987 :284), valence relations holding between



Verb serialization as a means of expressing complex events in Thai 

verbs and nouns are considered prototypical. There are nonprototypical valence relations holding between some elements other than verbs and nouns. An example of nonprototypical valence relations given by Langacker are those holding between nouns and nouns in noun compounds such as puppy dog and killer bee and in appositional expressions such as Jack the Ripper (Langacker 1987: 285). Another case of nonprototypical valence relations which I will introduce here is those holding between verbs in SVCs. In section 4.2, I will discuss valence relations in SVCs and how the notion of valence relation can be employed in accounting for the potentiality of verbs to be serialized.

4.2

Correspondences between verbs in SVCs

In accounting for constraints on serializability in terms of valence relations or correspondences, I will set out by discussing one of the basic assumptions of Cognitive Grammar concerning linguistic semantics. According to Cognitive Grammar, most lexical items have a considerable array of related meanings, which are represented in network form. The meaning of a lexical item, which is called the semantic structure, must be equated with the entire network. The semantic structures of lexical items are characterized relative to “cognitive domains,” which are cognitive entities such as mental experiences, representational spaces, concepts or conceptual complexes (Langacker 1987: 147). Linguistic semantics in Cognitive Grammar is held to be encyclopedic and open-ended in nature. This paper is concerned with verbs which co-occur in strings. Verbs are known to be relational in nature; they are not semantically complete in themselves. They must take some arguments in order to be semantically complete. According to Langacker (1987: 215), a relational predication puts interconnections in profile. Relations are conceptually dependent in nature. One cannot conceptualize interconnections without also conceptualizing the entities they interconnect. However, the verbs or verb phrases occurring in the SVCs under investigation are semantically complete in themselves. Serial verbs are apparently interconnected relational predications. An important question is what factor integrates two semantically complete verbs or verb phrases, which result in the composite structure called SVC. As discussed in section 4.1, valence relations involve the integration of two or more component structures to form a composite structure. They depend on correspondences established between substructures within the component elements. We might wonder what correspondences hold between two semantically complete verbs and verb phrases in our case. To account for such an integration requires that we have an encyclopedic conception of the semantics of verbs, which takes into consideration both the so-called linguistic and extralinguistic knowledge of the lexical items in question. In the first type of SVC in which primary action verbs co-occur with non-primary action ones exemplified in (14)-(19), the first verbs express the conventional means and manner of carrying out the actions indicated by the second verbs. The actions

 Kingkarn Thepkanjana

expressed by the second verbs are at the same time viewed as the objective in carrying out the actions expressed by the first verbs. Notice that the correspondences between the first and second verbs in the SVCs in (14)-(19) are not as prominent as those between verbs and nominal arguments, which constitute the prototypical valence relations. It requires extralinguistic knowledge to know that some actions can be seen as the means and manner of carrying out other actions whereas some actions are implemented by means of other actions. For example, only an inclusive semantic characterization of the first verb phayáknâa ‘nod’ in (17) can bring into the picture the action of he&ndûay ‘agree’ which can be placed in correspondence with phayáknâa ‘nod,’ thereby permitting the integration of the two verbs to form a coherent composite structure. In the second type of SVC exemplified in (20)-(26), the actions expressed by the first verbs are seen as the prerequisite actions in carrying out those expressed by the second ones. The latter actions are also regarded as the agents’ objectives in carrying out and completing the former ones. As in the first type of SVC, it requires encyclopedic knowledge to detect such nonprototypical correspondences between the two verbs in each SVC. In the third type of SVC shown in (27)-(33), the actions expressed by the first verbs are seen as the causing actions which give rise to the events expressed by the second verbs. An encyclopedic knowledge of verbs is necessary in identifying such correspondences between the verbs in each SVC. In the fourth type of SVC shown in (34)-(36), the actions expressed by the first verbs are seen as the agents’ postures while performing the actions denoted by the second verbs in the constructions. It requires some extralinguistic knowledge to identify the semantic relatedness or the correspondences between the two verbs in the SVCs. Such correspondences allow the integration of the two verbs into a coherent composite unit. In summary, the correspondences between the two verbs or verb phrases in the SVCs under investigation are not of the prominent type. The substructures of the component semantic structures overlap, not at the core, but at the periphery. This fact explains why the verbs in these SVCs are felt to be semantically complete to a large degree. It takes encyclopedic knowledge to detect such nonprominent correspondences between the two verbs in these SVCs. Now we can account for the ill-formedness of sentences (37)-(52) in terms of correspondences. Sentences (37)-(52) are ill-formed as long as we cannot find correspondences between the semantic structures of the two verbs in each sentence.

5. Conclusion In this paper, I have examined the following aspects of the basic nongrammaticalized type of SVC in Thai: (i) characteristics of events expressed by SVCs, (ii) relationships



Verb serialization as a means of expressing complex events in Thai 

between verbs in SVCs, and (iii) constraints on the serializability of verbs. The events expressed by SVCs are accounted for in terms of the Gestalt Principle, which essentially states that the whole is more than the sum of its parts. It is argued that each SVC expresses a single whole, complex event. An event can be complex in two ways. Firstly, it may consist of separate events which occur sequentially without any intervening time span, and which expresses a unitary, complex event with an additional purposive, causal or resultative meaning. Secondly, an event may consist of multiple event-facets which are not separable from one another in constituting a complete event. This paper represents another attempt which examines the nature of the event expressed by SVCs in Thai. However, it provides an in-depth analysis of the kinds of event expressed by SVCs in Thai, which will hopefully shed light on the investigation of the nature of events expressed by SVCs across languages. It is also found in this paper that verbs cannot be serialized freely. Verbs which can be serialized must have a common denominator in Wierzbicka’s terms, which corresponds to the notion of “correspondences” in Cognitive Grammar. This paper accounts for the serializability of verbs by analyzing semantic correspondences between verbs in SVCs. The correspondences between verbs in SVCs are considered nonprototypical valence relations. Discerning the nonprototypical valence relations requires encyclopedic knowledge. However, the analysis of the valence relations between serial verbs in SVCs is considered preliminary at this stage. Hopefully, the phenomenon of verb serialization will provide some insight into the further study of valence relations in Cognitive Grammar.

References Bruce, L. 1988. Serialization: From syntax to lexicon. Studies in Language 12 (1): 19–49. Chuwicha, Y. 1993. Clausehood in Serial Verb Constructions in Thai. Bangkok: Chulalongkorn University Ph.D dissertation. Croft, W. 2001. Radical Construction Grammar. Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Durie, M. 1997. Grammatical structures in verb serialization. In Complex Predicates, A. Alsina, J.Bresnan and P. Sells (eds), 289–354. Stanford: CSLI Publications. Enfield, N.J. 2002. Cultural logic and syntactic productivity: Associated posture constructions in Lao. In Ethonosyntax: Explorations in Culture and Grammar, N.J. Enfield (ed), 231–258. Oxford: Oxford University Press. Foley, W. and Olson, M. 1985. Clausehood and verb serialization. In Grammar Inside and Outside the Clause, J. Nichols and A.C. Woodbury (eds), 17–60. Cambridge: Cambridge University Press. Langacker, R.W. 1987. Foundations of Cognitive Grammar, Vol I: Theoretical Prerequisites. Stanford: Stanford University Press. ––– 1999. Correspondences, compositionality, and grammar. Manusya. Journal of Humanities. 2(2): 61–76.

 Kingkarn Thepkanjana Pawley, A. 1987. Encoding events in Kalam and English: Different logics for reporting experience. In Coherence and Grounding in Discourse, R.S. Tomlin (ed), 329–360. Amsterdam/ Philadelphia: John Benjamins. Stassen, L. 1985. Comparison and Universal Grammar. Oxford: Basil Blackwell. Vendler, Z. 1967. Linguistics in Philosophy. Ithaca: Cornell University Press. Wierzbicka, A. 1980. Lingua Mentalis: The Semantics of Natural Language. New York: Academic Press.

Notional asymmetry in syntactic symmetry Connective and accessibility marker interactions* Katsunobu Izutsu and Mitsuko Narita Izutsu

Hokkaido University of Education at Asahikawa / Sapporo University

The present paper argues that the distinction between symmetric- and asymmetric-event descriptions determines the pronominal reference in the conjoined structure “S1. discourse connective + S2.” with a pronoun in the S2. Symmetric-event descriptions presuppose an entity which does two things or to which two things are done. Asymmetric-event descriptions evoke an expectation that certain events/states will follow the event/state coded in S1. In asymmetric-event descriptions, the high-accessibility marker refers to the entity that lies inside the expectation, while the low-accessibility marker refers to the entity outside the expectation. We further show that the same phenomena with a different type of discourse connectives (Japanese suruto and Korean geureoja) can also be fully dealt with in the proposed account.

Keywords: discourse connective, pronominal reference, symmetric-event description, asymmetric-event description, expectation, stressed pronoun, topic marker, nominative marker, Japanese, Korean

1. Introduction The pronominal reference in the conjoined structure “S1. discourse connective + S2.” with a pronoun in the S2 has been accounted for in reference to “parallel function” (Sheldon 1974), “grammatical relations hierarchy” (Keenan & Comrie 1977 inter alia), and “thematic roles” (Stevenson et al. 1994; Stevenson et al. 2000). Cognitive Grammar recapitulates grammatical relations hierarchy as “salience” (Langacker 1991; van Hoek 1995, 1997). However, those accounts cannot deal fully with the phenomena as shown in sections 2.1 through 2.4. The sentences in (1) serve as litmus paper or touchstone for checking the explanatory power of the accounts. The sentence pairs (1a-b) and (1c-d) are syntactically parallel. One sentence in each pair has an unstressed pronoun as in (1a) and * We would like to express our gratitude to Yeongae Bin and Rebecca Dosch for their stimulating discussion and insightful comments on our Korean and English examples.

 Katsunobu Izutsu and Mitsuko Narita Izutsu

(1c), while the other has a stressed pronoun as in (1b) and (1d). Yet the pronominal reference in the first pair (John, Bill) is reversed in the second (Bill, John). An explanation for the pronominal interpretation in one pair would have to fail in the other pair. (1)

a. b. c. d.

John hit Bill. And he cried. (he=John) John hit Bill. And he cried. (he=Bill) John hit Bill. So he cried (he=Bill) John hit Bill. But he/?he cried. (he=John)

Section 3 demonstrates that the two events described in the structure “S1. discourse connective + S2.” can be divided into symmetric and asymmetric ones. Symmetric events presuppose that either one entity does two things or two things are done to it. Asymmetric events involve an expectation that certain events or states will follow the event or state coded in the first conjunct. Section 4 argues that discourse connectives serve to instruct the addressee whether the structure should be interpreted as describing symmetric or asymmetric events: and, in contrast, etc. (symmetric) and so, but, nevertheless, etc. (asymmetric). In symmetric-event descriptions, the entity which does two things or to which two things are done is the most plausible pronominal referent in the second conjunct. In asymmetric-event descriptions, the pronominal referent can be inside or outside the relevant expectation. The entities that lie inside and outside it respectively achieve high and low accessibility. The high-accessibility marker (unstressed pronouns) refers to the entity inside the expectation and the low accessibility marker (stressed pronouns) to the entity outside it. The second half of the section further shows that the distinction between the unstressed and stressed pronouns largely parallels that between pronouns with the topic marker (-wa and -neun) and the nominative marker (-ga and -ga/i) in Japanese and Korean. The account based on symmetric/asymmetric-event descriptions and high/low accessibility can deal with similar phenomena in Japanese and Korean such as (2) and (3). (2) a. b. c. d.

Jon-wa Biru-o tatai-ta. Sosite kare-wa nai-ta. (kare=‘John’) John-top Bill-acc hit-past and he-top cry-past ‘John hit Bill. And he cried.’ Jon-wa Biru-o tatai-ta. Suruto kare-wa nai-ta. (kare=‘Bill’) John-top Bill-acc hit-past and he-top cry-past ‘John hit Bill. And (thereupon) he cried.’ Jon-wa Biru-o tatai-ta. Sorede kare-wa nai-ta. (kare=‘Bill’) John-top Bill-acc hit-past therefore he-top cry-past ‘John hit Bill. So he cried.’ Jon-wa Biru-o tatai-ta. Sikasi kare-ga/?wa nai-ta. John-top Bill-acc hit-past but he-nom/?top cry-past ‘John hit Bill. But he cried.’ (kare=‘John’)



Notional asymmetry in syntactic symmetry 

(3) a. b. c. d.

Jon-eun Bil-eul chyeo-ss-da. Geurigo geu-neun uleo-ss-da. John-top Bill-acc hit-past-SE and he-top cry-past-SE ‘John hit Bill. And he cried.’ (geu=‘John’) Jon-eun Bil-eul chyeo-ss-da. Geureoja geu-neun uleo-ss-da. John-top Bill-acc hit-past-SE and he-top cry-past-SE ‘John hit Bill. And (thereupon) he cried.’ (geu=‘Bill’) Jon-eun Bil-eul chyeo-ss-da. Geureseo geu-neun uleo-ss-da. John-top Bill-acc hit-past-SE therefore he-top cry-past-SE ‘John hit Bill. So he cried.’ (geu=‘Bill’) Jon-eun Bil-eul chyeo-ss-da. Geureonde geu-ga/?neun John-top Bill-acc hit-past-SE but he-nom/?top ‘John hit Bill. But he cried.’ uleo-ss-da. (geu=‘John’) cry-past-SE

Section 5 points out that sentences like (2b) and (3b) instantiate a different type of asymmetric events, which, of course, the earlier studies do not intend to handle because they concentrate on English data. It is demonstrated that the proposed account also captures the pronominal reference in such a variant of the structure “S1. discourse connective + S2.” in the same line of account.

2. Earlier studies 2.1

Parallel-function account (Sheldon 1974; Cowan 1980)

Sheldon (1974) is the first to propose the concept of “parallel function” as a strategy for assigning antecedents to pronouns. The intent of the strategy is: “the pronoun in the second conjunct is interpreted as being coreferential with the NP that has the same grammatical function in the first conjunct” (ibid.: 280), as illustrated in (4). (4) a. John hít Billi, and then Sarah kícked himi. b. Johni hít Bill, and then hei kícked Sarah.

(Sheldon 1974: 280)

Parallel function strategy does not fully apply to the sentences in (1), let alone those in (2)–(3), because it would predict that John is the antecedent of the pronoun in all those sentences. Cowan (ibid.: 112), along with Sheldon (ibid.), suggests a supplementary rule: “the addition of contrastive stress reverses the assignment” of the antecedent. Although this might cover (1b), it still fails to account for (1c–d). The strategy, hand in

 Katsunobu Izutsu and Mitsuko Narita Izutsu

hand with the rule, predicts that the pronoun would refer to John in (1c) and that it should not have a stress on it in (1d). However, these predictions contradict the fact: the pronoun refers to Bill in (1c) and it is stressed in (1d). The inadequacy of parallel function strategy will further be shown by the pronominal interpretation of the sentences in (5). Sentences (5a-b) only differ in the connective used before the second conjunct. The pronoun is interpreted as being coreferential with Dave in (5a), while it is interpretable as being coreferential with either Simon or Dave in (5b). (5) a. b.

Simon bullied Dave. In contrast, Ken sympathized with him. (him=Dave) Simon bullied Dave. Nevertheless, Ken sympathized with him. (him=Simon or Dave) (Izutsu 2005: 34)

The Simon reading of the pronoun will be preferred with knowledge of ‘Simon’ to be a notorious school bully who has difficulty getting along with anyone else. However, the Dave reading will obtain with an assumption that ‘Simon’ and ‘Ken’ are close friends.

2.2

Reference-point account (van Hoek 1995, 1997)

Van Hoek (1997) proposes an account for pronominal anaphora based on the notion of “reference-point,” arguing that firstly the clausal subject, and secondly the object, serves as a reference point through which the addressee makes mental access to the referent of the following pronoun. Such a reference point is identified as the antecedent of the pronoun. Her model accounts for the felicitous interpretation of the pronoun in (1a) and (1d) but does not for the pronominal interpretation of (1b) and (1c). The subject is the most eligible for the reference point, since it prevails, by nature, over the object in prominence. The pronoun would invariably refer to John in all the four sentences. Van Hoek (ibid.: 58–59) claims that the linear order (as well as conceptual connectivity) helps to determine the appropriate pronominal referent (i.e., its reference point). Yet it only predicts that both the subject and object in the first clause are possible antecedents of the pronoun, without saying which is preferable as such. Following Lakoff (1971), van Hoek (ibid.: 62) makes a few preliminary observations on the role of contrastive or emphatic stress: “contrastive stress on a pronoun seems to signal that the speaker intends a nondefault reading of the pronoun.” 1 This line of explanation holds for (1b) but may contradict (1d). Furthermore, it cannot treat (1c) properly because it would expect a stressed pronoun to be used in the example. 1. Akmajian and Jackendoff (1970) observed earlier that there are some cases where interpretation of coreferentiality depends crucially on stress patterns.



Notional asymmetry in syntactic symmetry 

2.3

Thematic-role account (Stevenson et al. 1994; Stevenson et al. 2000)

Stevenson et al. (1994) discuss the “thematic role preferences” observable in the comprehension of pronouns. They demonstrate that pronouns are likely to be coreferential with an NP with the thematic role associated with the endpoint of the event (goal, patient, etc.) as exemplified in (6). (6) a. John seized the comic from Bill. He.... (he=John>Bill) b. John passed the comic to Bill. He.... (he=Bill>John) (cf. ibid.: 1994: 525–531) The goal- and patient-NP preferences as pronominal antecedents give no explanation for the interpretation of the pronoun in (1a) and (1d). Although they would appear to deal with (1b-c), the thematic role preferences do not account for the difference between the stressed and unstressed pronoun. Even cases like (1c) cannot be fully handled because the pronominal reference can be reversed in the second conjunct preceded by so as in (7). (7) a. John hit Bill. So he was blamed by everybody. (he=John) b. John hit Bill. So everybody blamed him. (him=John) Furthermore, Stevenson et al. (ibid.: 540) point out that the use of so reinforces the preference for the goal, Bill. This also runs counter to the fact in (7). A sufficient account must be able to describe when the pronoun refers to the subject and when it indicates the object.

2.4

Eclectic Account (Kehler 2002)

Grounded on the achievements of the earlier studies, Kehler (2002) proposes an eclectic account that utilizes three explanatory devices: “parallel function” (Sheldon 1974), “coherence” (Hobbs 1979), and “salience” (van Hoek 1997 among others). Coherence overlaps “causality” (Grober et al. 1978), while salience is in line with “reference point” (van Hoek 1997) and “thematic-role preferences” (Stevenson et al. 1994). Kehler (ibid.: 34) distinguishes three types of relations between two connected clauses (“resemblance,” “cause-effect,” and “occasion”), which “predict otherwise puzzling behavior of five diverse linguistic phenomena” including pronoun interpretation.2 This tripartite distinction traces back to Hume’s (1748) three principles of connection among ideas (Resemblance, Cause or Effect, and Contiguity in time or place). Discussing many pairs of clauses with a pronoun in the second, Kehler argues that different mechanisms operate in different types of clausal relations to determine the pronominal reference. “Parallel function” dominates the pronominal reference when 2. The phenomena handled in Kehler (2002) include VP-ellipsis, gapping, extraction, and tense interpretation as well as pronoun interpretation.

 Katsunobu Izutsu and Mitsuko Narita Izutsu

the connected clauses have a resemblance relation. “Coherence” determines the reference with a cause-effect relation.3 “Salience” instantiated as subject (the most salient grammatical relation), topic (the most salient discourse entity), and patient/goal (most salient thematic roles) serves for the pronominal reference in occasion relations (Kehler 2002: 168).4 Kehler’s approach might be worth pursuit in tying to differentiate conditions (“resemblance,” “cause-effect,” and “occasion”) in which each of the three mechanisms earlier proposed operate. However, the mechanisms and conditions do not work well. “Parallel function,” alleged to operate on “resemblance,” is applicable to many other examples classifiable as “occasion” relations like (1a). “Coherence,” which Kehler claims to determine the pronominal reference in examples with “cause-effect” relations like (1c-d), fails to specify in what case the pronoun refers to the subject of the first clause and in what case it refers to the object. Furthermore, Kehler’s notion of occasion relation itself is so vague that it will also subsume some cases he can treat as resemblance and cause-effect. Sentences (1a-b) will be treated as examples of occasion, in which “salience” accounts for the pronominal reference. However, those sentences can be viewed as instantiating resemblance and cause-effect, where the pronominal reference could be handled in terms of parallel function and coherence. He would also treat both (2a-b) and (3a-b) as instances of occasion. As demonstrated in section 5, however, a mechanism close to parallel function is responsible for the pronominal reference in (2a) and (3a), while some kind of coherence accounts for the reference in (2b) and (3b).

3. Symmetric- and asymmetric-event descriptions The conceptions coded in the structure “S1. discourse connective + S2.” can be divided into two: symmetric and asymmetric descriptions of the two events coded in S1 and S2 (henceforth symmetric- and asymmetric-event descriptions, respectively). In the symmetric-event description, the speaker and addressee presuppose that one and the same entity does two things or two things are done to it. In the asymmetric-event description, the speaker and addressee expect that certain events will happen if the event coded in the first conjunct occurs. Sentences like (8) are classifiable as symmetric-event descriptions. The speaker and addressee presuppose that one and the same entity does two things or two things are 3. His cause-effect relations include “result,” “explanation,” “violated expectation,” and “denial of preventer” (Kehler 2002: 21). 4. Kehler’s treatment of patient and goal on a par with subject seems inconsistent. Subject is by definition the most salient grammatical function, while patient and goal are generally assumed less likely than agent to be coded in the subject position. He appears to mix up different degrees or kinds of salience.



Notional asymmetry in syntactic symmetry 

done to it. In (8a), for example, John did two things: hitting Bill and crying. In (8b), two things were done to Bill. The two things can either occur sequentially as in (8a) or simultaneously as in (8d). In (8b-c), both sequential and simultaneous readings are possible. (8) a. b. c. d.

John hit Bill. And he cried. (he=John) [=1a] John hit Bill. And Sarah kicked him. (him=Bill) [≑4a] John hit Bill. And he kicked Sarah. (he=John) [≑4b] Simon bullied Dave. In contrast, Ken sympathized with him. (him=Dave) [=5a]

The symmetric-event descriptions can be diagramed as in Figure 1 (a). The crosshatched circle represents the referent of the pronoun, John in (8a) and Bill in (8b). It can either be in the subject as in (8a) or the object as in (8b) (including a prepositional object in the verbal complement as in (8d)). The two circles on the right side of the diagram are other entities involved: John and Sarah in (8b) and Bill and Sarah in (8c). One of the circles is inapplicable in (8a): crying needs only one participant. The lines labeled c1 and c2 correspond respectively to the events coded in the first and second conjuncts. They can be elaborated into right-headed arrows as in Figure 1 (b) when the cross-hatched circle induces the designated events. Or they can be specified as left-headed arrows as in Figure 1 (c) when the two circles on the right initiate the events. Yet the two arrows are in the same direction because the overall conception is symmetric: both rightward as in (b) or both leftward as in (c). (a)

(b)

(c)

c1 c2

Figure 1.  conceptual structures of symmetric-event descriptions

On the other hand, sentences like (9) are classed as asymmetric-event descriptions. The speaker and addressee share an expectation that certain events will (and others will not) happen if the event coded in the first conjunct occurs. (9) a. b. c.

John hit Bill. So he cried (he=Bill) [=1c] John hit Bill. But he cried. (he=John) [=1d] Simon bullied Dave. Nevertheless, Ken sympathized with him. (him=Simon or Dave) [=5b]

In (9a-b), for instance, the speaker and addressee have an expectation that Bill rather than John will cry if John hits Bill. The connective so informs the addressee that the

 Katsunobu Izutsu and Mitsuko Narita Izutsu

content of the second conjunct meets that expectation, whereas the connective but conveys that the content of the second conjunct is contradictory to the expectation. The asymmetric-event descriptions can be sketched as in Figures 2 (a)-(b): the conception of sentences like (9a) corresponds to Figure 2 (a) and that of sentences like (9b-c) corresponds to Figure 2 (b). The bold-faced structures designate the profiles of each sentence: the lines labeled c1 and c2 stand respectively for the events coded in the first and second conjuncts. The cross-hatched circle represents the entity coded in the pronoun: Bill in (9a) and John in (9b). The dotted line indicates correspondence between their connected entities. The smaller rectangle encloses the speaker and addressee’s expectation. In (9a), the content of the second conjunct, c2, lies within that expectation as seen in Figure 2 (a), but in (9b-c), the content is outside it as seen in Figure 2 (b). Specifically, the conception involved in (9a), (9b), and the Simon- and Dave-readings of (9c) are diagramed as in Figures 2 (e), (d), (j), and (h), respectively. (a)

(b) c2 c1

c1 c2

(c)

(e) (e.g., (9a))

(g)

(i)

(d) (e.g., (9b))

(f)

(h) (e.g., (9c) D-reading)

(j) (e.g., (9c) S-reading)

Figure 2.  conceptual structures of asymmetric-event descriptions

The pronoun occupies, and thus the cross-hatched entity is coded in, either the subject as in (9a-b) or the object as in (9c) (including a prepositional object in the verbal complement); accordingly, c2 is specified as either a right-headed arrow as in Figures 2 (c-f), exemplified in (9a-b), or a left-headed arrow as in Figures 2 (g-j), illustrated in (9c). Likewise, c1 is also specifiable as either a right-headed arrow as in Figures (c-d) and (i-j), exemplified in (9b), or a left-headed arrow as in Figures 2 (e-h), illustrated in (9a). The direction of the arrows depends upon which of their connected circles induces the designated event. The arrows that correspond to c1 and c2 are not necessarily



Notional asymmetry in syntactic symmetry 

in the same direction as observable in Figures 2 (e-f) and (i-j) because the overall conception is asymmetric by nature.

4. Connective-driven interpretations of pronouns Sentences discussed so far are all in a coordinate structure, thus all syntactically symmetric. However, they can notionally be further distinguished into symmetric- and asymmetric-event descriptions. Discourse connectives often serve to specify whether the sentences should be interpreted to be symmetric- or asymmetric-event descriptions. As mentioned earlier, symmetric-event descriptions presuppose that one and the same entity does two things or two things are done to it, and asymmetric-event descriptions involve an expectation that certain events will happen if the first conjunct event occurs. In reference to the presupposition or expectation, the addressee chooses an appropriate referent of the pronoun in the second conjunct. In this respect, the phenomenon discussed thus far can be characterized as connective-driven pronominal interpretations. Connectives like and and in contrast mark sentences (1a) and (8b-d) as symmetric-event descriptions, which presuppose that one and the same entity does two things or two things are done to it. In reference to the presupposition, the addressee chooses to identify the referent of the pronoun with that entity. (1)

a. b. c. d.

(8) b. c. d.

John hit Bill. And he cried. (he=John) John hit Bill. And he cried. (he=Bill) John hit Bill. So he cried (he=Bill) John hit Bill. But he cried. (he=John) John hit Bill. And Sarah kicked him. (him=Bill) John hit Bill. And he kicked Sarah. (he=John) Simon bullied Dave. In contrast, Ken sympathized with him. (him=Dave)

(9) c. Simon bullied Dave. Nevertheless, Ken sympathized with him. (him=Simon/Dave) The connective and functions as a symmetric-event description marker in (1b) as well. In this case, however, the stressed pronoun resists the connective-driven (symmetricevent) interpretation in that it rather favors another interpretation such that its referent is an entity with lower accessibility (Ariel 1988: 79). In this respect, the stressed pronoun might be characterizable as another asymmetric-event description marker. The speaker and addressee’s shared expectation distinguishes asymmetric-event from symmetric-event descriptions. Connectives like so, but, and nevertheless mark sentences (1c-d) and (9c) as asymmetric-event descriptions, where the speaker and addressee expect that certain events will happen if the event coded in the first conjunct

 Katsunobu Izutsu and Mitsuko Narita Izutsu

occurs. Connectives of cause or reason like so suggest that the content of the second conjunct lies within that expectation as diagramed in Figure 2(a), while concessive connectives like but and nevertheless imply that the content lies outside the expectation as diagramed in Figure 2(b). The entity located inside the expectation achieves high accessibility, whereas the one located outside it only has lower accessibility.5 Unstressed pronouns serve as high accessibility markers, while stressed pronouns function as lower accessibility markers (Ariel 1988). In reference to those understandings, the addressee identifies the referent of the unstressed pronoun with a participant in an event that lies inside the expectation as in (1c). He or she identifies the referent of the stressed pronoun as a participant in an event that lies outside the expectation as in (1d). Similar explanations also hold for the corresponding Japanese and Korean sentences (2a), (2c-d), (3a), and (3c-d). The speaker and addressee’s shared expectation distinguishes asymmetric-event from symmetric-event descriptions here as well. The connectives sosite and geurigo serve as symmetric-event description markers in (2a) and (3a). Connectives like sorede, geureseo, sikasi, and geureonde mark sentences (2c-d) and (3c-d) as asymmetric-event descriptions. The distinction between the topic marker (-wa and -neun) and the nominative marker (-ga and -ga/i) largely parallels that between the unstressed and stressed pronouns in English. The topic marker suggests that the referents of their preceding nominals have a higher-accessibility status, whereas the nominative marker implies that the referents have a lower-accessibility status. Sosite kare-wa nai-ta. (kare=‘John’) and he-top cry-past

(2) a. c. d.

Jon-wa Biru-o tatai-ta. John-top Bill-acc hit-past ‘John hit Bill. And he cried.’ Jon-wa Biru-o tatai-ta. John-top Bill-acc hit-past ‘John hit Bill. So he cried.’ Jon-wa Biru-o tatai-ta. John-top Bill-acc hit-past ‘John hit Bill. But he cried.’ (kare=‘John’)

(3) a. c.

Jon-eun Bil-eul chyeo-ss-da. Geurigo geu-neun uleo-ss-da. John-top Bill-acc hit-past-se and he-top cry-past-se ‘John hit Bill. And he cried.’ (geu=‘John’) Jon-eun Bil-eul chyeo-ss-da. Geureseo geu-neun uleo-ss-da. John-top Bill-acc hit-past-se therefore he-top cry-past-se ‘John hit Bill. So he cried.’ (geu=‘Bill’)

Sorede kare-wa nai-ta. (kare=‘Bill’) therefore he-top cry-past Sikasi kare-ga/?wa nai-ta. but he-nom/?top cry-past

5. The contrast between inside/outside-an-expectation statuses (i.e., “expected” and “unexpected”) can be safely equated with a kind of accessibility difference.



Notional asymmetry in syntactic symmetry 

d.

Jon-eun Bil-eul chyeo-ss-da. Geureonde geu-ga/?neun John-top Bill-acc hit-past-se but he-nom/?top ‘John hit Bill. But he cried.’ uleo-ss-da. (geu=‘John’) cry-past-se

In (2c) and (3c), the connective suggests that the pronominal referent lies inside the expectation and thus achieves higher accessibility; therefore, the reference is made by pronouns with a topic marker. In (2d) and (3d), by contrast, the connectives imply that the pronominal referent lies outside the expectation and thus has lower accessibility. Accordingly, pronouns with a nominative marker are used for the reference.

5. Another type of asymmetric-event description marker Our proposed approach to pronominal anaphora, based on the distinction between symmetric/asymmetric-event descriptions, does not only account for the data more or less problematic to the major earlier studies but also can solve another problem which still could not be fully dealt with by the earlier studies. This problem typically occurs with the Japanese and Korean discourse connectives suruto and geureoja, as illustrated in (2b) and (3b). (2) b. Jon-wa Biru-o tatai-ta. Suruto kare-wa nai-ta. (kare=‘Bill’) John-top Bill-acc hit-past and he-top cry-past ‘John hit Bill. And (thereupon) he cried.’ (3) b.

Jon-eun Bil-eul chyeo-ss-da. Geureoja geu-neun uleo-ss-da. John-top Bill-acc hit-past-se and he-top cry-past-se ‘John hit Bill. And (thereupon) he cried.’ (geu=‘Bill’)

Those connectives can be glossed as ‘and’ in English just like Japanese sosite and Korean geurigo shown in (2a-b) and (3a-b). However, they contrast with sosite and geurigo in their driven interpretation of the pronoun kare or geu. Sosite and geurigo indicate ‘John’ as the pronominal referent as seen in (2a) and (3a) but suruto and geureoja suggest ‘Bill’ as the referent as shown in (2b) and (3b). As far as the four connectives are treated likewise as meaning ‘and,’ any account fails to explain their driven different interpretations of the pronoun. In contrast, the symmetric/asymmetric distinction allows us to account for the pronominal interpretations. Suruto and geureoja differ in that they induce the addressee to make reference to an expectation that certain events will happen if the event coded in the first conjunct occurs. Therefore, they are identifiable as asymmetric- rather than symmetric-event description markers. Unlike the connectives of cause/reason (sorede and geureseo) and those of concession (sikasi and geureonde), they serve to

 Katsunobu Izutsu and Mitsuko Narita Izutsu

suspend the information of whether the content of the second conjunct lies inside or outside the expectation.6 In (2b) and (3b), just like (2c-d) and (3c-d), the speaker and addressee share an expectation that Bill rather than John will cry if John hits Bill. However, unlike the cases with sorede, geureseo, sikasi, and geureonde, the connectives suruto and geureoja do not specify whether the content of the second conjunct lies inside or outside the expectation. The topic marker or nominative marker helps to specify which. The topic markers -wa and -neun suggest that the referents of their preceding nominals have a higher accessibility status as noted in section 4. In (2b) and (3b), therefore, kare-wa and geu-neun refer to ‘Bill,’ a highly accessible entity in that he is the most likely to cry in accordance with the expectation. However, the pronouns kare and geu refer to John if it is followed by a nominative marker -ga and -ga/i as in (10a-b). As mentioned in section 4, the nominative markers imply that the referents of their preceding nominals have a lower accessibility status. Hence, kare-ga and geu-ga refer to John, whose accessibility is lower in that he is less likely to cry than Bill is.7 6. The connectives suruto and geureoja (asymmetric ‘and’) are natural in such narrative contexts as below, while sosite and geurigo (neutral, symmetric ‘and’) are not. Suruto and geureoja serve a crucial role of suspending whether the consequence is as expected or not, which sosite and geurigo do not evoke. i.

Oziisan-wa sono kozuti-o hurimasita. Suruto/??sosite old:man-top that wooden:hammer-acc shook and/and doo desyoo. Kozuti-o huru-tabi kin-no koban-ga zakkuzaku. how will:be wooden:hammer-acc shake-time gold-gen coin-nom clank ‘The old man shook the hammer. And what happened next? Each shake brought coins clanking out of the hammer.’

ii.

Halabeoji-neun geu-yosulbangmangi-reul heundeuleosseumnida. old:man-top that-bar-acc shook Geureoja/?geurigo oddeohge doeeosseulggayo. Yosulbangmangi-reul and/and how became bar-acc heundeuleul-dde-mada geumhwa-ga jjallangjjallang. shake-time-by gold:coin-nom clank ‘The old man shook the hammer. And what happened next? Each shake brought coins clanking out of the hammer.’

7. In the Korean example, the pronoun could be interpreted as referring to Bill as well as John. The Korean nominative marker -ga/i differs from the Japanese counterpart in that it can mark nominals with a near-topical status which can otherwise be marked by -neun/eun: i. a. Igeos-eun/i eos-i-mnigga? this-top/nom what-be-se ‘What is this?’ b. Kore-wa/??ga an-desu-ka? this-top/nom what-be-se ‘What is this?’



Notional asymmetry in syntactic symmetry 

(10) a. b.

Jon-wa Biru-o tatai-ta. Suruto kare-ga nai-ta. (kare=‘John) John-top Bill-acc hit-past and he-nom cry-past ‘John hit Bill. And thereupon he cried.’ Jon-eun Bil-eul chyeo-ss-da. Geureoja geu-ga uleo-ss-da. John-top Bill-acc hit-past-SE and he-nom cry-past-se ‘John hit Bill. And thereupon he cried.’ (geu=‘John/Bill’)

The suruto/geureoja type of connectives are unique. Although similar in meaning to symmetric-event description markers like sosite and geurigo, they behave as either the connectives of cause/reason or those of concession with respect to pronominal anaphora. We believe that our approach is advantageous over others in that it can give a full account for the pronominal interpretations mediated by the unique suruto/geureoja type as well as the interpretation driven by more common types of connectives like and, so, and but, etc. in English, Japanese and Korean.

6. Conclusion The present paper demonstrated that the pronominal reference in the conjoined structure “S1. discourse connective + S2.” with a pronoun in the S2 cannot be fully accounted for in reference to previously proposed notions like “grammatical relations hierarchy” (Keenan & Comrie 1977 inter alia), “parallel function” (Sheldon 1974), and “thematic roles” (Stevenson et al. 1994; Stevenson et al. 2000), and that even an eclectic approach like Kehler (2002) fails to deal fully with the phenomena. We argued that the whole phenomena are fully dealt with by an account based on the high/low-accessibility marking and the connective-driven distinction between symmetric- and asymmetric-event descriptions. In symmetric-event descriptions, the pronominal referent, which does two things or to which two things are done, serves as the most salient entity; accordingly, it is referred to by the high accessibility marker (unstressed pronouns in English). In asymmetric-event descriptions, the pronominal referent can be inside or outside the relevant expectation. The entities that lie inside and outside it respectively achieve high and low accessibility. The high accessibility marker refers to the entity inside the expectation and the low accessibility marker ii. a. b.

Geu-saram-eun/i gu-mnigga? that-man-top/nom who-SE ‘Who is that person?’ Sono-hito-wa/??ga re-desu-ka? that-man-top/nom who-be-SE ‘Who is that person?’

This nature might be responsible for the blurred interpretation of the pronominal referent in (10b).

 Katsunobu Izutsu and Mitsuko Narita Izutsu

(stressed pronouns in English) to the entity outside it. The distinction between the presence and absence of a stress on English pronouns largely parallels that between the topic marker (-wa and -neun) and the nominative marker (-ga and -ga/i) in Japanese and Korean. We further showed that sentences with suruto and geureoja instantiate another type of asymmetric events, where the information is suspended of whether the content of the second conjunct lies inside or outside the expectation.

References Akmajian, Adrian and Ray Jackendoff. 1970. Coreferentiality and Stress. Linguistic Inquiry 1: 124–126. Ariel, Mira. 1988. Referring and Accessibility. Journal of Linguistics 24: 65–87. Cowan J. Ronayne. 1980. The Significance of Parallel Function in the Assignment of Intrasentential Anaphora. Chicago Linguistics Society (ed.), Papers from the Pronouns and Anaphora, pp. 110–124. Chicago: Chicago Linguistics Societly. Grober, Ellen H., William Beardsley, and Alfonso Caramazza. 1978. Parallel Function Strategy in Pronoun Assignment. Cognition 6: 117–133. Gundel, Jeanette, Nancy Hedberg, and Ron Zacharsky. 1993. Cognitive Status and the Form of Referring Expressions in Discourse. Language 69: 274–307. Izutsu, Mitsuko Narita. 2005. Contrast, Concessive, and Corrective: A Unifying Analysis of Opposition Relations in English. Ph.D dissertation. Sapporo: Hokkaido University. Hobbs, Jerry R. 1979. Coherence and Coreference. Cognitive Science 3: 67–90. Hume, David. 1748 [1955]. An Inquiry Concerning Human Understanding. New York: The Liberal Arts Press. Keenan, Edward L. and Bernard Comrie. 1977. Noun Phrase Accessibility and Universal Grammar. Linguistic Inquiry 8: 63–99. Kehler, Andrew. 2002. Coherence, Reference, and the Theory of Grammar. Stanford: Center for the Study of Language and Information. Lakoff, George. 1971. Presupposition and Relative Well-Formedness. In D. D. Steinberg and L. A. Jakovovits (eds.), Semantics: An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology. Cambridge: Cambridge University Press. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar vol. 2: Descriptive Applications. Stanford: Stanford University Press. Sheldon, Amy. 1974. The Role of Parallel Function in the Acquisition of Relative Clauses in English. Journal of Verbal Learning and Verbal Behavior 13: 272–281. Stevenson, Rosemary J., Rosalind A. Crawley, and David Kleinman. 1994. Thematic Roles, Focus and the Representation of Events. Language and Cognitive Processes 9 (4): 519–548. Stevenson, Rosemary, Alistair Knott, Jon Oberlander, and Sharon McDonald. 2000. Interpreting Pronouns and Connectives: Interactions among Focusing, Thematic Roles and Coherence. Language and Cognitive Processes 15 (3): 225–262. van Hoek, Karen. 1995. Conceptual Reference Points: A Cognitive Grammar Account of Pronominal Anaphora Constraints. Language 71 (2): 310–340. van Hoek, Karen. 1997. Anaphora and Conceptual Structure. Chicago and London: The University of Chicago Press.

part 2

Subordination, nominalization, modification

Subordination in cognitive grammar Ronald W. Langacker

University of California, San Diego

This paper reexamines the treatment of subordination in Cognitive Grammar. Subordination is not susceptible to simple, categorical description. Understanding it requires the elucidation of numerous factors, with respect to which one can describe its many types, distinguish them from one another, and properly relate them to other phenomena. Among the relevant factors are asymmetries in prominence, including profiling, trajector/landmark organization, and status as the important content in a discourse. Special attention is devoted to how these interact with limited “windows of attention” in an integrated account of grammar and dynamic processing.

Keywords: subordination, complement clause, relative clause, adverbial clause, profiling, trajector/landmark, constituency, mental spaces, intonation units, attentional frame, grounding, nominal/finite clause, acquisition, attention, analyzability, processing, discourse My basic position on subordination was written some 15 years ago (Langacker 1991: 419): Despite their familiarity, … such standard terms as coordination, subordination, relative clause, etc. do not necessarily refer to notions that are clearly defined or thoroughly understood, nor can they be accepted as representing an optimal, revelatory, or even adequate classificatory system. Indeed, it is doubtful that any single classification could accommodate the actual diversity of multiclausal constructions together with the many kinds and degrees of similarity displayed by overlapping subgroups. A better strategy is to examine individually the various factors that figure in a full characterization of such constructions; a particular construction is then defined by a constellation of properties, each of which is shared by certain others.

I still subscribe to this view, recognizing the inadequacy and non-unitary nature of the traditional categories. In the same chapter, I nonetheless suggested that a single, unified characterization might be possible for the full range of clauses traditionally regarded as subordinate,

 Ronald W. Langacker

including complement, relative, and adverbial clauses: “A subordinate clause is … describable as one whose profile is overridden by that of a main clause” (Langacker 1991: 436). Strictly speaking, I do not believe this passage contradicts the preceding one. It is rather an attempt to explicate the traditional notion of subordination – to offer a reasonably precise definition based on an independently justified construct (profiling) – to the extent that this notion is descriptively useful. Still, it does run counter to the spirit of the earlier passage. More importantly, certain questions arise concerning the adequacy of the proposed characterization. It is thus worth taking a closer look. Though it will not lead to a simple, definitive analysis, it may at least expose some fundamental issues. The definition of subordination was based on certain assumptions about profiling in complex sentences. These are exemplified in (1), where boldface indicates a clause whose profile is said to be inherited by the sentence as a whole. Coordination represents the special case of multiple, co-equal profiles. Taken as a whole, sentence (1)a does not profile either clausal process – that of lending or that of wasting – to the exclusion of the other. Thus neither clause is subordinate to the other. In (1)b, the sentence as whole is said to designate the act of deciding, not the conjectural status of evolution; whatever one believes concerning evolution, the truth or falsity of the sentence as a whole is solely based on what the president decided. In (1)c, the profile of the relative clause they brought him is overridden even at the level of the subject nominal: the crown they brought him designates the crown, not the act of bringing. And in (1)d, the event of going to bed (rather than getting home) is profiled at the composite structure level. (1) a. I lent him some money and he simply wasted it. [conjoined clauses] b. The president has decided that evolution is only conjecture. [complement clause] c. The crown they brought him was too small for his head. [relative clause] d. I went straight to bed when I got home. [adverbial clause]

A number of questions can be raised about this analysis. There is first the question of whether adverbial clauses are really subordinate. The traditional label “subordinating conjunction”, applied to elements like because, since, when, and if, suggests ambivalence in this regard, as well as the absence of any clear boundary between subordination and coordination. In recent years, substantive arguments have been put forth by Diessel and Tomasello (2001), Thompson (2002), and Verhagen (2005) against the subordinate status of complement clauses, especially from the standpoint of acquisition and use in spoken discourse. There is also room for doubt concerning profiling. While the need for this descriptive construct and the details of its application seem fairly clear at lower levels of structural organization, they are far less evident at higher levels. And finally,



Subordination in cognitive grammar 

appeal to this construct may not be necessary. There may be other ways to account for the perceived asymmetries which motivate the term “subordination”. While the term subordination is not self-explanatory, neither is it arbitrary. It reflects a perceived asymmetry in the status of clauses in a complex sentence. My characterization in terms of profiling was an attempt to capture this asymmetry in a unified manner. But perhaps the asymmetries in question are not in fact all comparable (or are comparable only at a higher level of abstraction). Profiling is not the only possible source of a difference in status reasonably described impressionistically with labels like “main” vs. “subordinate”. Another possibility is trajector/landmark alignment, i.e. the difference between the primary and secondary focal participants in a profiled relationship. Trajector/landmark alignment is inherent in elements like when, while, before, after, if, unless, because, although, and since, which introduce adverbial clauses. If not due to profiling, the perceived asymmetry between an adverbial clause and the clause it modifies may thus be ascribable to the latter being the primary focal participant in the relationship serving to connect them. Suppose it is argued that the clauses in (1)d are equally prominent in terms of profiling; in that respect, (1)d would be analogous to the coordinate structure in (1)a. It would still be the case that I went straight to bed is primary, and I got home secondary, with respect to the connecting relationship expressed by when: they respectively elaborate its trajector and its landmark. While the profiled events are equally focused at the composite structure level, getting home is nonetheless invoked as a temporal landmark for going to bed. (a)

(b)

X CONN+Y tr X

tr

CONN+Y lm

tr

X CONN+Y tr tr

X

CONN+Y tr

tr

tr

lm CONN

tr Y

tr lm CONN

tr Y

Figure 1

The alternative analyses are presented in Figure 1, where X and Y are the component clauses, and CONN a connecting element. Diagram (a) represents the analysis I previously assumed as the basis for defining subordination. It implies that Y is a complement of CONN, which functions as the head (i.e. the profile determinant) at the lower

 Ronald W. Langacker

level of organization, while at the higher level the sequence CONN+Y serves to modify X, which is the head at that level and for the sentence as a whole. With respect to (1) d, it implies that when I got home profiles the when relationship, and that the entire sentence profiles the act of going to bed. The alternative is to claim that both events are profiled at the composite structure level. This implies the analysis sketched in diagram (b). The correspondences are the same as in diagram (a): in both structures, Y elaborates the landmark of CONN, and X its trajector. However, the sequence CONN+Y is presumed to profile the clausal process (not the connecting relationship). And at the higher level of composition, both components contribute their processual profiles to the composite structure – if, indeed, there is such a structure (a matter we will return to). Turning now to complement clauses, we can also find an alternative to profiling as the basis for perceived asymmetry. My characterization of subordination presupposed the analysis sketched in Figure 2(a) for a complement clause construction. The complement clause elaborates the trajector or the landmark of the process profiled by the matrix clause, which functions as profile determinant. In diagram (a), Doris left specifies the landmark of Chris knows, whose profile prevails at the composite structure level. Chris knows Doris left is said to designate the process of knowing, not the act of leaving. However, the component clauses display an evident asymmetry independently of any claims about profiling. The clause Doris left specifies a central participant in the relationship coded by Chris knows, but not conversely. At the conceptual level, this translates into the layering of mental spaces: know is a space-builder (Fauconnier 1985), and the event of Doris leaving occupies the space it evokes. Perhaps, then, a complement clause is subordinate to the matrix clause in the sense that the relationship it designates is embedded in a mental space which the matrix clause introduces. As shown in diagram (b), this layering constitutes a conceptual asymmetry even if one assumes that both clausal processes are profiled at the composite structure level. (a)

Chris knows Doris left lm tr C

C tr lm Chris knows

D

D tr Doris left

(b)

Chris knows Doris left tr C

C tr lm Chris knows

lm

D

tr

D tr Doris left

Figure 2

As in Figure 1(b), I have enclosed the composite structure in a dashed-line box because I have some doubt that we need to posit it as part of the grammatical structure



Subordination in cognitive grammar 

of such expressions. Perhaps in terms of grammar (though not in terms of meaning) we can simply omit the topmost structure both in Figure 1(b) and in 2(b). The matter hinges on issues that have barely begun to be explored, let alone resolved. At this stage I can only try to indicate the issues I consider to be most fundamental. Grammar, I claim, consists in assemblies of symbolic structures (form-meaning pairings). It is certainly not the case that these assemblies are always strictly hierarchical, in the manner of Figure 1(a). It is neither required nor assumed that the symbolic structures in an assembly form a single, consistent constituency hierarchy analogous to the classic tree structures of generative syntax. In particular, it is not assumed that a symbolic assembly is always graced by a single composite structure (at the “top”) subsuming all the other symbolic structures (directly or indirectly) as components. Constituency of the “classic” sort is non-essential and therefore often variable (Langacker 1995, 1997). A classic constituent emerges when a particular kind of conceptual grouping (based on a correspondence between focused elements) happens to be symbolized by a particular kind of phonological grouping (based on linear contiguity). A classic constituency hierarchy emerges when this occurs at multiple levels of organization, the grouped (or composite) structure at one level functioning as a component structure at the next. However, such hierarchies are not exhaustive of the structures and relationships inherent in complex expressions. One argument for this flexible view of constituency is that it straightforwardly accommodates the intonational groupings commonly observed in complex sentences, e.g. in (2). Almost to a person, syntacticians posit the constituency shown in (2)a, a strictly nested structure where each “embedded” clause is literally “contained” in the clause whose landmark it specifies. Phonologically, however, each clause is most naturally produced as a separate intonational unit, as shown in (2)b. To handle this discrepancy between the presumed syntactic organization and the observed phonological organization, Chomsky and Halle (1968: section 6.5) were forced to posit ad hoc “phonological adjustment rules”. But there is no need for such devices if grammatical structure resides in symbolic assemblies rather than rigid constituency hierarchies. One can simply say, instead, that the hierachical organization in (2)a is a matter of conceptual structure – the successive embedding of mental spaces – but is not reflected in the expression’s phonological or grammatical organization. (2) a. [Amy says [Bob thinks [Chris knows [Doris left]]]] b. Amy says / Bob thinks / Chris knows / Doris left It would be both gratuitous and descriptively problematic to assume that, in assembling symbolic structures to form a complex expression, we are limited to building a strictly hierarchical structure where an integrated composite conception predominates at each successive level. Another form a symbolic assembly can take is a chain-like structure, based on local connections between successive windows of attention. This kind of serial organization is characteristic of the “intonation units” discussed by Chafe (1994). While flexible in their application, these units tend to coincide with clauses. I

 Ronald W. Langacker

have suggested that such a unit can be characterized schematically as the symbolic pairing between a conceptual window of attention and a phonological intonation group (Langacker 2001a). The expression in (2)b can then be described as a chain-like symbolic assembly where each successive clause is identified with an attentional frame of this sort. This is sketched in Figure 3, where a box with rounded corners represents an attentional frame, and the arrow labeled T stands for processing time.

A

B

>

Amy says

C

>

Bob thinks

D

Chris knows

>

Doris left

T

Figure 3

I am not suggesting that this sequence of attentional frames represents everything that is going on, either linguistically or in processing terms. Language processing occurs simultaneously at multiple levels of organization, on different time scales, and with varying degrees of conscious awareness (Langacker 2001b). In proceeding through a sentence like (2)b, focusing in turn on each successive clause, we also apprehend each in relation to its predecessor, as shown in Figure 4(a). At least implicitly, moreover, we are building up a global conception of the full mental space configuration, as shown in 4(b). But there is no particular reason to assume that, at every step along the way, an overarching composite structure emerges which subsumes the content of all the preceding stages and imposes a single profile on the entire conception. And in general, I do not believe that any such structure is descriptively necessary. (a) A

>

A

>

B

B

>

C

C

D

(b) A

>

A

B

>

A

B

C

>

A

B

C

D

Figure 4

Should an account along these lines prove viable, it would still capture the asymmetry which motivates the traditional notion that a complement clause is “subordinate” to the matrix clause. Even if there is no constituency hierarchy with a single “highest” clause and a single overall profile, a complement clause is still subordinate in the sense



Subordination in cognitive grammar 

that it specifies a central participant in the matrix clause. Moreover, the matrix clause is the “main” clause in the sense that it represents the most accessible mental space, i.e. the one most directly connected to the speaker’s conception of reality. I should emphasize that this alternative does not constitute an abandonment of the notion profiling. The only thing at stake is the proper range of application for this notion – a single construct cannot and should not do everything. At lower levels of grammatical structure, it is generally fairly clear which element is profiled, according to the definition that an expression’s profile is what it designates (its conceptual referent). In English compounds, for example, the second element imposes its profile on the whole. Thus hairspray designates a kind of spray (not hair); hairspray can refers to the can; and within the conceptual complex it evokes, hairspray can label profiles the label. Similarly, Sally’s desk refers to the desk (not to Sally); on Sally’s desk profiles the locative relationship; while the lamp on Sally’s desk designates the lamp. And the process of liking is profiled by really like, really like it, and the entire finite clause most people really like it. By contrast, it is not at all self-evident (as assumed in Langacker 2001a) that a sentence like (2)b designates just the act of saying, or that conditionals of the form if X then Y profile just situation Y. Nor is it clear, even in principle, precisely what it means to say that complex sentences like these have a single overall referent. There has to be some upper limit on the size of the structures to which this notion is applicable (e.g. it cannot be applied to an entire discourse or a novel). Of course, the definition of profiling might prove to be a special case of a more abstract notion that would be applicable to higher-level structures. But that would require a specific proposal tangible enough to investigate. In short, while lower-level structures strongly support the adoption of profiling as a descriptive construct, the extent of its proper use is not apparent apriori, but has to be determined by careful investigation taking a wide range of considerations into account. Its use in defining subordination may very well prove to be erroneous. A natural limit for profiling – the domain in which its application is unproblematic – would be a single grounded structure, i.e. a full nominal or a finite clause. Grounding is itself a matter of reference: singling out a particular instance of a thing or process type in relation to the interlocutors and the speech event. If profiling represents a general notion of conceptual reference, grounding represents the special case of deictic reference. Admittedly, though, the definition of profiling is more impressionistic than it is a precise psychological characterization. That will require a much better understanding of the varied phenomena which go by such labels as “attention”, “prominence”, and “focusing”, and how they relate to one another. A nominal or a finite clause is a natural domain for profiling because it contains its own internal vantage point (the ground). As such, a grounded structure has the potential to be apprehended independently, in the manner of Figure 3, as opposed to being viewed solely in relation to another. It would however be too strong to claim that no grounded structure is ever incorporated as part of another, i.e. that its content is never

 Ronald W. Langacker

subsumed by a higher-level composite structure which imposes a different profile. That seems problematic for many structures, including the following. In possessives, such as Jack’s wife, the possessor is a full, grounded nominal, but so is the higher-level structure containing the noun it modifies, which imposes its profile on the whole. Since Jack is part of the nominal Jack’s wife, and the composite expression designates the wife, the profile of Jack is overridden at the higher level. Likewise, a relative clause can form a constituent with the noun it modifies, with the latter functioning as head (or profile determinant): the referent of the proposal I made is the proposal rather than the process of making it. And a finite clause profiles a process rather than a thing, even though it incorporates nominal arguments, e.g. the process admire in I admire her. A more defensible position is that a grounded structure is a natural candidate to be apprehended independently, in its own attentional frame, and that grounding diminishes the likelihood of a structure being incorporated as part of another with its profile overridden. This is a matter of degree, and certainly it interacts with other factors. In particular, clauses tend more strongly than nominals to retain their independence. This no doubt reflects their greater internal complexity, as well as their status as the basic units of discourse. Although nominals are conceptually autonomous, in and of themselves they do little by way of moving the discourse along – their typical role is rather to support the conception of relationships, which primarily fulfill this function. I am suggesting, then, that at higher levels of organization both constituency and profile determinance are variable. Consider the relative clauses in (3). Proceeding from (3)a to (3)e, the conceptual and structural integration of the relative clause and the noun it modifies becomes progressively looser. In (3)a, that book I read represents a canonical example of the standard syntactic account of relative clauses: it is a single constituent referring to a book which the relative serves to identify (hence it is quite analogous to that book on the floor). The relative construction in (3)b is also canonical, but here the relative clause is less in the shadow of the noun it modifies. It is more complex, it makes a greater, less predictable semantic contribution, and prosodically it is unreduced (in contrast to the relative in (3)a, which is unstressed and rhythmically compressed). The relative clause in (3)c has a greater measure of independence, in that its content is really not used to identify the book, but only to further characterize it. Indeed, its content is the main point of the sentence. In (3)d, the relative is still more independent, as it is not even adjacent to the noun it modifies. Hence they do not form a classic grammatical constituent. Finally, the second clause in (3)e is a canonical non-restrictive relative, which is generally accepted as being external to the nominal it characterizes. This is often likened to a coordinate construction, implying separate and equal profiles. (3)

a. b. c. d. e.

There were some really outrageous claims in that book I read. A book which makes outrageous claims is often a best-seller. I just read a book which makes some outrageous claims. I read a book last night which makes some outrageous claims. I just read a book, which makes some outrageous claims.



Subordination in cognitive grammar 

The extreme cases, the relative constructions in (3)a and (3)e, are roughly sketched in Figure 5. In the first case, the relative clause is subordinate as originally defined: it combines with the head to form a composite symbolic structure which profiles the book rather than the clausal process. In the second case, the components are clauses which retain their independence, as in Figure 3. Each constitutes a separate attentional frame, so there is no single window of attention which subsumes their content but selects just one component process for its profile. While I am not ready to venture a detailed characterization of all the intermediate cases, it may at least be apparent that relative clauses have variable implementation in terms of symbolic assemblies, constituency, and profiling. (a)

(b)

that book I read B

I

I tr I read

lm

B

I B tr lm I just read a book

wh C tr lm which makes some outrageous claims

that book

Figure 5

Comparable remarks can be made for adverbial and complement clauses. In (4)a-e, for example, the adverbial introduced by after becomes progressively more elaborate and more plausibly describable as occupying a separate attentional frame with its own independent profile. Sentence (4)c might be considered transitional, since the afterclause is finite but not set off as a separate intonational group by the slight pause usually written as a comma. Observe that both pronunciations in (5) are quite natural, depending on whether the second clause constitutes old or new information. Compared to (5)b, the after-clause in (5)a is prosodically compressed in several respects: it has no word with unreduced stress (small caps); it is lower in pitch; and it is pronounced more rapidly. We might speculate that this phonological compression correlates with, or even symbolizes, a conceptual subordination of the adverbial clause whereby its profile is overridden. The pronunciation in (5)b would then correlate with both clauses retaining their processual profiles at the composite structure level. (4)

a. b. c. d. e.

He took a nap after lunch. He took a nap after eating. He took a nap after he ate. He took a nap, after he ate lunch. After he ate an enormous lunch, he took a nap.

 Ronald W. Langacker

(5) a. He took a nap after he ate. b. He took a nap after he ate. As for complement clauses, I fully agree with Thompson (2002) and Verhagen (2005) that complements tend to be primary in the sense of providing the most important content. I further agree with Diessel and Tomasello (2001) that, in acquisition and to some extent in adult language use, the matrix clause tends to be apprehended holistically and only as an appendage to the complement, so that only the complement expresses a proposition. They recognize, however, that these are matters of degree and subject to variation – a point that is argued quite forcefully by Boye and Harder (2007). Whatever the proportion, it would be as wrong to claim that a matrix clause never expresses the proposition of primary interest as it is to claim that it always does. The problem is to come up with a unified description that accounts in a natural and revealing way for the full range of options. The examples in (6) afford some idea of the range. In (6)a, it is certainly the matrix clause that expresses the proposition of primary interest. One can plausibly maintain that the complement’s profile is overridden by that of the matrix clause in a composite conception viewed in a single window of attention. In (6)b, the clauses would seem to be of roughly equal status as informative propositions. In (6)c the balance tips in favor of the complement, and in (6)d I think is clearly an appendage to it, an epistemic afterthought. It is phonologically reduced to the point that it can hardly be taken as representing a separate window of attention. For (6)d I would posit an overall composite structure in which the profile of think is overridden. (6) a. Is evolution only conjecture? Well, the president has definitely decided that it is. b. Bush has conclusively demonstrated that evolution is only conjecture. c. I suspect that evolution is only conjecture. d. Evolution is only conjecture, I think.

Here cases like (6)c may be transitional. Note the alternate pronunciations in (7): the matrix clause can either be fully realized phonologically, with normal focus stress, or else compressed in the same manner as the adverbial clause in (5)a and the appendage in (6)d. Possibly this marks the transition between a composite conception where both clauses retain their profile and one where the matrix clause profile is overridden. (7) a. I suspect that evolution is only conjecture. b. I suspect that evolution is only conjecture. Wherever the transition occurs, and whether it is abrupt or gradual, at some point along the spectrum the matrix clause loses any vestige of prominence and becomes decidedly secondary to the complement. Thompson describes the matrix as having an “epistemic/evidential/evaluative” function. Diessel and Tomasello describe it (for young children) as an “epistemic marker”, an “attention getter”, or a “marker of illocu-



Subordination in cognitive grammar 

tionary force”. Accepting that something along these lines properly characterizes one end of the spectrum, the ultimate descriptive challenge is to provide an explicit account of all the varied structures that constitute it, and how they relate to one another synchronically, diachronically, and developmentally. Part of the challenge is to determine precisely what descriptive constructs are necessary and how to optimally apply them. In particular, we need to ascertain – for those cases where the matrix is clearly secondary to the complement – just what kind of prominence the latter has by virtue of which the matrix is secondary to it. I would hypothesize that the central construct is indeed to be identified as profiling. Depending on the factors I have mentioned, and no doubt others, a matrix clause and its complement can be integrated in various ways, as shown in Figure 6. Either clause can impose its profile on the composite conception, or alternatively, each can maintain its status as a momentary focus of attention. It is ultimately a matter of timing and how much content is subsumed in a single window of attention. Since it is hard to deal with simultaneous foci of attention, it is easiest for both clauses to be profiled when each occupies its own attentional frame, as in diagram (b). While an integrated conception may well emerge, the primary focus is at the component structure level. If two processual profiles are squeezed into a single attentional frame, e.g. for expressions like (7)a, I presume they are nonetheless accessed in sequence. Diagrams (a) and (c) show the primary focus at the composite structure level. It is easier for the full, integrated conception to fit in a single window of attention if only one component process retains its profile at this level. (a)

(b)

(c)

Figure 6

If a finite clause is a natural candidate to be apprehended independently, in its own attentional frame, the configurations in Figures 6(a) and (c) are less than fully optimal in this respect. I suggest, however, that both are quite natural in another respect: each is the analog, at a higher level of organization, of a basic aspect of the structure which finite clauses canonically exhibit internally (cf. Achard 1998; Langacker 2004). In the case of 6(a), the multi-clause configuration approximates the single-clause structure of a verb and its nominal arguments. To varying degrees, the complement is reduced in size and becomes less clause-like, to the point that it is almost like a pronoun, as seen in (8).

(8) a. Bush most certainly believes it’s only conjecture.

 Ronald W. Langacker

b. Bush most certainly believes it is. c. Bush most certainly believes it. As for Figure 6(c), I perceive a similarity to clausal grounding. Internally, a finite clause profiles the grounded process, not the grounding relationship. I have argued (Langacker 2002) that this is also true of the grammaticized grounding elements (for English, tense and the modals). In diagram (c), where each component is itself a grounded clause, the composite conception shows the analogous organization in which an unprofiled relationship serves to epistemically qualify a profiled process. For instance, the relation which may bears to like in (9)a is analogous to that which I think bears to he likes her in (9)b. (9) a. He may like her. b. I think he likes her. If the analogy is valid, it provides an argument for the hypothesis that profiling is the pivotal construct for describing different positions along the spectrum, in the manner of Figure 6. We could then go on to define the notions “main” and “subordinate clause” in discourse terms: a main clause is one whose profile is maintained as such, and a subordinate clause one whose profile is overridden, in a window of attention. Perhaps ironically, these are essentially the same definitions we started with, the only difference being that profiling is considered in relation to a variable window of attention instead of a composite structure presumed to be invariant. But if the definitions are the same, the extensions are not. The clauses identified as “main” and “subordinate” – in discourse terms – coincide with those traditionally labeled as such only toward one end of the spectrum. The discourse definitions accord quite well with the characterization offered by Thompson and by Diessel and Tomasello for expressions at the opposite end. Nor do I seriously disagree with their characterization of the matrix as being formulaic and unanalyzable. I have been assuming, for sake of discussion, that the matrix clause retains its analyzability. But this is undoubtedly a matter of degree. In the case of acquisition, originally holistic elements become analyzable to various extents as the child gradually masters the grammar of complex sentences. Diachronically, we might posit the opposite development, where independent clauses gradually lose their independence and their analyzability. In the acquisition process they would then be analyzed to a lesser and lesser degree, eventually losing their clausal status altogether. Whatever the details, acquisition and grammaticization are essential components of an overall account (cf. Boye and Harder 2007). In sum, I have proposed a slightly revised account of subordination in Cognitive Grammar. While it still identifies profiling as the pivotal notion, complex sentences are seen as being flexible in regard to what is profiled, as determined by discourse and processing factors. These proposals are at best exploratory. They might well be criticized for an inappropriate mixing of considerations that properly belong to the sepa-



Subordination in cognitive grammar 

rate domains of grammar, processing, and discourse. I am certainly mixing these considerations, but I doubt very much that this is inappropriate. I do not believe that grammar, in all its actual complexity, is coherently describable independently from processing and discourse, or that those are neatly distinct from one another. I believe we need an integrated account of all three domains for any of them to be properly understood.

References Achard, M. 1998. Representation of Cognitive Structures: Syntax and Semantics of French Sentential Complements [Cognitive Linguistics Research 11]. Berlin/New York: Mouton de Gruyter. Boye, K. and Harder, P. 2007. Complement-taking predicates: Usage and linguistic structure. Studies in Language 31: 569–606. Chafe, W. 1994. Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago/London: University of Chicago Press. Chomsky, N. and Halle, M. 1968. The Sound Pattern of English. New York: Harper & Row. Diessel, H. and Tomasello, M. 2001. The acquisition of finite complement clauses in English: A corpus-based analysis. Cognitive Linguistics 12: 97–141. Fauconnier, G. 1985. Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, MA/London: MIT Press. Langacker, R.W. 1991. Foundations of Cognitive Grammar, vol. 2, Descriptive Application. Stanford: Stanford University Press. Langacker, R.W. 1995. Conceptual grouping and constituency in cognitive grammar. In Linguistics in the Morning Calm 3, I.-H. Lee (ed.), 149–172. Seoul: Hanshin. Langacker, R.W. 1997. Constituency, dependency, and conceptual grouping. Cognitive Linguistics 8: 1–32. Langacker, R.W. 2001a. Discourse in cognitive grammar. Cognitive Linguistics 12: 143–188. Langacker, R.W. 2001b. Dynamicity in grammar. Axiomathes 12: 7–33. Langacker, R.W. 2002. Deixis and subjectivity. In Grounding: The Epistemic Footing of Deixis and Reference [Cognitive Linguistics Research 21], F. Brisard (ed.), 1–28. Berlin/New York: Mouton de Gruyter. Langacker, R.W. 2004. Aspects of the grammar of finite clauses. In Language, Culture and Mind, M. Achard and S. Kemmer (eds.), 535–577. Stanford: CSLI Publications. Thompson, S.A. 2002. ‘Object complements’ and conversation: Towards a realistic account. Studies in Language 26: 125–164. Verhagen, A. 2005. Constructions of Intersubjectivity: Discourse, Syntax, and Cognition. Oxford: Oxford University Press.

Asymmetric events, subordination, and grammatical categories Sonia Cristofaro University of Pavia

The notion of asymmetric events plays a central role in many theories of subordination. Subordination has been associated with various types of event asymmetry, such as foreground-background, figure-ground, asserted vs. nonasserted information, and presence vs. absence of an autonomous cognitive profile for individual events. This paper discusses the consequences of this view for the definition of subordination as a grammatical category. It is shown that event asymmetry as such does not correspond to any consistent constructional class, either cross-linguistically or within individual languages. As a result, event asymmetry should be kept distinct from subordination as a grammatical category. Event asymmetry is a component of a number of different constructions and provides a sound basis for comparing these constructions both cross-linguistically and within individual languages. However, these

constructions correspond to different grammatical categories, any of which may be arbitrarily labeled as subordination. These categories are languagespecific and construction-specific.

Keywords: subordination, assertiveness, grammatical categories, cognitive profile, foreground-background, figure-ground, cross-linguistic comparison, typology, constructions, embedding

1. Introduction The notion of asymmetric events plays a central role in many theories of subordination. For example, it is often argued that subordination involves a pragmatic asymmetry between the events expressed by two linked clauses, in that these events correspond to either asserted vs. nonasserted information (Harris and Campbell 1995: chap. 9, among others), or backgrounded vs. foregrounded information respectively (Reinhart 1984, Thompson 1987, and Tomlin 1985). Talmy (2000: chaps. 5–6) and Croft (2001: chap. 9) argue that subordination is typically associated with perceptual asymmetry between events, as manifested in the figure-ground distinction. Langacker (1997) and

 Sonia Cristofaro

Cristofaro (2003) view subordination as a situation of cognitive asymmetry where an event lacks an autonomous cognitive profile, and is construed in the perspective of another event1. This paper discusses the consequences of this view for the definition of subordination as a grammatical category. By grammatical category is meant here, based on Dryer’s definition of grammatical relations (Dryer 1996 and 1997), not just any category that could be defined for descriptive purposes, but a category that plays a role in the grammar of particular languages and is arguably part of a speaker’s knowledge of their language. In some of the aforementioned proposals, event asymmetry is a feature of particular sentence types, which are regarded as instances of subordination independently of event asymmetry as such. The criteria on which these sentence types are classified as instances of subordination, however, do not make it possible to define an internally consistent constructional class, either cross-linguistically or within individual languages. In addition to that, individual sentence types may not always encode asymmetric events. This challenges the assumption that there is an independently definable grammatical category of subordination that is connected with event asymmetry. Other proposals assume that subordination should be identified with event asymmetry, independently of the sentence types in which the latter is manifested. Crosslinguistic as well as intralinguistic investigation shows however that there is no particular morphosyntactic feature that is systematically associated with event asymmetry as such from one construction to another. In fact, while some languages use different morphosyntactic structures to encode the same semantic relation between events depending on whether asymmetric or symmetric events are involved, other languages make no such distinction at the morphosyntactic level, and the symmetric vs. asymmetric nature of the relevant events is inferred from the context. Thus, event asymmetry as such does not correspond to any consistent constructional class, either crosslinguistically or within individual languages. This suggests that there is no cross-linguistic and cross-constructional grammatical category that can be defined in terms of event asymmetry. The paper will argue that the logical consequence of these facts is that event asymmetry should be kept distinct from subordination as a grammatical category. Event asymmetry is a component of a number of different constructions and provides a sound basis for comparing these constructions both cross-linguistically and within individual languages. However, just like what has been argued by Dryer (1996, 1997), Van Valin and LaPolla (1997: chap. 6), and Croft (2001) for a number of grammatical relations and categories, these constructions correspond to different grammatical 1. The term `event’ will be used in this paper as a cover term for a number of entities that, depending on their semantic features, are usually indicated as `events’ proper, `states’, `situations’ and the like. In fact, `event’ will be used as a synonim of the Functional Grammar term `state of affairs’, i.e. the conception of something that can be the case in some world, as defined by a nuclear predication consisting of a predicate and its terms (Dik 1997).



Asymmetric events, subordination, and grammatical categories 

categories, any of which may be arbitrarily labeled as subordination. These categories are language-specific and construction-specific.

2. Event asymmetry as a feature of particular sentence types Over the past decades, a number of proposals have been put forward that associate subordination with pragmatic or perceptual asymmetry between events. Although these proposals differ in their details, they all appear to share a basic assumption that the events encoded by subordinate clauses, contrary to those encoded by independent clauses, represent information that is not introduced in discourse in its own right and is not open to challenge. For example, it is usually argued that complement, adverbial, and relative clauses, contrary to main clauses, are not asserted. In some cases, this refers to the semantics of the sentence, that is, the fact that the propositions encoded by these clauses are not within the scope of the speaker’s commitment. As a result, the truth value of these clauses is not affected by sentential negation or interrogation (Harris and Campbell 1995: chap. 10, among others). In other cases, this refers to the pragmatics of the sentence, that is, the fact that complement, adverbial, and relative clauses encode information that is not open to challenge because the hearer is expected to know or take it for granted at the moment the sentence is uttered. For instance, Lambrecht (1994: 51) argues, by uttering the sentence

(1) I finally met the woman who moved in downstairs

the speaker assumes that the addressee already knows that someone moved in downstairs, and wishes to inform the addressee that s/he finally met this person (for an illustration of the differences between the semantic and the pragmatic sense of the notion of assertion, see Lambrecht 1994: 61–5 and Levinson 1983: 199–225). The nonasserted status of complement, adverbial and relative clauses is revealed by the fact that these clauses lack illocutionary force, that is, the ability to function as an autonomous speech act. This can be proven by means of standard tests for illocutionary force such as for example the application of sentential negation, sentential questioning, or tag-questions (Lehmann 1988, Dik 1997: 299–307, Cristofaro 2003: chap. 2). Adverbial clauses have also been associated with the pragmatic opposition between foreground and background. This opposition pertains to the structure of narrative texts. Foregrounded material represents the backbone or skeleton of the text, while backgrounded material conveys accessory information. Foreground material always has to be temporally ordered, while background material is typically out of the timeline. Foreground and background are reflected cross-linguistically by clusters of morphosyntactic phenomena. These include for example the presence of special particles, particular word order, and particular tense, aspect, and mood distinctions, such as the distinction

 Sonia Cristofaro

between completed and uncompleted actions, which tend to be associated with foreground and background respectively (Hopper 1979, Hopper and Thompson 1980). Based on extensive analysis of English narrative texts, Reinhart (1984), Haiman (1985: section 2.1 and chap. 4), Tomlin (1985), and Thompson (1987) argue that adverbial clauses tend to encode backgrounded information, while main clauses encode foregrounded information. Another opposition that has been associated with adverbial clauses is that between figure and ground. The notions of figure and ground originate from Gestalt psychology, and pertain to a perceptual asymmetry between two linked entities. One of these entities, the ground, is construed as a cause, precondition, or reference point, for the other, the figure. Talmy (1978, 2000: chaps. 5–6) has argued that the figure-ground asymmetry is manifested in adverbial constructions, that is, the events encoded by adverbial clauses are the ground for the events encoded by main clauses. For example, in a sentence such as

(2) He dreamed while he slept

sleeping is the ground for dreaming, because the act of dreaming is contingent on the act of being asleep. Similarly, Croft (2001: chap. 9) investigates the figure-ground opposition with respect to a wide variety of constructions, and argues that adverbial constructions prototypically encode figure-ground relations. In most of these proposals, pragmatic and perceptual asymmetry between events are regarded as a feature of particular sentence types, those that are traditionally considered as instances of subordination based on European languages. Since these sentence types usually encode asymmetric events, event asymmetry is assumed to be a distinguishing feature of subordination2. The various features displayed by these sentence types, however, do not occur in exactly the same combinations, either cross-linguistically or within individual languages. This casts doubt on the possibility to define an internally consistent grammatical category of subordination, and consequently on the existence of a relationship between such a category and event asymmetry. To illustrate this point, let us consider what is perhaps the most widely recognized criterion for subordination, clausal embedding. Clausal embedding is a syntactic property whereby a clause functions as a constituent of another clause. Clausal embedding is usually identified in terms of a number of syntactic phenomena, such as the fact that one clause can be placed either before or after the other clause, it can be inserted into this clause, which becomes discontinuous, and it may include cataphoric reference to this 2. An exception to this is represented by Croft 2001: chap. 9, where the relation between perceptual asymmetry between events and subordination is viewed in terms of prototypes. Perceptually asymmetric (figure-ground) relations are prototypically associated with subordination, while perceptually symmetric (complex-figure) relations are prototypically associated with coordination. However, this association may be reversed in nonprototypical situations.



Asymmetric events, subordination, and grammatical categories 

clause (Roberts 1988, Haspelmath 1995). Complement, adverbial and relative clauses in European languages turn out to be embedded under at least some of these criteria, as shown by the Italian examples in (3) (the embedded clause is in square brackets). Italian (3) a.

Io [ di vedere un film del genere] non ho I of see:INF a movie such not have:PRES:IND:1SG proprio voglia really desire `I really don’t want to see such a movie’

b. [ Di vedere un film del genere] io non ho of see:INF a movie such I not have:PRES:IND:1SG proprio voglia really desire `I really don’t want to see such a movie’ c.

Io non ho proprio voglia [ di vedere un I not have:PRES:IND:1SG really desire of see:INF a film del genere] movie such `I really don’t want to see such a movie’

d. Luii, [ chiedendo-gliei-lo], ti direbbe hei ask:PRES:GER-to.himi-it to:you say:PRES:COND:3SG probabilmente di non fare domanda probably of not apply:PRES:INF `If you asked him, he would probably tell you not to apply’ e.

Il libro [ che sto leggendo] non è the book that AUX:PRES:IND:1SG read:PRES:GER not is particolarmente interessante particularly interesting `The book I am reading is not particularly interesting’

Embedded clauses, however, do not form an internally consistent class, either crosslinguistically or within individual languages. For example, they may either involve the same verb forms used in independent declarative clauses ((3e)), or verb forms with reduced inflection, such as infinitives, gerunds, and the like ((3a-d)). Also, not all clauses respond in the same way to the various tests for embedding. Individual clauses may show some, but not all of the properties that are regarded as distinctive for embedding, and these properties may not be the same from one clause to another, both cross-linguistically and within individual languages. For example, (3a)-(3c) show that certain types of complement clauses in Italian can be placed before, after, or within the main clause. The English equivalent of these clauses, however, can only be placed after the main clause, hence the

 Sonia Cristofaro

ungrammaticality of (4a) and (4b). Similarly, while (3d) and (5a) show that in Italian certain types of adverbial clause can be either inserted into the other clause or placed before it, (5b) shows these same clauses cannot usually be placed after the other clause. Finally, while relative clauses in Italian and English are inserted within the other clause ((3e)), they cannot be moved before or after this clause, hence the ungrammaticality of (4c), (4d), (5c), and (5d). (4)

a. b. c. d.

*I [to see that movie] really don’t want *[To see that movie] I really don’t want *[That I am reading] the book is not particularly interesting *The book is not particularly interesting [that I am reading]

Italian (5) a. [`Chiedendo-gliei-lo], luii ti direbbe ask:PRES:GER-to.himi-it hei to:you say:PRES:COND:3SG probabilmente di non fare domanda probably of not apply:PRES:INF `If you asked him, he would probably tell you not to apply’ b. ? Luii ti direbbe probabilmente di non he to:you say:PRES:COND:3SG probably of not fare domanda, [chiedendo-gliei-lo] apply:PRES:INF ask:PRES:GER-to.himi-it `If you asked him, he would probably tell you not to apply’ c. * Il libro non è particolarmente interessante [ che the book not is particularly interesting that sto leggendo] AUX:PRES:IND:1SG read:PRES:GER `The book I am reading is not particularly interesting’ d. *[ Che sto leggendo] il libro non è that AUX:PRES:IND:1SG read:PRES:GER the book not is particolarmente interessante particularly interesting `The book I am reading is not particularly interesting’ These facts show that the clauses that are usually regarded as embedded differ with regard to both their morphosyntactic properties in general, and the specific properties that are considered as distinctive for embedding. Thus, one can include them within the same category only by arbitrarily disregarding their differences. This procedure may be justified in descriptive terms to the extent that use of the same label (subordination) for a variety of clauses reflects the fact that these clauses either share particular properties, or display different properties that are regarded as manifestations of the same phenomenon (as is the case with the various properties that are regarded as



Asymmetric events, subordination, and grammatical categories 

distinctive for embedding). As these are different clause types, however, there is no evidence that they belong to a single grammatical category that is a part of a speaker’s knowledge of their language, nor that they instantiate the same grammatical category from one language to another3. This means that the fact that these clauses are associated with event asymmetry cannot be taken as evidence that event asymmetry is a distinctive feature of a general grammatical category of subordination to which all of these clauses belong. An additional problem is that embedded clauses (whatever the criterion used to define embedding) are not always associated with particular types of asymmetry between events. For example, Haiman (1985: chap. 4) mentions a number of languages where interrogative operators apply simultaneously to an embedded clause and the clause in which it is embedded. This is illustrated in (6) (the embedded clause displays the medial verb ending -na, and its embedded status is revealed by the fact that it can be surrounded by material from the other clause: Haiman 1985: 206–7): Hua (Trans-New Guinea, Papua New Guinea) (6) [ P-mi-na] de-ve? them-give-3SG.ANT eat-3SG:INT `Did he give it to them and eat?’ (Haiman 1985: 205) While in such cases the embedded clause may be backgrounded (Haiman 1985: 228–9), the fact that it can be questioned suggests that it is asserted, just like the clause in which it is embedded4. This challenges the assumption that there is a connection between subordination (defined in terms of embedding) and the assertiveness/nonassertiveness asymmetry. Similarly, as is shown in Croft 2001: chap. 9, complement sentences and some types of adverbial sentences, which usually involve embedded clauses (under at least some embedding criteria, see the discussion above), encode percepetually symmetric relations between events, which he labels complex figure relations. This casts doubt on the existence of a connection between subordination and the figure-ground asymmetry. Similar observations apply to the properties of complement, adverbial, and relative sentences other than embedding. For example, these sentences often contain 3. The necessity to distinguish between the description of a language and the description of the grammatical knowledge of the speakers of that language has been repeatedly advocated by Dryer (1996, 1997, 2006, to appear), as well as Haspelmath (2004). 4. Haiman and Thompson (1984: 517-8) and Haiman (1985: 204-6) argue that the ability of a clause to fall within the scope of an interrogative or negative operator in another clause, as in (6), is a sign of its being embedded into the other clause. This property reveals however that the clause is asserted. This is in contrast with the other properties that are usually regarded as distinctive for embedding, in that these properties are usually found in clause that are not asserted. This is another instance of how individual criteria for embedding do not define internally consistent clause classes.

 Sonia Cristofaro

inflectionally reduced verb forms, such as infinitives, gerunds and the like (see e.g. the Italian and English examples in (3) above). Inflectional reduction, or nonfiniteness, is usually regarded as a hallmark of subordination. As illustrated in detail in the literature on nonfiniteness (see among others Joseph 1983: chap. 1, Koptjevskaja-Tamm 1993, Nikolaeva 2007), however, inflectional reduction is manifested in different ways and combines with different properties from one verb form to another, both cross-linguistically and within individual languages. For example, while some verb forms are not inflected for a subset of the categories relevant to verbs in the language, other verb forms are not inflected for any of these categories. Some forms display nominal or adjectival properties, while others don’t. This means that there is no evidence that inflectionally reduced forms correspond to a single grammatical category in a speaker’s knowledge of their language, nor that they instantiate the same grammatical category from one language to another. In addition to that, inflectionally reduced verb forms may be also be used in sentences encoding perceptually or pragmatically symmetric events. This casts doubt on the existence of a connection between these forms and event asymmetry. For example, the complex figure relations discussed by Croft (2001: chap. 9), which involve perceptually symmetric events, are often encoded by means of inflectionally reduced verb forms. Similarly, medial verb forms in Hua, which are not marked for tense, are used in sentences where all of the encoded events are asserted (Haiman 1976: example (6) above). This is the case also with Italian gerunds, as can be seen from (7) below, where both the indicative and the gerund clause fall within the scope of the interrogative operator vero5: Italian (7) Hanno vissuto in diverse città, finendo poi AUX:PRES:IND:3PL live:PAST:PTCPL in various towns end.up:PRES:GER then a Londra, vero? in London true `They lived in several towns and eventually ended up in London, didn’t they?’ In all of the cases discussed so far, the assumption that there is a relation between subordination and event asymmetry is challenged by the fact that the individual features 5. In fact, there is robust cross-linguistic evidence that the use of inflectionally reduced verb forms is related to information recoverability, that is, these forms tend to be used when the information encoded by verbal inflection is either recoverable from the context or irrelevant, and therefore need not be specified overtly. For example, verb forms not inflected for person, such as Italian gerunds, tend to be used cross-linguistically when the linked clauses share an argument, so this argument need not be expressed overtly in both clauses. Similarly, verb forms not inflected for tense tend to be used when the time reference of the relevant clause can be recovered from the context (Haiman 1985, Noonan 1985, Cristofaro 2003). This principle is completely independent of event asymmetry, which accounts for why sentences with inflectionally reduced verb forms may be used to encode both symmetric and asymmetric events.



Asymmetric events, subordination, and grammatical categories 

that are regarded as distinctive for subordination do not combine with the same other features from one sentence type to another, both cross-linguistically and within individual languages. For example, different sentence types may respond in the same way to individual tests for embedding, but encode asymmetric and symmetric events respectively. Particular verb forms may all share particular morphosyntactic properties, but they may differ with respect to other morphosyntactic properties, and may not always be associated with asymmetric events. As a result, if a category of subordination is defined in terms of any of these features, individual members of this category will not have the same properties. Rather, they will share the relevant feature, but differ in their other properties, including event asymmetry itself. This creates two problems. First, there appears to be no principled reason to posit a grammatical category of subordination in the first place, because there is no evidence that the entities that would fall within such a category are actually included within the same category in a speaker’s knowledge of their language. Second, even if such a category is posited (based on arbitrary selected features of the relevant sentence types), the members of such a category will not always be associated with event asymmetry. In the next section, an alternative approach to subordination and event asymmetry will be discussed. In this approach, subordination is identified with event asymmetry, independently of the specific features of particular sentence types.

3. Subordination as event asymmetry Langacker (1987: 183–9) argues that any semantic structure has two fundamental components, the base and the profile. The base, or scope of predication, corresponds to the aspects of a scene (or a subset thereof) that are specifically included in a particular predication. The profile is a substructure within the base that is selected for designation. The profile is the part of a scene that is obligatorily accessed, and accorded special cognitive prominence. For instance, the base for a predication like circle is the basic domain of two-dimensional space, while the profile is a set of points within this domain. On the other hand, a predication like arc has as its base the two-dimentional configuration designated by circle (in that the conception of an arc presupposes that of a circle), and profiles a segment of that configuration.

 Sonia Cristofaro

According to Langacker (1991: 435–9), subordination is a situation of cognitive asymmetry between the processes described by two linked clauses6. The process described by the subordinate clause lacks an autonomos cognitive profile, and is construed in the perspective of the process described by the main clause. As a result, the sentences involving a subordinate clause designate just one process, that described by the main clause. For example, a typical complement sentence like

(8) I know she left

designates the process of knowing, not of leaving. Likewise,

(9) Alarms ringing, the burglar fled

designates the process of fleeing, not of ringing, and (10) The skirt she bought was too tight designates the skirt, not somebody’s buying it. On the other hand, in a coordinate structure like (11) The Cubs won and the Padres lost neither clausal profile overrides the other, and each of the two processes described by the two clauses has an autonomous cognitive profile. The cognitive definition of subordination provided by Langacker is also used in Cristofaro’s (2003) large scale typological study of subordination. This study is based on the socalled Asymmetry Assumption (Cristofaro 2003: chap. 2). Under this assumption, subordination is identified with a situation of asymmetry between the events described by two linked clauses, whereby one event lacks an autonomous cognitive profile and is construed in the perspective of the other event. This situation is universal across languages, and can be equated with the pragmatic opposition between asserted and nonasserted information. This opposition, as defined by Lambrecht (1994: 51–65), refers to an asymmetric communicative organization of the sentence. The asserted part of the sentence corresponds to what the speaker wants to communicate, that is, what the hearer is expected to know or take for granted as a result of hearing the sentence uttered. The nonasserted part of the sentence, on the other hand, corresponds to what the hearer is expected to know or take for granted at the moment the sentence is uttered, and is introduced in discourse for accessory purposes, such as e.g. help the hearer activate asserted information by relating it to some already given piece of knowledge (see the discussion of example (1) above). 6. In Langacker’s model (Langacker 1987a, 1987b) , a process is a cognitive entity that is scanned sequentially through a span of conceived time, in a state-by-state, non-cumulative fashion. The notion of process overlaps with the notion of event used in this paper in that both refer to a predication consisting of a predicate and its terms and denoting something that can be the case in some world. The notion of event as such, however, has no implications about the way in which this entity is conceptualized. In fact, a process can be regarded as a sequentially scanned event.



Asymmetric events, subordination, and grammatical categories 

If a clause is profiled, the event described by that clause corresponds to (part of) the information that the sentence is meant to communicate, i.e., the assertional part of the sentence, while the event described by a nonprofiled clause corresponds to the nonassertional part of the sentence. This appears to be in accordance with Langacker’s own discussion of assertion (Langacker 1991: 496–501). As a result, for all languages, subordination can be identified by means of standard assertiveness tests on the pragmatic content of the sentence. These tests are basically of two types. Some may test for which part of the sentence is open to challenge by explicitly denying it. This is the case of sentential negation, as illustrated by (12) It is not the case that, alarms ringing, the burglar fled What is being denied in this sentence is that the burglar fled, not that alarms were ringing. It is quite clear why only the asserted part of a sentence is open to challenge. If an assertion is what the speaker wishes to communicate, it may be relevant for the hearer to challenge it, while it would make no sense for the hearer to challenge that part of the sentence that the speaker does not wish to communicate. The second type of assertiveness tests change the illocutionary force of the sentence. This is the case of sentential questioning and tag-questions, as illustrated by (13) Is it the case that, alarms ringing, the burglar fled? (14) Alarms ringing, the burglar fled, didn’t he? Once again, what is being questioned in these examples is that the burglar fled, not that alarms were ringing. The relation between assertion and illocutionary force is also quite clear. By making an assertion, or designating a process, a speaker performs a particular kind of speech act, one aiming at making the addressee know or take for granted some part of the sentence. Illocutionary force is the property whereby a sentence can function as a speech act (Dik 1997: 299–307). If some part of the sentence lacks illocutionary force, it cannot represent a speech act, so it is not asserted. These tests show that the situation of pragmatic/cognitive asymmetry identified with subordination under the Asymmetry Assumption underlies all of the sentence types that are traditionally identified as instances of subordination, namely complement, adverbial, and relative sentences (in fact, these sentence types are usually argued to involve nonasserted clauses, see Section 2). In addition to that, subordination encompasses any sentence type that encodes the same conceptual situation, independently of the morphosyntactic properties of that sentence type. For example, the function of relative sentences in European languages is to identify a relevant discourse participant within a set of possible referents by describing an event in which this participant is involved. This event corresponds to nonasserted information (see the discussion of example (1)). While European languages express this conceptual situation by means of sentences involving an embedded clause, a number of languages use sentences consisting of two juxtaposed clauses, neither of which is

 Sonia Cristofaro

embedded into the other. Depending on the context, one of these clauses can also be used to provide a specification concerning the circumstances under which the event described by the other clause takes place, very much like embedded adverbial clauses in other languages. This is illustrated in (15). Walbiri (Australian, Paman) (15) ŋatjulu-u a yankiri pantu-u, kutja-lpa ŋapa ŋa-u I-ERG AUX emu spear-PAST COMP-AUX water drink-PAST `I speared the emu which was / while it was drinking water’ (Hale 1976: 78) Insofar as they perform the same function (i.e., encode the same conceptual situation) as relative and adverbial sentences in European languages, these sentences also count as instances of subordination under the Asymmetry Assumption (see Cristofaro 2003: chap. 2 for a detailed discussion of this issue). In Langacker’s and Cristofaro’s approach, subordination is a conceptual situation independent of the morphosyntactic properties of any particular sentence type. In particular, Cristofaro (2003: chap. 2) explicitly argues that subordination should be defined independently of the morphosyntactic properties of the sentence types found in any particular language. This view originates from a major theoretical assumption of the functional-typological approach to the study of language, namely that the structure of linguistic expressions can be at least partially explained in terms of the conceptual situations they express. As a result, all of the possible ways in which the relevant conceptual situations are encoded across languages should be taken into account, because they could provide either counterexamples or supporting evidence for the proposed connections between conceptual situations and linguistic structure. Only in this way can generalizations be made that are valid for all languages. In practice, this means that no language should be left out of the investigation, because individual conceptual situations are universal across languages, that is, all languages have a means to encode them. If the phenomenon under investigation is defined in terms of particular morphosyntactic features, however, a number of languages will have to be left out of the investigation where these features are not found. As a result, functional-typological investigation should be based on conceptual situations, rather than morphosyntactic features. For example, an investigation of relativization should be based on the conceptual situation encoded by particular relative clause types, rather than on those clause types as such. In fact, the various sentence types that count as instances of subordination under the Asymmetry Assumption appear to obey a number of universal implicational hierarchies, that are motivated in terms of the conceptual situations they encode (Cristofaro 2003: chaps. 5–10).



Asymmetric events, subordination, and grammatical categories 

4. The grammatical encoding of asymmetric events The idea that subordination should be identified with event asymmetry, regardless of the morphosyntactic features of individual sentence types, does not make it possible to define subordination as an internally consistent grammatical category. From this point of view, the same problems are encountered as when subordination is defined in terms of individual features of particular sentence types. This was already implicit in the discussion in Section 2, where it was shown that a number of sentences encoding asymmetric events, because of their morphosyntactic differences, cannot be argued to be part of the same grammatical category of subordination, either cross-linguistically or within individual languages. The same situation is found when all of the sentence types encoding asymmetric events, not just those that are assumed to belong to a particular grammatical category, are considered. These sentence types display different morphosyntactic properties, and, more generally, there appears to be no particular morphosyntactic feature that is systematically associated with asymmetric events as such. As a result, there is no evidence that these sentence types correspond to a grammatical category that is a part of a speaker’s knowledge of their language, nor that these constructions instantiate the same grammatical category from one language to another. This is shown for example by the contrast between the relative sentence in (3e) and the sentence in (15). Insofar as they encode the same situation of event asymmetry, these two sentence types both count as instances of subordination under the Asymmetry Assumption. As they have different morphosyntactic properties, however, there is no evidence that they instantiate the same grammatical category in the relevant languages. In fact, they would not be regarded as instances of the same grammatical category under most syntactic analyses, (3e) falling within the domain of subordination and (15) within the domain of coordination. Similar contrasts are also found within the same language. For example, purpose sentences in European languages are used to encode a semantic situation whereby one event is brought about in order to obtain the realization of another event. The latter corresponds to nonasserted information, as is shown for example in (16), where the tag question cannot apply to the clause encoding this event: (16) Your parents sent you to school in order that you learn new things, didnt’t they?/ *don’t you? In Cubeo, this conceptual situation may be encoded by two different sentence types. The first one, exemplified in (17a), involves two clauses with two independent verbs. The second one, exemplified in (17b), is used with motion verbs. The motion verb takes affixes agreeing with the subject in gender and number, and is encliticized to the verb expressing the event whose realization corresponds to the purpose of motion:

 Sonia Cristofaro

Cubeo (Tucanoan, Central Tucanoan; Colombia) (17) a. j-RE xejo-k-RE epe-Ibã xi-b’k-wA 1SG-OBJ child-M.SG-OBJ put-N/H.3PL 1SG.POSS-adult-PL bue-I-Rõ-I xi bãxi-kijepe teach-STVZ-IN.SG.NMLZ-LOC 1SG.POSS know-PUR.DS a-Iw ape-Rõ xi say-NFUT.PL.NMLZ other-IN.SG.NMLZ 1SG.POS bãxi-bE-Rõ-RE know-NEG-IN.SG.NMLZ-OBJ `From (when I was) a child, my parents put me in a school in order that I learn something that I didn’t know’ (Morse and Maxwell 199: 175) b. oko-kobe-I î-ko=R -bIko oko-RE water-CLS:hole-LOC get-NFUT:F:SG:NMLZ=go-3F:SG water-OBJ `She recently went to get water at the well’ (Morse and Maxwell 1999: 65) Insofar as these two sentence types encode the same relation between events, and this relation is the cognitively/pragmatically asymmetric one defined by the Asymmetry Assumption, they can both be regarded as instances of subordination under this assumption. As they differ considerably in their morphosyntactic properties, however, there is no evidence that they are instances of the same grammatical category in Cubeo. In fact, in their grammar of Cubeo, Morse and Maxwell (1999) describe the sentence in (17a) in the chapter on subordination, and that in (17b) in the chapter on independent clauses. In fact, languages may not even have sentence types specifically dedicated to the expression of asymmetric events. In a number of languages, different sentence types are used to express the same semantic relation between events depending on whether these events are construed as cognitively/pragmatically symmetric or asymmetric. For example, in Lango, a process of sensory perception of an ongoing event can be expressed by means of two different sentence types. The first one consists of the paratactic juxtaposition of two clauses ((18a)), and is used when both the process of sensory perception and the perceived event are asserted.The second one involves clausal embedding by means of a complementizer ((18b)), and is used when only the process of sensory perception is asserted (Noonan 1992: 203–5). Lango (Nilo-Saharan) (18) a. àtîn ànέnò lócә òŋàŋò pàlà child 3:SG-see-PERF man 3:SG-blunt-PERF knife `The child saw the man, he blunted the knife’ (Noonan 1992: 203) b. àtîn ànέnò [ nî lócә òŋàŋò pàlà] child 3:SG-see-PERF COMP man 3:SG-blunt-PERF knife `The child saw that the man blunted the knife’ (Noonan 1992: 203)



Asymmetric events, subordination, and grammatical categories 

In a number of languages, however, the same sentence type can be used to encode either asymmetric or symmetric events depending on the context. This is the case with converbs in Burushaski. Sometimes, the events expressed by sentences involving a converb are all asserted, as demonstrated by the fact that they can all be questioned. This corresponds to the first reading of the sentence in (19a). In other cases, one of the two events is presupposed, and cannot be questioned. This corresponds to the second reading of the sentence in (19a). The likelihood of either reading depends on the context. For example, the reading where one of the two events is presupposed is the most likely one for the sentence in (19b), although in principle both readings would be possible (Tikkanen 1995: 506–15). Burushaski (Isolate) (19) (a) Má-a guttaš-o dú-čú-n já-a díš-ulo you-ERG corpse:PL 3HM.PL.DO-bring-CP 1-GEN place-INESS bése yáar-e ό-čá-an? why down-LOC 3HM.PL.DO-do(DUR)-AUX-HM.PL.SUBJ `Why do you bring your corpses and bury them on my land? / Why do you bury your corpses on my land, after bringing them?’ (Tikkanen 1995: 509, quoted from Lorimer 1935: 112) (b) Ún-e čái n-ét-an ité-er bayú thou-ERG tea CP-do-CP it-DAT rock.salt e-é-waš-č-á-a? NEG-3MSG.DO-throw-DUR-AUX-2SG.SUBJ `After making tea, do you not put rock salt into it? * Do you not make tea and put rock salt into it?’ (Tikkanen 1995: 12) A similar phenomenon is attested in Mandarin Chinese, where serial verb sentences may have different readings depending on the context, as can be seen from example (20a) below (Li and Thompson 1973). Under one reading, all of the events encoded by the sentence are asserted. This is shown for example by the fact that in this case, when negated, the sentence takes the negative marker búshi, that has scope over all of the linked events ((20b)). Under another reading, only one event is asserted. This is shown by the fact that the sentence takes the negative marker bú, that has scope over only one of the linked events ((20c)). Mandarin Chinese (Sino-Tibetan, Sinitic) (20) a. Nˇi guì-xialai qiú Zháng-san you kneel.down beg Zhang-san `You knelt down and begged Zhang-san / You knelt down in order to beg Zhang-san’ (Li and Thompson 1973: 98) b. Woˇ búshi guì-xialai qiú Zháng-san I not kneel.down beg Zhang-san `It is not the case that I kneel down and beg Zhang-san’ (Li and Thompson 1973: 101)

 Sonia Cristofaro

c. Woˇ bú guì-xialai qiú Zháng-san I not kneel.down beg Zhang-san `I do not kneel down to beg Zhang-san’ (Li and Thompson 1973: 101) Insofar as they can encode cognitively/pragmatically asymmetric events, the Lango sentence in (18b), the Burushaski sentence in (19) and the Mandarin Chinese sentence in (20) can all be regarded as instances of subordination under the Asymmetry Assumption. It is quite clear, however, that there is a substantial difference between sentence types specifically dedicated to the expression of asymmetric events, such as the Lango one, and sentence types where event asymmetry is inferred at the contextual level. In this respect, there is no principled ground to assume that these sentence types are all instances of the same grammatical category in the relevant languages.

5. Concluding remarks: Asymmetric events and subordination as a grammatical category Event asymmetry is regarded in the literature as a distinguishing feature of subordination, either in the sense that it is a feature of particular sentence types that can be identified as instances of subordination on independent grounds, or in the sense that subordination should be identified with event asymmetry. However, asymmetric events are encoded by morphosyntactically different sentence types, individual morphosyntactic features may be associated with both symmetric and asymmetric events, and there appears to be no particular morphosyntactic feature that is systematically associated with asymmetric events as such. Only by arbitrarily overlooking these facts can one argue that event asymmetry corresponds to a specific grammatical category, e.g. subordination, either cross-linguistically or within individual languages. These facts are very much the same as those described in typological literature with regard to grammatical relations and other grammatical categories such as parts of speech. This can be illustrated with the well-known case of subject. Subject is traditionally regarded as an argument alignment pattern that plays a role in a number of grammatical phenomena, such as case marking, verbal agreement, or argument omission in clause linkage. These phenomena, however, may not define the same alignment patterns from one language to another. For example, case marking defines an A + S alignment pattern in nominative languages, but an S + O alignment pattern in ergative languages. Also, different morphosyntactic phenomena may define the same alignment patterns in one language, but different alignment patterns in another. For example, case marking and argument omission in clause linkage both reflect an A + S alignment pattern in nominative languages, but in ergative languages they may reflect an S + O and an A + S alignment pattern respectively.



Asymmetric events, subordination, and grammatical categories 

Similar problems are encountered within individual languages, in that individual phenomena that are taken as distinctive for subject may not define the same alignment patterns from one context of occurrence to another. For example, in socalled active languages, such as Acehnese (Durie1985), case marking or verbal agreement may work on an ergative or an accusative basis depending on the semantic properties of the verb. Finally, not all of the phenomena used to define subject are present in all languages, which means that subject cannot be defined in the same way from one language to another. For example, since many languages do not have inflectional case systems or verbal agreement, a different way to define subject should be found for these languages (for detailed discussion of these issues, see, among others, Dixon 1994, Palmer 1994, and Croft 2001). These facts have been variously accomodated for in the literature on grammatical relations, just like the diversity of the morphosyntactic features associated with asymmetric events has been variously treated in the literature on subordination. For example, in a number of approaches subject is defined either in terms of a particular alignment pattern (A + S), independently of the morphosyntactic phenomena that define this pattern, or in terms of particular morphosyntactic phenomena, independently of the alignment patterns defined by these phenomena (Anderson 1976, Dowty 1991). These approaches are akin to those that define subordination either in terms of particular morphosyntactic phenomena, independently of the other features of the sentence types in which these phenomena are manifested, or in terms of event asymmetry, independently of the sentence types in which it is manifested. In yet another approach (Keenan 1976, Comrie 1988 and 1989, Givon 1995), subject is defined in terms of a variety of morphosyntactic, semantic, and pragmatic properties that may not occur in exactly the same combinations cross-linguistically. Particular combinations of phenomena are regarded as the prototypical realization of subject, and languages displaying different combinations are argued to have nonprototypical subjects. In this case too, there is a similar approach to subordination, one in which subordination is defined in terms of a variety of properties (including pragmatic asymmetry between the events being encoded) that may not combine in the same way from one sentence type to another, both cross-linguistically and within individual languages (see e.g. Lehmann 1988). In all of these cases, the relevant categories (subject, subordination) do not correspond to an internally consistent class, but rather to a variety of entities that are similar with respect to some features, but differ with respect to other features. For example, if a category of subject is defined in terms of particular morphosyntactic features, the members of this category will not involve the same alignment patterns, and if this category is defined in terms of particular alignment patterns, its members will not display the same morphosyntactic properties. Similarly, if a category of subordination is defined in terms of particular morphosyntactic features, it will encompass sentence types that differ in their other morphosyntactic features, and may encode either asymmetric or symmetric events (Section 2). If this category is defined in terms of particular

 Sonia Cristofaro

conceptual situations, such as event asymmetry, it will encompass sentence types that differ in their morphosyntactic properties (Section 3). If subject or subordination are defined in terms of a number of features that may not combine in the same way from one construction to another, the members of these categories will obviously display different properties. Dryer (1996, 1997) and Croft (2001) have extensively argued that, if the entities falling within a putative grammatical relation or category do not display the same range of properties from one context of occurrence, or construction, to another, there is no reason to assume that one is dealing with the same grammatical relations or categories in each case. Contrary to what is often assumed in both generatively and functionally oriented approaches, there are no universal grammatical relations and categories that are manifested across different constructions in all languages. Grammatical relations and categories are language-specific, because the entities that belong to a particular grammatical relation or category in a language do not display the same range of properties in another language. Also, grammatical relations and categories are construction-specific, because the entities that belong to a particular grammatical relation or category in one context fail to display the same range of properties in another context. The language-specific and construction-specific grammatical relations and categories may be all be called in the same way (for example, subject), to emphasize their similarities, but this is a terminological issue that does not bear on the fact that these are different grammatical relations and categories. On this view, for example, ergative and nominative languages have different grammatical relations, even if these are defined by the same morphosyntactic phenomena, because these phenomena are associated with different alignment patterns in the relevant languages. If distinct morphosyntactic phenomena (such as case marking and argument omission in clause linkage) are associated with different alignment patterns in a language, one should assume that these phenomena define different grammatical relations in that language, and this even if the relevant phenomena are all associated with the same alignment pattern in another language. These grammatical relations will be construction-specific, because they can only be defined with respect to particular morphosyntactic phenomena. Similarly, if a particular morphosyntactic phenomenon is associated with different alignment patterns in a language depending on the context (for example, depending on the semantic properties of the verb), one should assume that that morphosyntactic phenomenon defines different, construction-specific grammatical relations in that language. Similar arguments apply to the encoding of asymmetric events. Event asymmetry is a conceptual situation manifested in a number of different constructions. Insofar as they encode the same conceptual situation, these constructions can be compared, and a number of generalizations can be formulated concerning the relation between the relevant conceptual situation, or particular aspects thereof, and particular morphosyntactic features (as is done in Cristofaro 2003). As event asymmetry is combined



Asymmetric events, subordination, and grammatical categories 

with different properties in each construction, however, these constructions are best assumed to instantiate different grammatical categories. On a similar line of reasoning, constructions sharing particular morphosyntactic features (such as for example particular nonfinite verb forms, or particular embedding properties), but encoding asymmetric and symmetric events respectively, should be assumed to instantiate different grammatical categories, because they differ (at least) with respect to event asymmetry. It follows that there are three things that should be kept distinct when addressing the issue of the relationship between subordination and event asymmetry: the conceptual situation of event asymmetry, the constructions encoding this conceptual situation, and the grammatical categories manifested in these constructions, including the category of subordination. Event asymmetry as such does not appear to correspond to any cross-constructional and cross-linguistic grammatical category. Rather, it should be regarded as a component of a number of distinct grammatical categories, that are manifested in the constructions encoding asymmetric events. The label `subordination’ can be (arbitrarily) used to designate the conceptual situation of event asymmetry, or any of the constructions encoding this situation. Alternatively, the label `subordination’ can be used to designate any construction that displays particular morphosyntactic features, independently of whether or not that construction is used to encode asymmetric events. However, this is a merely terminological issue that should not obscure the fact that, in each case, we only have evidence for distinct grammatical categories. Each of these categories is manifested in particular constructions in particular languages, that is, is construction-specific and language-specific.

Abbreviations ANT AUX CLS COMP COND CP DO DS DUR ERG GEN GER HM

anticipatory desinence auxiliary classifier complementizer conditional conjunctive participle direct object different subject durative ergative genitive gerund human

 Sonia Cristofaro

IMPF IN IND INESS INF INT M NEG NFUT N/H NMLZ PRES PUR SG STVZ SUBJ

imperfect inanimate indicative inessive infinitive interrogative masculine negative nonfuture nonrecent past/ present habitual nominalizer present purpose ingular stativizer subject

References Anderson, S. R. (1976). On the notion of subject in ergative languages. In C. N. Li (Ed.), Subject and topic, pp. 1–23. New York: Academic Press. Comrie, B. (1988). Passive and voice. In M. Shibatani (Ed.), Passive and voice, pp. 9–23. Amsterdam and Philadelphia: John Benjamins. Comrie, B. (1989). Language universals and linguistic typology. 2nd Edition. Oxford: B. Blackwell. Cristofaro, S. (2003). Subordination. Oxford: Oxford University Press. Paperback edition 2005. Croft, W. (2001). Radical Construction Grammar. Oxford: Oxford University Press. Dik, S. C. (1997). The Theory of Functional Grammar: Part 1: The Structure of the Clause. Berlin and New York: Mouton de Gruyter. Dixon, R. M. W. (1994). Ergativity. Cambridge: Cambridge University Press. Dowty, D. (1991). Thematic Proto-roles and Argument Selection. Language 67, 547–619. Dryer, M. (1996). Grammatical Relations in Ktunaxa (Kutenai): The Belcourt Lecture Delivered Before the University of Manitoba on 24 February 1995. Voices of Rupert’s Land, Winnipeg, Canada. Downloadable at http://linguistics.buffalo.edu/people/faculty/dryer/dryer/dryer.htm. Dryer, M. (1997). Are grammatical relations universal? In J. Bybee, J. Haiman, and S. A. Thompson (Eds.), Essays in language function and language type, pp. 115–43. Amsterdam: John Benjamins. Dryer, M. (2006). Descriptive Theories, Explanatory Theories, and Basic Linguistic Theory. In F. Ameka, A. Dench, and N. Evans (Eds.), Catching Language: The Standing Challenge of Grammar Writing. Berlin and New York: Mouton de Gruyter.



Asymmetric events, subordination, and grammatical categories  Dryer, M. (To appear). Functionalism and the Metalanguage – Theory Confusion. MS, University at Buffalo, downloadable at http://wings.buffalo.edu/socsci/linguistics/people/faculty/ dryer/dryer/dryer.htm. Durie, M. (1985). A Grammar of Acehnese. Dordrecht: Foris. Givón, T. (1995). Functionalism and grammar. Amsterdam and Philadelphia: John Benjamins. Haiman, J. (1976). Presuppositions in Hua. In Papers from the Twelfth Regional Meeting of the Chicago Linguistics Society, pp. 258–70. Haiman, J. (1985). Natural syntax. Cambridge: Cambridge University Press. Haiman, J. and S. A. Thompson (1984). `Subordination’ in universal grammar. In Proceedings of the tenth annual meeting of Berkeley Linguistic Society, pp. 510–23. Hale, K. (1976). The adjoined relative clause in Australia. In R. M. W. Dixon (Ed.), Grammatical categories of Australian languages, pp. 78–105. Canberra: Australian Institute of Aborigenal Languages. Harris, A. C. and L. Campbell (1995). Historical syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. Haspelmath, M. (1995). The converb as a cross-linguistically valid category. In M. Haspelmath and E. König (Eds.), Converbs in cross-linguistic perspective, pp. 1–56. Berlin and New York: Mouton de Gruyter. Haspelmath, M. (2004). Does linguistic explanation presuppose linguistic description? Studies in Language 28, 554–79. Hopper, P. (1979). Aspect and foregrounding in discourse. In T.  Givón (Ed.), Discourse and Syntax, Syntax and Semantics 12, pp. 213–41. New York: Academic Press. Hopper, P. and S. A. Thompson (1980). Transitivity in grammar and discourse. Language 56, 251–99. Joseph, B. D. (1983). The synchrony and diachrony of the Balkan infinitive. Cambridge: Cambridge University Press. Keenan, E. L. (1976). Towards a universal definition of ``subject’’. In C. N. Li (Ed.), Subject and topic, pp. 303–33. New York: Academic Press. Koptjevskaja-Tamm, M. (1993). Finiteness. In R. E. Asher and J. M. Simpson (Eds.), Encyclopedia of Language and Linguistics, pp. 1245–8. Oxford and Aberdeen: Pergamon Press and Aberdeen University Press. Lambrecht, K. (1994). Information structure and sentence form. Cambridge: Cambridge University Press. Langacker, R. W. (1987a). Foundations of Cognitive Grammar. Vol. I: Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, R. W. (1987b). Nouns and verbs. Language 63, 53–94. Langacker, R. W. (1991). Foundations of Cognitive Grammar. Vol. II: Descriptive Applications. Stanford: Stanford University Press. Langacker, R. W. (1997). Generics and Habituals. In A. Athanasiadou and R. Dirven (Eds.), On Conditionals Again, pp. 191–222. Amsterdam: John Benjamins. Lehmann, C. (1988). Towards a typology of clause linkage. In J. Haiman and S. A. Thompson (Eds.), Clause Combining in Grammar and Discourse, pp. 181–225. Amsterdam and Philadelphia: John Benjamins. Levinson, S. C. (1983). Pragmatics. Cambridge: Cambridge University Press. Li, C. N. and S. A. Thompson (1973). Serial verb constructions in Mandarin Chinese: Coordination or Subordination? In You Take the High Node and I will Take the Low Node: Papers from the Comparative Syntax Festival, Chicago Linguistics Society, pp. 96–103.

 Sonia Cristofaro Lorimer, D. L.  R. (1935). The Burushaski Language. Oslo: Institute for Comparative Cultural Studies. Morse, N. L. and M. B. Maxwell (1999). Cubeo grammar. Studies in the Languages of Colombia 5. Summer Institute of Linguistics and The University of Texas at Arlington. Nikolaeva, I. (Ed.) (2007) Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press. Noonan, M. (1985). Complementation. In T.  Shopen (Ed.), Language Typology and Syntactic Description, Volume 2: Complex constructions, pp. 42–140. Cambridge: Cambridge University Press. Noonan, M. (1992). A Grammar of Lango. Berlin and New York: Mouton de Gruyter. Palmer, F. (1994). Grammatical Roles and Relations. Cambridge: Cambridge University Press. Reinhart, T. (1984). Principles of Gestalt perception in the temporal organization of narrative texts. Linguistics 22, 779–809. Roberts, J. R. (1988). Amele switch-reference and the theory of grammar. Linguistic Inquiry 19, 45–63. Talmy, L. (1978). Figure and ground in complex sentences. In J. H. Greenberg, C. A. Ferguson, and E. A. Moravcsick (Eds.), Universals of human language, Volume 4: Syntax, pp. 625–49. Stanford: Stanford University Press. Talmy, L. (2000). Toward a cognitive semantics. Vol. I: Concept structuring systems. Cambridge, Massachusetts and London: MIT Press. Thompson, S.  A. (1987). ``Subordination’’ and narrative event structure. In R.  Tomlin (Ed.), Coherence and Grounding in Discourse, pp. 435–54. Amsterdam and Philadelphia: John Benjamins. Tikkanen, B. (1995). Burushaski converbs in their South and Central Asian context. In M. Haspelmath and E. König (Eds.), Converbs in cross-linguistic perspective, pp. 487–528. Berlin and New York: Mouton de Gruyter. Tomlin, R. (1985). Foreground-background information and the syntax of subordination. Text 5, 85–122.

Asymmetry reversal Frantisek Lichtenberk University of Auckland

Subordination is sometimes said to be characterized by a cognitive asymmetry between a matrix clause and its complement clause: the matrix clause is the profile determinant (Langacker 1991); the state of affairs expressed in the matrix clause but not the one expressed in the subordinate clause is (pragmatically) asserted (Cristofaro 2003). This paper discusses a situation in Toqabaqita, an Austronesian language, where erstwhile bi-clausal structures have been restructured as mono-clausal ones. Has the original asymmetry undergone a change? And how could one tell? There is a danger in relying exclusively on linguistic structures when dealing with cognitive asymmetry between states of affairs. There should be other kinds of evidence whether such asymmetry does or does not exist.

Keywords: assertion, associative, asymmetry, Binding Scale, Cristofaro, desiderative (marker), Givon, grammaticalization, Heine, Hooper, Kuteva, Lambrecht, Langacker, phasal (marker), predetermination, Ross, subordination, Toqabaqita, “undesiderative” (marker)

1. Introduction1 From a cognitive perspective, subordination is sometimes said to be characterized by an asymmetry between a main/matrix clause and its subordinate/embedded clause. For example, for Langacker (1991: 436) a subordinate clause is “one whose profile is overridden by that of a main clause”. The main clause “lends its profile” (ibid.) to the construction as a whole; it is the profile determinant. It is the main clause that designates 1. An earlier version of this paper was presented at the 9th International Cognitive Linguistics Conference, July 17–22, 2005, Seoul. Later versions were presented at colloquia at the University of Melbourne and at the University of Leiden in 2006. I am grateful to a number of participants in the Asymmetric Events theme session at the International Cognitive Linguistics Conference and to members of the audience at the colloquia for valuable comments on the versions presented on those occasions. My work on the Toqabaqita language has been supported by grants from the University of Auckland Research Fund, which are gratefully acknowledged.

 Frantisek Lichtenberk

the overall state of affairs. To use Langacker’s example of a sentence with a complement clause, the sentence in (1) designates the state of affairs of the speaker’s knowing, not that of the woman’s leaving:

(1) I know she left.

Cristofaro (2003) recasts Langacker’s notion of a clause being or not being the profile determinant in terms of a clause expressing or not expressing an assertion. She adopts Lambrecht’s (1994: 52) definition of (pragmatic) assertion, as “[t]he proposition expressed by a sentence which the hearer is expected to know or take for granted as a result of hearing the sentence uttered”. A clause that in Langacker’s terminology is a profile determinant encodes an asserted state of affairs, while a clause that lacks an autonomous profile encodes a non-asserted state of affairs. Thus, in I know she left what is asserted is the speaker’s knowing [that she left]. There is then cognitive asymmetry between a main clause and its subordinate clause: the former is asserting, and the latter non-asserting. In fact, Cristofaro (2003: 33) uses this kind of asymmetry to define subordination: By subordination will be meant a situation whereby a cognitive asymmetry is established between linked SoAs [states of affairs; F.L.], such that the profile of one of the two (henceforth, the main SoA) overrides that of the other (henceforth, the dependent SoA). This is equivalent to saying that the dependent SoA is (pragmatically) non-asserted, while the main one is (pragmatically) asserted.

The question that will be addressed in this study is the possibility of a reversal of asymmetry through a grammaticalization process. Specifically, an erstwhile complementtaking verb may cease to function as a verb, becoming a particle or an auxiliary, while what used to be a complement-clause structure becomes, as a consequence, a noncomplement clause structure, because an erstwhile two-clause structure has become mono-clausal. Has asymmetry reversal taken place in such a case? By “asymmetry reversal” I mean a situation where what used to be a higher, assertion-expressing clause has lost its status as a clause and so has, presumably, lost its assertion-making status vis-à-vis the erstwhile complement clause; while the erstwhile complement, non-assertion-expressing clause has been reinterpreted as part, possibly the major part of the sole clause that contains the new particle/auxiliary and hence, presumably, is now assertion-expressing. But how do we know whether such asymmetry reversal has indeed taken place? Because of lack of relevant information on the language to be discussed in what follows, this question can only be raised here, without any definite answers. Nevertheless, the question is worth raising. The data come from Toqabaqita, an Austronesian language spoken in the Solomon Islands.



Asymmetry reversal 

2. Desiderative and phasal structures in Toqabaqita 2.1

Finite complement clauses

Toqabaqita has a fairly large number of complement-taking verbs of various semantic types, such as manipulative, perception, knowledge and acqusition-of-knowledge, utterance, propositional attitude, desiderative and phasal. Of relevance here are two desiderative verbs and one phasal verb, because they exhibit certain properties not found with any of the other complement-taking verbs. The desiderative verbs are thathami ‘want, like’ and lalakwa ‘not want, not like’, and the phasal verb is thafali ‘begin, start’. The desiderative verb thathami ‘want, like’ and the phasal verb thafali ‘begin, start’ are transitive and can take non-clausal direct objects. This is illustrated in (2) and (3), respectively. (2) Nau ku thathami-a ta si fanga. 1sg 1sg.nfut want-3.obj some prtt food ‘I want some (of the) food.’ (3) … keka thafali-a laqu boqo ta si kula faalu. 3pl.seq begin-3.obj add asrt some prtt place be.new ‘[When people have started harvesting food from a garden,] they begin/ start a (lit.: another) new place [to make it into a garden for the next crop].’ The other desiderative verb, lalakwa ‘not want, not like’, is intransitive and takes oblique objects: (4) Nau ku lalakwa qana wane qeri. 1sg 1sg.nfut not.like genp man that ‘I don’t like that man.’ All three verbs can take clausal complements, and from now on our attention will be focused on these verbs with clausal complements. For reasons that will become clear as we proceed, the transitive desiderative verb ‘want, like’ and the transitive phasal verb ‘begin’ will usually be dealt with together, and the intransitive desiderative verb ‘not want, not like’ separately. Example (5) shows ‘want, like’ with a finite complement clause. The subject of the higher, desiderative verb and the subject of the verb in the complement clause are different: first person singular and second person dual, respectively. The complement clause is in square brackets. (5) Iu, ma ni nau ku thathami-a yes and profore 1sg 1sg.nfut want-3.obj

 Frantisek Lichtenberk

[ kamaroqa moki baqo-a alinga-ku] …. 2du 2du.fut pierce-3.obj ear-1sg.pers ‘Yes, and I want you two to pierce my ears ….’ (Lit.: ‘Yes, and I want [you two will pierce my ears] ….’) The verb ‘want’ carries the third person suffix -a, which indexes the complement clause. The same suffix is also used to index lexical direct objects; see baqo-a alinga-ku ‘pierce my ears’ in the second line of (5), and examples (2) and (3). The complement clause is finite, and its structure is no different from those found in independent clauses/sentences. The verb baqo-a ‘pierce’ has its own subject marker, moki, which simultaneously marks second person dual and future tense. And there is also a subject noun phrase, the second person dual independent pronoun kamaroqa. In (6) the verb ‘want’ also has a clausal complement, but there the subject of the main verb and the subject of the verb in the complement clause are coreferential: (6) Nau ku thathami-a [ kwai fanga]. 1sg 1sg.nfut want-3.obj 1sg.fut eat ‘I want to eat.’ Here too the higher verb carries the object suffix -a, and the complement clause is finite. The verb in the complement clause has its own future-tense subject marker. In (7) it is the phasal verb ‘begin’ that has a clausal complement. With this verb, there must be identity of subjects in the main and the complement clauses. The verb ‘begin’ carries the object suffix -a, which indexes the complement clause. The complement clause is finite, and the verb has a future-tense subject marker: (7) Nau ku thafali-a [ kwai uqunu naqa]. 1sg 1sg.nfut begin-3.obj 1sg.fut narrate prf ‘I am about to begin to tell a/the story.’ Finally, example (8) shows the intransitive verb ‘not want, not like’ with a finite complement clause: (8) Ku lalakwa [ qoki sifo uri qamali]. 1sg.nfut not.want 2sg.fut descend all sea ‘I don’t want you to go down to the sea.’ The subjects are different; in fact, they have to be different. There cannot be identity of subjects. Here too the subject-tense marker in the complement clause is future. All three verbs can, then, take finite complement clauses. The tense in the complement clause is future. With ‘begin, start’ there must be identity of subjects between the main and the complement clauses; with ‘not want, not like’ there cannot be identity of subjects; while with ‘want, like’ there may, but need not, be identity of subjects.



Asymmetry reversal 

2.2

Reduced structures

When there is identity of the relevant participants, alternative constructions are available or in one case required. By the “relevant participants” I mean the participant that desires that a state of affairs obtain, or the participant that does not desire that a state of affairs obtain, in desiderative contexts, and the participant that begins to bring a state of affairs about, in a phasal context. There is always identity of the relevant participants in phasal contexts, as pointed out in section 2.1. And there may, but need not, be identity of the relevant participants in desiderative contexts, depending on the verb. For simplicity, the relevant participants will be designated as Actors, because most commonly those participants do indeed have the thematic (macro)role Actor. The reason for speaking of identity, or non-identity, of Actors rather than identity/non-identity of subjects will become clear later on. And, also for simplicity, the term “event verb” will be used to signify the verb that encodes the state of affairs that is desired, not desired or begun, and “event clause” will designate the clause in which the event verb occurs. In one of the alternative constructions, the otherwise transitive lexemes ‘want, like’ and ‘begin’ do not carry the object-indexing suffix -a. Instead, they carry the suffix -qi, which functions with some verbs as a “detransitivizer”. The suffix will be discussed in some detail in sections 4 and 5. For the time being, it will be glossed detr(ansitivizer), but the question of its status will be revisited in section 5. Furthermore, in this alternative construction the event verb is not, and cannot be, (directly) preceded by a (futuretense) subject marker. This alternative structure and the one discussed further below will be referred to as “reduced”. The reduced alternatives to sentences (6) and (7) in section 2.1 are given in (9) and (10), respectively: (9) Nau ku thathami-qi fanga. 1sg 1sg.nfut want-detr eat ‘I want to eat.’ (10) Nau ku thafali-qi uqunu naqa. 1sg 1sg.nfut begin-detr narrate prf ‘I am about to begin to tell a/the story.’ The element ‘not like, not want’ too occurs in a reduced structure, except that it does not carry the suffix -qi, because the verb ‘not want, not like’ is not transitive. However, just as is the case with ‘want, like’ and ‘begin’ in their reduced structures, there is no subject-tense marker (directly) preceding the event verb. (11) Nau ku lalakwa qani-a fanga naqi, … 1sg 1sg.nfut not.like eat-3.obj food this ‘I don’t like to eat this (kind of) food, ….’ While with ‘want, like’ and ‘begin’ either the construction with a finite complement clause or the reduced structure is possible when there is identity of Actors, with ‘not want, not like’ only the reduced structure is grammatical under Actor identity.

 Frantisek Lichtenberk

The reduced structures are possible only if there is identity of Actors. If there is not identity of Actors, only the constructions with finite complement clauses are grammatical. Compare (12a) and (12b), where the verb is ‘want’: (12) a. Nau ku thathami-a qoki qadomi nau. 1sg 1sg.nfut want-3.obj 2sg.fut help 1sg ‘I want you to help me.’ b. *Nau ku thathami-qi qoki qadomi nau. 1sg 1sg.nfut want-detr 2sg.fut help 1sg (‘I want you to help me.’) In the reduced structures, the event verb cannot have a (future-tense) subject marker occurring between it and the desiderative or the phasal element; see (12b) above for ‘want, like’ and (13b) below for ‘begin’, and the same is true in the presence of the form ‘not want, not like’. The structure in (13a), without a subject marker (immediately) before the event verb, is grammatical, while (13b), with a subject marker immediately before the event verb, is ungrammatical: (13) a. Qusungadi kuki thafali-qi kwaqe-a masuqu loo. tomorrow pl(incl).fut begin-detr cut-3.obj bush upward ‘Tomorrow we will begin to clear the bush up there.’ b.

*Qusungadi kuki thafali-qi kuki tomorrow pl(incl).fut begin-detr pl(incl).fut kwaqe-a masuqu loo. cut-3.obj bush upward (‘Tomorrow we will begin to clear the bush up there.’)

In fact, nothing can intervene between the desiderative or the phasal element and the event verb in the reduced structures, illustrated in (9), (10) and (11). This will be further demonstrated later on. It appears, at first sight anyway, that in the reduced structures the event verbs have been deranked (Cristofaro 2003). They cannot be (immediately) preceded by a subject-tense marker; that is, they are, or at least appear to be, unmarked for subject and tense. While subject markers are occasionally, although not frequently, omitted in independent clauses, apart from second-person singular imperatives, where they are usually omitted, in the reduced structures under discussion here they are ungrammatical. Nevertheless, such apparently deranked verbs can be inflected with the object-indexing suffix. This can be seen in (11) further above with the verb qani-a ‘eat’, where the object suffix indexes ‘this food’; in (13a) above with the verb kwaqe-a ‘cut’, where the suffix indexes the object ‘the bush up there’; and in (14) below with the verb kuqu-fi-a ‘drink’, where the suffix indexes the interrogative phrase ‘what?’.



Asymmetry reversal 

(14) Taa na qo thathami-qi kuqu-fi-a? what? foc 2sg.nfut want-detr drink-tr-3sg.obj ‘What do you want to drink?’ Toqabaqita does have a construction with deranked subordinate clauses where the verb in the deranked clause occurs without a subject marker, but this construction is significantly different from the reduced structures discussed here. Deranked subordinate clauses will be illustrated and briefly discussed in section 3.

3. Constructions with finite complement clauses and with reduced structures compared At this point we need to consider some other aspects of Toqabaqita grammar. First, verbs can be accompanied by a variety of particles, such as directionals, intensifiers, aspect and mood markers and so on. Some of these follow the verb, while others precede the verb. Example (15) contains the preverbal immediate past/immediate future particle biqi and a combination of two post-verbal particles, assertive bo=2 and andative kau: (15) Kera biqi lae bo=kau. 3pl.nfut imm go asrt=and ‘They have just left.’ The postverbal particles come before lexical direct objects. This is shown in (16), which contains the assertive marker bo= and the intensifier naqa. The combination of the two elements bo=naqa occurs between the verb ‘bail out’ and the direct object ‘the water down there’: (16) … qe nalu-fi-a bo=naqa kafo fuu. 3sg.nfut bail.out-tr-3.obj asrt=ints water downward ‘… he busily bailed out the water down there.’ Similarly, the postverbal particles precede finite complement clauses. The three verbs ‘want, like’, ‘not want, not like’ and ‘begin’ can take some of those particles. This is shown for ‘want’ in (17), which too contains the assertive marker bo= and the intensifier naqa: (17) Keka thathami-a bo=naqa keki thau-ngi-a qaburu qeri. 3pl.seq want-3.obj asrt=ints 3pl.fut kill-tr-3.obj ogre that ‘They wanted very much to kill the ogre.’ 2. The assertive marker serves to emphasize that a given state of affairs is indeed such and such or to assert a request for information. The term “assertive” has a different significance here from “assertion” in Lambrecht’s (1994) sense. The equal sign signifies phonological fusion of two or even three words, in which case one or both/all of the words appear(s) in a reduced form. The full form of the assertive marker is boqo.

 Frantisek Lichtenberk

While postverbal particles may intervene between the verbs ‘want, like’, ‘not want, not like’ and ‘begin’ and their finite complement clauses, in the reduced structures they can come only after the event verb. They cannot come between the desiderative or the phasal elements and the event verb. In (18) the particles are the two intensifiers qasia and naqa, and they can occur only after the verb uqunu ‘converse’: (18) ... ma keka thathami-qi uqunu qasia naqa. and 3pl.seq want/like-detr converse ints ints ‘... and they wanted/liked very much to converse (with each other).’ Second, Toqabaqita has a completementizer that is occasionally, although not very frequently and never obligatorily, used to introduce complement clauses. (The same element also introduces relative clauses; see [46] in section 6. It also introduces the clause of presupposition in the focus construction; see [14] in section 2.2.) In (19) the complementizer, whose form is na, occurs with a clause embedded under the verb ‘see’: (19) ... laalae moka riki-a na ni nau kwai nofi naqa .... when 2du.seq see-3.obj comp profore 1sg 1sg.fut die prf ‘... when you see that I am dying/about to die ....’ In (20) it is the verb ‘begin’ that has a complement clause with the complementizer: (20)

Nau ku thafali-a na kwai uqunu 1sg 1sg.nfut begin-3.obj comp 1sg.fut narrate suli-a tha Bariqi. prol-3.obj persmkr Bariqi ‘I am beginning to tell the story about Bariqi.’

And in (21) the complementizer occurs with a clause that functions as a complement of the intransitive verb ‘not want, not like’: (21) Nau ku lalakwa na qoki ngata bia thaari neqe. 1sg 1sg.nfut not.want comp 2sg.fut speak com girl this ‘I don’t want you to speak with this girl.’ It is only finite complement clauses that can be introduced by the complementizer. The complementizer is ungrammatical in the reduced structures. There is one more type of difference between constructions with finite complement clauses and the corresponding reduced structures. In the former there may be a slight rise in intonation at the end of the matrix clause and there may also be a slight pause between the matrix and the complement clauses. On the other hand, the reduced structures are characterized by single intonation contours, without a rise in intonation on the positive or negative desiderative element or on the phasal element, and without a pause between those elements and the following material. There is one type of construction that involves subordination where the verb in the subordinate clause occurs without a subject-tense marker. The construction



Asymmetry reversal 

involves a deranked adverbial subordinate clause. The subordinate clause does not express a state of affairs additional to that expressed in the main clause. Rather, it expresses a circumstance or a characteristic of the state of affairs expressed in the main clause. The verb in the deranked clause has no subject marker. Its notional subject is coreferential with the subject or the direct object of the main clause (or, much less commonly, with another element in the main clause). The sentences in (22) and (23) contain deranked subordinate clauses. The verbs in the deranked clauses are in bold. In (22) the notional subject in the deranked clause is coreferential with the subject of the preceding, main clause: (22)

Kere takwe ba-da i laa one, 3pl.nfut stand lim-3pl.pers loc in beach maqasi-a faka. wait.for-3.obj ship ‘They just stood on the beach, waiting for a ship.’

In (23) the notional subject of the deranked clause is coreferential with the direct object of the main clause: (23) Kere qili-a rara kali-a biqu kera. 3pl.nfut build.wall-3.obj wall surround-3.obj house 3pl ‘They built a wall around their house.’ Various kinds of material can intervene between the verb in the deranked clause and the preceding verb. In (22) the limiter particle ba-da and the prepositional phrase i laa one ‘on the beach’ intervene between the two verbs, and in (23) the direct object rara ‘wall’ does. Furthermore, there may be a slight pause between the main clause and the verb in the deranked clause, indicated by the comma in (22). Nothing like this is possible with the reduced structures under discussion. We can now return to the constructions with finite complement clauses and the corresponding reduced structures. From now on, the term “desiderative (element)” will specifically designate the positive desiderative element ‘want, like’, while the negative desiderative element ‘not want, not like’ will be referred to as “undesiderative”. The clusterings of the properties of the two types of structure are summarized in (24a) and (24b): (24) a. structures with finite complement clauses: if the higher verb is transitive, the complement clause is indexed on the verb by the object suffix -a; the event verb has its own future-tense subject marker; there may be postverbal particles between the higher verb and the complement clause; the complement clause may be introduced by the complementizer; with ‘want, like’ as the higher verb, there may but need not be identity of Actors; with ‘not want, not like’, there cannot be identity of Actors; with ‘begin’, there is always identity of Actors; there may be a

 Frantisek Lichtenberk

rise in intonation at the end of the matrix clause, and there may be a pause between the matrix and the complement clauses; b. reduced structures: the desiderative and the phasal elements carry the suffix -qi, not the object-marking suffix -a; the event verb cannot be directly preceded by a subject-tense marker; nothing can occur between the desiderative, the phasal, or the undesiderative element and the event verb (no postverbal particles, no complementizer, or any other material); such structures are grammatical only when there is identity of Actors; there is a single intonation contour. Compared to the structures with finite complement clauses, in the reduced structures there is a much closer formal bond between the desiderative, the undesiderative or the phasal element on the one hand and the event verb on the other. The verbs ‘want, like’, ‘not want, not like’ and ‘begin’ are the only verbs in the language that take finite complement clauses for which there exist reduced-structure counterparts. While there is no intrinsic reason why that should be the case, it is not surprising for it to be so. First, with all three verbs there is predetermination of the temporal status of the state of affairs encoded in the finite complement clause and predetermination of the participants, in Cristofaro’s (2003) sense. There is predetermination of the temporal status of a state of affairs desired or not desired and a state of affairs begun. When these verbs take finite-clause complements, the verbs in the complements can only be marked for future tense. (The future tense in Toqabaqita is relative: it signals posteriority with respect to the time of reference, past, present or future.) A state of affairs that is wanted, desired or not wanted, not desired (in the relevant sense) can only obtain at a later time, and an event that has been begun can continue only after having been initiated. There is also predetermination of the participants. In a phasal context, the participant that begins to bring a state of affairs about is, of necessity, the same participant as the one who causes it to come into existence (although he/she/it is not necessarily the same participant as the one that continues it and/or completes it). In a desiderative or an undesiderative context, the participant that desires that a certain state of affairs obtain or the participant that does not desire that a certain state of affairs obtain may but need not be the same participant as the one who brings it about. When it is so, the reduced structure is possible. In fact, the presence of the reduced structure signals identity of Actors. And second, there is a relatively high degree of conceptual connectedness between a desire that a state of affairs obtain or not obtain and its obtaining or not obtaining, and there is conceptual inseparability between the initial stage of a state of affairs and its existence. Cristofaro (2003) speaks of “(semantic) integration”, but integration is not always the appropriate notion. What Cristofaro terms “semantic integration” was originally discussed by Givón (1980, 1990) in relation to his Binding Scale. With



Asymmetry reversal 

respect to the phasal element ‘begin’, one cannot speak of integration of two states of affairs, because there are not two states of affairs. Rather, there is only one state of affairs that is at an initial stage of its development. In Givón’s (1984) terminology, ‘begin’ is an implicative verb: if X began to V, then X V-ed. “Conceptual dependence” is a more appropriate notion here than “semantic integration”. The situation is different with the desiderative and undesiderative elements. ‘Want’ and ‘not want’ are not implicative: what one wants to be the case does not necessarily become so, and what one does not want to be the case may turn out to be the case. There is a relatively low degree of integration here, a looser causal link, more independence between the two states of affairs. Nevertheless, it can be argued that, all other things being equal, states of affairs over which we do, in principle, have control are more likely to materialize when they are desired by somebody than when they are not desired by anybody, and states of affairs are less likely to materialize when they are not desired. Thus, there is some integration, connectedness. There is another respect in which the (positive) desiderative constructions discussed here exhibit a relatively high degree of connectedness between the two states of affairs, this one being specific to Toqabaqita. The language makes a grammatical distinction between desiderative situations where there is no reason to assume that the desired state of affairs will not obtain on the one hand and situations where there is some doubt, uncertainty that the desired situation will indeed obtain, or when the desired situation did not, in fact, obtain, that is, when there is counterfactuality. Both constructions use the verb thathami ‘want’. The constructions with this verb given thus far all express the former type of situation: a state of affairs is desired and there is no reason to assume that it will not obtain (even though it may eventually turn out not to obtain). The sentence in (25) is another example: (25)

Ku thathami-a wela nau kai oli 1sg.nfut want-3.obj child 1sg 3sg.fut return na=mai fasi-a qi Honiara. prf=vent abl-3.obj loc Honiara ‘I want my child to come back from Honiara now/straightaway.’

The sentence in (25) could be used when the speaker expects his child to come back, for example because he has asked him to. There is an expectation that the event will take place. On the other hand, if a person wishes that a state of affairs obtain but there is no particularly strong expectation that it will, the clause embedded under thathami ‘want’ contains, in addition to a future-tense subject marker, the irrealis marker sa. Compare (25) above and (26) below. In the situation expressed in (26), the speaker is not at all

 Frantisek Lichtenberk

certain his child will come back from Honiara any time soon, perhaps because the child does not even know about his father’s wish: (26)

Ku thathami-a wela nau sa kai oli 1sg.nfut want-3.obj child 1sg irr 3sg.fut return na=mai fasi-a qi Honiara. prf=vent abl-3.obj loc Honiara ‘I wish my child would come back from Honiara now/straightaway.’

The next pair of examples involves sentences where there is identity of Actors between the main and the embedded clauses. In the situation expressed in (27) the speaker has no reason to think that he/she will not go: (27) Nau ku thathami-a kwai lae naqa. 1sg 1sg.nfut want-3.obj 1sg.fut go prf ‘I want to go now.’ On the other hand, in the situation expressed in (28) the desired event did not take place, as is made explicit in the final clause. The clause that expresses the desired event contains the irrealis marker sa. (28) Nau ku thathami-a sa kwai lae, keka lufi nau. 1sg 1sg.nfut want-3.obj irr 1sg.fut go 3pl.seq prevent 1sg ‘I wanted to go, [but] they prevented me.’ Sentence (27), without the irrealis marker in the complement clause, has a reduced counterpart of the type discussed earlier: (29) Nau ku thathami-qi lae naqa. 1sg 1sg.nfut want-detr go prf ‘I want to go now.’ On the other hand, there is no corresponding reduced counterpart to (28), with the irrealis marker, even though there is identity of Actors. The irrealis marker can only occur in finite clauses, in the presence of a subject-tense marker. The reduced structure, then, is possible with the desiderative element only when there is no indication that the desired state of affairs might not come about or that it, in fact, did not come about. All other things being equal, it is expected that the desired state of affairs will obtain, there being no indication to the contrary. That is, there is a relatively high degree of connectedness between the desire and the desired situation, relatively high compared to cases where there is uncertainty about the desired situation coming about and compared to cases of counterfactuality.



Asymmetry reversal 

4. The suffix -qi In the reduced structures the desiderative element and the phasal element carry the suffix -qi, rather than the object-indexing suffix -a. The suffix -qi is used elsewhere with just a handful of verbs. It functions as a “detransitivizer”: it is added to verbs that are otherwise transitive, reducing their transitivity status. For instance, the verb riki ‘see’ can take the detransitivizing suffix. Example (30) shows riki as a fully transitive verb, with the object-indexing suffix -a: (30) Qo riki-a wela baa? 2sg.nfut see-3.obj child that ‘Did you see the child?’ The verb ‘see’ takes the detransitivizing suffix when it has an incorporated object (but this is not true of nearly all other verbs with incorporated objects). In (31) the incorporated object is fula-a ‘people arriving (at a place)’, a deverbal noun based on the verb fula ‘arrive’. The verb and the noun fula-a form an indivisible unit. While postverbal particles come between verbs and their lexical objects (see example [16] in section 3), they can only come after verb – incorporated-object units. This is the case with the self-benefactive marker qa-kuqa in (31): (31) Lae kwai riki-qi fula-a qa-kuqa. go 1sg.fut see-detr arrive-dvn sben-1sg.pers ‘I’ll go see a group of arrivals/groups of arrivals.’ (Lit.: ‘I’ll go and see me/for myself a group of arrivals/groups of arrivals.’) The suffix is also used with the verb sore ‘say’, which can take direct-speech or indirectspeech complements. However, for reasons we need not go into, speech complements are not direct objects. With the verb sore the detransitivizing suffix has the form -qe, because the final vowel of the verb base is non-high. (For the form -qe see also [35] further below.) (32)

Wela qeri ka sore-qe, “Thaina-ku bia child that 3sg.seq say-detr mother-1sg.pers and maka nau kera lae qana uusia.”. father 1sg 3pl.nfut go genp market ‘The child said, “My mother and father went to the market.”.’

What is historically the same etymon as the detransitivizing suffix occurs on the head noun in associative noun phrases. In an associative noun phrase the head noun is modified by a non-referential noun phrase. The modifiers in associative noun phrases

 Frantisek Lichtenberk

have various semantic values. In one function, they characterize the referent of the whole associative noun phrase as to its type, as in (33): (33) ifu-qi thaqaro feather-assoc bird ‘bird feather(s)’ Compare that to the possessive noun phrase in (34), where the modifying noun phrase is referential. There the possessum noun carries a “personal” suffix that indexes the possessor. (34) ifu-na thaqaro naqi feather-3.pers bird this ‘this bird’s feathers’ The modifier in an associative noun phrase may designate the type of contents of the container designated by the overall noun phrase. In (35) the suffix has the form -qe because the final vowel of the head noun is non-high. (35) baeke-qe kafara bag-assoc copra ‘bag(ful) of copra’ There are some parallels between associative noun phrases and the reduced desiderative and phasal structures, even though, as will become clearer later, in the present-day language the parallels are primarily diachronic in nature. In associative noun phrases the associative suffix occurs on the phrase head, and similarly in the desiderative and the phasal structures the suffix occurs on the desiderative or the phasal element, which corresponds to the head of the verb phrase in the constructions with finite complement clauses. The reason for not referring to the desiderative and the phasal elements as “heads” when they take the suffix -qi in the reduced structures will become apparent shortly. Historical evidence tells us that the associative function of the suffix -qi/-qe was historically primary (Hooper 1985, Ross 2001), and that the detransitivizing function was a later development. Of the two forms of the suffix, ­-qi and -qe, the former is historically primary.

5. An alternative analysis In the discussion thus far, the reduced constructions have been implicitly assumed to be bi-clausal: there is a higher verb — ‘want, like’, or ‘begin’ — that carries the detransitivizing suffix -qi, or the intransitive verb ‘not like, not want’, and there is a complement clause with a deranked-like verb. But there is another possibility, which is this: the reduced structures, rather than being bi-clausal, are, in fact, mono-clausal.  The



Asymmetry reversal 

elements that have the desiderative, the undesiderative and the phasal meaning, respectively, are not complement-taking verbs but preverbal particles. And what looks like a deranked event verb in a complement clause is the only verb. The event verb takes a (preverbal) desiderative, undesiderative or inceptive particle. That is, rather than the structures given in (36a), (37a) and (38a), the structures are those given in (36b), (37b) and (38b), respectively. The suffix -qi in the (b) structures in (36) and (37) has, for convenience, been given the gloss qi rather than detr, because if/since the desiderative, the undesiderative and the phasal elements are not verbs, the suffix does not have a detransitivizing function here. (36) a. = (9) [ Nau ku thathami-qi 1sg 1sg.nfut want-detr ‘I want to eat.’ b. [ Nau ku thathami-qi 1sg 1sg.nfut desid-qi ‘I want to eat.’

[ fanga]]. eat fanga]. eat

(37) a. = (13a) [ Qusungadi kuki thafali-qi [ kwaqe-a tomorrow pl(incl).fut begin-detr cut-3.obj masuqu loo]]. bush upward ‘Tomorrow we will begin to clear the bush up there.’ b.

[ Qusungadi kuki thafali-qi kwaqe-a tomorrow pl(incl).fut incep-qi cut-3.obj masuqu loo]. bush upward ‘Tomorrow we will begin to clear the bush up there.’

(38) a. = (11) [ Nau ku lalakwa [ qani-a fanga naqi,]….] 1sg 1sg.nfut not.like eat-3.obj food this ‘I don’t like to eat this (kind of) food, ….’ b. [ Nau ku lalakwa qani-a fanga naqi, ….] 1sg 1sg.nfut undesid eat-3.obj food this ‘I don’t like to eat this (kind of) food, ….’ If the reduced structures are mono-clausal, the properties that make them different from the structures with finite complement clauses fall out automatically. The properties of the reduced structures were listed in (24b) in section 3. For convenience they are repeated in (39), arranged somewhat differently: (39) Properties of the reduced structures: a. the otherwise transitive desiderative and phasal elements carry the suffix -qi, not the object-marking suffix -a; b. the event verb cannot be directly preceded by a subject-tense marker;

 Frantisek Lichtenberk



c. nothing can intervene between the desiderative, the phasal, or the undesiderative element and the event verb; d. there must be identity of Actors; e. there is a single intonation contour.

The positive desiderative and the phasal elements do not carry the object-indexing suffix -a because there is no complement clause. The event verb cannot be directly preceded by a subject-tense marker because the subject-tense markers are positioned before the preverbal desiderative, undesiderative and phasal particles; cf. examples (15) in section 3 and (40)–(42) further below. The event verb in a reduced structure does have a subject-tense marker, but the marker occurs before the desiderative, the undesiderative or the phasal marker. There is only one subject marker because there is only one verb. The subject marker belongs to the event verb, not to the desiderative, the undesiderative or the phasal element. What appears to be a deranked verb in a complement clause is the only verb, and it is not deranked. Since the structure is mono-clausal and so there is no complement clause, there can be no complementizer. And no postverbal particles can intervene between the desiderative, the undesiderative or the phasal element and the event verb. The identity of Actors is only apparent. Since there is only one clause, there is only one Actor role. And since there is only one clause, there is only one intonation contour. The analysis of the desiderative and the phasal elements in the reduced structures as particles rather than verbs is further supported by the fact that there are (other) preverbal particles in the language that contain the suffix -qi/-qe. It so happens that with all/most? of the particles the suffix has the form -qe, because the vowels in the preceding syllables are non-high. Unlike the desiderative and the phasal elements, those particles do not have verbal counterparts. They cannot form predicates by themselves and can only occur with a verb. One of these particles is the “attenuative” marker thafa/thafe-qe. The suffix -qe is optional. When it is present, the a in the final syllable of the base changes to e. This is a regular phonological process in the presence of the suffix -qe that takes place in associative noun phrases as well; see (35) further above: the basic form of the noun ‘bag’ is baeka. The attenuative particle signals that the event is (to be) performed in an attenuated way, calmly, slowly, quietly or carefully: (40) Qoko { thafa / thafe-qe} ngata. 2sg.seq atten / atten-qi speak ‘Speak calmly/slowly/quietly.’



Asymmetry reversal 

The suffix is also found with the “restrictive” particle tofa/tofe-qe, which signals that the situation encoded in the clause is restricted to that particular one, when some other state of affairs might be expected in addition: (41) Qo { tofa / tofe-qe} qono ba-mu neri! 2sg.nfut restr / restr-qi sit lim-2sg.pers npast.here ‘You just sit here (doing nothing)!’ The suffix is also most likely historically part of another particle, one which marks ineffectuality of events: contrary to expectation, hope or desire, a given state of affairs does not result in another one: (42) Nau ku marungeqe raa ba-kuqa. 1sg 1sg.nfut ineff work lim-1sg.pers ‘I worked in vain.’ ‘I worked, and nothing.’ (For example, the speaker did work for somebody but got no reward in return, or only a very small reward; or he or she worked hard in the garden, but the result, in terms of the crops, was nil or negligible.) In spite of its phonological bulk, marungeqe is a particle. It can only be used together with a verb. Historically, the suffix, in the form qi, is probably also present in the immediatepast/immediate-future marker biqi: (43) = (15) Kera biqi lae bo=kau. 3pl.nfut imm go asrt=and ‘They have just left.’ The fact that some of the particles clearly take the suffix is evidenced by the fact that the suffix is optional; see (40) and (41) above. And, in fact, the suffix is also optional with the desiderative and the inceptive elements: (44) Nau ku thathami fanga. 1sg 1sg.nfut desid eat ‘I want to eat.’ Compare the synonymous sentence in (36b) further above with the suffix. (45) Diapana ka thafali lae mai uri fanu kia …. Japan 3sg.seq incep go vent all country pl(incl) ‘Japan/The Japanese began to come to our country ….’ (The suffix is also occasionally absent in associative noun phrases.)

 Frantisek Lichtenberk

6. Asymmetry reversal? There is then strong evidence that the Toqabaqita desiderative, undesiderative and phasal elements are bi-functional. They can function as verbs, in which case they take finite complement clauses, as we saw in 5, 6, 7 and 8 in section 2.1. And they can function as preverbal particles. Grammaticalization of a verb ‘want’ into a desiderative marker would not be at all unusual cross-linguistically. And Heine and Kuteva (2002) mention the development of verbs meaning ‘begin’ into inceptive markers, although they have not found many examples. I have not come across another instance of grammaticalization of a verb ‘not want’ into an undesiderative marker. When the three etymons function as verbs, they occur in matrix clauses that express assertions. On the other hand, when they function as particles associated with verbs, one would expect, following Cristofaro (2003), the clause/sentence that contains the verb to express an assertion. We would, then, have asymmetry reversal: a verb in an assertion-making clause in one construction corresponds to a particle in the other construction; and a verb in a non-assertion-making, complement clause in the former type of construction corresponds to the head of the predicate in an assertion-making clause in the latter type of construction. Obviously, this is not unique to this particular case. One would need to assume that every time, in any language, a higher complement-taking verb grammaticalizes into a particle or an auxiliary such a reversal takes place. But does it necessarily take place, and how does it take place? When does an erstwhile verb cease to serve to make an assertion? Is it simply a consequence of its becoming a grammatical element? Conversely, asymmetry reversal can be used as evidence that a grammaticalization process has occurred, when an element — a verb — that used to serve to make an assertion loses this function. As I was working on this paper, I came across a comment I had made to myself during one of my field trips to Toqabaqita. I had recorded a text where the speaker had produced the following construction with the verb ‘begin’ with the object suffix -a and a finite complement clause: (46) Manga na keki thafali-a keki qani-a oqola kera qeri, .... time rel 3pl.ipfv begin-3.obj 3pl.fut eat-3.obj garden 3pl that ‘When (lit. the time that) they begin to eat (the food from) those gardens of theirs, ....’ (See [3] in section 2.1 for partial continuation.) However, as I was transcribing the recording of the text, the speaker said, several times, as he was listening to the recording and dictating to me, what is in (47), with a reduced structure: (47) ... keki thafali-q qani-a …. 3pl.ipfv incep-qi eat-3.obj



Asymmetry reversal 

‘... they begin to eat....’ (In [47] the vowel of the suffix -qi has been deleted before a word that starts with q [glottal stop]. Optional deletion of the word-final vowel from a syllable that contains a consonant identical to the initial consonant of the following word is not restricted to the suffix -qi/-qe.) What was on tape and what the consultant dictated to me were two different constructions: one with a finite complement clause and the other one with the reduced structure. But, evidently, to him they were semantically identical, or sufficiently close in their meanings. Was there asymmetry reversal?3 There is a danger in relying exclusively on linguistic evidence when dealing with asymmetry between states of affairs, a cognitive phenomenon. There should be other kinds of evidence that such asymmetry exists and which way it goes, rather that relying exclusively on linguistic structures, such as the presence or absence of subordination. When a grammaticalization process reaches a certain point, a major semantic change, such as asymmetry reversal, may occur, but in the earlier stages of the process the relation between the structure and the meaning need not be fully fixed. One has to be careful about relying exclusively on linguistic evidence in drawing conclusions about conceptualization.

Abbreviations 1 2 3 abl add all and asrt assoc atten com comp desid detr du

first person second person third person ablative additive allative andative assertive associative attenuative comitative complementizer desiderative detransitivizer dual

3. Incidentally, the reduced structures show that, pace Cristofaro (2003), translation is not a reliable guide in deciding whether a given structure involves subordination or not.

 Frantisek Lichtenberk

dvn foc fut genp imm in incep incl ineff ints ipfv irr lim loc nfut npast.here obj pers persmkr pl prf profore prol prtt qi rel restr sben seq sg tr undesid vent

deverbal noun focus future general preposition immediate past/immediate future inessive inceptive inclusive ineffectual intensifier imperfective irrealis limiter locative nonfuture nonpast here object personal person marker plural perfect pronominal foregrounder prolative partitive -qi suffix relative restrictive self-benefactive sequential singular transitivizer undesiderative ventive



Asymmetry reversal 

References Cristofaro, S. 2003. Subordination. Oxford: Oxford University Press. Givón, T. 1980. The binding hierarchy and the typology of complements. Studies in Language 4: 333–377. Givón, T. 1984. Syntax: A Functional-typological Introduction, vol. I. Amsterdam/Philadelphia: John Benjamins. Givón, T. 1990. Syntax: A Functional-typological Introduction, vol. II. Amsterdam/Philadelphia: John Benjamins. Heine, B. and Kuteva T. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hooper, R. 1985. Proto-Oceanic *QI. In Austronesian Linguistics at the 15th Pacific Science Congress, A. Pawley and L. Carrington (eds), 141–167. Canberra: Pacific Linguistics. Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Langacker, R.W. 1991. Foundations of Cognitive Grammar, vol. II, Descriptive Application. Stanford: Stanford University Press. Ross, M.D. 2001. Proto Oceanic *i, *qi, and *-ki. In Issues in Austronesian Morphology: A Focusschrift for Byron W. Bender, J. Bradshaw and K.L. Rehg (eds), 259–278. Canberra: Pacific Linguistics.

Transparency vs. Economy How does Adioukrou resolve the conflict? Kaoru Horie, Prashant Pardeshi, and Guy Kaul

Tohoku University, Kawauchi / Kobe University / Kobe University

The tripartite distinction of subordinate clauses, i.e. relative, complement, and adverbial clause, is arguably motivated by the differing type of conceptual asymmetry between matrix clause and subordinate clause events encoded by each clause type. The three types of subordinate clauses are also known to manifest cross-constructional continuity in constructional space (Croft 2001). From cognitive-typological viewpoint, this paper will explore the polyfunctionality of a ‘versatile’ subordinator èké (‘which/that/if ’) in Adioukrou, a Kwa language within the Niger-Congo language family spoken in Côte d’Ivoire. Èké encodes three major subtypes of subordinate clauses, i.e. relative clause, complement clause, and conditional adverbial clause, and thus apparently neutralizes the tripartite distinction. This paper will also address the issue of how Adioukrou responds to a possible tension between two cognitivefunctional principles governing form-meaning correspondence in a language, i.e. transparency (isomorphism) and economy.

Keywords: Adioukrou, conceptual space, cross-constructional continuity, economy, Japanese, polyfunctionality, Radical Construction Grammar (RCG), subordination, transparency (isomorphism), ‘versatile’ subordinator

1. Introduction Polysemy or polyfunctionality is among the most productively explored topics in Cognitive Linguistics (Lakoff 1987, Nerlich et al. 2003, Croft and Cruse 2004). Polyfunctionality of grammatical markers is the norm rather than the exception cross-linguistically, as has been well demonstrated by case syncretism phenomena (e.g. English to encoding goal and dative recipient, Japanese ni encoding goal, dative recipient, and stative location). One of the linguistic regions where polyfunctionality is widely observed is the Southeast Asian linguistic area, as discussed by scholars like Matisoff (1991) and

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

Enfield (2002) (see also Noonan 1997). In the neighboring region of East Asia, polyfunctionality is also prevalent, as reported by Horie (1998a). The extent to which a language allows for polyfunctionality is arguably related to the issue of whether that language prioritizes faithfulness in form-meaning correspondence (transparency) at the cost of proliferation in forms, or economical use of finite forms at the cost of proliferation in surface ambiguity (Hawkins 1986, Horie 2002a). This study will explore the nature of a ‘versatile’ subordinator èké in Adioukrou, a Kwa language (Greenberg 1966, Stewart 1989) within the Niger-Congo language family spoken in Côte d’Ivoire. What is noteworthy of this morpheme is the fact that it serves to encode three major subtypes of subordinate clauses, i.e. relative clause, complement clause, and adverbial clause (Cristofaro 2003). Subordinate clauses entail an asymmetric relation between a matrix clause event and a subordinate clause event. Three types of subordinate clauses differ in terms of the nature/degree of conceptual asymmetry, which are reflected in their differential degrees of morpho-syntactic subordination (cf. Givón 1980, Croft 2001). Hence the cross-linguistic tendency for these three types of subordinate clauses to be distinguished, with the precise dividing lines determined by particular languages. It is thus remarkable that Adioukrou apparently ‘neutralizes’ the tripartite distinction. We will address this issue in greater detail. We will also address how Adioukrou responds to a possible tension between two cognitive-functional principles governing form-meaning correspondence in a language, i.e. transparency (isomorphism) and economy. The organization of this paper is as follows. Section 2 briefly introduces crosslinguistic studies of subordination relevant to the current study. Section 3 presents a cognitive-typological analysis of the èké-construction in Adioukrou. Section 4 explores a possible cognitive-functional principle motivating the use of èké in Adioukrou. Section 5 presents the conclusion and prospects.

2. Subordination: Typology and cross-linguistic variation Subordination is a complex morpho-syntactic and semantic phenomenon that has been studied by functional-typologists (e.g. Foley and Van Valin 1984, Haiman 1985, Cristofaro 2003). Together with coordination, subordination constitutes so-called complex clauses. Traditional definitions of subordination include a three-way subdivision into relative clause, complement clause, and adverbial clause, respectively illustrated in (1)-(3):

(1) (relative clause) the man [who bought the book]



(2) (complement clause) John thinks [{that/0} his wife told a lie].



(3) (adverbial clause) [If John thinks so], that’s fine.



Transparency vs. Economy 

The traditional notion of subordination has been criticized for being unable to exhaust all possible subtypes, leading to a proposal of ‘cosubordination’ (Foley and Van Valin 1984), or being too restrictive for the purpose of cross-linguistic comparison (Cristofaro 2003). In this paper, however, we will adopt the traditional three-way subdivision of subordinate clause into relative, complement, and adverbial clause, as it has proven useful in highlighting cross-linguistic coding variations in previous typological works (e.g. Comrie and Horie 1995, Dixon 1995). Comrie and Horie (1995) observe that languages differ in terms of whether complement clauses with a nominal head are treated similarly to complement clauses with a verbal head, or similarly to relative clauses. In languages like English, complement clauses with a nominal head and those with a verbal head pattern are alike, while relative clauses can be coded differently as in (4): (4) a. b. c.

the fact [{that/*which} John bought the book] John thinks [{that/*which} Bill bought the book] the book [{0/which}the man bought]

It should be noted, however, that relative clauses can be marked by the same marker that which codes complement clauses, as in (5):

(5) the book [that John bought for me]

This type of syncretism, which is by no means uncommon (e.g. the French que) appears to indicate conceptual continuity between the two types of subordinate constructions, relative clauses and complement clauses. That these two types of subordinate clauses are marked alike is arguably attributable to the fact that they both regularly occur in argument positions. Complement clauses, by definition, occur in the argument position, as in (4b). Relative clauses, which modify head nouns, regularly occur in the argument position as part of the noun phrase constituents, as in (6): (6) a. I lost [NP the book [that John bought for me]]. (direct object) b. [NP The man [who bought the book]] is a murder suspect. (subject) In fact, in some instances, the formal resemblance of two types of subordinate clauses is maximal, as in (7): (7) a. I forgot the fact [that John mentioned] (relative clause) b. I forgot the fact [that John mentioned it]. (complement clause) A less common type of syncretism is that between adverbial clauses and complement or relative clauses. With adverbial clauses being a non-argument (adjunct), it is natural that many languages formally distinguish them from the other two types of subordinate clauses that regularly occur in the argument position, as in (1)-(3) and Japanese examples (8):

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

(8) a. [ Taroo-ga kita no/*toki]-o atode sit-ta. Taro-NOM came COMP/when-ACC later know-PAST ‘I later found out that Taro came.’ (complement clause) b. [ Taroo-ga kita toki/*no], dare-mo ie-ni inakat-ta. Taro-NOM came when/COMP who-also home-at not-exist-PAST ‘When Taro came, no one was at home.’ (adverbial clause) However, there are cases where adverbial and non-adverbial clauses are coded in a like manner, as in (9): (9) a. [ Taroo-ga kita no]-ni kizuka-nakat-ta. Taro-NOM came COMP-DAT become aware-NEG-PAST ‘(I) wasn’t aware that Taroo had come.’ (complement clause) b. [ Taroo-ga kita noni] kizuka-nakat-ta. Taro-NOM came though become aware-NEG-PAST ‘Though Taro came, (I) wasn’t aware of it.’ (adverbial clause) These two structures are superficially indistinguishable. Furthermore, the interpretations given to (9a) and (9b) do not appear to be very different. This suggests the possibility that languages do not necessarily distinguish between adverbial clause and non-adverbial clause formally, especially when the interpretations encoded therein are not very distinct. That adverbial and non-adverbial clauses are coded by the same subordinating marker (though differentiated through position) is observable in other languages like English: (10) a. b.

(complement clause) John doesn’t know [whether Mary will attend the party]. (concessive adverbial clause) [Whether Mary attends the party or not], John will definitely show up.

(11) a. b.

(complement clause) John doesn’t know [if Mary will attend the party]. (conditional adverbial clause) [If Mary attends the party], John will definitely show up.

The conceptual continuity between (10a) and (10b) is rather straightforward in that both types of subordinate clauses encode two alternative possibilities. In contrast, the conceptual continuity between (11a) and (11b) is less straightforward. However, it remains true that both types of subordinate clauses encode the presence of uncertainty. While it is not uncommon for two types of subordinate clauses (e.g. relative-complement, complement-adverbial) to share linguistic coding, it is cross-linguistically rather rare for three types of subordinate clauses viz. relative, complement, and adverbial clauses to be marked similarly. This is indeed the case with the subordinator èké in Adioukrou. We will closely examine this phenomenon in Section 3.



Transparency vs. Economy 

3. ‘Versatile’ subordinator èké in Adioukrou We draw particular attention to the morpheme èké used in (12), which functions as a relativizer. The same morpheme is also used to code complement and adverbial clauses, as in (13) and (14): (12) (relative clause) ígŋ/j jfw à èké ìm dàbú man/woman the EKE AOR-go Dabou ‘the man/woman who went to Dabou’ (13) (complement clause) mέl dâd -m èké lí b-ôw-n-ím bâŋn Mel PPERF-say me EKE he FUT-come village ‘Mel told me that he will go to the village.’ (14) (conditional adverbial clause) èké m- f\´ EKE I

os

mí -b-ow-m-eb êl

PPERF-have money I

FUT-buy

house

‘If I have money, I will buy a house.’ The following examples culled from the bible translation corpus collected by Guy Kaul attest the three subordinating functions èké is put to serve: (15)

ké krîst ét∫í dŠàm ùs εs èké Then Christ POSS back follow father EKE f\ ôb à/jεdŠí sós έm íŋn èké AOR-have thing the also body in HORT-animate EKE \àmf\ ìs és à sósJέm / God HORT-take bring Verb. P the because ‘And the followers of Jesus who have wealth, they should glory in being brought low.’

(16)

ow ìt´ èké lí- k-úw á\ èké ìn ét∫í it suffice EKE it DES-know face EKE he POSS ôb f\ à èl-m ` lów kaakâ thing have the PPERF NEG-be affair nothing wùs áf áŋá / Earth on here ‘What it would be good to know is that the wealth of the rich (will last no longer) is nothing here on earth.’

(17) èhè él εgŋ èké b-íŋm kríkrí εbm Happiness HORT-be man EKE HAB-stand firm suffice

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

εkn ówíj εm à ét∫í see business in the POSS ‘Blessed be anyone who perseveres when trials come.’ (Lit.: Happiness for the one who stands firm in the temptation.’ (18)

èké n-ì\m és kríkrí a sósjεm \am EKE heAOR-stand Verb. P firm the reason God b-ów- fŋ -n sél a èké lFUT-give him prize/breath the EKE he εwar agŋ èké erur -(í)r a AOR-promise people EKE AOR-love him the ‘Such a person is of proven worth and will win the prize of life, the crown that the Lord has promised to those who love him.’ (Lit.: if he stood firm, God will give him the life that he has promised to people who love him.)

(19)

èké mεdŠu ow´ εgŋ εm à / ké líEKE lie PPERF-come man in the that he k^^-dad´ lí- k-εsε / \am ànâ ké PROHIB-say he PROHIB-say God it is that εdŠu- m  à / aŋké \am mεdŠu ówíj/ PPERF-lie me the then God lie affairs ànŋ-m ìn εm PPERF NEG-walk he/she/it in ‘Never, when you are being put to the test, say, ‘God is tempting me’, God cannot be tempting by evil.’ (Lit.: If you are telling lies, don’t say ‘God lied to me’. For God is not in the lie.)

As confirmed in examples (15)-(19), we can appreciate that the èké construction can be interpreted as a relative clause, complement clause, or conditional adverbial clause as it occurs in the actual context of usage. It is an intriguing question how Adioukrou native speakers resolve the potential ambiguity of èké-marked clauses contextually. Our preliminary analysis indicates that iconic word order (cf. Haiman 1985a, b, Tai 1985) crucially serves to distinguish the ‘conditional adverbial’ (adjunct) interpretation (see example 19), on the one hand, and the ‘relative/complement’ (argument) interpretations, on the other (see examples 15–17). Specifically, the former (‘adjunct’) interpretation is iconically signaled by the sentence-frontal (non-embedded) position, while the latter (‘argument’) interpretations are licensed by the sentence-medial (embedded) positions. Similar distributional patterning of argument vs. adjunct clauses is not uncommon, as observed with English subordinating markers whether and if in (10) and (11). This iconically motivated patterning of relative/complement vs. conditional adverbial clauses is confirmed by the token frequency of èké in Adioukrou bible and tale texts collected by Guy Kaul, as shown in Table 1.



Transparency vs. Economy 

Table 1.  Distribution of subordinate clause marker èké in Adioukrou texts

Relative Clause Complement Clause Conditional Adverbial Clause

Sentence-initial

Sentence-medial

0% 0% 46% (100%)

184 (100%) 28 (100%) 0%

The token frequency patterns presented in Table 1 clearly demonstrate that the ordering of èké within a sentence is sensitive to the subordinating function it is put to serve, i.e. whether èké codes a relative/complement clause or a (conditional) adverbial clause. The fact that the type of polyfunctionality manifested by èké is rather rare crosslinguistically is confirmed by the following informal survey reported in Table 2 below: Table 2.  Cross-linguistic coding variation of relative, complement, and conditional adverbial clauses Languages

Subordinating marker Relativizer

Complementizer

Conditional Adverbializer

Adioukrou Chinese

èké de

èké –

èké

English French German Japanese Korean Marathi Mongolian Spanish

which, that etc. qui, que etc. der/die/das (tokoro) – – – que, quien

that, if que daB koto/no/tokoro/to kes/ko – – que

rugua o if si wenn nara/tara/ba myen dzar avbal si

On the basis of the data presented in Table 2, we arrive at a semantic map in Figure 1, which captures the cross-linguistic variation in the shared coding patterns of relative, complement, and conditional adverbial clauses: We can see from Figure 1 that languages differ in terms of the extent to which subordinated clauses are marked differently (or not). German and Korean, which use different markers for all three types of subordinate clauses, are located at one end of the scale. Adioukrou, which uses the same marker for all three types of subordinate clauses, is located at the other end. In between these two poles are languages that manifest partial syncretism such as French, Japanese, and English.

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

Complement clause

Relative clause

Conditional Adverbial clause

German der/die/das

daß

wenn

kes/ko

myen

Korean – French que

si

Japanese nara, tara, etc

tokoro English

if

that Adioukrou èké

Figure 1.  Cross-linguistic distribution of subordinate clause marking

cosubordination

complements

adverbial clauses

ive

se

o rp

pu

s

se

co rre lat

u cla rel at

ive

internally headed relative clauses

cla u

se

s

adjoined relative clauses

serial verbs, paratactic clauses, speech complements

coordination

relative clauses

Figure 2.  The continuum of complex sentence types (Croft 2001: 322)



Transparency vs. Economy 

How can we interpret this three-way polyfunctionality of èké? We argue that this apparently ‘rampant’ polyfunctionality of Adioukrou èké is by no means anomalous in view of Radical Construction Grammar (henceforth RCC; Croft 2001). RCC proposes the continuous conceptual space (or continuum) of complex sentence types covering four pivotal constructions, i.e. relative clauses, complements, coordination, and adverbial clauses, as shown in Figure 2: What is unique about the Adioukrou èké-construction is that, unlike constructions documented in Croft (2001) that span the region between two pivots in Figure 2 (for example, complements and relative clauses), the èké-construction appears to cover a wider conceptual space. This space spans the region demarcated by three construction types, i.e. relative clauses, complements, and adverbial clauses, thereby confirming that these three constructions do indeed form a continuum, as suggested by RCC. Constructions like èké-construction appear to be rare cross-linguistically. The cross-constructional continuity (Horie 1993) between these four construction types can be manifested in varying ways cross-linguistically. Adioukrou èké-construction presents one manifestation type covering three contiguous constructions on the conceptual space viŠ. complement, relative and adverbial clauses. Other languages show a different type of manifestation covering two pivotal constructions mediated by an ‘intermediate’ construction. Japanese is a case in point. As extensively discussed by Horie (1993, 1998b, 2002a, b, 2003), Japanese has a construction known as ‘internally headed relative clauses’ that mediates between the two pivotal constructions ‘relative clauses’ and ‘adverbial clauses’, as in (20) and (21): (20) [Ringo-ga tukue-no ue ni oite aru no]-o tabe-ta. apple-NOM desk-GEN top-LOC be put COMP-ACC eat-PAST ‘I ate an apple (which/as) it was put on the desk.’ (21) [Doroboo-ga heya-kara detekita tokoro]-o tukamae-ta. thief-NOM room-from came out PLACE-ACC catch-PAST ‘I caught a thief (who/as) he came out of the room.’ It appears that some languages favor this type of cross-constructional continuity to a greater extent than other languages. Japanese belongs to the former group of languages, as demonstrated by constructions shown in (9), (20), and (21). In contrast, in spite of superficial structural similarity to Japanese, its neighboring language Korean doesn’t favor such cross-constructional continuity, as suggested by Horie (1993, 1998b, 2002a, b, 2003). The translational equivalent of (20) in Korean, for instance, is not fully acceptable: (22) *[ Sakwa-ka chayksag wi-ey nohye iss-nun kes]-ul mek-ess-ta. apple-NOM desktop-LOC be put exist-ADN:PRES-ACC eat-PAST-DECL ‘I ate an apple (which/as) it was put on the desk.’ It is interesting to note that tokoro (‘place’) in Japanese, which spans the region between complement clauses and adverbial clauses, partially shows similar distribution to èké-construction in covering three pivotal constructions (relative clauses,

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

complements, and adverbial clauses). Though relative clauses in Japanese are normally formed by gap strategy without any overt relative pronoun, in some highly formal written registers, tokoro can be used to code a relative clause, as in (23) (the relativizer tokoro is thus put in parentheses in Table 2): (23) Kore-wa [watasi-ga kat-ta (tokoro-no)] hon desu. this-TOP I-NOM buy-PAST place-GEN book COP ‘This is the book that I bought.’ There are two crucial differences between Adioukrou èké and Japanese tokoro. First, unlike tokoro, èké doesn’t exhibit skewing in any of the three functions it is employed for, as observed in examples (15)-(19). Second, èké doesn’t appear to attest (at least at the current stage of investigation) ‘intermediate’ constructional use, unlike tokoro.

4. What motivates the use of èké in Adioukrou? We noted in Section 3 that languages can differ in terms of the extent to which they allow for cross-constructional continuity and suggested that Adioukrou, like Japanese, arguably belongs to those languages that favor cross-constructional continuity. Complement clause

Relative clause

Conditional Adverbial clause

German der/die/das

daß

wenn

kes/ko

myen

Transparency

Korean – French que

si

Japanese nara, tara, etc

tokoro English that

if

Adioukrou èké

Economy

Figure 3.  Cross-linguistic distribution of subordinate clause marking in relation to the principles of Transparency and Economy



Transparency vs. Economy 

It is widely agreed among functionally and typologically oriented linguists (e.g. Du Bois 1985, Haiman 1985a, b, Hawkins 1986) that there is a tension between Isomorphism (transparency) and Economy, two fundamental cognitive-functional principles governing form-meaning mapping cross-linguistically. In Horie (1998b, 2002a, b, 2003), inspired by collective insight, it was suggested that Japanese and Korean respectively prioritize Economy and Isomorphism. Among the linguistic manifestations of this contrast is the higher vs. lower likelihood of cross-constructional continuity as exemplified in a pair of sentences (16) and (18), suggesting that Japanese speakers show a greater preference to use a single form/construction to have multiple meanings and to disambiguate potential ambiguity contextually. This point is demonstrated by Figure 3, which is extended from Figure 1 in Section 3. It presents cross-linguistic variation in subordinate clause marking as it relates to the two cognitive-functional principles. We can see that Adioukrou is located at one end where Economy is prioritized, while languages like German and Korean are located at the other end where Transparency (Isomorphism) is the norm. The cross-linguistic variation presented in Figure 3 accords with the findings by Hawkins (1986) and Horie (1998b, 2002a, b, 2003). According to these studies, German and Korean are characterized as languages prioritizing transparency in form-meaning mapping in contrast to English and Japanese, which prioritize economy and are more tolerant of economy. Is the versatile/economical use of èké an isolated phenomenon? Though our investigation is at a preliminary stage, we have additional linguistic evidence suggesting that this is not the case, i.e. ‘versatile’ demonstratives na& and âmwa. Adioukrou demonstratives na& (singular) and âmwa (plural) are used to refer to humans as well as non-humans. The demonstratives na& and âmwa are used as demonstrative determiners in (24a), (24b), (25a) and (25b), and as demonstrative ‘identifiers’ in (24c) and (25c). These demonstratives do not indicate the gender distinction between masculine and feminine, nor do they indicate the distinction between ‘proximal’ and ‘distal’ reference. This is observed in examples (24) and (25). (24) na& a. [ígŋ naa] εadŠù-m  εgŋ kaakâ SG man DEM HAB-NEG-lie mankind no ‘This/that man does not lie to anyone.’ b. [ jów na&] εadŠù-m  εgŋ kaakâ SG woman DEM HAB-NEG-lie mankind no ‘This/that woman does not lie to anyone.’ c. naa jów /ígŋ /ndèj / kpókó / líkŋ DEM SG woman/SG man/SG meat/SG plate/SG tree ‘This/that is a woman/man/animal/plate/tree.

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul

(25) âmwa a. [ έgŋ âmwa] âm-kôk sób àp a màmn fεŋ man DEM PROG-do thing offer good all ‘These/those men are doing all that is good.’ b. [ fjów âmwa] âm-kôk sób àp amamn fεŋ women DEM PROG-do thing offer good all ‘These/those women are doing all that is good.’ d. âmwa fjów/égŋ/mndèj/síkpókpó/ékŋ DEM PL woman/PL man/PL meat/PL plate/PL tree ‘These/those are women/men/plates/trees.’ (Lit.: These/those ones are women/men/plates/trees.’) These ‘versatile’ morphemes arguably suggest that Adioukrou speakers favor the principle of Economy in using a single form/construction with potentially multiple interpretations, the precise identification of which is made in actual context of use.

5. Conclusion The cross-constructional continuity as advocated by Radical Construction Grammar offers an explanation in the understanding of the ‘versatile’ subordinator èké in Adioukrou, which encodes relative, complement, and conditional adverbial clauses. Its use is arguably motivated by the principle of Economy and contextual disambiguation, which appears to be a preferred communicative strategy used by Adioukrou native speakers. Further cross-linguistic investigation into cross-constructional continuity phenomena is expected to uncover the nature of intricate cognitive-functional motivations for the use of ‘intermediate’ constructions in those languages in which they are attested.

Abbreviations ACC ADN AOR COMP COP DAT DECL DEM DES

Accusative Adnominal Aorist Complementizer Copula Dative Declension Demonstrative Desiderative



Transparency vs. Economy 

FUT GEN HAB HORT LOC NEG NOM NP PAST POSS PPERF perfect PROHIB TOP Verb. P

Future Genitive Habitual Hortative Locative Negative Nominative Nominal phrase Past Possessive Past PRES: Present Prohibitive Topic Verbal particle

Acknowledgements This study was supported in part by the Tohoku University 21st Century Center of Excellence (COE) Program (Humanities) (Ministry of Education, Culture, Sports, Science and Technology). Our thanks go to an anonymous reviewer, the participants of the ICLC 2005 theme session “Asymmetric Events” (July 2005, Seoul, Korea), the participants of CSDL 7 (October 2004, Alberta, Canada), Andrew Barke, Chris Wade, and, most of all, to the editor of this volume, Barbara Lewandowska-Tomaszczyk, for her advice and encouragement.

References Comrie, Bernard, and Kaoru Horie. 1995. Complement Clauses versus Relative Clauses: SomeKhmer Evidence. In: Abraham, Werner, et al.  (eds.), Discourse Grammar and Typology. Amsterdam: John Benjamins, 65–75. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Croft, William. 2001. Radical Construction Grammar. Oxford: Oxford University Press. Croft, William, and Alan Cruse. 2004. Cognitive Linguistics. Cambridge: Cambridge University Press. Dixon, Robert W. 1995. Complement Clauses and Complement Strategies. In: Palmer, F. R. (ed.), Meaning and Grammar. Cambridge: Cambridge University Press, 174–220.

 Kaoru Horie, Prashant Pardeshi, and Guy Kaul Du Bois, John. 1985. Competing Motivations. In: Haiman, John (ed.) Iconicity in Syntax. Amsterdam: John Benjamins, 343–365. Enfield, Nick J. 2002. Linguistics Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia (Curzon Asian Linguistics). London: Routledge Foley, William and Robert Van Valin Jr. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Greenberg, J.H. 1966. The Languages of Africa. Berlin: Mouton. Haiman, John. 1985a. (ed.) Iconicity in Syntax. Amsterdam: John Benjamins. Haiman, John. 1985b. Natural Syntax. Cambridge: Cambridge University Press. Hawkins, John A. 1986. A Comparative Typology of English and German. Berlin: Croom Helm. Horie, Kaoru. 1993. A Cross-linguistic Study of Cognition and Perception Verb Complements. A Cognitive Perspective. Unpublished doctoral dissertation, University of Southern California. Horie, Kaoru. 1998a. On the Polyfunctionality of the Japanese Particle No: From the Perspectives of Ontology and Grammaticalization. In: Ohori, Toshio. (ed.), Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives. Tokyo: Kurosio Publishers, 169–192. Horie, Kaoru. 1998b. Functional Duality of Case-marking Particles in Japanese and its Implications for Grammaticalization: A Contrastive Study with Korean. In: Silva, David. (ed.), Japanese/Korean Linguistics 8. Stanford: CSLI [distributed by Cambridge University Press], 147–159. Horie, Kaoru. 2002a. Verbal Nouns in Japanese and Korean: Cognitive Typological Implications. In: Kataoka, Kuniyoshi, and Ide Sachiko (eds.) Culture, Interaction, and Language. Tokyo: Hizuzi publishers, 77–101. Horie, Kaoru. 2002b. A Comparative Typological Account of Japanese and Korean MorphoSyntactic Contrasts. Eonehag 32 (The Linguistic Society of Korea): 9–32. Horie, Kaoru. 2003. What Cognitive Linguistics Can Reveal about Complementation in non-IE languages: Case Studies from Japanese and Korean. In: Eugene H.Casad and Gary B. Palmer (eds.), Cognitive Linguistics and Non-Indo European Languages. Berlin: Mouton de Gruyter, 363–388. Lakoff, George. 1987. Women, Fire, and Dangerous Things. Chicago: The University of Chicago Press. Matisoff, James A. 1991. Areal and Universal dimensions of Grammaticalization in Lahu. In: Traugott, Elizabeth Closs, and Bernd Heine (eds.), Approaches to Grammaticalization. Vol. II. Amsterdam: John Benjamins, 383–452. Nerlich, Brigitte, Zazie Todd, Vimala Herman, and David D. Clarke (eds.) 2003. Polysemy. Flexible Patterns of Meaning in Mind and Language. Berlin: Mouton de Gruyter. Noonan, Michalel. 1997. Versatile Nominalizations. In: Bybee, Joan, John Haiman, and Sandra A. Thompson. (eds.) Essays on Language Function and Language Type. Amsterdam: John Benjamins, 373–394. Stewart, John M. 1989. Kwa. In: Bendor-Samuel John (ed.) The Niger-Congo Languages. Lanham, MD: University Press of America, 216–245. Tai, James H. Y. 1985. Temporal Sequences and Chinese Word Order. In: Haiman, John (ed.) Iconicity in Syntax. Amsterdam: John Benjamins, 49–72.

Relating participants across asymmetric events Conceptual constraints on obligatory control* Klaus-Uwe Panther University of Hamburg

This chapter discusses a specific type of asymmetric events, namely a configuration where a non-expressed participant in a subordinate event must be identical with a participant in the superordinate event of which it is a part. The phenomenon is known as “obligatory control” in generative grammar and illustrated by sentences such as John begged Mary to play computer games with him, in which the main clause argument Mary determines (“controls”) the reference of the understood subject in the infinitive clause. Traditionally, obligatory control has often been regarded as a purely formal relation, describable in terms of a Minimal Distance Principle or some other syntactic constraint. Approaches that rely on the notion of thematic role have also been proposed. In contrast, the present contribution, following work by Köpcke and Panther, develops a conceptual-pragmatic account of obligatory control. It is claimed that control involves a matching process between conceptual-pragmatic roles of a participant in the superordinate clause and the understood subject of the embedded non-finite clause. Two Pragmatic Role Identity Principles account for the basic facts of obligatory control into non-finite complement clauses in English and German. These principles provide a unified account of a large class of cases of obligatory control phenomena involving syntactic subject control, object control, and control shift.

Keywords: conceptual-pragmatic roles, control shift, English, German, object control, obligatory control, Pragmatic Role Identity Principles, subject control

1. Introduction When two events are coded in the same sentence, the question arises how these two events are related to each other and how this relation is coded in language. One * This research is associated with Francisco Ruiz de Mendoza’s project HUM200405947-C02-01/FILO (University of La Rioja, Spain).

 Klaus-Uwe Panther

possibility is that the relation between the two events is asymmetric: one event is conceptualized epistemically or ontologically as being dependent on another event. This conceptual configuration can be coded through hypotaxis: the dependent event is syntactically expressed in the subordinate clause; the conceptually superordinate event is expressed in the main clause. In this paper I consider a specific class of asymmetric events in which one and the same participant partakes in both the superordinate event and in the subordinate event. One may then ask how the referential identity of this participant across the embedding event and the embedded event is coded by speakers and how it is retrieved by hearers on the basis of their grammatical competence and, in particular, their conceptual, encyclopedic and contextual knowledge. A specific case of participant identity across events is known as ‘obligatory control’, a relation of referential dependence between one argument in a main clause and a nonexpressed argument in an embedded non-finite infinitival or gerund clause.1 In syntactic terms, this latter silent argument is often called the “understood subject” or “null” subject of the subordinate clause – in generative grammar known as the “big” PRO.2 Obligatory control is generally contrasted with “arbitrary” or “free” control. The difference between these two kinds of control is illustrated by sentences (1) and (2), respectively:

(1) Johni begged Maryj [Øj to play computer games with him]. (obligatory control)



(2) [Øgen waiting in crowded airports] is a drag. (non-obligatory control)

In (1) the person supposed to play computer games with John is Mary – in grammatical terms, the matrix clause object. This argument is called the controller; the argument in the subordinate clause whose reference depends on the controller is, accordingly, named the controllee. Coreference between controller and controllee is indicated by means of identical subscripts. In (2) there is no argument in the superordinate clause that functions as controller; hence the silent argument in this sentence exemplifies what is called ‘non-obligatory’ or ‘free’ control, and the subscript gen on Ø marks it as ‘generic’.3 Control verbs include, among others, the English verbs promise, ask, persuade, advise, recommend, thank and their respective counterparts in German, viz. versprechen, 1. Jackendoff and Culicover (2003) and Culicover and Jackendoff (2005: ch. 12) prefer the term ‘unique control ’to ‘obligatory control’. 2. The terms ‘understood subject’ and ‘null subject’ suggest that these arguments represent phonetically empty syntactic functions. I will argue, however, that they are really arguments that are present only in conceptual structure and find no expression on the syntactic level. This is also the line of thought of Jackendoff and Culicover (2003: 520) and Culicover and Jackendoff (2005: 419–20). In what follows I will use the empty set symbol Ø for non-syntacticized arguments instead of PRO. 3. Jackendoff and Culicover (2003) and Culicover and Jackendoff (2005) distinguish between unique control, free control, and nearly free control. Since I am only concerned with unique or obligatory control I can ignore the distinction between free and nearly free control here.



Relating participants across asymmetric events 

bitten, überreden, raten, empfehlen, and danken. In English, control verbs are used in constructions such as (3) and (4): 4

(3) NPi V (P) NPj [Ø to VPaction]



(4) NPi V (P) NPj (P) [Ø VP-ingaction]

Construction types (3) and (4) are exemplified by sentences (5) and (6), respectively:

(5) Maryi asked Johnj [Øj to support her application]



(6) Maryi thanked Johnj for [Øj for supporting her application]

The German equivalents would normally be rendered by means of infinitival constructions such as (7) or nominal constructions like (8): (7) Mariai bat Hansj [Øj ihre Bewerbung zu unterstützen]. Mariai asked Hansj [Øj her application to support] ‘Mary asked Hans to support her application’ (8) Mariai dankte Hansj für die Unterstützung ihrer Bewerbung. Mariai thanked Hansj for the support of.her application ‘Mary thanked Hans for supporting her application’ The above-listed control verbs have one common property. They prototypically select a complement clause that denotes an action performed by the referent of the silent argument of the complement clause. This chapter is concerned exclusively with control into non-finite action complements. As the subsequent discussion will reveal, obligatory control constitutes a complex conceptual and pragmatic problem that, to my knowledge, has hardly ever been touched upon by cognitive linguistics.5 The control problem becomes especially intriguing when the matrix event involves two participants, as in (1). The hearer’s interpretive task is then to select one of the two available matrix arguments as the intended controller of the silent argument in the non-finite subordinate clause. I will illustrate the conceptual complexities of control

4. In fact, control may also be exerted by nominals such as in the plea by John to Athena to take care of herself (Culicover and Jackendoff 2005: 441) (controller: Athena) or adjectives as in John is grateful to Greta for helping him out with a loan (controller: Greta). This paper is exclusively concerned with control verbs. 5. Exceptions are Panther and Köpcke (1993), Panther (1994), Panther (1997), Köpcke and Panther (2002). Jackendoff and Culicover (2003) and Culicover and Jackendoff (2005: Ch. 12), who work in a generative framework, also present a conceptual solution to the control problem.

 Klaus-Uwe Panther

with two languages, English and German, and show that their conceptual control properties are basically the same.6

2. General hypotheses The claims I want to defend in this paper are as follows: – Obligatory control in non-finite complement clauses is primarily motivated by conceptual and pragmatic factors, i.e. only minimally by syntactic factors. – English and German abide by the same general conceptual-pragmatic constraints on control. In Section 3, I provide evidence for the claim that control is governed by conceptual and pragmatic principles, arguing against a syntactic solution to the control problem. Section 4 draws some conclusions.

3. Syntactic and semantic-pragmatic determinants of control 3.1

Syntactic constraints on control

In the generative literature there is a tradition that views control (analogously to anaphoric and pronominal binding) as determined by purely syntactic conditions. The first generative linguist to assume such structural conditions was Rosenbaum (1967, 1970), who proposed a Minimal Distance Principle (MDP), according to which the argument closest to the controllee (in terms of the number of branches in a phrase structure tree) is chosen as the controller. This principle correctly predicts the behavior of object control verbs (in active matrix clauses) such as in (5): of the two possible controllers the matrix clause object argument John is chosen. But it makes the wrong prediction with regard to control structures with commissive control verbs like promise, vow, agree (with), etc.:

6. Culicover and Jackendoff (2005: 417) find it likely “that control behaves crosslinguistically in much the same fashion […]”, although they concede that there are “within limits” some language-specific differences. In fact, as Panther and Köpcke (1993), Panther (1994), Panther (1997) and Panther (2001) have demonstrated, there are a number of differences in control behavior between English and German, which lack of space prevents me from discussing in this paper. Possibly, these differences correlate with other typological contrasts between English and German (see e.g. Hawkins 1986, 1987).



Relating participants across asymmetric events 



(9) John promised Susan to take care of himself. (Culicover & Jackendoff 2005: 433) 7

(10) John vowed to/pledged to/agreed with/is obligated to Susan to take care of himself. (Culicover & Jackendoff 2005: 433) In both (9) and (10) the controller is John, not Susan, as (wrongly) predicted by the MDP. The control properties of verbs denoting commissive speech acts have thus to be marked as exceptional in Rosenbaum’s framework. In addition, as pointed out e.g. by Panther and Köpcke (1993), Panther (1994) and Köpcke and Panther (2002), Rosenbaum’s MDP selects the wrong controller in cases of control switch (also called control shift): (11) Jimmy asked his mother (to be allowed) to go the playground. In (11), the referent of the silent (subject) argument in the infinitival clause is most likely the subject of the matrix clause, Jimmy. Rosenbaum’s MDP however selects the object his mother as controller. These few examples suffice to demonstrate that the MDP is an implausible candidate for an adequate account of obligatory control. Another putative syntactic constraint has been proposed by Manzini (1983), who claims that the controller must be located in the immediately containing matrix clause. This constraint does not say anything about which of the available arguments is selected, but restricts the clausal domain in which they have to occur. As an illustration of Manzini’s constraint, consider (12): (12) Maryi promised Johnj [that Evek would ask her nephewl [Ø*i/*j/*k/l to take care of the matter]]. In the preceding example neither Mary nor John are possible controllers of Ø, but the controller has to be selected from the immediately superordinate clause, in this case the object argument her nephew. One can however question the syntactic nature of this constraint; it might turn out to be a consequence of the semantics and pragmatics of the control verbs promise and ask, respectively. A second constraint that, at least at first sight, does not seem to be conceptually or pragmatically motivated is illustrated by the following pair of sentences: (13) Barbarai told Klausj [Øj/*k to submit his paper by the end of February]. (14) Barbarai told Klausj that hej/k should submit his paper by the end of February. Sentences (13) and (14) differ in an important respect: in (13) the understood complement clause subject must be coreferential with a matrix argument (here: the matrix 7. Although promise is often used with two matrix arguments in syntactic textbooks and scholarly articles, as in Mary promised John to exercise regularly, many native speakers find this construction unacceptable and would prefer a finite complement clause in such cases: Mary promised John that she would exercise regularly. In actually occurring language use, there is a strong tendency for the indirect object of the matrix clause to remain unexpressed when promise is used with an infinitival complement: e.g. Mary promised to exercise regularly.

 Klaus-Uwe Panther

object), whereas in (14) the pronoun he may be coreferential with a matrix argument, but may also refer back to a third person not mentioned in the clause but retrievable from the linguistic or extralinguistic context (here indexed by the subscript k). The constraints on coreference are thus much tighter for the non-expressed subject in nonfinite complement constructions than for pronouns in finite complement constructions (see Köpcke and Panther (2002: 198) for more detailed discussion of this point). In fact, non-finite complement clauses with control verbs of the sort listed above require the presence of a controller in the matrix clause. Examples like (15), where Ø has no coreferential antecedent in the matrix clause, are consequently ill-formed: (15) *Lindai asked Maryj [Øk to send her an update on recent developments in American politics]. Culicover and Jackendoff (2005: 434) show convincingly that obligatory control does not depend on particular syntactic constructions. For example, the sentence John ordered Susan to take care of herself has the same control properties as John gave Susan some order to take care of herself or the noun phrase the order to Susan from John to take care of herself. One can even think of discourse structures such as Take care of herself; that’s what John ordered Susan to do. In all of the above examples, the controller is Susan and this has more to do with the semantics and pragmatics of the predicate order than with its word class (verb vs. noun) or the syntactic configuration it occurs in. In conclusion, control is, if at all, only minimally governed by structural factors. In the following section, I argue that the main burden of controller identification is borne by the conceptual system, which, in Cognitive Linguistics, includes information about communicative function and inferences about utterance meaning.8

3.2

Conceptual and pragmatic constraints on obligatory control

3.2.1 Prototypical control In this section, I develop the thesis that the main factors influencing control behavior are conceptual and pragmatic (see also Panther and Köpcke 1993, Panther 1994, Köpcke and Panther 2002). In particular, I assume that the following parameters guide controller choice: i. the conceptual and pragmatic properties of the control verb; ii. the conceptual content of the non-finite subordinate clause; iii. the pragmatic roles assumed by the arguments in the superordinate and the subordinate clause; 8. The importance of semantic factors in control theory has been recognized by a number of linguists working in the generative and formal semantics tradition. A non-exhaustive list of important works includes Comrie (1984, 1985), Ružička (1983, 1999), Farkas (1988), Ladusaw and Dowty (1988), Wegener (1989), Sag and Pollard (1991), Pollard and Sag (1994), and more recently, Jackendoff and Culicover (2003) and Culicover and Jackendoff (2005).



Relating participants across asymmetric events 

iv. additional encyclopedic knowledge about the arguments in the matrix clause, e.g. assumptions about social relations among participants; v. the linguistic and extralinguistic context. For reasons of space I limit myself to the discussion of parameters (i)–(iii), especially, pragmatic roles. For some discussion of factors (iv) and (v), see Köpcke and Panther (2002). The significance of pragmatic roles in the determination of control relations has been advocated over a period of fifteen years by Panther and Köpcke (1993), Panther (1994,) Köpcke and Panther (2002). Many of the matrix verbs that prototypically occur with actional complement clauses are speech act verbs – most of them illocutionary and some of them perlocutionary in kind. The notion of pragmatic role used as an analytical tool in this paper is derived from speech act theory. As is well known, speech acts depend for their appropriate performance on the fulfillment of a set of felicity conditions (see e.g. Searle 1969). Consider the speech act of promising. In a felicitous promise, the promisor undertakes a commitment to a future action that benefits the promisee. I claim that this pragmatic information has to be integrated in a proper account of control and that a theory of control that exclusively relies on thematic roles misses an important generalization.9 To begin with, consider a standard textbook example like (16): (16) Mary promised Brian to finish her thesis by May. This sentence exhibits what is referred to as ‘subject control’ in the generative literature: The subject of the matrix clause determines the reference of the understood subject of the non-finite complement clause (see Figure 1). Theta Roles THEME AG

ACTION

REC

AG

ACTION

Mary

promised

Brian

Ø

to finish her thesis by May.

AG

ACTION

BEN

AG

ACTION

Pragmatic Roles

Figure 1.  Semantic and pragmatic roles in the promise frame 9. Ružička (1983) presents a control theory that, for the most part, relies on thematic roles. Ružička argues that subject control (control verbs of the promise class) exhibit thematic role identity between the controller and the controllee; object control is supposed to abide by a principle of thematic distinctness between controller and controllee.

 Klaus-Uwe Panther

In Figure 1, two kinds of roles are distinguished. First, there are traditional semantic roles, also known as thematic roles (theta roles) in Government-and-Binding and Principles-and-Parameters Theory. The verb promised has three arguments that receive theta roles; Mary is the agent (ag) of promised, Brian might be classified as the recipient (rec) of the control verb; finally, there is a propositional argument syntacticized as an infinitive clause in (16), which receives the neutral theta role theme.10 Furthermore, there is an independent assignment of theta roles within the infinitive clause; here I only indicate the agent role assigned by the infinitive finish to the understood argument Ø. Second, the arguments in (16) are assigned pragmatic roles. Thematic roles will prove to be dispensable in the formulation of control relations. However, pragmatic roles are essential to state important generalizations about how controllers are identified. One limitation of theta roles is that they are “local”; each individual predicate assigns its own argument roles. As a consequence, an account of control relations in terms of theta roles does not explain the fact that Mary in (16) is not only the agent of promised but also the agent of the prospective action coded in the complement clause, viz. finish her thesis by May. Furthermore, theta roles do not say anything about the relation between the action denoted by the complement clause and the indirect object of the matrix clause. However, it is certainly of relevance that Brian is the intended beneficiary (ben) of the promised action in (16). As I will argue below, the relation between the beneficiary and the proposition denoted by the complement clause must be taken into account in a theory that aims at predicting the right control relations. Consider now an example exhibiting ‘object control’: (17) Mary asked Brian to finish his thesis by May. In Figure 2, sentence (17) is represented in terms of theta roles and pragmatic roles. From this example it becomes even more transparent that it is necessary to distinguish between theta roles and pragmatic roles. Again, I claim that the theta role approach ignores important conceptual and pragmatic information. In terms of theta role assignment, the verb asked allocates the role agent to the matrix clause subject Mary; the direct object of the matrix clause Brian may be categorized as patient (pat), and the infinitive clause can be regarded as a kind of abstract goal.11 In (17) the patient Brian controls the reference of the understood subject of the non-finite subordinate clause. Thus, sentences (16) and (17) seem to exemplify quite distinct types of control.

10. There is no consensus about how many theta roles there are and how they should be categorized. It does not matter for my argument, however, what kind of role should be allocated to the propositional argument of (16). 11. Again it does not matter for the point to be made what the nomenclature for thematic roles is.



Relating participants across asymmetric events 

Theta Roles GOAL AG

ACTION PAT

Mary

asked

Brian

BEN

ACTION AG

AG Ø AG

ACTION to finish his thesis by May. ACTION

Pragmatic Roles

Figure 2.  Semantic and pragmatic roles in the ask frame

However, for an adequate account of the control properties of (17), the theta roles assigned to the arguments are irrelevant. I claim that what is really relevant for the determination of control relations in (17) is what might be called the speech act scenario prompted by the sentence (here: the “request scenario”). This scenario includes information about felicity conditions and the pragmatic roles of the participants (see e.g. Thornburg and Panther 1997, Panther and Thornburg 1998). Among other things, the request scenario contains the important pragmatic information that the asker (requester) is the prospective beneficiary of the requested action and that the requestee is the prospective agent of the action. In terms of theta roles, Mary in (17) is the agent of asked, but, what is more important for the assignment of control relations is that Mary has the pragmatical role of beneficiary (of the requested action). Brian is assigned the theta role patient, but in pragmatic terms Brian is the prospective agent of the requested action. To repeat the main point, it is the pragmatic roles of the participants that are crucial in the determination of control relations. A comparison of the two examples (16) and (17) diagrammed in Figures 1 and 2, respectively, shows that in both cases the controller has the pragmatic role agent and the controllee also has the pragmatic role agent.12 It follows that both subject and object control abide by one general principle: controller and controllee have the same pragmatic role. We thus arrive at the following generalization:13

12. For the controllee the distinction between theta roles and pragmatic roles is neutralized. I will assume that Ø is both semantically and pragmatically an agent (ag). 13. See also the detailed discussion of what follows in Panther and Köpcke (1993), Panther (1994), and Köpcke and Panther (2002).

 Klaus-Uwe Panther

First Principle of Pragmatic Role Identity (agent Identity) If the pragmatic role of the understood subject is agent, search for an antecedent in the matrix clause that also has the pragmatic role agent and interpret this argument as controller. Sentence (18), whose pragmatic role configuration is represented in Figure 3, illustrates the First Principle of Pragmatic Role Identity with an example that involves subject control. Since theta roles are, by hypothesis, irrelevant to the assignment of control relations, they have been omitted in Figure 3. Coreference between controller and controllee is symbolized by a double-headed arrow. (18) Mary vowed to Susan to stop smoking. Maryi AG

vowed to Susan j Øi to stop smoking. BEN

AG

Figure 3.  agent Identity Principle: subject control

Sentence (19), which is analyzed in Figure 4, exemplifies the principle of agent Identity in a case of object control. (19) John asked his boss to raise his salary. Johni BEN

asked his boss j AG

Øj to raise his salary. AG

Figure 4.  agent Identity Principle: object control

German functions exactly like English in this respect. The following two examples (20) and (21) (see Figure 5), one with the subject control verb versprechen ‘promise’ and the other with the object control verb bitten ‘ask’, demonstrate that the First Principle of Pragmatic Role Identity holds for this language as well. (20) Marie versprach Paul, sein Gehalt zu erhöhen. ‘Marie promised Paul that she would raise his salary’ (21) Marie bat Paul, ihr Gehalt zu erhöhen. ‘Marie asked Paul to raise her salary’



Relating participants across asymmetric events 

Mariei versprach Paul j Mariei promised Paul j

Øi sein Gehalt zu erhöhen. Øi his salary to raise

AG

AG

BEN

Mariei bat Paul j Mariei asked Paul j

Øj ihr Gehalt zu erhöhen. Øj her salary to raise

BEN

AG

AG

Figure 5.  agent Identity Principle: German examples

To conclude, for predicates selecting propositional arguments with an actional meaning, the basic conceptual and pragmatic principle of agent identity appears to hold for both English and German, and, most likely, for other languages as well. The predicate of the matrix clause (which denotes the superordinate event) implies a pragmatic agent that is identical to the agent performing the action expressed in the complement clause (the subordinate event). Identity between agents can be regarded as the “prototypical” or “unmarked” case of obligatory control into non-finite complement clauses. However, as I show in the following section, it is not strictly necessary for the complement clause to overtly code an action. This fact requires the stipulation of an additional principle of pragmatic role identity, which, however, will turn out to follow quite naturally from the First Principle of Pragmatic Role Identity. 3.2.2 Non-prototypical control As shown above, control verbs like ask, promise, thank, etc., prototypically select an action complement. But what happens when these verbs are combined with a complement clause that does not denote an action? Consider the following sentence: (22) Mary asked Barbara to be invited to the party. Sentence (22) contains a passive complement clause and thus does not correspond to the prototypical control scenario discussed in Section 3.2.1. Furthermore, the most natural interpretation of (22) seems to be that Mary (the matrix clause subject) wants to be invited to the party. In other words, Mary is the controller of the silent argument of the complement clause. Sentence (22) exemplifies a switch from object control (the prototypical controller with directive control verbs) to subject control. The distribution of pragmatic roles in (22) is diagrammed in Figure 6 below.

 Klaus-Uwe Panther

Maryi asked Barbara j BEN

AG

Øi to be invited to the party. BEN

Figure 6.  Control switch: object to subject control

As can be seen from Figure 6, the First Principle of Pragmatic Role Identity is not applicable in this case, since the pragmatic agent argument is not matched by a subject argument with the same role in the complement clause. However, by virtue of the felicity conditions of the requestive speech act ask, there is a matrix argument with the role beneficiary (Mary), and the silent argument of the complement clause can also pragmatically be construed as being a beneficiary. Being invited to a party is certainly, in most circumstances, beneficial to the invited party.14 There is thus a role configuration with two beneficiaries that can be linked through a coreference relation. On the basis of such as data as (22), I suggest that there is a second control principle for the class of control verbs analyzed in this paper, which accounts for those cases in which the First Principle (agent identity) does not hold. Second Principle of Pragmatic Role Identity (beneficiary Identity) If the pragmatic role of the understood subject is not agent, but beneficiary, search for an antecedent in the matrix clause that also has the pragmatic role beneficiary and interpret this argument as controller. This principle also applies in German, as can be seen from the following examples. Let us first consider a case of control switch from subject to object control: (23) Barbara versprach Maria eingeladen zu werden. ‘Barbara promised Maria that she (= Maria) would be invited’ Figure 7 represents the pragmatic role configuration in (23). The indirect object Maria (the promisee) is a potential beneficiary of the propositional content of the promise and the same role can be assigned to the silent argument of the infinitival clause. Barbara i Barbara i AG

versprach Maria j Øi eingeladen zu werden. promised Maryj Øi invited to be BEN

BEN

Figure 7.  beneficiary Identity Principle: Control switch from subject to object control

14. Since ben is a pragmatic role, it is, in principle, defeasible. For some reason, Mary might not benefit from the party that is referred to in (22) at all.



Relating participants across asymmetric events 

In contrast to (23), sentence (24) exemplifies control switch from prototypical object control to subject control: (24) Barbara bat Maria eingeladen zu werden. ‘Barbara asked Marie to be invited’ The relevant role constellation is diagrammed in Figure 8. Barbara i Barbara i BEN

bat Maria j Øi eingeladen zu werden. asked Mary j Øi invited to be AG

BEN

Figure 8.  beneficiary Identity Principle: control switch from object to subject control

The examples from English and German discussed in this section contain non-actional complement clauses. I have shown that the Principle of Pragmatic Role Identity can also be applied to these cases. The coreference relation obtains between two arguments that are both pragmatically categorized as beneficiary. However, a closer semantic and pragmatic examination of these examples reveals that even if the complement clause is not coded as an action, an implied actional interpretation of it is mandatory. There is thus a mismatch between syntactic structure and conceptual structure in sentences (22)–(24), to which I turn in the following section. 3.2.3 The role of metonymic coercion As pointed out by various scholars (e.g. Sag and Pollard 1991, Panther 1994, Jackendoff and Culicover 2003, Culicover and Jackendoff 2005), a control sentence like (22) repeated here as (25) Mary asked Barbara to be invited to the party. has an interpretation such as ‘Mary asked Barbara to do something to the effect so that Mary would be invited to the party’. The reading of the sentence is that Mary asks Barbara to do something, although the requested action is not coded in the utterance. What is coded is the result of this unspecified action. The conceptual phenomenon at hand is known as coercion (Pustejovsky 1993) or semantic shift (Talmy 2000: 324). I argue that the coercive process of interpreting the coded content of the complement clause in (25) as the result of an action is induced by a pervasive conceptual principle

 Klaus-Uwe Panther

in English, the result for action metonymy.15 The relevant pragmatic role configuration is depicted in Figure 9.

Mary i asked Barbara j [Øi BEN

AG

to be invited to the party]]

BEN

RESULT FOR ACTION

Maryi asked Barbara j [Øj CAUSE [Øi to be invited to the party]] BEN

AG

AG

BEN

Figure 9.  Control and metonymic coercion

The German sentences (23) and (24) can be interpreted in an analogous way: both versprechen ‘promise’ and bitten ‘ask, request’ conceptually involve an action performed by the promisor and the requestee, respectively. The same kind of conceptual conflict as for (25) arises: The complement clause does not code an action, but a meaning that can be interpreted as the result of some unspecified action. For (23), repeated here as (26), the metonymic shift from result to action can be diagrammed as in Figure 10. (26) Barbara versprach Maria eingeladen zu werden. ‘Barbara promised Maria that she (= Maria) would be invited’ The conceptual structure of (26) is enriched by a predicate of causation and a causing agent, and it is this argument that is coreferential with the agentive argument of the matrix clause. On the conceptual-pragmatic level, there are, as in the case of (25) (see Figure 9), two coreference relations, only one of which shows up syntactically – the coreference link between the two arguments labeled ben. “Control shift”, i.e. the “non-prototypical” or “marked” case of control, thus turns out to be a phenomenon of grammatical form; on the level of conceptual structure, the class of control predicates I have been considering in this chapter enforces an actional interpretation of the complement clause. To summarize, the actional interpretation of non-finite complement clauses is not always explicitly coded in languages like English and German. Nevertheless, the 15. The role of the result for action metonymy in coercive processes has been described e.g. by Panther and Thornburg (1999), Panther and Thornburg (2001), Panther (2005: 363–365), Ruiz de Mendoza and Pérez (2001), and Ruiz de Mendoza and Otal (2002: 85–91).



Relating participants across asymmetric events 

control properties of the sentences discussed in this section ultimately follow from their meaning – not from their syntax. An important result of the discussion in this section is then that, on the conceptual level, the First Principle of Pragmatic Role Identity holds for both the prototypical cases of control as well as for the non-prototypical (coerced) cases. The latter, it turns out, are only apparently exceptional.

Barbarai Barbarai AG

versprach Mariaj [Øj promised Mariaj [Øj BEN

eingeladen invited

zu werden] to be]

BEN

RESULT FOR ACTION

Barbarai Barbarai AG

versprach Mariaj [Øi CAUSE [Øj eingeladen zu werden]] promised Mariaj [Øi CAUSE [Øj invited to be]] BEN

AG

BEN

Figure 10.  Control and metonymic coercion: versprechen ‘promise’

4. Conclusion In this chapter I have argued that the phenomenon known as control in the generative literature has a conceptual and pragmatic basis. Soon after Rosenbaum proposed his Minimal Distance Principle in the late 1960s, its flaws became obvious and many control theorists began to look for a semantic solution to the control problem. The present contribution is in line with this thinking but goes one step further in suggesting that the notion of pragmatic role is crucial to an adequate account of obligatory control. Pragmatic roles have to be sharply distinguished from ‘semantic cases’ as proposed by Fillmore (1968), which were later adopted as ‘thematic roles’ by many generative linguists under the influence of Chomsky’s Government-and-Binding framework. I have demonstrated that a theory that relies on pragmatic roles has one important advantage over an approach based on thematic roles: It allows a unified account of a large class of cases of obligatory control phenomena involving syntactic subject control, object control, and control shift.

 Klaus-Uwe Panther

References Comrie, Bernard. 1984. Subject and Object Control. In Brugman, Claudia et al. (eds.) Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society, February 17–20, 1984. Berkeley, CA: Berkeley Linguistics Society, 450–464. Comrie, Bernard. 1985. Reflections on Subject and Object Control. Journal of Semantics 4, 47–65. Culicover, Peter W. / Jackendoff, Ray. 2005. Simpler Syntax. Oxford: Oxford University Press. Farkas, Donka F. 1988. On Obligatory Control. Linguistics and Philosophy 11, 27–58. Fillmore, Charles. 1968. The Case for Case. In Bach, Emmon / Harms, Robert (eds.) Universals in Linguistic Theory. New York: Holt, Rinehart and Winston, 1–88. Hawkins, John A. 1986. A Comparative Typology of English and German: Unifying the Contrasts. London: Croom Helm. Hawkins, John A. 1987. German. In Comrie, Bernard (ed.) The World’s Major Languages. London & Sidney: Croom Helm, 110–138. Jackendoff, Ray / Culicover, Peter W. 2003. The Semantic Basis of Control. Language 79, 517–556. Köpcke, Klaus-Michael / Panther, Klaus-Uwe. 2002. Zur Identifikation leerer Subjekte in infinitivischen Komplementsätzen – ein semantisch-pragmatisches Modell. Folia Linguistica 36, 191–217. Ladusaw, William A./ Dowty, David R. 1988. Towards a Non-grammatical Account of Thematic Roles. In Wilkins, Wendy (ed.) Thematic Relations. San Diego: Academic Press, 62–73. Manzini, Rita. 1983. On Control and Control Theory. Linguistic Inquiry 14, 421–446. Panther, Klaus-Uwe. 1994. Kontrollphänomene im Englischen und Deutschen aus semantischpragmatischer Perspektive. Tübingen: Gunter Narr. Panther, Klaus-Uwe. 1997. An Account of Implicit Complement Control in English and German. In Verspoor, Marjolijn / Lee, Kee Dong / Sweetser, Eve (eds.) Lexical and Syntactical Constructions and the Construction of Meaning. Amsterdam / Philadelphia: Benjamins, 417–432. Panther, Klaus-Uwe. 2001. Syntactic Control. In Smelser, Neil J. / Baltes, Paul B. (eds.) International Encyclopedia of the Social & Behavioral Sciences. New York: Elsevier Science, 15397–15401. Panther, Klaus-Uwe. 2005, The Role of Conceptual Metonymy in Meaning Construction. In Ruiz de Mendoza Ibañez, Francisco / Peña Cervel, M. Sandra (eds.) Cognitive Linguistics: Internal Dynamics and Interdisciplinary Interaction. Berlin / New York: Mouton, 353–386. Panther, Klaus-Uwe / Köpcke, Klaus-Michael. 1993. A Cognitive Approach to Obligatory Control Phenomena in English and German. Folia Linguistica 27, 57–105. Panther, Klaus-Uwe / Thornburg, Linda L. 1998. A Cognitive Approach to Inferencing in Conversation. Journal of Pragmatics 30, 755–769. Panther, Klaus-Uwe / Thornburg, Linda L. 1999. Coercion and Metonymy: The Interaction of Constructional and Lexical Meaning. In Lewandowska-Tomaszczyk, Barbara (ed.) Cognitive Perspectives on Language (Polish Studies in English Language and Literature 1). Frankfurt am Main: Peter Lang, 37–52. Panther, Klaus-Uwe / Thornburg, Linda L. 2001. The effect for cause metonymy in English Grammar. In Barcelona, Antonio (ed.) Metaphor and Metonymy at the Crossroads: A Cognitive Perspective (Topics in English Linguistics 30). Berlin: Mouton de Gruyter, 215–231.



Relating participants across asymmetric events 

Pollard, Carl / Sag, Ivan A. 1994. Head-Driven Phrase Structure Grammar. Chicago: Chicago University Press. Pustejovsky, James. 1993. Type Coercion and Lexical Selection. In Pustejovsky, James (ed.) Semantics and the Lexicon. Dordrecht: Kluwer, 73–96. Rosenbaum, Peter. 1967. The Grammar of English Predicate Complement Constructions. Cambridge, MA: MIT Press. Rosenbaum, Peter. 1970. A Principle Governing Deletion in English Sentential Complementation. In Jacobs, Roderick A. / Rosenbaum, Peter S. (eds.) Readings in English Transformational Grammar. Waltham, MA: Ginn and Company, 20–29. Ruiz de Mendoza, Francisco / Otal, José L. 2002. Metonymy, Grammar and Communication. Granada: Comares. Ruiz de Mendoza, Francisco J. / Pérez, Lorena. 2001. Metonymy and Grammar: Motivation, Constraints and Interaction. Language and Communication 21, 321–357. Ružička, Rudolf. 1983. Remarks on Control. Linguistic Inquiry 14, 309–324. Ružička, Rudolf. 1999. Control in Grammar and Pragmatics: A Cross-linguistic Study. Amsterdam / Philadelphia: Benjamins. Sag, Ivan I. / Pollard, Carl. 1991. An Integrated Theory of Complement Control. Language 67, 63–113. Searle, John. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Talmy, Leonard. 2000. Toward a Cognitive Semantics, Vol. II: Typology and Process in Concept Structuring. Cambridge, MA and London: The MIT Press. Thornburg, Linda / Panther, Klaus. 1997. Speech Act Metonymies. In Liebert, Wolf-Andreas / Redeker, Gisela / Waugh, Linda (eds.) Discourse and Perspective in Cognitive Linguistics (Current Issues in Linguistic Theory 151). Amsterdam / Philadelphia: Benjamins, 205–219. Wegener, Heide. 1989. Kontrolle – semantisch gesehen. Deutsche Sprache 17, 206–228.

The Portuguese inflected infinitive and its conceptual basis Augusto Soares da Silva Catholic University of Braga

This paper offers a conceptual analysis of the Portuguese inflected (or personal) infinitive. The analysis makes use of a few concepts of Cognitive Grammar. The inflected infinitive is described as an objective construal of the main participant of a process type. This conceptual import is consistent with the grammatical behavior of the inflected infinitive. After giving an overview of the structural properties of the inflected infinitive in modern Portuguese, the paper will describe the continuum of event asymmetry in which the inflected infinitival clause holds an intermediate position between the subordinate finite clause and the infinitival clause. The meaning of the inflected infinitive is therefore characterized by imagery dimensions that are specific to subordinate constructions, namely independence, specificity, grounding, prominence and objectivity/subjectivity. Finally, some aspects relating to the distribution of inflected infinitival (complement, adverbial and subject) clauses will be analyzed in terms of semantic compatibility.

Keywords: inflected infinitive, Portuguese, Langacker, Cognitive Grammar, viewing arrangement, objective construal, grounding, event asymmetry, subject clause, complement clause, adjunct clause, adverbial clause, word order, imagery, uninflected infinitival construction, inflected infinitival construction, finite construction, event integration/asymmetry, sequential scanning, summary scanning, grounding, personal inflection, focal prominence, grounding, objectification, clausal grounding, nominal grounding, person category, subjective construal, subjectification, indicative/subjunctive alternation, volition verbs, modal verbs, factive verbs, nominalization, declarative and epistemic verbs, causative and perception verbs, causation, Though Raising construction, adverbial constructions, impersonal constructions

 Augusto Soares da Silva

1. Introduction1 The inflected infinitive is an old and typical phenomenon of the Portuguese language and almost always appears when the infinitive has an overt subject. The infinitive inflection is marked for person and indicates subject-verb agreement. The inflected (or personal) infinitive and the uninflected infinitive both exist and have quite a complex distribution. In an influential article, Raposo (1987) takes on Chomsky’s Principles and Parameters framework and builds up an analysis of the syntax of the Portuguese inflected infinitive. His analysis is based on the crucial idea that in inflected infinitival clauses the functional category Infl (specifically, Agr(eement) in Infl) can assign nominative case to the subject only if it is itself assigned case. The reasoning behind this hypothesis is that the functional category Agr is nominal by nature in a null subject language, like Portuguese, and therefore must be case-marked. In finite clauses, Agr is case-marked by tense. Based on the assumption that infinitives do not have tense, Raposo states that infinitival Agr must ‘look for’ an external case assigner. Thus, inflected infinitival clauses are expected to occur only in case-marked positions.2 This study will attempt to show that the inflected infinitive is not a meaningless element controlled by syntactic rules, but a meaningful element of which the grammatical behavior reflects its conceptual import. Within the framework of Langacker’s (1987, 1991, 1999) Cognitive Grammar it will be shown that the notions of viewing arrangement, objective construal and grounding are crucial to the understanding of the meaning and use of the Portuguese inflected infinitive. Firstly, an overview of the structural and distributional properties of the inflected infinitive will be given. Secondly, the conceptual continuum of event asymmetry starting from infinitival clauses to finite (subordinate) clauses will be analyzed and then a description of the inflected infinitive semantic import will be attempted. Finally, a few aspects concerning the distribution of inflected infinitival clauses will be examined.

1. I am grateful to Ronald Langacker, Ricardo Maldonado and Michel Achard for their valuable comments and suggestions at the early stage of this study. Short versions of this paper were presented at the International Conference “From Gramm to Mind: Grammar as Cognition” in Bordeaux and the 9th International Cognitive Linguistics Conference (in the theme session about “Asymmetric events”) in Seoul, in May and July 2005, respectively. I am grateful to the respective audiences for their comments. This research was funded by the Portuguese Foundation for Science and Technology and by the Orient Foundation of Portugal. 2. Later studies within the generative framework and more specifically within the Minimalist program, like Ambar’s (1994, 2000) and Mensching’s (2000) studies, have pointed out a few critical points in Raposo’s account. See also Martins’ (2001) generative study about the complex issue of the origin of the Portuguese inflected infinitive.



The Portuguese inflected infinitive and its conceptual basis 

2. Grammatical properties and distribution of the inflected infinitive The inflected infinitive in Modern Portuguese shares some morphophonological and syntactic properties with both finite and uninflected infinitival forms of the verb. Inflected infinitival forms provide the infinitival morpheme -r(e)- (INF) with subjectagreement suffixal morphemes (encoding person and number, P/N) that are present in most finite paradigms of the verb. In Table 1, the paradigm of the inflected infinitive can be compared to the uninflected infinitive one-form paradigm and to the present and pluperfect indicative paradigms (VR indicates the verbal root, TV the thematic vowel and T the tense morpheme). Table 1.  Inflection paradigms – verb estudar ‘to study’

1sg 2sg 3sg 1pl 2pl 3pl

inflected infin. vr-tv-inf-p/n estud-a-r estud-a-re-s estud-a-r estud-a-r-mos estud-a-r-des estud-a-re-m

uninflected infin. vr-tv-inf estud-a-r

present indic. vr-tv-p/n estud-o estud-a-s estud-a estud-a-mos estud-a-is estud-a-m

pluperfect indic. vr-tv-t-p/n estud-a-ra estud-a-ra-s estud-a-ra estud-a-ra-mos estud-a-re-is estud-a-ra-m

As for finite clauses, inflected infinitival clauses take lexical subjects marked as nominative and display overt subject-verb agreement morphology. Compare the inflected infinitival clause in (1) to the finite clause in (2); in addition, contrast the inflected infinitival clause in (1) with the uninflected infinitival clause in (3). (1) Será difícil eles resolverem o problema. will be difficult they solve-infl.infin-3pl the problem ‘It will be difficult for them to solve the problem.’ (2) Será difícil que eles resolvam o problema. will be difficult that they solve-3pl the problem ‘It will be difficult that they solve the problem.’ (3) * Será difícil eles resolver o problema. will be difficult they solve-infin the problem Contrary to finite clauses and like uninflected infinitival clauses, inflected infinitival clauses do not occur as independent or main clauses but are restricted to embedded clauses and as such they never take the complementizer que ‘that’. This is illustrated in (4), as opposed to (5), where relevant clauses are finite. (4) a. * Eles resolverem o problema. they solve infl.infin-3pl the problem

 Augusto Soares da Silva

b. * Será difícil que eles resolverem o problema. will be difficult that they solve-infl.infin-3pl the problem. (5) a. b.

Eles resolveram o problema. ‘They solved the problem.’ Será difícil que eles resolvam o problema. ‘It will be difficult for them to solve the problem.’

Examples (1) and (2) show the two construction possibilities that exist in Portuguese when the verb person is made explicit. In the very same context, the other Romance languages only allow the finite complement construction, illustrated in (2). The inflected infinitive occurs in a great number of syntactic structures of many different types and its use is optional in various contexts. This makes any description about the use of inflected infinitive rather difficult. Various attempts have been made by grammarians and linguists but without any success.3 For this reason, some grammarians have asserted that there are no dependable rules governing the use of the inflected infinitive in Portuguese, while others argue that the choice of the inflected infinitive is more properly a stylistic issue rather than a grammatical one. This is a complex issue in the sense that various factors contribute to the use of the inflected infinitive, which therefore cannot be predicted in any absolute way. Three different types of factors can be identified: structural, semantic and pragmatic. Syntactically, the inflected infinitive can only occur in a biclausal construction (see example (6) where the construction is monoclausal). In a recent study, Scida (2004) suggests this property as the main condition for the use of the inflected infinitive. But this necessary condition is not, however, sufficient. Semantically, the personal meaning of the inflected infinitive is coded grammatically, that is to say, it has its own subject which may or may not be lexically expressed and may or may not be the same as that of the governing verb. Even though the uninflected infinitive does not refer to any subject, it can be revealed by the context as in (7), or it can be manifested by some other function like in (6b). Finally, pragmatic or communicative factors like the need to avoid ambiguity can determine the use of the inflected infinitive, including cases where structural or semantic factors would not lead one to expect it. Example (8) shows that the unexpected inflection of the second infinitive (termos ‘have-1pl’) is due to the distance from the main verb. (6) a. * A Maria fez/deixou/viu correrem os miúdos Mary made/let/saw run-infl.infin-3pl the children b. A Maria fez/deixou/viu correr os miúdos. Mary made/let/saw run-infin the children ‘Mary made/let/saw the children run.’

3. Among these attempts Sten’s study (1952) stands out and above all Maurer’s patient work (1968) in which long lists of rules about the use of both infinitives are displayed along with examples.



The Portuguese inflected infinitive and its conceptual basis 

(7) Queremos sair. want-1pl go-infin out ‘We want to go out.’ (8) Não esperávamos encontrá-lo aqui e ainda menos termos a honra da visita. no expected-1pl find-infin him here and even less have-infl.infin-1pl the honor of the visit ‘We didn’t expect to find him here and even less to have the honor of his visit.’ Regular use of the inflected infinitive in Modern Portuguese when the infinitive comes with an overt subject results from these properties – see examples (1) and (3) again. There is only a minor exception to this sufficient, but not necessary, condition which occurs with causation/perception verbs, when, as in (9), the infinitive logical subject (os miúdos ‘the children’/-os ‘them’) is encoded as a direct object of the main verb (Soares da Silva 2004). In Old Portuguese, however, uninflected infinitival forms can be found with an overt subject. In contrast, the case of a subject, known from the context but non-overt, favors the possibility of using both infinitives. (9) A Maria fez/deixou/viu os miúdos /-os correr. Mary made/let/saw the children/-them run-infin Consider the distribution of the inflected infinitive in Modern Portuguese. Inflected infinitival clauses occur in three major syntactic environments. Firstly, they occur as subject clauses, both in impersonal and non-impersonal constructions, including constructions introduced by a determiner as exemplified in (10). Secondly, they appear as complement clauses subcategorized by different types of verbs, namely factive, epistemic, and declarative verbs as in (11), and as causation/perception verbs as well as prepositional complement clauses of verbs, adjectives and nouns, as in (12). Thirdly, they occur as adjunct clauses (mainly adverbial clauses) introduced by a preposition, as in (13). (10) a. b.

(Eles) ganharem as eleições significa uma mudança radical. (they) win-infl.infin-3pl the elections means a change radical ‘The fact they won the election means a radical change.’ O/O facto de (eles) ganharem as eleições significa uma mudança radical. the/the fact of (they) win-infl.infin-3pl …

(11) a. b.

O João lamenta eles não terem resolvido o problema. John regrets they no have-infl.infin-3pl solved the problem ‘John regrets they haven’t solved the problem.’ O João pensa/afirma terem eles encontrado a solução. John thinks/affirms have-infl.infin-3pl they found the solution ‘John thinks/says they found the solution.’

(12) a. Aconselhou os amigos a estudarem matemática. advised-3sg the friends to study-infl.infin-3pl mathematics ‘He advised his friends to study mathematics.’

 Augusto Soares da Silva

b. Eles estão ansiosos por encontrarem a solução. they are anxious for find-infl.infin-3pl the solution ‘They are anxious to find the solution.’ (13) a. b.

Regressaram sem conseguirem resolver o problema. came back-3pl without manage-infl.infin-3pl solve the problem ‘They came back without having solved the problem.’ Trabalharam muito para conseguirem este resultado. worked-3pl a lot for obtain-infl.infin-3pl this result ‘They worked a lot to get this result.’

Inflected infinitival clauses, however, are not allowed as complements of volition verbs, see (14), and of causation/perception verbs when these are immediately followed by the infinitive, i.e the monoclausal construction exemplified above in (6). They also cannot occur as embedded relative/interrogative clauses, see (15), and as independent or main clauses, as in (4) above. (14) * Eles querem resolverem o problema. they want solve-infl.infin-3pl the problem ‘They want to solve the problem.’ (15) * Nós não sabemos quem convidarmos para a festa. we not know who invite-infl.infin-1pl for the party ‘We don’t know who to invite to the party.’ There is also another aspect of distribution that will be analyzed further, namely the internal word order of inflected infinitival clauses. The most relevant fact is that epistemic and declarative verbs do not allow a preverbal (canonical) position of the subject of the infinitive.

3. From the infinitival clause to the finite clause: clines of imagery Infinitival and finite subordinate clauses represent different but well-related continuum clines and the inflected infinitival clause holds an intermediate point in these clines. Table 2 systematizes relevant clines. They represent imagery parameters (Langacker 1987) specific to finite/infinitival subordinate constructions. Sentences (a), (b) and (c) from (16)-(18) exemplify these contrasts: (a) uninflected infinitival construction, (b) inflected infinitival construction, and (c) finite construction. (16) and (17) are samples of complement constructions and (18) exemplifies adverbial constructions. (16) a. b.

É preciso fazer uma pausa. is necessary make-infin a break É preciso fazermos uma pausa. is necessary make-infl.infin-1pl a break



The Portuguese inflected infinitive and its conceptual basis 

c. É necessário que façamos uma pausa. is necessary that make-1pl a break ‘We need to take a break.’ (17) a. b. c.

Eles reconheceram ter errado. they admitted have-infin wrong Eles reconheceram terem errado. they admitted have-infl.infin-3pl wrong Eles reconheceram que erraram. they admitted that mistook-3pl ‘They admitted they were wrong.’

Table 2.  Clines of imagery more

less independence

independent V: all markings no raising specified+nominative subject biclausal structure instance of a process type sequential scanning temporal more verbal expression full grounding

full prominence

independent V: many markings no raising specified+nominative subject biclausal structure process specificity (personal) process type summary scanning holistic and atemporal grounding minimum (personal) grounding prominence trajector prominence

independent V: no marking raising, cl.climbing/bare inf. unspecified/raised subject monoclausal structure process type summary scanning holistic and atemporal more nominal expression no grounding

no event prominence

conceptual distance and control more distance, less control + objectivity: event + subjectivity (subj. axis) finite (subordinate) clause

less distance, more control objectivity/subjectivity objectivity: trajector inflected infinitival clause

(18) a. Alegram-se por ver o pai. rejoice themselves for see-infin the father

– objectivity: process – subjectivity (subj. axis) uninflected infinitival clause

 Augusto Soares da Silva

b. c.

Alegram-se por verem o pai. rejoice themselves for see-infl.infin-3pl the father Alegram-se porque vêem o pai. rejoice themselves because see-3pl the father ‘They are cheerful because they are seeing their father.’

The clines of Table 2 will not be described in detail as this is beyond the scope of this study. Only a few short remarks will be made. – Independence of the subordinate event with respect to the main event. The inflected infinitival clause codifies a greater conceptual independence and a minor event integration/asymmetry than the alternative uninflected infinitival clause. Hence, the inflected infinitival verb has many of the markings of an independent verb: it has person and number markings, it preserves all its argument structure such as the subordinate finite verb (there is no argument raising nor clitic climbing); furthermore, its subject is marked in the nominative, just as the subject of an independent or dependent finite clause. Moreover, the inflected infinitival clause can vary with the subordinate finite clause in almost any context. Notice that the ‘independence parameter’ properties indicated in Table 2 for the uninflected infinitival construction are only found in complement clauses in which the infinitive and the main verb compound into a single complex verb (see Soares da Silva (2004) in relation to causation/perception constructions); these properties cannot be found in example (17a). – Specificity of the subordinate process. By definition, an infinitive merely codes a process type while a finite clause describes a specific elaboration (or grounded instance) of a process type; in between the inflected infinitive adds specification of its main participant to the process type meaning, allowing a less general reading of that process. Thus, both infinitives do not profile the temporal determination of a process (or sequential scanning), but only give a holistic view of that process provided by a summary scanning and impose in this way an atemporal construal on a processual base (Langacker 1987: chap. 7). Nevertheless, tense and mood markers can be found in an infinitival clause, as in (17a, b). The infinitival clause might also represent a specific process, as in examples (a) de (16)-(18), but the finite clause always codifies a more elaborated and specific situation. – Grounding or relation to the speech situation (Langacker 1987, 1991). By definition, the finite clause (more precisely, the indicative clause) is fully grounded4, when the uninflected infinitival clause is not grounded (because the uninflected infinitive codes a process type). It is interesting that the Portuguese inflected infinitive, even though it does not provide any information about the epistemic status of the conceptualized process (location in reality and time relative to the speech 4. An indicative clause is ‘fully grounded’, because it includes tense predications that locate the conceptualized process within reality (see Achard’s account 1998: chap. 6, more particularly the distinction between indicative and subjunctive clauses as regards levels of grounding).



The Portuguese inflected infinitive and its conceptual basis 

act), does include some kind of grounding predication: indeed, personal inflection establishes a relation between the main participant in the conceptualized subordinate event and the speech act participants (speaker/hearer), and according to Achard (1998: 50), this can be described as grounding predication. This means that the Portuguese language can split what other languages conflate in a single inflection, thus it is possible in Portuguese to separate the person category from the tense and mood categories and then identify who is actually involved in the event, even if that event is not actually considered in respect to reality.5 Anyway, the idea of personal grounding predication is not unproblematic, as will be referred to later. – Prominence: the inflected infinitival clause has got a higher degree of prominence than the uninflected one for two reasons: first, because it provides information about the subject of the subordinated event, either through inflection only or through the lexical presence of the subject as well; second, because this participant is the trajector of the event (by definition, a trajector has got a high degree of inherent cognitive relevance – Langacker 1987). Naturally, prominence is greater when the trajector is explicitly referred to. – Conceptual distance and control: greater distance (iconically marked by greater formal complexity) generally corresponds to minor subject control on the subordinate scene, and vice-versa. – The objectivity/subjectivity parameter that will be analyzed in the next section.

G

G

G

Figure 1.  finite (subordinate) Figure 2.  inflected infinitival Figure 3.  uninflected infinitival clause clause clause

5.

I am grateful to Michel Achard for this point.

 Augusto Soares da Silva

Some of the differences between these three constructions are illustrated in Figures 1, 2 and 3. The rectangles only represent the subordinate clause. The differences shown in these figures include the presence/absence of temporal determinations or sequential scanning (shown by the heavy-line bar along the inner arrow), the degree of relationship with the ground (G) (see the outer lines), specific points of this relationship or objects of per/conception (oblique lines), and focal prominence differences of the subordinate trajector (inner circles).

4 The meaning of inflected infinitive: grounding and objectification In regard to the specific conceptual import of the inflected infinitive, is personal inflection a grounding predication? On the one hand, the person category relates the main participant in the conceptualized event to the speech situation. Following the position expressed in Achard (1998: 50; 2002: 200), this is grounding predication. On the other hand, as Langacker (1991, 2005) points out, clausal grounding involves an elaborated process type and a connection to reality (tense and modality): Langacker (2005) explains that the primary epistemic concern is the existence (i.e. occurrence) of an event (while for nominal grounding the primary epistemic concern is the identification of an object). Anyway, the personal infinitive allows the conception of the main participant actually involved in the event, even without the event being epistemically assessed, that is, without the event being actually considered with respect to reality. There might be at least here a reason not to treat the person category as parasitic on mood and tense categories. So, what is the specific conceptual import of the inflected infinitive? It is a matter of viewing arrangement (the relative positions of the subject and object of per/conception: Langacker 1985, 1990, 2001) and it involves an objective construal of the infinitival trajector. Compared with the simple infinitive, the inflected infinitive indicates an objectification of the conception of the infinitival trajector. Take for example (16)-(18) again and compare inflected infinitival clauses in (b) to uninflected ones in (a). The inflection – on its own, as in examples (b), or together with the overt subject –, renders the infinitival trajector more prominent within conception as it provides information about it. The inflection also increases the extent to which the infinitival trajector is construed objectively rather than subjectively as it refers to the main participant explicitly. In other words, the inflection is going to put the main participant “onstage” as a more specific focus of attention, as an object of conception. The personal inflection would then tend to distance the infinitival clause and its main participant from the conceptualizer, so that the respective event gains certain independence and can be seen from “outside”. This way, personal inflection increases the observer/observed asymmetry and therefore creates the conditions so that an optimal viewing arrangement almost occurs (maximally observer/observed asymmetry). As Langacker (1985, 1991: 445–449) emphasizes, there is a correlation between the objective construal of an



The Portuguese inflected infinitive and its conceptual basis 

entity and its explicit mention and also a certain conceptual distance from the conceptualizer: explicit mention and conceptual distance have an objectifying impact. In contrast, the simple infinitive leaves its trajector implicit and hence unprofiled, which can be revealed by the context, as in example (a) in (16)-(18), or be truly undetermined. The simple infinitive leaves the trajector “offstage”, inherent in the very process of conception without being its target. The unique object of conceptualization is now the process type profiled by the infinitive (see above Figure 3 in contrast with Figure 2), and the main participant in this process is indeed construed subjectively. Putting it another way, the lack of explicit mention of the infinitival trajector follows from its subjective construal, and the lack of person inflection of the verb follows from the lack of explicit mention of its subject. To be more specific, it is in cases where the subject of the uninflected infinitive is known from the context, i.e example (a) in (16)-(18), that the term subjectification can be used. On the other hand, notice that the conceptualizing role of the infinitive subject is possible only in infinitival complement clauses that also have a conceptualizer as a main subject and when both subjects are coreferential, as in (17a): the conceptualizing main subject construes the complement event from the vantage point of the subordinate subject (see Achard’s account (1998: chap. 5) on conceptualizing subject constructions). Finally, the finite subordinate clause is in this continuum the highest point of objectification of the subordinated event with respect to the objective axis of per/conception: it is here that all the aspects of the subordinate scene are or can be put onstage; it is here that the greatest asymmetry occurs between the conceptualizer and the object of conceptualization. However, it is also the finite subordinate clause that involves the greatest subjectivity with respect to the subjective axis of per/conception: it is here that all clausal grounding predications are (or can be) present, and all these relationships to the ground are aspects of subjective construal, in that they are the expression of the active role that the speaker plays in structuring the perceived scene. The finite clause is the place where the epistemic status of the process is assessed and where the indicative/ subjunctive alternation occurs – the codified distinction between an established event in reality, and therefore independently accessible to any conceptualizer (indicative) and an established event in a specific and local mental space, separate from reality (subjunctive) (see Achard’s account, 1998: chap. 6). The finite clause is the place where mental space elaborations and transfers take place. For example, while the infinitival adverbial clause in (18a, b) can only indicate a cause relationship, the finite adverbial clause in (18c) can also indicate a more subjective relationship of an inferential type. In this sense of subjectivity/subjectification (along the subjective axis), the inflected infinitival clause involves a certain degree of subjectivity expressed in the personal grounding predication, that is, the epistemic activity from speech act participants (speaker/hearer) identifies the subordinate main participant

 Augusto Soares da Silva

5. Semantic (in)compatibility and the distribution of the inflected infinitival clause The distribution of the inflected infinitive can be explained in the light of its conceptual import. This distribution is a matter of compatibility between the meaning of the main verb or construction and the meaning of the inflected infinitive. This study will focus on infinitival complement constructions as they show the most obvious distributional restrictions. Consider now the inflected infinitival complement clauses and examples (19)-(23). Why is it the case that volition verbs cannot be followed by an inflected infinitival complement (see 19) when factive verbs (like in 21) are compatible with the inflected infinitive? Why is it not the case that declarative and epistemic verbs can be followed by the infinitive subject (see 22) when factive verbs allow both positions of the infinitive subject (see 21)? (19) *Eles querem resolverem o problema. they want solve-infl.infin-3pl the problem (20) *Eles podem resolverem o problema. they can solve-infl.infin-3pl the problem (21) a. b. c.

O João lamenta eles não terem resolvido o problema. John regrets they no have-infl.infin-3pl solved the problem O João lamenta não terem eles resolvido o problema. John regrets no have-infl.infin-3pl they solved the problem O João lamenta o/o facto de eles não terem John regrets the/the fact of they no have-infl.infin-3pl resolvido o problema. solved the problem ‘John regrets (the fact that) they didn’t solve the problem.’

(22) a. b. c.

O João pensa/afirma terem eles encontrado a solução. John thinks/affirms have-infl.infin-3pl they found the solution *O João pensa/afirma eles terem encontrado a solução. John thinks/affirms they have-infl.infin-3pl found the solution *O João pensa/afirma o /o facto de terem John thinks/affirms the/the fact of have-infl.infin-3pl eles encontrado a solução. they found the solution ‘John thinks/says/affirms (that) they found the solution.’

(23) a. b.

A Maria fez/mandou/deixou/viu os miúdos correrem. (VSV) Mary made/ordered/let/saw the children run-infl.infin-3pl A Maria fez/mandou/deixou/viu os miúdos correr. (VOV) Mary made/ordered/let/saw the children run-infin



The Portuguese inflected infinitive and its conceptual basis 

c. A Maria fez/mandou/deixou/viu correr os miúdos. (VV) Mary made/ordered/let/saw run-infin the children ‘Mary made/let/saw the children run / ordered the children to run.’ Volition verbs create mental spaces that are relevant only to their subjects and so impose an inherently subjective construal (an egocentric viewing arrangement) on the complement scene: the content of their complements has no necessary independent existence outside the conceptualization of their subjects; their clausal complements are indeed minimally distinct from the main clause and its subject and, conversely, volition verbs themselves are dependent on them, and hence act as semi-auxiliaries (Cornillie 2007). Consequently, the trajector of a volition verb infinitival complement, necessarily coreferential to the main subject, cannot be (and does not need to be) referred to explicitly or put onstage – hence the impossibility of personal inflection in the infinitive. The same can be said about modal verbs, as in (20), with the difference that these verbs are speaker-oriented verbs and, as such, do not allow finite complements. In contrast, factive verb complements have a factual status that is independent from the main clause, since these verbs suppose the truth of their complements. This not only explains the compatibility of factive verbs with the inflected infinitive in both (preverbal and postverbal) positions of the infinitive subject, but also explains the possibility to nominalize the inflected infinitival clause by using the definite article o ‘the’ or the expression o facto de ‘the fact of ’, as in (21c). In the last case, nominalization increases conceptual autonomy and objectification of the presupposed complement event taken holistically as a unitary entity. Declarative and epistemic verbs, as in (22), hold in some way an intermediate position. Their compatibility with the inflected infinitive results from the fact that the content of its complement is considered with respect to reality. Their incompatibility with the preverbal position of the infinitive subject (see 22b) may have to do with their tendency to select sentential objects and therefore create complex verbal unities with the infinitive. Moreover, declarative and epistemic verbs create a mental space. Consequently, these verbs do not allow the nominalization of the infinitival complement (see 22c). Finally, causative and perception verbs, see (23), can be followed by three infinitival complements – VSV, VOV and VV constructions, the first one (VSV) indicating the inflected infinitive – and can also be followed by finite complements. As was shown in another paper (Soares da Silva 2004), Portuguese causation/perception constructions display a rich imagery in relation to the conceived complement event. The choice of the inflected infinitival complement from the three infinitival complements available expresses a greater independence of the complement event and of its main participant and a less direct interaction between the main subject and the subordinate subject. It means that with causative verbs, such a choice indicates a less direct causation, in which the causee has more autonomy and control in the induced process, while the same choice with perception verbs indicates a perception with two focal points (the process and its main participant).

 Augusto Soares da Silva

Sentence (24a) exemplifies the so-called complex adjectival construction or Though Raising construction (“object to subject raising”). Why does the complement construction not allow the inflected infinitive? Probably because of syntactic changes involved in this complex construction and, in particular, by the fact that the inflected infinitive always recovers the trajector, not the landmark – which justifies its occurrence in the passive voice, as in (24b). Nevertheless, this construction demands a more careful analysis. That reason can also explain the infelicity of the inflected infinitive in embedded relative/interrogative clauses, as in (15) above. (24) a. b.

*Esses relógios são difíceis de arranjarmos. these watches are difficult of fix-infl.infin-1pl Esses relógios são difíceis de serem arranjados. these watches are difficult of be-infl.infin-3pl fixed ‘These watches are difficult to fix / to be fixed.’

Looking briefly at adverbial constructions, why is the inflected infinitive generally more common in adverbial clauses rather than the uninflected infinitive? As seen in examples (25)-(27), it is usually more natural and spontaneous to say (a) rather than (b). The reasons are not difficult to understand. Firstly, a greater independence of the adverbial clause. Secondly, the inexistence of valence restrictions on the verb and a greater need to identify the infinitival trajector. Thirdly, the preposition that always introduces the infinitival adverbial clause has a nominalizing (objectifying) effect. Finally, in cases of coreferentiality, the inflected plays an anaphoric (or cataphoric) role taking up the trajector again from the main clause.6 (25) a. b.

Apanhámos uma maçã ao sairmos do pomar. picked up-1pl an apple as leave-infl.infin-1pl of the orchard Apanhámos uma maçã ao sair do pomar. picked up-1pl an apple as leave-infin of the orchard ‘We picked-up an apple as we left the orchard.’

(26) a. b.

Acho que eles te chamaram para te exporem o think that they you called to you expose-infl.infin-3pl the plano. plan Acho que eles te chamaram para te expor o plano. think that they you called to you expose-infin the plan ‘I think they called you to show you the plan.’

(27) a. Afinal, foram embora, sem conseguirem encontrá-lo. after all, went away, without manage-infl.infin-3pl find him 6. Vesterinen (2006) offers a Cognitive Grammar oriented analysis about the alternation between finite and infinitival adverbial clauses in Portuguese, which includes some observations about the inflected/uninflected infinitive alternation.



The Portuguese inflected infinitive and its conceptual basis 

b. Afinal, foram embora, sem conseguir encontrá-lo. after all, went away, without manage-infin find him ‘After all, they left without being able to find him.’ Finally, the remaining contexts of the inflected infinitive distribution should be mentioned. With prepositional complement clauses (of verbs, adjectives and nouns), there usually is a free alternation between both infinitives: the use of the inflected infinitive depends on the conceptual import referred to above and/or pragmatic factors. The same can be said about impersonal constructions (existential in nature, as they pertain to the presence of a certain event in reality, Achard 1998: chap. 7), as in (16) above7 and to the other infinitival subject clauses.

6. Conclusion Three conclusions can be drawn from the present study. First, the use of the inflected infinitive in modern Portuguese is motivated by different factors (conceptual, structural, pragmatic) and therefore, in contrast to Raposo’s (1987) and other formalist (generative) proposals, it cannot be predicted in any absolute way. Second, the inflected (or personal) infinitive is revealed to be meaningful. Its conceptual import consists of the objective construal of a process type trajector (the main participant of an event not epistemically determined). Such a meaning is consistent with distributional restrictions of the inflected infinitive, particularly its incompatibility with volitional verbs and internal word order constraints in the context of declarative and epistemic verbs. It is also consistent with the great variety of syntactic environments in which the inflected infinitive can appear. It is in this way that the inflected infinitive can be, if not predicted, at least highly expected. Third, Portuguese finite/infinitival subordinate clauses reveal a more elaborate continuum of event asymmetry and a richer imagery relating to the conceived subordinate event than other languages. More specifically, uninflected and inflected 7. The case of the epistemic verb parecer ‘to seem’ is rather more complex: beyond the finite construction, this verb allows infinitival constructions with or without raising in which some restrictions occur regarding the use of both infinitives: (i) Parece terem eles chegado. seem-3sg have-infl.infin-3pl they arrived (i’) ?Parece eles terem chegado. seem-3sg they have-infl.infin-3pl arrived (ii) Eles parece terem / *ter chegado. they seem-3sg have-infl.infin-3pl / have-infin arrived (iii) Eles parecem ter /*terem chegado. they seem-3pl have-infin / have-infl.infin-3pl arrived ‘They seem to have arrived.’

 Augusto Soares da Silva

infinitives allow the language to distinguish grammatically general statements from more specific ones without assessing these statements epistemically. As a final note, the Portuguese inflected infinitive challenges cognitive (and noncognitive) linguists to refine the rich Cognitive Grammar notion of grounding.

References Achard, M. 1998. Representation of Cognitive Structures. Syntax and Semantics of French Sentential Complements [Cognitive Linguistics Research 11]. Berlin/New York: Mouton de Gruyter. Achard, M. 2002. The meaning and distribution of French mood inflections. In Grounding. The Epistemic Footing of Deixis and Reference, F. Brisard (ed.), 197–249. [Cognitive Linguistics Research 21]. Berlin/New York: Mouton de Gruyter. Ambar, M. 1994. ‘Aux to COMP’ and lexical restrictions on verb movement. In Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne, G. Cinque, J. Koster, J.-Y. Pollock and L. Rizzi (eds.), 1–21.Washington, D.C.: Georgetown University Press. Ambar, M. 2000. Infinitives vs. participles. In Portuguese Syntax. New Comparative Studies, J. Costa (ed.), 14–30. Oxford: Oxford University Press. Cornillie, B. 2007. Evidentiality and Epistemic Modality in Spanish (Semi-) Auxiliaries. A Cognitive-functional approach [Applications of Cognitive Linguistics 5]. Berlin/ New York: Mouton de Gruyter. Langacker, R. W. 1985. Observations and speculations on subjectivity. In Iconicity in Syntax, J. Haiman (ed.), 109–150. Amsterdam: John Benjamins. Langacker, R. W. 1987. Foundations of Cognitive Grammar, Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press. Langacker, R. W. 1990. Subjectification. Cognitive Linguistics 1: 5–38. Langacker, R. W. 1991. Foundations of Cognitive Grammar, Vol. 2: Descriptive Application. Stanford: Stanford University Press. Langacker, R. W. 1999. Grammar and Conceptualization [Cognitive Linguistics Research 14]. Berlin/New York: Mouton de Gruyter. Langacker, R. W. 2001. Viewing and experiential reporting in cognitive grammar. In Linguagem e Cognição: A Perspectiva da Linguística Cognitiva, A. Soares da Silva (ed.), 19–49. Braga: Associação Portuguesa de Linguística and Universidade Católica Portuguesa. Langacker, R. W. 2005. Enunciating the parallelism of nominal and clausal grounding. Paper presented at the International Conference From Gramm to Mind: Grammar as Cognition. Université Michel de Montaigne, Bordeaux, 19–21 May 2005. Martins, A. M. 2001. On the origin of the Portuguese inflected infinitive. A new perspective on an enduring debate. In Historical Linguistics 1999, L. J. Brinton (ed.), 207–222. Amsterdam: John Benjamins. Maurer Jr., T. H. 1968. O Infinitivo Flexionado Português (estudo histórico-descritivo) [The Portuguese Inflected Infinitive (a historical and descriptive study)]. São Paulo: Companhia Editora Nacional. Mensching, G. 2000. Infinitive Constructions with Specified Subjects. A Syntactic Analysis of the Romance Languages. Oxford: Oxford University Press.



The Portuguese inflected infinitive and its conceptual basis  Raposo, E. 1987. Case Theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18: 85–109. Scida, E. 2004. The Inflected Infinitive in Romance Languages. New York & London: Routledge. Soares da Silva, A. 2004. Imagery in Portuguese causation/perception constructions. In Imagery in Language. Festschrift in Honour of Professor Ronald W. Langacker, B. Lewandowska-Tomaszczyk and A. Kwiatkowska (eds.), 297–319. Frankfurt/Main: Peter Lang. Sten, H. 1952. L’infinitivo impessoal et l’infinitivo pessoal en portugais moderne. Boletim de Filologia 13: 83–142, 201–256. Vesterinen, R. 2006. Subordinação Adverbial. Um estudo cognitivo sobre o infinitivo, o clítico SE e as formas verbais finitas em proposições adverbiais do Português Europeu [Adverbial Subordination. A cognitive study on the infinitive, the clitic SE and finite verb forms in European Portuguese adverbial clauses]. Ph.D dissertation, Stockholm University.

The periphrastic realization of participants in nominalizations Semantic and discourse constraints* Liesbet Heyvaert

Fund for Scientific Research – Flanders and University of Leuven

This paper discusses the inclusion of periphrastic participants in deverbal nominalizations. While traditionally the clause-like properties of explicitly realized participants are focused on, it zooms in on deverbal event nominalizations that lack one or more participants and points to the importance of identifying the precise nominal status of their periphrastic participants accurately. In particular, referring to Willemse’s (2005) analysis of of-phrases as ‘reference points’, it suggests, firstly, that periphrastic participants primarily be analyzed as nominal reference points of the deverbal head noun, functioning as autonomous discourse referents, and, secondly, that those cases where participants are not included but ‘understood’ and retrievable from the discourse context are cases of what one might classify as a type of ‘nominal ellipsis’. In a second part, the existence of event nominalizations of which the participants are not retrievable in the discourse context is pointed out and it is argued that such nominalizations are comparable to a clausal type of ‘events’, i.e. to nonfinite events, which are not temporally grounded and do not necessarily take a Subject-entity.

Keywords: deverbal nominalization; reference point; complementation; modification; ellipsis

* I would like to thank Barbara Lewandowska-Tomaszczyk and Hubert Cuyckens for their encouragement while I was writing out this paper; Peter Willemse for the many discussions we had about his reference point analysis of binominal NPs with of; and Kristin Davidse for her general support.

 Liesbet Heyvaert

1. Introduction This paper will deal with deverbal nominalizations that contain genitives, possessives and/or of-phrases specifying the participant roles of the base verb. The analysis suggested in it applies to nominalizations of the types illustrated in (1) to (3), i.e. so-called ‘derived’ nominalizations, e.g. in -ion (as in (1)), ‘action’ -ing nominalizations (Lees 1960) as in (2), and the participant-profiling system of -er nominalization, as in (3).1

(1) a. ... the Government’s temporary abolition of stamp duty on house purchases … (CB)



(2) a. ... the opening of a monthly account... (CB)



(3) a. John Travolta was... a smart deliverer of lines. (CB)

Ever since Lees (1960), the literature on the (periphrastic) inclusion of participants in deverbal nominalizations has been mainly preoccupied with the participants’ status as clausal or clause-like units. Nominalized structures such as the ones in (1a) to (3a) are thus almost invariably analyzed in terms of their resemblances with the clauses in (1b) to (3b):

(1) b. the Government (has) temporarily abolished stamp duty on house purchases



(2) b. open a monthly account



(3) b. he delivered lines smartly

In early transformational accounts of word formation, the participants figuring in nominalizations were said to have been ‘inherited’ from the clausal realization of the nominalization’s base verb, nominalizations being viewed as transformations of (deep structure) clauses. Later, following Chomsky’s Remarks on Nominalization (1970), the resemblances between structures like those in (1a) and (1b) were brought down to the fact that nominalizations and their base verbs have the same participants in their lexical entry (Chomsky 1970; Rappaport 1983; Rappaport Hovav and Levin 1992) – derived nominals in this approach inherit the participant roles of their base verb. The differences between the nominal and clausal realizations, it was suggested, can be attributed to the fact that nominalizations of the types in (1) to (3) “have the internal structure of noun phrases” (Chomsky 1970: 189) and thus “differ crucially in the grammatical functions made available to them for the purposes of mapping arguments onto syntactic structure” (Rappaport 1983: 113). The internal nominal status of derived nominalizations was thus explicitly brought into the picture. Strikingly, however, in descriptive accounts of deverbal nominalizations, the import of the nominal paradigm on the lexicogrammatical status and semantics of the base verb’s participants has often 1. The examples marked ‘CB’ have been extracted via remote log-in from the COBUILD corpus. They are reproduced here with the kind permission of Collins COBUILD.



On the status of periphrastic participants in nominalizations 

been ignored. Most of the existing descriptive accounts of deverbal nominalizations are merely aimed at identifying the ‘linking conventions’ that exist for participants and syntactic functions at clausal and nominal level (e.g. Chomsky 1970; Fraser 1970; Rappaport 1983), resulting in well-known generalizations and restrictions such as: (4) a. clausal Agents are typically mapped onto nominal Possessors, as in: The Romans destroyed Carthage → the Romans' destruction of Carthage b. clausal Patients may be mapped onto nominal Possessors if they designate affected entities The city was destroyed → the city's destruction The problem was perceived → *the problem's perception That little or no attention is paid to the role of the nominal environment in the realization of participants is also illustrated by the fact that the periphrastic markers ‘s and of have tended to be ignored or conceived as making no specific contribution to the lexicogrammar and semantics of the nominalization, i.e. as ‘empty’ (see also Langacker 1991: 35). (In Grimshaw’s (1990) book-length analysis of ‘argument-taking’ nominalizations, for instance, they are hardly mentioned.) In addition to being linked up with related clausal realizations, nominalizations that include periphrastic participants have been assigned specific semantic properties. They have variously been described as ‘minimal’ (Strang 1968), ‘entirely predictable’ (Quirk et al. 1985) or ‘partial’ (Mackenzie 1990) ‘syntactic recategorizations’ (Motsch 1979, as cited in Kastovsky 1986 :411) and they have been said to designate ‘events’ (Levin and Rappaport 1988) or ‘complex events’ (Grimshaw 1990). Levin and Rappaport (1988: 133, 135) define ‘event’ -er nominalizations such as a smart deliverer of lines in (3a) as ‘referring to someone who has actually participated in the event of V-ing’ and event derived nominalizations like the government’s abolition of stamp duty as describing ‘the actual event of V-ing’. Grimshaw (1990: 49) merely specifies that complex event nominals have an associated event structure “which provides them with an internal event analysis”. In the rest of the paper I will be using the term ‘event’ nominalizations for them. Event nominalizations have been distinguished from so-called ‘bare’ deverbal nominalizations (Roeper 1987) or nominals that cannot include periphrastic participants (as in (2c) and (3c)). The latter have been labelled as ‘concrete’ (Lees 1960) or ‘full’ nominalizations (Mackenzie 1990) that carry a ‘non-event’ meaning (Levin and Rappaport 1988) and designate ‘products’ or ‘results’ rather than ‘actions/processes/ states’ (Quirk et al. 1985; Huddleston and Pullum 2002). (Note that this distinction cuts across the distinction between ‘participant’ and ‘nonparticipant-profiling’ nominalization systems.)

(2) c. Planes could take off by speeding along covered runways before being launched into the air from an opening in the hillside. (CB)

 Liesbet Heyvaert



(3) c. Business Post is not the cheapest deliverer, but it claims total efficiency. (CB)

Nominalizations in -ion likewise have variants with and without periphrastically realized participants: (5) a. (...) this stipulation has since been altered to permit the acquisition of pictures at least thirty years old. (CB) b. This season the Bullets show off its fine new acquisition, 23 year old Jason Chatwin (...). (CB)

Even though the link between the periphrastic realization of participants and the event reading has been repeatedly referred to in the literature (e.g. Roeper 1987; Levin and Rappaport 1988; Grimshaw 1990; Mackenzie 1990; Kastovsky 1995), many questions about it remain unanswered. It is, for instance, unclear precisely how the inclusion of participants gives rise to an event reading. It also remains to be seen whether (explicitly realized) periphrastic participants are a precondition for an event reading. Finally, the role of the nominal, periphrastic status of the participants in deverbal nominalizations deserves to be looked at in more detail. In this paper I will point to two aspects of the analysis of event nominalizations which in my view shed new light on the periphrastic realization of participants in them. In a first part, I will consider more closely the (nominal) status of periphrastic participants in the nominalized NP and its impact on their discourse behavior. I will argue that periphrastic participants differ from participants at clause level in that, even when they are obligatory at clause level, they are optional in the NP. Importantly, however, the optional inclusion of participants in nominalizations cannot be put on a par with that of ordinary NP modifiers such as on the table in the fruit on the table. Because of their close conceptual link with the deverbal noun, periphrastic participants in event nominalizations are always understood, even if they are not explicitly included in the NP. Unlike ordinary modifiers, moreover, they are typically retrievable from the discourse context and, if retrievable, can always be explicitly added to the event nominalization. Event nominalizations that do not explicitly realize their participants, in other words, are cases of what one might call ‘nominal ellipsis’, in which participants that are left unrealized are very much ‘understood’ (see Halliday and Hasan’s definition of nominal ellipsis, 1976: 143). The periphrastic participants in deverbal nominalizations cannot just be classified as ‘complements’ either (Langacker 1991). As suggested in Willemse’s (2005) analysis of binominal NPs with of-phrases as ‘reference-point’ constructions, nominalizations with a genitive/possessive and/or an of-phrase realize a ‘reference-point’ relationship or a phoric relationship of identification between two NPs or discourse referents. Significantly, this reference point analysis also sheds new light on the discourse behavior of periphrastic participants. While some non-realized participants are cases of ellipsis and are retrievable in the discourse context, others are not included because they are not part of the event semantics



On the status of periphrastic participants in nominalizations 

which the language user has in mind. In the second part of the paper, I will briefly elaborate on the semantics of event nominalizations and propose a semantic distinction which in my view elucidates the language user’s choice for explicitly realized participants, implied but anaphorically retrievable participants or non-retrievable participants. I will start by bringing together a number of observations that have been made in the literature on the ‘obligatory’ inclusion of participants in event nominalizations. I will confront these observations with the lexicogrammatical status of periphrastic participants as components of nominalized NPs and briefly point to Willemse’s (2005) analysis of periphrastic participants as reference points. In Section 3, I will confront the event reading as defined in Rappaport Hovav and Levin (1992) and Grimshaw (1990) with instances of event nominalizations that do not designate actual events and discuss the impact of the latter category’s semantics on the realization of the Subjectlike participant in the nominalization. The analysis of periphrastic participants in deverbal nominalizations, it will be concluded, cannot be restricted to identifying corresponding clausal participant roles, but necessarily includes a discourse and a semantic perspective, both of which are rooted in a careful componential analysis of the nominalized NP (see also Heyvaert 2003).

2. Nominalizations as asymmetrical units? 2.1

Obligatory vs. optional realization of participants

Grimshaw (1990: 45) argues that the distinctive feature of complex event nominalizations is that they “have an argument structure” and “are systematically like verbs in their argument-taking capacities” (1990: 47). She adds that the “argument structure of such a noun must be satisfied, hence the obligatoriness of its arguments” (Grimshaw 1990: 45). When the nominalization examination in (6) is thus used as a concrete or result nominalization (as in (6a)), it can be construed without a complement. When used with a complement, as in (6b), on the other hand, it receives an event reading (Grimshaw 1990: 48–49). (6) a. John’s examination / The examination / The exam was long. b. John’s examination of the patients took a long time. Event nominalizations or nouns that have an argument structure, Grimshaw concludes, “show the same argument-taking properties as verbs in that their argument[s] are fundamentally obligatory” (1990: 70). Rappaport Hovav and Levin (1992: 136) point out that Grimshaw’s claim that “a nominal derived from a verb that takes obligatory complements must express the complements if the nominal is to receive an event interpretation” also holds for deverbal -er nominalizations: -er nominals derived from verbs taking obligatory complements “must express the complements if the nominal is

 Liesbet Heyvaert

to receive an event interpretation” (Rappaport Hovav and Levin 1992: 136). The examples they give to prove their point are the following: (7) a. b.

She devours fresh fruit – a devourer of fresh fruit *she devours – *a good devourer She eats fresh fruit – an eater of fresh fruit she eats – a good eater

In short, in their analyses of event derived and -er nominalizations, Grimshaw and Rappaport Hovav and Levin suggest that there is a kind of ‘automatic and direct transfer’ of clausal relations of participation to nominalization. Significantly, however, neither of them makes use of natural data or takes into account the discourse context. Yet, when event nominalizations are considered in their discourse context, the reduction in “’actual valency’, i.e. the number of arguments that receive overt expression in attested discourse” is quite obvious (Mackenzie 1985: 29). As Mackenzie (1985: 32) points out, ... observation of the actual occurrence of nominalizations in texts shows that there is an overwhelming tendency for nominalizations of all types not to be structurally parallel to their non-nominalized counterparts. The difference resides in the fact that the terms associated with a nominalized verbal predicate are much more likely not to be expressed than is the case with non-nominalized verbal predicates.

He gives the following example to illustrate his point (Mackenzie 1985: 32):

(8) A pilot usually requires a horizon so that he can orient himself. He should still scan his instruments from time to time, but if he cannot see a horizon, then this scanning becomes much more important.

In her typological study of deverbal or ‘action’ nominalizations, Koptjevskaja-Tamm (1993: 15) points out that in many languages it is common that the “semantic argument ai of the action nominal X could, in principle, be expressed in the given sentence, but is omitted in accordance with the communicative goal of the sentence”. She continues by saying that “all of the semantic arguments of an action nominal are seldom present in ANCs [action nominalizations]” because “they are either clear from the context or irrelevant” (Koptjevskaja-Tamm 1993: 15). Following Mackenzie (1985), Koptjevskaja-Tamm (1993: 261) further argues that the argument reduction that is found among nominalizations is closely tied up with their main function “to be a constituent part in the backgrounded portions of discourses”, i.e. in “that part of discourse which does not immediately and crucially contribute to the speaker’s goal, but which merely assists, amplifies, or comments on it” (Hopper and Thompson 1980: 280). Corpus data (based on Van den Bosch 2004 and De Roover 2004) confirm the observations made by Mackenzie and Koptjevskaja-Tamm and make clear that event



On the status of periphrastic participants in nominalizations 

nominalizations can indeed be used without the participants of their base verb, even if at clause level these participants are obligatorily realized:

(9) On July 9th a prominent Afghan writer, Mr Abdul Rahim Chinzai, was abducted in Peshawar. This abduction appears politically motivated. (CB)

(10) The 12- or 13-year old is assimilating an enormous amount of new (...) experiences. While all this absorption is going on (...) the young person is in a more or less perpetual state of disequilibrium. (CB) (11) Accordingly the claim to privilige failed because Parliament had abrogated the privilige in respect of Theft Act offences and in the circumstances, including the fact of that abrogation, there was no privilige in respect of non-Theft Act offences. (CB) (12) Words are not coined in order to extract the meaning of their elements and compile a new meaning from them. The new meaning is there first, and the coiner is looking for the best way to express it without going to too much trouble. (Bolinger 1975, as cited in Mackenzie 1990: 132) (13) A few thought they had noticed someone resembling the man in the picture. I wasted two days tracking one of the supposed resemblers. (Kastovsky 1986: 410) (14) (...) where a person is acquiring shares in a company, it is not lawful for the company to give financial assistance (...) for the purpose of that acquisition or at the same time as the acquisition takes place. (CB) The examples in (9) to (14) show that there is no automatic transfer from clause to NP as far as obligatory participants are concerned: all of the nominalizations in them are based on verbs with obligatory patients and yet they do not include of-phrases specifying them. Moreover, Mackenzie (1985) and Koptjevskaja-Tamm (1993) turn out to be right when they argue that it is the discourse context that makes explicit realization of the participants in the NP superfluous: the participants that have not been realized in the nominalized NP are all present in the preceding discourse context. However, the question remains why it is that participants which are obligatory at clause level can be left out in event nominalizations and how the event reading can in that case be preserved.

2.2

Modification, complementation or phoric identification?

Let us have a closer look at the periphrastic integration of the base verb participants in nominalizations. What is the linguistic status of the periphrastic participants in the NP? Grimshaw (1990: 56) claims that possessives and of-postmodifiers (and by-phrases) have ‘argument structure-related meanings’ with complex event nominals (see (15a)) and ‘modifier-like interpretations’ with result nominalizations (as in (15b)). When a noun is ambiguous between a ‘complex event’ and a ‘result’ reading (as is

 Liesbet Heyvaert

examination in (15c)), she argues, “an associated possessive is also ambiguous between the modifier reading found with concrete nouns and the a-structure-related reading in which the possessive provides information about a position in the argument structure of the noun” (Grimshaw 1990: 48). (15) a. the city’s development of inexpensive housing; the development of inexpensive housing by the city b. John’s foundation; a painting by Picasso; a decanter of port/ the dispenser of the machine c. John’s examination Notice that the use of the periphrastic markers ‘s, of and by in complex event nominalizations is ignored: the focus is on the clause-like meaning of the periphrastic participants only. Grimshaw’s analysis thus violates one of the basic tenets of a symbolic approach to language, as taken in, for instance, Cognitive Linguistics, i.e. the tenet “that so-called ‘grammatical markers’ invariably have some kind of conceptual import, which may be quite abstract but is nonetheless essential to their function” (Langacker 1991: 35). Langacker’s (1991, 1992) own analysis of the integration of base verb participants in nominalizations is based on the claim that two levels of constituency have to be distinguished in nominalized constructions (Langacker 1987: 304–310; Langacker 1991: 35–43; Langacker 1992): whereas at clause level there is merely complementation between the verb (i.e. the conceptually dependent head or profile determinant) and its participants (complements or conceptually autonomous modifiers), in nominalizations, Langacker argues, we find both modification and complementation. First, a relationship of complementation exists between ‘s, of and by (conceptually dependent heads), on the one hand, and the NPs that accompany them (which are conceptually autonomous elements), on the other hand: “Grammatically, therefore, Zelda and the contract are not the subject and object of sign(ing) in Zelda’s signing of the contract (as they are in the corresponding finite clause), but rather complements of the relational predications ‘s and of. The net result of these periphrastic constructions is nevertheless to identify them semantically with the appropriate participants” (Langacker 1991: 35–36). At the second level of constituency, then, we find the modification relationship between the head noun (conceptually autonomous head or profile determinant) and its modifiers (conceptually dependent elements such as possessives, ofphrases and by-phrases). Unlike Grimshaw, Langacker thus analyzes the periphrastic participants in deverbal nominalizations (e.g. Zelda’s and of the contract) as modifiers of the head noun (signing) which themselves contain a relationship of complementation between the relational predications ‘s and of, on the one hand, and Zelda and the contract, on the other. In addition to this two-level constituency analysis, Langacker (1992: 491) suggests that of-phrases also function as complements to the head noun because of signals an ‘intrinsic relationship’ between two NPs and elaborates a schematic notion of relationship invoked by the first NP. More particularly, in the case of of-phrases used in



On the status of periphrastic participants in nominalizations 

nominalizations, the of-phrase is argued to designate a conceptually autonomous participant NP which elaborates the conceptually dependent nominalized head or first NP: participants are conceptually autonomous because it is “possible to conceive of them independently of any event conception” (Langacker 1992: 494), whereas the nominalized verb is conceptually dependent and cannot be conceptualized “without in some way invoking the conception of its participants (if only in vague, schematic terms)” (Langacker 1992: 495). Of-phrases are in this respect different from prepositional phrases such as on the bench and under the tree in the woman on the bench and the bench under the tree, which introduce a relationship that is not invoked by the head noun (Langacker 1992: 491) and therefore function as pure modifiers. Unlike pure modifiers, Langacker argues, nominalizations with of-phrases thus also display a relationship of complementation between the nominalized head noun and its of-phrase participant, the relationships of modification and complementation in them being “superimposed” and “merg[ing] into one at the composite-structure level” (Langacker 1992: 491). Strikingly, both Grimshaw and Langacker thus assign the periphrastic participants in nominalizations special status because of the conceptual relationship which they have with the base verb of the nominalized head. Unlike Grimshaw, Langacker also emphasizes that, structurally speaking, periphrastic participants do function as modifiers of the head noun. I suggest that event nominalizations in which the participants are not periphrastically realized within the NP support an alternative analysis of possessive and of-phrase participants which gives credit to the conceptual relationship between the nominalized verb and its periphrastic participants. My analysis builds on Willemse’s (2005) analysis of of-phrases in which it is argued that, just like possessive NPs, binominal NPs with of-phrases are a type of reference point construction, viz. one “in which the reference point – the referent of the second component NP (NP2) – follows the target – the referent of NP1 – within the same binominal NP” (Willemse 2005: 204). Like possessive reference point constructions, Willemse (2005) claims, binominal NPs with of establish a phoric identification relationship between two distinct discourse referents which is based on a salient conceptual relationship that exists between the two referents. Crucially, as reference point constructions, binominal NPs with of thus indicate that the referent of NP1 is identifiable on the basis of its conceptual relationship with the referent of NP2 (see, for instance, (16) and (17), based on Willemse 2005: 266). As such, Willemse (2005) argues, they have to be distinguished from, on the one hand, binominal NPs with modifiers such as restrictive relative clauses and prepositional phrases, which realize only one discourse referent, and, on the other hand, appositive constructions (e.g. the idea of him being there), which consist of two nominals that are basically identical in reference but differ in degree of abstraction (Willemse 2005: 291–299). (16) Confusion over the televising of the Australian cricket team’s tour of the West Indies in March began last July. (CB)

 Liesbet Heyvaert

(17) It was unlikely that the explosion of a bomb and the sudden disappearance of Stauffenberg would long remain unconnected. (CB) How can deverbal nominalizations in which the participants have been left unrealized adduce evidence in favor of an analysis of event nominalizations as establishing a phoric relationship of identification? At first sight, possessives and/or of-phrases in deverbal nominalizations are like structural modifiers in that they are optional rather than obligatory. It could be argued that, while at clause level they are typically obligatorily realized as complements of the verb, in nominalizations they are ‘adapted’ to the nominal paradigm and take on more peripheral nominal modifier functions. And because the nominalized verb functions as the conceptually autonomous profile determinant within the nominalized NP, it could be argued, it can function on its own, leaving its ‘dependents’ or modifiers implicit. However – and here the ‘special status’ of periphrastic participants (as also identified in Grimshaw (1990) and Langacker (1991)) comes in – unlike ordinary modifiers such as on the bench and in the park, periphrastic participants have a close conceptual relationship with the base verb and are central rather than peripheral or ‘extrinsic’ specifications of the nominalization’s base verb. In particular, according to Willemse (2005), they function as reference points for the deverbal noun and, unlike modifiers, are themselves autonomous discourse referents. Secondly, because they contain vital information for the identification of the nominal head, and again unlike ordinary modifiers, periphrastic participants cannot be merely left out: even if they are not included in the NP they are ‘understood’ and often their identity is recoverable from the discourse context. As the examples in (9) to (14) show, the intranominal reference point mechanism of identification is in that case replaced by extranominal or anaphoric links (through referential markers such as demonstratives and determiners) between the nominalized NP and its participants in the discourse context. Thirdly, unlike NPs with ordinary modifiers, ‘reduced valency’ event nominalizations can always be replaced by their full, non-elliptical equivalents if their participants are retrievable from the context:

(9) On July 9th a prominent Afghan writer, Mr Abdul Rahim Chinzai, was abducted in Peshawar. This abduction appears politically motivated. (CB) [the abduction of Chinzai]

(10) The 12- or 13-year old is assimilating an enormous amount of new (...) experiences. While all this absorption is going on (...) the young person is in a more or less perpetual state of disequilibrium. (CB) [the absorption of new experiences] (11) Accordingly the claim to privilige failed because Parliament had abrogated the privilige in respect of Theft Act offences and in the circumstances, including the fact of that abrogation, there was no privilige in respect of non-Theft Act offences. (CB) [the abrogation of the privilige...]



On the status of periphrastic participants in nominalizations 

(12) Words are not coined in order to extract the meaning of their elements and compile a new meaning from them. The new meaning is there first, and the coiner is looking for the best way to express it without going to too much trouble. (Bolinger 1975, as cited in Mackenzie 1990 :132) [the coiner of the word] (13) A few thought they had noticed someone resembling the man in the picture. I wasted two days tracking one of the supposed resemblers. (Kastovsky 1986 :410) [the resemblers of the man in the picture] (14) (...) where a person is acquiring shares in a company, it is not lawful for the company to give financial assistance (...) for the purpose of that acquisition or at the same time as the acquisition takes place. (CB) [a person’s acquisition of shares] Notice that this is not necessarily possible with ordinary modification (because there the conceptual relationship is less salient), nor with the ‘concrete’ or ‘result/product’ variants of deverbal nominalizations. In the nominalization in (18), for instance, addition of the of-phrase of Jason Chatwin turns the nominalization acquisition into a process or event nominalization (see also Rappaport Hovav and Levin 1992): whereas the result nominalization acquisition designates the patientive participant (i.e. ‘he that has been acquired’) and implies an agentive participant (just as, for instance, bread also implies a baker), the participants implied in the event nominalization acquisition in (14) are much more salient and can therefore be realized explicitly as nominal reference points. (18) This season the Bullets show off its fine new acquisition, 23 year old Jason Chatwin (...). (CB) Event nominalizations in which the periphrastic participants are not explicitly realized can thus be argued to represent cases of nominal ellipsis. As with other cases of ellipsis, there is no implication in them that “what is unsaid is not understood; on the contrary, ‘unsaid’ implies ‘but understood nevertheless’” (Halliday and Hasan 1976: 142). Like other cases of nominal ellipsis, moreover, they are ‘cohesive’ in that the presupposed participants are present in the preceding text (Halliday and Hasan 1976: 144). Summarizing, then, being nominal and not clausal constituents, periphrastic participants are optional rather than obligatory. Unlike what has been claimed in the literature (Grimshaw 1990 and Langacker 1991), I have argued that they are not asymmetrically related to the deverbal noun, like ordinary complements or modifiers. Instead, following Willemse (2005), I have argued that they have to be analyzed as reference points which are conceptually closely linked to the deverbal noun and play a crucial role in the identification of the deverbal nominal head. Lexicogrammatically speaking, I have shown, they are either realized as explicit nominals within the reference point construction, or they are ‘ellipsed’. Note that the analysis of periphrastic participants as possibly elliptical reference points has interesting implications for their discourse behavior. In the literature on

 Liesbet Heyvaert

nominalization, it has tended to be emphasized that nominalizations are often used to summarize what has been said before, turning a clausal or clause-like unit into a nominal constituent that can itself be commented on. The analysis of periphrastic participants as optionally realized reference points makes clear, however, that the alternative relationship established in nominalizations between a process and its participants does more than summarize or rephrase its clausal ‘agnate’ construction. More particularly, it opens up a range of basically ‘nominal’ lexicogrammatical options which allow the language user to turn a clausal complementation relation (with the verb as conceptually dependent head) into a reference point relationship with the identification of the deverbal noun dependent on the identification of the reference point(s) or periphrastic participants. And because the language user can make use of anaphoric discourse relationships between the nominalized NP and previously mentioned participants, he or she can opt for ellipsis and leave out certain or all participants in the NP while being certain that they are ‘understood’ by the hearer.

3. The ‘event’ semantics and its impact on the periphrastic inclusion of participants Up till now, I have only discussed event nominalizations of which the periphrastic participants are realized in the NP or are retrievable in the discourse context via anaphoric reference. In Koptjevskaja-Tamm’s (1993: 15) words, I have only dealt with those cases of ‘reduced valency’ in which the participants are not included in the NP because they “are (…) clear from the context”. However, participants may also not be included because they are, for one reason or another, considered “irrelevant” (Koptjevskaja-Tamm 1993: 15). It is interesting to also briefly consider this type of event nominalizations somewhat more closely and fit it in with the status of periphrastic participants in deverbal nominalizations as identified in Section 2. Consider the examples in (19) and (20): (19) By a good character you surely mean the voluntary adoption of good habits. (CB) [voluntarily adopting good habits] (20) Grade’s major reason for going to the festival is to call for the abolition of the Thatcher-inspired legislation that makes his channel hand over 50 million pound a year of its profits to the ITV network. (CB) [to abolish the Thatcherinspired legislation…] The nominalizations in (19) and (20) both lack a possessive/genitive that specifies what at clause level would have been realized as the Subject, i.e. the Agent of the base verb. Rather than classifying with the elliptical type of deverbal nominalizations, however, I suggest that they do not take an agentive participant simply because the language user



On the status of periphrastic participants in nominalizations 

has no Agent/Subject in mind. Observe the difference with nominalizations as in (21) and (22), which explicitly specify the Agent of the base process: (21) Marxist socialist were prominent in the unrest that led to the abdication of the Kaiser shortly after the cessation of hostilities. (CB) [the Kaiser abdicated] (22) … his amassing of artefacts was so great that an Imperial Roman villa was built in California to accommodate it. (CB) [he amassed artefacts] The nominalizations in (21) and (22) both refer to a specifically grounded event (in the past). As such, they clearly meet Rappaport Hovav and Levin’s description of ‘event’ nominalizations and designate an ‘actual’ event (1992: 135). A closer look at corpus data of derived nominalizations, however, reveals that only some of them designate events grounded in time and person (i.e. through the Subject, see Davidse 1997) (Van den Bosch 2004). In many other cases, the nominalization designates a different kind of event, i.e. one which at clause level would be realized by means of what I have elsewhere labelled an ‘atemporalized clausal head’ (Heyvaert 2003: 73–76), i.e. with -ing or to-infinitive. Atemporalized or non-finite clauses may or may not be instantiated by a Subject. Interestingly, event nominalizations designating non-finite events are typically found with specific verbs (e.g. call for, avoid), in clauses with future reference (e.g. the intent is, will) and as complement in relational clauses (e.g. mean, represent) (Van den Bosch 2004). They can be paraphrased either in terms of a to-infinitive or by means of an -ing form and may or may not take or imply a Subject-like participant: (23) The County is now negotiating the acquisition of a Lease from the Developer of the site. (CB) [to acquire a Lease] (24) He said: (...) the intent is an accommodation of what the bankers would like to have, what the shareholders want and what the company needs. (CB) [to accommodate...] (25) According to the professor, removal of the monarchy on this plan would represent much less of a constitutional change than our accession to Europe under the Maastricht treaty. (CB) [our/us acceding to Europe] (26) Whatever the outcome of the election MX believes that other influences will make the eventual adoption of PR inevitable. (CB) [eventually adopting PR] The distinction that thus appears to exist among deverbal event nominalizations is, in short, remindful of the distinction that exists at clause level between finite and non-finite clauses. Whereas finite clauses are always grounded in the speech event, linked to a Subject-entity and situated in time, atemporalized clausal heads do not necessarily take Subjects and are not temporally grounded. I suggest that the semantic distinction between finite and non-finite clauses also exists within the event category of nominalizations, where it constitutes yet another factor that influences the periphrastic inclusion of participants, and in particular of the Subject-like entity.

 Liesbet Heyvaert

4. Conclusion This paper set out to examine the periphrastic realization of participants in deverbal nominalizations and the event reading that has often been associated with it. In particular, it zoomed in on two aspects of the analysis of periphrastic participants which were felt to be neglected in the literature so far, i.e. the nominal status of the periphrastic units within the nominalized NP and the event semantics. I showed that the status of the periphrastically realized participants in the NP is different from what has been suggested thus far. Against Grimshaw (1990) and Rappaport Hovav and Levin (1992), I argued that there is no automatic or direct transfer of participants from the clause to the nominalization. Rather, periphrastic participants are optional and whether or not they are explicitly realized within the NP depends on the discourse effect which the language user wants to bring about and on the semantic type of event he or she is referring to. Secondly, I have argued that periphrastic participants in deverbal nominalizations should not only be described in terms of their clause-like properties, but also in terms of their status as nominal components. Against Grimshaw (1990) and Langacker (1991, 1992), I have argued that they are ordinary complements nor modifiers. Rather, following Willemse 2005, it was suggested they be analyzed as nominal reference points of the deverbal head noun. Importantly, being reference points, the periphrastic participants can be said to function as autonomous discourse referents, which can be ‘ellipsed’ or left out while still being ‘understood’. In the latter case, the event reading is preserved because the nominalized NP, through anaphoric referential links, makes clear that the identity of the participants in relation to which the deverbal noun or target is to be identified is to be retrieved not intra- but extranominally, i.e. in the preceding discourse context. Finally, I also pointed to the existence of event nominalizations in which the Subject-like participant is left unrealized, not so much because it is ellipsed (and ‘understood’), but because it realizes a kind of ‘non-finite’ option within the event semantics, in which the Subject-like participant is not necessarily identified.

References Chomsky, N. 1970. Remarks on Nominalization. In Readings in English Transformational Grammar, R.A.Jacobs and P.S.Rosenbaum (eds.), 184-221. Waltham, Massachusetts: Ginn. Davidse, K. 1997. The Subject-Object versus the Agent-Patient asymmetry. Leuven Contributions in Linguistics and Philology. 86 (4): 413-431. De Roover, C. 2004. A systematic corpus analysis of deverbal -er + of -constructions. Unpublished Master Thesis, University of Leuven. Fraser, B. 1970. Some Remarks on the Action Nominalization in English. In Readings in English Transformational Grammar, R.A.Jacobs and P.S.Rosenbaum (eds.), 83-98. Waltham, Massachusetts: Ginn. Grimshaw, J. 1990. Argument Structure. Cambridge: MIT Press.



On the status of periphrastic participants in nominalizations  Heyvaert, L. 2003. A Cognitive-functional Approach to Nominalization in English. Berlin/New York: Mouton de Gruyter. Halliday, M.A.K. and R.Hasan. 1976. Cohesion in English. London and New York: Longman. Hopper, P.J. and Thompson, S.A. 1980. Transitivity in grammar and discourse. Language 56:251-299. Huddleston, R. and Pullum, G.K. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Kastovsky, D. 1985. Deverbal nouns in Old and Modern English: from stem-formation to wordformation. In Historical Semantics: Historical Word-Formation, J.Fisiak (ed.), 221-261. Berlin/New York: Mouton de Gruyter. Kastovsky, D. 1986. Diachronic word-formation in a functional perspective. In Linguistics Across Historical and Geographical Boundaries. In Honour of Jacek Fisiak on the Occasion of His Fiftieth Birthday. Volume 1: Linguistic theory and historical linguistics, D. Kastovsky and A. Szwedek (eds.), 409-421. Berlin: Mouton. Koptjevskaja-Tamm, M. 1993. Nominalizations. London and New York: Routledge. Langacker, R.W. 1991. Foundations of Cognitive Grammar. Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald W. 1992. The symbolic nature of cognitive grammar: the meaning of of and of of-periphrasis. In Thirty Years of linguistic Evolution. Studies in Honour of René Dirven on the Occasion of his Sixtieth Birthday, M.Pütz (ed.), 483-502. Amsterdam: Benjamins. Lees, R.B. 1960/1968. The Grammar of English Nominalizations. Bloomington: Indiana University Research Center in Anthropology, Folklore and Linguistics, publication 12. Levin, B. and Rappaport, M. 1988. Nonevent -er nominals: a probe into argument structure. Linguistics 26: 1067-1083. Mackenzie, L.J. 1990. First Argument Nominalization in a Functional Grammar of English. Linguistica Antverpiensia 24:119-147. Quirk, R.; Greenbaum, S.; Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Rappaport, M. 1983. On the nature of derived nominals. In Papers in Lexical-Functional Grammar, L.Levin, M.Rappaport and A.Zaenen (eds.), 113-142. Bloomington: Indiana University Linguistics Club. Rappaport Hovav, M. and Levin, B. 1992. -ER nominals: implications for the theory of argument structure. In Syntax and Semantics. Volume 26: Syntax and the Lexicon. T. Stowell and E. Wehrli (eds.), 127-153. New York: Academic Press. Roeper, T. 1987. Implicit Arguments and the Head-Complement Relation. Linguistic Inquiry 18(2):267-310. Strang, B.M.H. 1968. Swift’s agent-noun formations in -er. In Wortbildung, Syntax und Morphologie: Festschrift zum 60. Geburtstag von Hans Marchand. H. Brekle and L. Lipka (eds.), 217-229.The Hague: Mouton. Van den Bosch, C. 2004. Semantic Types of -(at)ion Nominalizations. Unpublished Master Thesis, University of Leuven. Willemse, P. 2005. Nominal Reference-Point Constructions: Possessive and Esphoric NPs in English. Unpublished PhD thesis, Department of Linguistics, University of Leuven.

Asymmetries in participial modification Barbara Lewandowska-Tomaszczyk University of Lodz

The paper focuses on differences in the construal relations between participial constructions as used postnominally and prenominally in English and Polish. Their meaning is taken to be a function of both conceptual content of an expression and its construal. It is assumed that construal principles involved in building alternate profiles of expressions determine their conceptualizations. Consecutive sections of the paper present a discussion of, first, present participle modification in English and Polish, then past participle modification in English and Polish and establish a system of cognitive contrastive parameters which can serve as tertium comparations in the analysis.

Keywords: adjective, aspect, conceptual content, construal, contrastive analysis, English, equivalence, figure/ground relations, grounding, modification (prenominal -, postnominal -), participle (passive -, past -, present -) Polish, profiling, scanning (sequential -, summary -), schematicity, tertium comparationis, verb, voice (active -, middle -, passive -)

1. Tertium comparationis and cross-linguistic equivalents Within a single linguistic system and cross-linguistically, syntactic structures function as vehicles which convey the speaker’s conceptualization of an outside scene or event. Linguistic systems provide the speakers with distinct conventional imagery, distinct analysability basis and the construal principles, which cover, inter alia, the degree of prominence of a scene and action parameters, grounding, different figure/ground relations, different degrees of schematicity (cf. the coarse- vs. fine- grained picture), scope of predication, or force-dynamic relations. The concept of cross-language equivalence evolves in Cognitive Grammar as a dynamic notion, represented as a cline gradually exhibiting more and more diversification (cf. Lewandowska-Tomaszczyk 1999). The degree of equivalence can be measured in terms of the reference categories such as the typology of basic level and prototypical category members, image-schemata and their extensions as well as the construal relations of various types.

 Barbara Lewandowska-Tomaszczyk

The present discussion involves differences in the construal relations between participial constructions as used postnominally and prenominally from the English and Polish perspectives. Their meaning is taken to be a function of both conceptual content of an expression and its construal. Also, following Langacker (2000 :45–46), it is assumed that alternate profiles of expressions can evoke the same conceptual content as in paint – painting – painted, where the expressions’ profiles determine the grammatical categories of the words.

2. Sentence, clause, phrase Langacker (2000: 75, 214) makes a distinction between complements (e.g. the bench under the tree) and modifiers (e.g. the father of the bride). He proposes that “a complement is a component structure which elaborates a salient substructure of its head, whereas a modifier is a component structure a salient substructure of which is elaborated by the head.” However, at the higher level of conceptual organization, complements can also function as modifiers. The distinction is not absolute but is rather a matter of degree. In the case of participles, their function satisfies the condition of modification, i.e. their structures are salient and elaborated by the head noun. There exists an asymmetry between the events represented in the basic part of the sentence built around the main verb on the one hand and a participial construction on the other, which changes its character from a verbal to successively more and more adjectival. In subordination, the main clause shapes the asymmetry. In modification, the asymmetry is not controlled by the main clause but rather by the head Noun. It is a consequence of juxtaposing of two (or more) events as used in finite clauses as opposed to the same events, which can be described in terms of a more adjectival category. Participial modifiers can participate in final clauses as weakly-grounded or nongrounded forms. In such constructions the participial event (represented in the modifier) is either attributivized with reference to the modified nominal, (it is most frequently used prenominally in such cases), or left in its more verbal shape and then most often following the noun. The degree of adjectivization can vary in one language and cross-linguistically and is evident in a number of morphosyntactic parameters and their interrelationship. The parameters are the nominal [articles and gender, case, number in inflectional languages] as opposed to verbal [person, tense, aspect] grammatical categories, the position of the modifier with respect to the modified noun and the presence of the complementary sentence elements such as the original direct object or adverbial(s). The paper addresses the conceptual asymmetries expressed structurally in the position of nominal modification with reference to the nominal. The modifier can be put either pre- or post-nominally. The focus of the paper are asymmetries in the semantic interpretation of different morphosyntactic structures of participial modifiers and their sentence positions, involving prenominal and postnominal participial modifiers in



Asymmetries in participial modification 

English and Polish. Most of the examples in the paper are taken from natural language corpora, BNC – for English, and the PELCRA (University of Łódź) corpus for Polish.1

3. Participial modifiers Verbs – to quote Langacker’s definition (1991) – designate a process. Langacker divides English verbs into two aspectual classes (2000 :223), perfective (e.g. learn, eat), which occur in the progressive constructions, and imperfective (know) which typically do not. There are also cases of ambiguity as in the case of the verb see, which can be interpreted as either of the two. In Polish, a completed action is marked by the perfect aspect2, expressed in terms of a rich range of affixes. A perfective process is internally bounded within the immediate temporal scope, i.e. it has an onset and offset, while bounding is not relevant for an imperfective process. Relational units such as adjectives, adverbs, prepositions, infinitives, and participles designate an atemporal relation. The relations are simple (stative) or complex (comprising multiple component states) and comprise a holistic viewing of a scene (summary scanning). Adjectives are thus typically used to designate stative properties, profiling an atemporal relation between a thing and the property, more permanent when used pronominally, and occasional when following the noun or clause. Langacker proposes (2000: 11–12) that adjective is a relational landmark, evoked with a trajector, a thing, prototypically a Noun, in particular its active zone. ADJECTIVES then are the relational landmarks associated with the trajector through their active zones. VERBS on the other hand “profile a process – the change is plotted through time in the active zone’s location.” PARTICIPLES are the linguistic elements which retain the sequential scanning of an original verb, but which are typically profiled in terms of atemporal, summary scanning. The prototypical adjectival position in English and Polish is that preceding the Noun. However, in both languages, adjectival forms, in particular participial modifiers, can also be used postnominally. There is a class of post-predicate to-clause controlling adjectives e.g. unaccustomed or used (cf. Biber et al. 1999 :716), whose route of adjective-forming derivation is more complex, frequently from the Noun, via denominal verb, participle, and, eventually, adjective. In a few cases, as suggested by Biber et al. (1999 :530), e.g. in the case of the adjectives interested or crowded, it is the noun rather than verb which seems to function as the original base form in the derivation. Some of the negative participial modifiers are not formed directly from the verb either (some of such negative verbs do not exist in English or Polish), but from the (positive form of) the corresponding adjective (e.g. uninteresting ← interesting, cf. *uninterest or 1.

Corpus materials quoted in this paper are given in italics.

2. The terms perfect – imperfect are used here for the Polish morphologically marked category, retaining the terms perfective – imperfective in the original senses posited by Langacker.

 Barbara Lewandowska-Tomaszczyk

unread ← read, cf. *unread). In still other cases, there exists a semantic relation between the adjective and the base verb in English (outstanding, ashamed), but the relationship between the two is not so straightforward as in other instances. Participles belong, as noted before, to the category of atemporal relations. And yet, they clearly maintain in their base a dynamics of the verb they derive from. In English there are basically two types of participial modifiers: present or active (-ing) participles and past (-ed) participle. A subcategory of the latter is the passive participle, i.e. the participle of transitive verbs. Some of the modifiers can function both attributively as prenominal and postnominal modifiers, and predicatively, while some others cannot be used in one or some of these positions. While taken in absolute terms, attributive uses of participial adjectives in English outnumber predicative uses for both types of participle (cf. Biber et al. 1999 :530), and their position is not semantically neutral. The situation is similar in Polish. However, both conceptually and syntactically, English and Polish represent a different picture. Polish permits some complex participial premodification in the attributive use, which cannot occur in English. An important observation by Huddleston and Pullum (2002: 1644) refers to a polysemy between the main –ing categories in English, those of adjectives, gerunds and gerund participles, and present and past participle forms of the verb such as in: I. amusing, boring, entertaining, stunning, tiring, worrying II. amused, bored, stunned, spoiled, tired, worried Huddleston and Pullum give examples of verbs which do not have adjectival polysemous forms such as entertained (verbal forms only) and spoiling (verbal forms only). Biber et al. (pp. 66,67) also note the borderline cases of participial class membership used as a premodifier. An ambiguity arises between permanent and occasional uses as well as between adjectival and nominal (gerund) instances as in the travelling public in the sense ‘the public which is travelling’, or (more likely) ‘the public which travels’ or a travelling rug meaning ‘a rug for travelling (with)’. Other cases of ambiguity, between Noun and Adjective are e.g. living standards as opposed to living creatures or dancing classes versus the dancing children.

4. Present participle The participial construction in English, used postnominally, is ambiguous. It can be interpreted as a present participial clause, adverbial in character as in (1a), where the participial subject is coreferential with the main clause subject (a boy), or as an adjectival noun complement, where the subject of the participle is coreferential with a girl (1b).

(1) A boy looked at a girl reading a book [noun complement present participle (present participial clause)] [a. I-I, b. I-Mary]



Asymmetries in participial modification 

In Polish, the first interpretation involves the use of an adverbial participle and is rendered as (2a), while the modification reading, equivalent to (2b), is expressed by the adjectival participle, which is inflected for Case, Number and Gender, similarly to Polish Nouns and Adjectives. (2a) Chłopiec spojrzał na dziewczynę czytając [adv, coreferential with main subject] książkę (2b) Chłopiec spojrzał na dziewczynę czytającą [adj Acc Sg], noncoreferential with main subject] książkę Some of the participial modifiers in English can be used both predicatively and attributively, some others cannot (cf. Huddleston & Pullum (2002: 78)), or involve an ambiguity between adjectival and participial forms (e.g. (5) below). In Polish, even though the position of participial modification is generally more relaxed, the sentences below, in particular (6a) are only marginally acceptable: (3)

Spojrzałam na czytającą (książkę) Marysię

(3a) *I looked at reading a book Mary (4)

I got rid of this smoking chimney-pipe. [present participial adjective in attributive position].

(4a) Pozbyłam się tego dymiącego komina (5) This chimney-pipe is smoking (acceptable only as the present participle of the Present Continuous tense, not as a modifying (adjectival) participle) (5a) ? Ten komin jest dymiący (6)

* This chimney-pipe is very/too/more smoking (cf. Huddleston & Pullum (2002) for the verb/adjective tests)

(6a) ?Ten komin jest bardzo/zbyt/bardziej dymiący As originally noted by Biber et al. (1999 :530), the participial adjectives such as smoking above (similarly to the Polish dymiący) or following Pol. ‘następujący’ and alleged Pol. ‘rzekomy’ do not posses all the characteristics typical of adjectives: they are not gradable and cannot occur freely with all copular verbs (cf. also examples (33)-(35). Others, such as surprised ‘zdziwiony’ or interesting ‘interesujący’ are not so restricted in either of the two languages. It means that the forms such as surprised are more adjectivized i.e. their profile is fully atemporal, in contradististinction to e.g. following, in which the verbal (sequential) profle retained seems much more salient. In English, as seen in (3a) above (cf. also Huddleston and Pullum (2002)), the adjectival participial modifier cannot take an object:

(7) This was a frightening scream



(8) This was a frightening * us all scream

 Barbara Lewandowska-Tomaszczyk

In Polish such constructions are acceptable: (7a) To był przerażający krzyk (8a) To był przerażający nas wszystkich krzyk Langacker observes that the –ing constructions have an imperfectivizing or iterative function with open-ended initial and final state verbs, even inherently perfective e.g. He’s kicking the ball (many times) (cf. Figure 1 after Langacker 1987 :26, 2000 :226). Imperfective processes do not presuppose onset and offset of a process in their conceptual model. What is profiled onstage is, in a mass-like fashion, only a portion of an otherwise unbounded process, in a sequential manner. Perfective processes have both a beginning and an ending profiled onstage. The –ing participle, as in a stopping car, imperfectivizes the process of stopping by excluding an offset point in the profile. At the same time, the sequential scanning is retained only indirectly. What is eventually profiled is a static atemporal property associated with the participant.

Figure 1.

The adjectival, in particular prenominal, uses of such participles involve the imperfectivizing function. A dancing girl is a girl that performs the action of dancing iteratively to the extent that this property is seen as a permanent feature, adjectival in character. The construal of such an adjectival phrase is clearly atemporal and represents summary scanning, similar to nominalization (cf. Figure 1(c)). The difference between a gerundial –ing form as in a dancing party and the participial adjective in a dancing girl is that the gerund involves reification (thing) conceptualization, while the adjective expresses a stative property. A kicking player is possibly a less plausible construction, if understood as a player performing one act of kicking, cotemporal with the act of speaking, but more probable in the sense of, say, a player whose steady job is to perform kicking (viz. a dancing partner). There can also be observed an interdependency between word order and transitivity in both languages, which will be discussed in 4.1. below. In Polish the position of the modifying participle in a nominal phrase is weakly constrained but an interdependency between the position of the participle before or after the noun and the type of scanning can be identified in some such cases. The phrases tańcząca (w barze) dziewczyna lit. ‘dancing in bar girl’ or palący mężczyźni lit.



Asymmetries in participial modification 

‘smoking men’ or dziewczyna tańcząca (w barze) lit. ‘girl dancing (in bar)’ and meżczyźni palący lit. ‘men smoking’ may be interpreted both in terms of sequential scanning, coextensive with the speech event, or with atemporal summary scanning, as an adjective, extending beyond the time of speaking. These positions convey thus either a descriptive, non-restrictive interpretation or a generic/identifying reading, if unaccompanied by other modifiers, as will be further discussed. However, there exists an interesting regularity in Polish which can be observed in all adjectival modifiers. Generally, the Polish prenominal modifier tends to convey a descriptive interpretation as in e.g. brunatny niedźwiedź ‘brown bear’, while the postnominal modifier introduces rather a generic (genus) interpretation as in niedźwiedź brunatny lit. ‘bear brown’ (cf. Fisiak et al. 1978). I observe that this regularity is also present in participial modifiers: (9) Grupa tańcząca zbiera się tu, grupa śpiewająca – tam. ‘The dancing group [lit. group dancing] is gathering here, the singing group [Lit. group singing] there’. [The italicized phrases refer to the groups of people not dancing or singing at the moment of speaking] (10) Tańcząca grupa młodzieży widoczna była z daleka. ‘The dancing group of young people was seen from a distance’ [The phrase refers to a group of people dancing, time coreferential with the moment of speaking]. (11) (12)

Mięso ugotowane – pokrój. Surowe – włóż do lodówki ‘The cooked meat [lit. meat cooked] – slice, while the raw – put to the fridge’ Gotuj pół godziny. Ugotowane mięso pokrój. ‘Cook for half an hour. The cooked meat slice’.

As seen in the above examples, the generic type of modifiers, including basic or derived adjectives, is put more frequently in the postnominal position in Polish. This position signals imposition of a broader temporal scope, which does not necessarily coincide with the time of speaking. It expresses a fixed property and involves a contrastive identification of an object (one type of person/thing as opposed to another one). The prenominal modifier on the other hand, including the participial one, expresses a descriptive attribute, frequently occasional and time coreferential with the moment of speaking. It should be mentioned at this point that the stative property is also fairly regularly rendered in both languages in the form of Agentive nominalization as in tancerka ‘dancer’ or palacz ‘smoker’. A number of contextual grammatical markers in both languages can introduce more constrained interpretation of participial modifiers. Polish and English corpus data confirm this intuition. The indefinite article, the category of Number or particular lexical classes (e.g. the verb zastała – found in (15) below) express either a temporal overlap between the main event and the event expressed by a participle. They impose a more static, adjectival interpretation, when the Plural number of the Noun, adverbial

 Barbara Lewandowska-Tomaszczyk

phrases of frequency or verbs semantically marked for stativity or duration are used (cf. 15). Compare a more holistic interpretation of: (13) listonosz czytający cudze listy ‘a postman reading other people’s letters’ [iterative/habitual marked by the Plural form listy ‘letters’ and the imperfective/ unmarked czytać ‘read’] (14) myślimy o bystrym i czytającym nas od dawna czytelniku lit. ‘we’re thinking of a bright and reading us since long reader’ [generic Singular/habitual marked by the Singular czytelniku ‘reader’ and the temporal Adverbial od dawna lit. ‘since long’]. as opposed to instances expressing more sequential scanning: (15) zastała Carlę czytającą jakąś odręczną notatkę lit. ’she found Carla reading a hand-written note’ [time overlapping between found-reading, the semantics of find, the Singular note] (16) powiedziała do czytającego gazetę ojca lit. ‘she said to [her] reading a newspaper father’ [time overlapping between said-reading, the semantics of say, the Singular newspaper] The English sentence: (17) Eh? What? I’m not a reading man, but it sounds all right does not show an aspectual or temporal correlation with the corresponding relative construction He is the man who is reading. Neither is the interpretation of the modified NP in the latter synonymous with that present in (17). The modified NP in the latter designates a temporal overlap between reading and the moment of speaking, while (17) presupposes rather that He is the man who reads (books, etc.), and conveys a habitual or iterative interpretation. The effect of –ing in (17) then is to imperfectivize the inherently bounded, event though atelic, process of reading. The sequential scanning and an immediate temporal scope overlapping with the time of speaking is rendered by a postnominal modifier (i.e. complement): (18) It was not him, that old man reading whom he wanted to kill. This contrast is clearly visible in the examples below. In (19) the prenominal, adjectival property involves a summary scanning of opening, where all the subsequent phases of the process are present in the construal of the scene, even though the profile is stative and atemporal: (19) he closed the windows and doors, the doors of his mind flew open one after another, like living inside the heart of an opening flower.



Asymmetries in participial modification 

In (20) and (21), the contrast between the (postnominal) sequential and (prenominal) summary scanning is particularly strong: (20) returning home with baskets of yams from a distant farm across the stream when they had heard the voice of an infant crying in the thick forest. (21) Now she was surrounded by girls, girls, girls – hundreds of them, working, playing, singing, laughing, crying, dancing, studying, lazing, praying. It started that way: laughing children, dancing men, crying women and then it got mixed up. Similarly to Polish, there can be mentioned a number of contextual factors that contribute to the habitual as opposed to temporally coextensive and deperfectivized interpretations in English. Such factors involove the information of the generic status of nouns used (marked by the article and the grammatical Number) or the semantic content of the verb (e.g. keep in (23), which imposes a (progressive) aspectual constraint on the act/state): (22) To make promises to himself like a man looking at travel brochures of exotic places. No, more a man reading about polar explorers living off snow, moss and boot-leather. (23) He kept a smiling face to hide the strain he felt

4.1

Transitivity and word order (cf. Dziwirek & Lewandowska-Tomaszczyk forthcoming)

In both languages verb transitivity is observed to play a role in the position of noun modifiers. The frequency of transitive present participles, which tend to follow the head Noun (24) is higher than that of intransitive ones, which are more often observed to precede it (25): (24) chłopiec czytający książkę the boy reading the book (25) śpiące dziecko the sleeping child One area of difference are participial modifiers based on intransitive verbs containing adverbial expressions, which in Polish may act as modifiers and precede their heads, just as a simple participle does in (25). This structure/word order is not possible in English, where we find the complex modifiers with the nominal head followed by the participle and the adverbial expression, or relative clauses instead. (26) śpiące na podłodze dziecko lit. sleeping on the floor child the child (who is) sleeping on the floor

 Barbara Lewandowska-Tomaszczyk

(27) odchodzący do rezerwy zażywny podpulkownik Wiktor lit. leaving for the reserves jovial lieutenant-colonel Victor the jovial lieutenant-colonel Victor (who is/was) leaving for the reserves (28) kontrastujący z całym otoczeniem budynek lit. contrasting with the entire surrounding building the building (which is/was) contrasting with the entire surrounding (29) obowiązujący do końca sezonu 1999/2000 kontrakt lit. binding to the end of the 1999/2000 season contract the contract (which is/was) binding till the end of the 1999/200 season (30) siedzący obok mnie starszy dżentelmen lit. sitting next to me older gentleman the older gentleman (who is/was) sitting next to me

5. Past participle Huddleston and Pullum (2002  :79) observe that in English objects and predicative complements are found with verbs but not with adjectives. The following forms are therefore verbal not adjectival participial modifiers: (31) He quickly spent the money given him by his uncle (Verb + Indirect object) (32) They sacked those considered guilty of bias (Verb + Predicative complement) Another test, involving the verb seem, is passed by adjectives but not by all participial modifiers: (33) The picture seems excellent (34) ?The boss seemed considered guilty of bias A similar test can be applied to the Polish material as well, where (34a) is an equivalent to (34). Some such constructions are peripherally acceptable in Polish. (34a) ?Szef wydawał się uważany za winnego (posiadającego uprzedzenia) (35) Książka wydawała się?przeczytana/czytana przez niego – *the book seemed read / (by him) The English language has a polysemous past participle form. Langacker (1987 :203) proposes that one of these constructions he calls PERF1 involves stative adjectival forms, derived from the participle formed on intransitives The windshield is cracked (cf. The windshield cracked). PERF2 on the other hand is a stative adjectival form, based on the participle formed from transitive verbs This car is stolen [nonpassive]; A tornado left the town totally devastated. Even though all the phases of the process are present in their conceptualizations, what is profiled is the final state of the process.



Asymmetries in participial modification 

Fig. 2a

Fig. 2b

Langacker argues for the adjectival sense of these stative participial constructions by referring to their base, in which the processual predication designates “a continuous series of states distributed through time...The stative participle, however, designates only the final state in the overall process...The evolution of the situation through time is thus a prominent facet of the participle’s meaning, but it is confined to the base and left unprofiled” (Langacker 1987: 221). The difference between a prototypical adjective and a participial adjective relates exactly to the temporal dimension, which is absent in a typical adjective. A process is certainly interpreted sequentially and temporarily, and can be conceived in terms of a state-change sequence, where the input state must be kept counterfactual with reference to each successive step (cf. Lewandowska-Tomasz­ czyk 1996), while in the adjectival participle this change sequence remains at the background, is confined to the base and left unprofiled. In English, change-of-state participles can be used prenominally: (36) I saw a broken glass (adj participle in attributive position) However, in the contexts where the past participle is used with the Agent, and is thus more verbal in character, the only option available is the postnominal noun complement modification, expressed by the passive participle construction: (37) I saw a glass broken by Tom [past participial noun complement (clause) – passive participle]

 Barbara Lewandowska-Tomaszczyk

An important parameter which contributes to prenominal modification in English and involves past participles is the status of the nominal element of the phrase, which is more likely to enter such a construction if the object has the status of an affected object. A difference can be observed between the frequent forms such as a written document or spoken language, in which the document is a product of writing and the language – a product of speaking, as opposed to much less likely ? a read document, or ? a seen man which presuppose the existence of a document or a man, not in fact affected by the reading or seeing. In the phrase a painted house, in both the senses – as a product of painting, or as an object of painting – both types of object are affected entities and undergo a change, from non-existence to existence in the former or from not painted to painted in the latter case. Similarly affected is the object in a broken arm, which informs us of the condition of the affected entity, i.e. an arm. In other words, premodification is more likely in English in the case of past participles if the state change affects a modified object. Prenominal modification in English is also frequent in the cases where the completeness of the process is additionally marked by a preposition, which conveys a clearly perfectivizing funcion. While then the sequence (in non-contrastive setting) such as ‘this is washed laundry’ is not annotated in the corpus, examples such as (38) and (39) below are abundant: (38) She was wearing a washed-out lilac print dress which hung limply upon her (39) It was an overcooked dinner A more confined interpretation can be expressed in English by a variety of adverbial phrases modifying the participial adjectival phrase. Such phrases play a perfectivizing role in otherwise non-perfective, atelic verbs such as write: (40) It was a hurriedly written letter As noted in Huddleston & Pullum (2002: 78) the Agent marking cannot appear with the adjectival participle in the pronominal position: (41) It was *a hurriedly written by her friend thank you letter In Polish, similarly to the cases of the present participle, the situation is only partly analogous to English. On the one hand, there are fewer constraints on the past participles in their usage in the pronominal position, which could signify the higher degree of their adjectivalization. On the other, the participles retain a number of verbal (sequential) properties such as the Agent marking, even when used in the pronominal position. Compare (41a), which is equivalent to the English (41): (41a) To był pospiesznie napisany przez jej przyjaciółkę (dziękczynny) list / list z podziękowaniem.



Asymmetries in participial modification 

The problem with participial adjectives in Polish would then be that, differently to what Langacker (2000  :10) suggests for prototypical adjectives, viz. “an adjective or adverb has just one participant with focal prominence”, the prenominal participial adjective in Polish can retain the original Verb semantic roles (participants). Its verbal properties are thus clearly profiled in Polish. In both languages the past participle can also be used in premodifying and postmodifying positions. In English the prenominal position can be almost exclusively occupied by the passive participle, i.e. by the participle formed of the transitive verb, more often with the generic or permanent interpretation, reinforced by other relevant grammatical markers such as Articles and Number: (42) When the market for a printed book declines, the presses stop rolling but thousands of copies may remain. (43) It should be noted, however, that not all advertised products are specials; (44) Yet he remains a forgotten figure. The objects with less permanent duration, of temporal coincidence or more occasional ones, have the participial modification used postnominally: (45) He had a deep hole dug in the floor of his room. (46) If we could immobilise a small area of that slope and the upper face of a hole dug into the sand, we might be able to put down a shaft to the pack. (47) Scrappy notes of lectures piled up unread kept pouring in. As has been already noted above, the past participles of some change-of-state verbs can be used prenominally in both languages: ‘a broken arm/złamana ręka’. This position is also usually reserved to a participle with more narrow modification as in: (48) making unintended errors of judgement. (49) widely shared errors of judgment (50) devastatingly combed hair (51) they found a stream running down from the hills behind in a deeply dug bed with steep slippery sides overhung with brambles.

 Barbara Lewandowska-Tomaszczyk

(52) She was a half-caste and had a lovely coffee coloured skin.3 In Polish the word-order constraints are weaker: (53) Popraw zrobione błędy lit. ‘Correct the made errors’ – ‘Correct the errors you (or somebody else) made’. (54) 500 specjalnie na tę okazję przygotowanych koszulek piknikowych lit. ‘500 specially for this occasion prepared shirts picnic [Adj].’ Passive participles may alternate with transitive relative clauses in both languages (cf. Dziwirek & Lewandowska-Tomaszczyk forthcoming): (55) Praca, którą wykonywały te kobiety/ Praca wykonywana przez te kobiety Work which was performed by these women/Work performed by these women The intransitive past participles are extremely rare in the prenominal modifier position in English (escaped criminal/animal; fallen tree). Their sense is frequently perfective in such contexts, similar to that of the transitive change-of-state verbs (a broken glass). They are readily extended in the range of meaning and incorporate metaphorical senses as in the fallen woman, the fallen angel, or as in the poet John Betjeman’s never-sent letter to his despised tutor “that white unlived-in room of yours in New Buildings... which always depressed me” (after Priestman 2006 :23). The participle unlived-in in the last example is especially interesting in that this expression profiles a state, constrained by the negative prefix un- and additionally grounded by the preposition/ adverb in. In Polish, more flexibility in this respect is observed, in particular in fixed expressions, frequently presenting obsolete grammatical forms such as padłe zwierzę ‘fallen [dead] animal’, rosły chłopak lit. ‘grown [tall] boy’, or in prefixal perfective verbs 3. cf. Biber et al. (2001) for numerous examples of participial compounds in English, very frequent in journalistic language. Compare also Polish equivalents, in which various construals are used to convey original English compounding processes such as Prepositional Phrases, Case marking, Adverbs, etc.: Noun - Noun compound: glass-topped ‘ze szklanym (pokryciem, dachem, szczytem...)’ lit. ‘with a glass (cover, roof, top...[Instrumental]’ Adverb/Noun – Verb compound: sexually-transmitted ‘przenoszony drogą płciową [Instrumental]’, psychologically-disturbed ‘z zaburzeniami psychicznymi [Preposition+Instrumental]]’, classroom-based ‘w klasie (szkolnej)/klasowy [Preposition/Adjective+ Locative]’, world-renowned ‘o swiatowej sławie [Preposition+Instrumental]’. Adv+ ed-participle ill-suited ‘niepasujący’ [Negation] internationally-acclaimed ‘o międzynarodowym uznaniu’ [Preposition+Instrumental] new-born ‘nowo narodzony’ [Adverb] Adverb + ing-participle: brightly-shining ‘jaskrawo/jasno świecące’ [Adverb] constantly-changing ‘stale/ciągle zmieniające się’ [Adverb]



Asymmetries in participial modification 

(opadłe liście lit. ‘around + fallen leaves’) or wyrośnięty chlopak lit. ‘out + grown boy’), which provide an additional perfectivizing dimension.

6. Process/state Interesting cases with the adjectival past participle interpretation can be observed in such verbs, which have both past (passive) participial and adjectival forms derived from the same base. Such is the case of the pair opened and open. In the form opened [V – process] what is present in the base is a change of state. It seems that what is profiled is not only the final resultant stage but the initial phase notopen as well.4 This is especially conspicuous when compared with the fully adjectivized open. (56) Cordelia rummages in the fridge and brings out an opened package of store doughnuts. (57) Friday afternoon when I came home, unlocked my door and found an opened Globe and Mail pushed under my door.

4. A similar conception of the semantics of CHANGE in language was proposed by Jackendoff (1972). Out of the three major semantic ‘proto-verbs’ characteristic of verbal concepts, namely BE, CAUSE, and CHANGE, CHANGE is taken to presuppose CAUSE and involves a shift from one state to another, i.e. from an initial to a final state (examples below after Jackendoff 1972:41). This shift can also be expressed as a succession of steps alternating between the presence and absence of a state, i.e. between the positive and the negative.

(1) The door opened open +V +NP CHANGE Physical (NP, NOT OPEN, OPEN)



(2) John opened the door open +V +NP ___ NP CAUSE(NP,___CHANGE Physical (NP, NOT OPEN, OPEN)

 Barbara Lewandowska-Tomaszczyk

In the form open [Adj – state], what is profiled is only the final stage of the whole process of opening, conceptualized in terms of an atemporal state: (58) I loved the floods: every year I’d pray in the dark for more rain, more rain. One year in high apostate excitement, I steered a canoe through an open window of the church itself; rowed around the altar and round the organ and floated out a window on the opposite side. In the Polish language the dichotomy between the state and process can be rendered in the form of two participial forms: otwarte-a ‘opened, open’, which designates a state with the final state profiled, and otwierane lit. ‘being opened’, which conveys a sequential, processual conceptualization. In both, the base contains all the successive phases of the process together with the negative state of the beginning of an action and the resultant one. The former is ambiguous between the participial and adjectival sense. Under the first (participial, dynamic) interpretation what is profiled is the state change, under the other (stative, adjectival) interpretation, profiled is the final state alone. In Polish there are regular – morphologically distinguished – perfective and imperfective verbs – jeść ‘to eat’/zjeść ‘ to have eaten’. I use the terms perfect and imperfect for the Polish aspectual distinctions to avoid overlaps with Langacker’s perfective-imperfective terms as the two are not entirely synonymous with each other. Both the (grammatically) perfect and imperfect verbs in Polish can be bounded (perfective in Langacker’s terminology) or unbounded (stative, imperfective). The Past Participle, formed with the suffixes –ony/-a/-e and – any/-a/-e is declined for grammatical Number, Case and Gender. The Past Participle of the perfect verbs conveys the meaning of a completed action as in zjedzony obiad ‘an eaten (up) dinner’, which in fact conveys a stative adjectival sense.5 The Past Participle of the imperfect verb renders the durative and progressive meaning as in jedzony obiad lit. ‘a being eaten dinner’. The adjectival uses of past participles with both Langacker’s perfective as well as imperfective verbs are much more frequent in Polish than in English, e.g. kopnięta piłka lit. ‘a kicked ball, and e.g. zrozumiane polecenie lit. ‘an understood order’, (u)słyszana piosenka lit. ‘(having) heard song’. The prenominal participle position also occurs in Polish in the case of inherently imperfective and/or iterative verbs (e.g. czytywane czasopisma lit. ‘read [iterative] magazines’). The reasons for the Polish/English contrasts in this case may be conceptual and semantic. The English eaten dinner would be ambiguous between a completed and progressive reading. The participial –ed suffix then seems to have no sufficient perfectivizing effect in such cases. The adjective is characterized as atemporal because, as Langacker (2000 :10) proposes “evolution through time is not intrinsic or central to the profiled relationships’s characterization”. The reason why the English participial adjective is considered atemporal is that in the process of its construal what is profiled is not a temporal sequence 5. The meaning of the phrase is certainly much more complex as it involves a verb of cessation or annihilation, which implicates non-existence after the action is completed.



Asymmetries in participial modification 

of states or events but an atemporal state, and more precisely the final resultant stage of an action. In Polish this would be true of perfect participles, while participles of imperfect verbs retain their temporal dimension. This property of the imperfect past participle in Polish overlaps with that of the present participle (which are formed for imperfect verbs only). Both profile a process. The phrases gotowana zupa ‘soup being cooked’ and gotująca się zupa ‘cooking itself soup [Middle reading]’ profile processes par excellence, and, due to their incremental nature, they can be perceived as a series of gradual changing states of the same character, with no final offset stage profiled. Taken from this perspective, the participles are only partly adjectival, even though, similarly to prototypical Adjectives in both languages, they can be used before the Noun, more frequently in Polish. These facts would explain why the imperfective participles in English would be very rarely used in the adjectival prenominal function. The inherent imperfective (unbounded) meaning of such verbs would contradict their possible use as a fixed property, complete by implication. Only when some additional modifiers are used to perfectivize their senses (especially productive is the negative prefix here as in unheard of stories, unseen shapes, misunderstood prophecy, etc.) would the past participles of inherently imperfective and atelic verbs get perfectivized (i.e. provide a mental boundary) and be used in their adjectival prenominal function as in (59, 60, 63): (59) a half-eaten [?an eaten] sandwich ‘(na wpół) (z)jedzona kanapka‘ (60) a moth-eaten [?an eaten] Beatles wig on his head. ‘zjedzona (przez mole) peruka Beatelsów na głowie’ (61) ?a drunk beer ‘(wy)pite piwo6 (62) ?a seen/heard man ‘widziany/słyszany człowiek’ (63) ?a read book ‘(prze)czytana książka’ a half-read book ‘na pół przeczytana książka’ an unread article ‘nieprzeczytany arytkuł’7 (64) ?an enjoyed performance ‘przyjemne [lit. pleasant] przedstawienie’8 What is important here is the perfectivity status of the verb and a corresponding participle, in particular the aspect of inherent or imposed bounding, duration and permanence. The verbs such as see, eat, are semantically imperfective verbs in English, which do not lend themselves to adjectivization. In the examples more narrowly defined 6. a drunk person – Pol. ‘pijany’ {deverbal lexicalized Adj} is an outcome of a different semantic process 7. In a discussion with Dirk Geerearts, he enquired about a possibility of using some of the examples, not attested in the corpora, in a contrastive setting as e.g.: ‘These are read books on the left, and unread ones on the right’. It seems that clearly contrastive contexts may allow more varied uses. 8. There is no one-word equivalent of the verb enjoy in Polish.

 Barbara Lewandowska-Tomaszczyk

(half-eaten, moth-eaten) the perfectivity parameter, imposing mental bounding, is (more) clearly signalled. Such a boundary acts similarly to the state change, which, in English, enables participles to adjectivization in the pronominal position. This seems to be a reason for the acceptability of such examples. In Polish – the reference to the whole category of objects or to its unique member [generic/unique distinction] also plays a role, in English it is mostly a distinction between the permanent steady/state property and an occasional one.

7. Passive and middle sequential scanning The -ing participial adjectives in English, discussed in previous sections, present cases of conceptualization ambiguity, which is systemically disambiguated in Polish. The English participle yields two possible readings, one – medio-passive (middle) reading, and one – passive reading, as in (65–66) below. In Polish, the equivalent constructions can have either the middle (reflexive) form with the particle/clitic się lit. ‘oneself ’ (ex. 65) or the imperfective (durative) form, equivalent in meaning to the English passive progressive form (ex. 66): (65) Poirot heard distinctly: `Bonne nuit, Madame,’ and a closing door [zamykające się drzwi lit. ‘closing itself door’/zamykane drzwi lit. ‘having been closing door’]. (66) Maybe this one time she could stop dead still in the middle of a cooking meal [gotującego się posiłku lit. ‘cooking itself meal’/ gotowanego posiłku lit. ‘having being cooked meal’] The Polish forms, with an unexpressed Agent, profile thus either successive stages of the process (middle reflexive) or a progressive on-going process (passive continuous). The middle reflexive form is a clear case of intrasitivizing of a transitive verb, while passive continuous involves the imperfectivization of the process.

8. Conclusions The participle modification in English and Polish involves the following parameters (cf. Table 1 and Table 2): 1. Present participle modification Imperfecivizig function English: steady/state (prenominal)– occasional (postnominal) prenominal middle vb – durative Polish: descriptive/nonidentifying (prenominal) – generic/identifying (postnominal)



Asymmetries in participial modification 

Prenominal: intransitivizing middle – reflexive or passive durative English: multi-unit (MU) participles – only postnominal Polish: multi-unit (MU) participles – pre/post nominal Table 1.  Present participle modification in English and Polish PRESENT PARTICIPLE MODIFICATION IMPERFECTIVIZING FUNCTION PRENOMINAL Summary Scanning

POSTNOMINAL Sequential scanning

Eng.

steady/state Middle Vb – durative

occasional

Pol.

descriptive/nonidentifying intransitivizing middle reflexive passive durative

generic/identifying

Eng.

Multi Word Participle [MWP]

Pol.

MWP

2. Past participle modification Perfectivizing function Prenominal English: transitive/intransitive – perfective Polish: transitive – perfective, intransitive/transitive – passive progressive English: multi-unit (MU) participles – only postnominal Polish: multi-unit (MU) participles – pre/post nominal

sequential scanning

MWP

 Barbara Lewandowska-Tomaszczyk

Table 2.  Past participle modification in English and Polish PAST PARTICIPLE MODIFICATION PERFECTIVIZING FUNCTION PRENOMINAL Summary scanning

POSTNOMINAL Sequential scanning

Eng.

transitive/intransitive – perfective [bounded] [bounded] state change – affected object

transitive perfective

Pol.

transitive perf. passive imperf. progressive [unbounded]

sequential scanning

Eng.

Adverbial Past Prt N

MUPs

Pol.

MUPs

MUPs

The aim of the paper was to identify cognitive conditions on the use of some participial modification forms. Parameters which constrain the use of the participles pronominally and postnominally appear to cover the following characteristics of the processes and events: temporal relations (coextensiveness), connected with the scanning type (process/summary), generic/unitary interpretation, profiling (scope), bounding (perfectivity), transitivity, and a range of auxiliary modification. What is of a special role here is the aspectual status of the verb and a corresponding participle, in particular its bounding, duration and coextensiveness as well as the category of grammatical voice and associated semantic roles. What matters in Polish is the reference to either the whole category of objects or to its unique member. A version of this parameter also plays a role [generic/unique] in English with reference to a permanent/occasional property of an object.

References Biber, D., Johansson, S., Leech, G., Conrad, S. & E. Finegan. 1999. Longman Grammar of Spoken and Written English. Harlow: Longman/Pearson Education Ltd. Dziwirek, K. & B. Lewandowska-Tomaszczyk (forthcoming). Complex Emotions and Grammatical Mismatches. Berlin/New York: Mouton de Gruyter. Fisiak, J., Lipińska-Grzegorek, M. & T. Zabrocki. An Introductory Polish-English Contrastive Grammar. Warszawa: PWN. Huddleston, R. & G. K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Jackendoff, R. S. 1972. Semantic Interpretation in Generative Grammar, Cambridge, Mass.: MIT Press.



Asymmetries in participial modification  Langacker, R. W. 1987, 1991. Foundations of Cognitive Grammar vols. 1 and 2, Stanford, Calif.: Stanford University Press. Langacker, R. W. 2000. Grammar and Conceptualization. Berlin/New York: Mouton de Gruyter. Lewandowska-Tomaszczyk, B. 1996. Depth of Negation: A Cognitive-Semantic Study. Łódź: Łódź University Press. Lewandowska-Tomaszczyk, B. 2004. Gramatyka angielska na autentycznym materiale jezykowym [English Grammar on Authentic Language Material]. Łódź: Wyższa Szkoła Studiów Międzynarodowych.

Sources British National Corpus Pelcra Polish and English Corpora [Department of English Language, University of Łódź, Poland] Priestman, Judith. 2006. The dilettante and the dons. Oxford Today Trinity 2006: 20–23.

Author index A Achard, M.  59, 85, 147, 149, 227, 235, 236, 237, 241, 242 Akmajian, A.  134 Ambar, M.  242 Anderson, S. R.  167, 170 Ariel, M.  129, 130, 134 Arjona, M.  54, 85 Aygen, G.  100, 101 B Bentivoglio, P.  54, 81, 83, 85 Bhatt, R.  101 Biber, D.  280 Boretti de Macchia, S.  53, 85 Boye, K.  146, 148, 149 Bruce, L.  51, 104, 114, 119 C Campbell, L.  151, 153, 171 Cano Aguilar, R.  53, 74, 85 Casad, E. H.  25, 29, 33, 35, 40, 46, 51, 86, 208 Chafe, W.  141, 149 Chomsky, N.  141, 149, 223, 228, 246, 247, 258 Chung, D.  101 Chung, S.  101 Chuwicha, Y.  104, 107, 119 Comrie, B.  121, 133, 134, 167, 170, 197, 207, 224 Conrad, S.  280 Cornillie, B.  227, 239, 242 Ronayn, C. J.  134 Cristofaro, S.  88, 101, 151–153, 160, 162, 168, 170, 173, 174, 178, 182, 190, 193, 196, 197, 207 Croft, W.  114, 119, 151, 152, 154, 157, 158, 167, 168, 170, 195, 196, 202, 203 Cruse, A.  195 Culicover, P. W.  88, 91, 101, 214, 221 D Davidse, K.  85, 257, 258

De Roover, C.  250, 258 del Moral, G.  54, 86 Delbecque, N.  53, 60, 68, 75, 85, 86 DeMello, G.  86 Diessel, H.  138, 146, 148, 149 Dik, S. C.  153, 161, 170 Dixon, R. M. W.  167, 170, 171, 197, 207 Dowty, D.  167, 170, 224 Dryer, M.  152, 168, 170, 171 Du Bois, J.  205, 208 Durie, M.  104, 119, 171 Dziwirek, K.  280 E Enfield, N. J.  108, 119, 196, 208 F Fauconnier, G.  140, 149 Fillmore, Ch.  223, 224 Finegan, E.  280 Fisiak, J.  280 Foley, W. A.  11, 22, 88, 94, 104, 196, 197 Francis, G.  4, 14, 23 Fraser, B.  247, 258 G García, E.  54, 71, 82, 86 Geeraerts, D.  6, 23 Givón, T.  97, 101, 171, 182, 183, 193, 196 Goldsmith, J.  88, 101 Good, J.  51, 56, 58, 70, 78, 82, 101, 199, 206, 250, 256 Goodall, G.  88, 101 Gómez Torrego, L.  55, 56, 74, 86 Greenbaum, S.  259 Gries, S. Th.  23 Grimshaw, J.  247–255, 258 H Haiman, J.  89, 101, 154, 157, 158, 170, 171, 196, 200, 205, 208, 242

Hale, K.  162, 171 Halle, M.  141, 149 Halliday, M. A. K.  259 Harder, P.  51, 146, 148 Harris, A. C.  151, 153, 171 Haspelmath, M.  88, 94, 101, 155, 171, 172 Hawkins, J. A.  29, 51, 196, 205, 208, 224 Heine, B.  19, 23, 190, 193 Hernández Cabrera, C. E.  86 Herskovits, A.  61, 86 Heylen, K.  6 Heyvaert, L.  245, 249, 257, 259 Hobbs, J. R.  125, 134 Hooper, R.  173, 186, 193 Hopper, P. J.  154, 171, 250, 259 Horie, K.  195–197, 203, 205, 207, 208 Huck, G.  88 Huddleston, R.  5, 21, 23, 247, 259 264, 265, 270, 280 Hume, D.  125, 134 Hunston, S.  4, 14, 23 I Izutsu, M. N.  121, 124, 134 J Jackendoff, R. S.  88, 91, 101, 134, 213, 214, 221, 224, 280 Johansson, S.  280 Johnson-Laird, P. N.  51 Joseph, B. D.  158, 171 Juilland, A.  86 K Kany, Ch.  53, 86 Kastovsky, D.  247, 248, 251, 255, 259 Keenan, E. L.  121, 133, 134, 167, 171 Kehler, A.  88, 101, 125, 126, 133, 134 Kemmer, S.  86 Köpcke, K.-M.  209, 213, 214, 215

 Asymmetric Events Koptjevskaja-Tamm, M.  158, 171, 250, 251, 256, 259 Kuteva, T.  19, 23, 190

Newman, J.  23 Nikolaeva, I.  158, 172 Noonan, M.  164, 172, 196, 208

L Lakoff, G.  88, 101, 124, 134, 195, 208 Lambrecht, K.  153, 160, 171, 173, 174, 193 Lane, J.  11, 23 Langacker, R. W.  15, 17, 23, 25–31, 35, 38, 39, 41, 42, 45, 46, 48, 49, 51, 104, 111, 115–117, 119, 121, 134, 137, 138, 141–143, 147–149, 151, 159–162, 171, 173, 174, 193, 227, 228, 232, 234–236, 242, 243, 247, 248, 252–255, 258, 259, 262, 263, 266, 270, 271, 273, 276, 281 LaPolla, R.  4, 23, 152 Leech, G.  259, 280 Lees, R. B.  246, 247, 259 Lehmann, C.  153, 167, 171 Levin, B.  246–250, 255, 257–259 Levinson, S. C.  153, 171 Levy Podolsky, P.  75, 86 Lewandowska-Tomaszczyk, B.  207, 224, 243, 261, 269, 274, 280, 281 Li, C. N.  165, 166, 171 Lichtenberk, F.  5, 23, 173 Lin, J.  23 Lipińska-Grzegorek, M.  280 Longacre, R.  89, 102 Lorimer, D. L. R.  165, 172

O Olson, M.  11, 22, 104

M Mackenzie, L. J.  247, 248, 250, 251, 255, 259 Manning, C. D.  23 Manning, E.  4, 14, 23 Manzini, R.  213, 224 Marr, D. N.  51 Martins, A. M.  242 Matisoff, J. A.  195, 208 Matsumoto, Y.  47, 51 Maurer Jr., T. H.  242 Maxwell, M. B.  164, 172 Mensching, G.  242 Miller, G. A.  51 Morse, N. L.  164, 172 N Na, Y.  88 Nerlich, B.  208

P Palmer, F.  167, 172, 207, 208 Panther, K.-U.  23, 209, 213–215, 217, 221, 224, 225 Pawley, A.  11, 23, 105, 106, 108, 120, 193 Pollard, C.  221 Portner, P.  100 Pottier, B.  61, 86 Pullum, G. K.  5, 21, 23, 247, 264, 265, 270, 280 Pustejovsky, J.  221, 225 Q Quirk, R.  259 R Rabanales, A.  55, 56, 86 Raposo, E.  227, 228, 241, 243 Rappaport-Hovav, M.  246–250, 255, 257–259 Reinhart, T.  151, 154, 172 Rice, S.  23 Roberts, J. R.  155, 172 Roeper, T.  247, 248, 259 Rohde, A.  23 Rosenbaum, P.  212, 213, 223, 225, 258 Ross, J. R.  87, 88, 91, 102, 173, 186, 193 Ross, M. D.  193 Rudnitskaya, E.  89, 99, 102 Ruiz de Mendoza, F.  225 S Sadock, J. M.  88 Sag, I. I.  221 Samper Padilla, J. A.  86 Schmerling, S.  88, 102 Schwenter, S. A.  54, 71, 86 Scida, E.  230, 243 Searle, J.  215, 225 Serradilla Castaño, A. M.  53, 86 Sheldon, A.  121, 123, 125, 133, 134 Shim, S.-W.  100, 102 Soares da Silva, A.  227, 231, 234, 239, 242, 243 Sohn, H.-M.  88–91, 95, 102

Stassen, L.  114, 120 Stefanowitsch, A.  4, 23 Sten, H.  243 Stevenson, R.  134 Stewart, J. M.  196, 208 Strang, B. M. H.  247, 259 Subirats-Rüggeberg, C.  75, 86 Svorou, S.  52 T Tai, J. H. Y.  200, 208 Talmy, L.  42, 52, 101, 151, 154, 172, 221, 225 Thompson, S. A.  138, 146, 148, 149, 151, 154, 165, 166, 170– 172, 208, 250, 259 Thornburg, L.  217 Tikkanen, B.  165, 172 Tomasello, M.  138, 146, 148, 149 Tomlin, R.  120, 151, 154, 172 Traugott, E. C.  86, 208 Tummers, J.  6, 23 V Van den Bosch, C.  250, 257, 259 Van Hoek, K.  124 Van Valin, R.  4, 23, 88, 94, 152, 196, 197 Vendler, Z.  110, 111, 120 Verhagen, A.  26, 52, 138, 146, 149 Vesterinen, R.  243 Vimala H.  208 W Wälchli, B.  88, 102 Wegener, H.  225 Whitman, J.  100, 102 Wierzbicka, A.  114, 119, 120 Willemse, P.  245, 248, 249, 253, 254, 255, 258, 259 Wulff, S.  4, 23 Y Yi, E.-Y.  89, 102 Yoon, J.  89, 102 Yuasa, E.  88, 102 Z Zabrocki, T.  280 Zanuttini, R.  100, 102 Zazie T.  208

Subject index A Ablative  53, 61 accessibility  121, 122, 129–134 acquisition  134, 137, 138, 146, 148, 149, 257 actual ground  53, 59, 66, 81 Adioukrou  195–206 adjective  33, 55, 261, 263–267, 271, 273, 276 adjunct clause  200, 227, 231 adverbial clause  137–139, 145, 146, 154, 156, 162, 195, 195–206, 227, 231, 237, 240, 243 adverbial constructions  154, 227, 232, 240 analyzability  137, 148 argument structure  234, 249, 251, 252, 258, 259 aspect  5, 6, 50, 85, 110, 153, 171, 261–263, 277 assertion  40, 54, 55, 153, 161, 173, 174, 190 assertiveness  151, 157, 161 associative  173, 185–189 asymmetry  3–8, 21, 22, 25–31, 45, 50, 87, 90, 94, 100, 102, 121–134, 139–142, 151–169, 173, 174, 190–196, 207–210, 236, 237, 258, 262 attention  31, 137, 141–148 attentional frame  137, 142, 144, 145, 147 auxiliary verb  3 B Binding Scale  173, 182 Burushaski  165, 166, 172 C causality  87, 125 causation  222, 227, 231–234, 239, 243 causative verbs  227, 239 clausal grounding  148, 227, 236, 237, 242

clause chain  87, 88, 89, 100 clause complementation  53, 62, 71 clause linkage  87, 88, 90, 100, 166, 168, 171 Cognitive Grammar  23, 59, 103, 104, 111, 115–121, 134, 137, 148, 149, 171, 193, 227, 228, 242, 259, 261, 281 cognitive profile  151, 152, 160 coherence  87, 101, 120, 125, 126, 134, 172 complement clause  55–62, 65, 137–149, 155, 173–191, 195–203, 207, 209, 211–222, 227, 231, 234, 237, 238, 241 complementation  54, 55, 57, 65, 66, 68, 71–75, 82, 85, 86, 172, 208, 225, 245, 251–256 complex event nominalizations  249, 252 complex matrix  3, 15 complex sentence  138, 139, 141, 143, 148, 172, 202, 203 conceptual content  31, 214, 261, 262 conceptual space  195, 203 conceptual-pragmatic roles  209 conceptualizer  27, 28, 29, 30, 39, 53, 54, 59–66, 71, 236, 237 conceptualizing subject  53, 59, 60, 61, 62, 63, 64, 73, 75, 84, 237 constituency  137, 141, 142, 144, 145, 149, 252 construal  28, 51–87, 97–99, 234, 236, 261, 262, 266, 268, 276 contrastive analysis  261 control shift  209, 213, 223 Coordinate Structure Constraint (CSC)  87 coordination  5, 21, 22, 87, 88, 89, 90, 92, 94, 95, 100, 101, 102, 137, 138, 163, 171, 196, 203 Cora  25

corpus  3–10, 12, 14, 21–23, 55, 86, 149, 199, 250, 257, 258, 263, 267, 272, 281 correspondences  103, 104, 115–119, 140 cross-constructional continuity  195, 203–206 cross-linguistic comparison  151, 197 Cubeo  163, 164, 172 D declarative verbs  227, 238, 239, 241 deictic center  53, 59, 60, 64, 67, 73 desiderative  173, 175, 177–190 deverbal nominalization  245–249, 252, 254–258 deverbal noun  185, 248, 254, 255, 256, 258 discourse  26, 51, 60–73, 80–85, 101, 120–138, 143, 144, 148, 149, 153, 160–172, 193, 207, 208, 214, 225, 245–259 discourse connective  121, 122, 123, 126, 131, 133 E ellipsis  245, 248, 255, 256 embedding  90, 92, 93, 98, 141, 151, 154, 155, 156, 157, 159, 164, 169, 210 English  3–15, 21–23, 48, 51, 55, 71, 79, 88–92, 99, 100–105, 120, 123, 130–134, 143, 148–158, 195–201, 205, 208–212, 218–225, 258 -281 epistemic verbs  227, 238, 239, 241 equivalence  68, 261 event  1, 15, 18, 23, 28, 48, 49, 50, 59, 62, 90–98, 103–111, 115, 119–122, 125–133, 138, 140, 143, 151–169, 172, 177–188, 196, 209–211, 219, 227, 228, 234–237, 239–262, 267, 268

 Asymmetric Events event asymmetry  151, 152, 154, 157–159, 163, 166–169, 227, 228, 241 event integration/asymmetry  227, 234 evidentiality  53, 54, 86, 242 expectation  121, 122, 127–134, 183, 189 F face preserving strategy  53, 69, 82 factive verbs  227, 238, 239 figure-ground  53, 61, 151, 154, 157 finite construction  227, 232 focal prominence  227, 236, 273 focus of attention  28, 48, 147, 236 foreground-background  151 frequency  7, 9, 10, 13, 53, 55, 71, 75–79, 86, 200, 201, 268, 269 G Genitive  53, 61, 248, 256 German  201, 205, 208–212, 218–224 grammatical categories  151, 166, 169, 262 grammaticalization  3, 5, 23, 53, 54, 86, 103, 173, 174, 190–193, 208 ground  25–31, 38, 40, 42, 45, 49, 50, 53, 61, 62, 67, 74, 143, 154, 166, 172, 236, 237, 261 grounding  29, 39, 43, 48, 120, 137, 143, 144, 148, 149, 172, 227, 228, 234–237, 242, 261 H Hua  157, 158, 171 I imagery  51, 227, 232, 233, 239, 241, 243, 261 impersonal constructions  227, 231, 241 indicative/subjunctive alternation  227, 237 inferences  53, 54, 55, 62, 71, 80, 81, 84, 214 infinitival clause  213, 220, 227–239 inflected infinitive  227–243

intentionality  3, 16, 18, 19 Italian  155, 156, 158 J Japanese  51, 121, 122, 130–134, 189, 195, 197, 201–205, 208 K Kaititj  3, 11, 12 Kalam  3, 11, 12, 23, 105, 120 ko construction  87 Korean  87–89, 90–95, 99, 100–102, 121, 122, 130–134, 201, 203, 205, 208 L Lango  164, 166, 172 locative adverbs  25, 31, 35, 49 locative particles  25, 26, 28, 29, 31, 32, 35, 46, 47, 50, 51 locative verbal prefixes  25, 26, 50 M Mandarin Chinese  165, 166, 171 markedness  53, 65, 73, 74, 75, 76 mental spaces  53, 59, 61, 66, 67, 71, 137, 140–143, 149, 237, 239 middle voice  261 modal verbs  227, 239 modification  135, 245, 251–255, 261, 262, 265, 271–280 N nominal grounding  227, 236 nominal clause  137 nominalization  135, 227, 239, 246, 247, 248, 249, 250–259, 266, 267 nominative marker  121, 122, 130–134 O object control  209, 212, 216–221, 223, 224 object of perception  26–28, 37–41, 46 objectification  46, 227, 236, 237, 239 objective construal  26, 37, 39, 60, 227, 228, 236, 241 obligatory control  209–214, 219, 223, 224 onstage conceptualizer  53, 69

P part/whole alternation  53 participle  261–280 partitioning  53, 54, 75 perception verbs  227, 239 periphrastic participants  245, 247–256, 258 person category  227, 235, 236 personal inflection  227, 235, 236, 239 phasal marker  173–178, 180, 182–190 point of reference  25, 26, 30–34, 36, 38, 39, 40, 41, 44, 49 Polish  224, 261–281 polyfunctionality  195, 196, 201, 203, 208 Portuguese  71, 227–231, 234, 235–243 Pragmatic Role Identity Principles  209 pragmatics  5, 53, 54, 71, 81, 113, 151–154, 160, 161, 167, 174, 209–223, 230, 241 predetermination  173, 182 Principle of Cognitive Economy  75, 195, 196, 204–206 processing  51, 66, 80, 82, 137, 142, 148, 149 profile  17, 25–28, 35–45, 51, 54, 59, 61–66, 80, 116, 117, 138–148, 159, 160, 173, 174, 234, 252, 254, 263–278 profiling  50, 137, 138, 139, 140, 143, 145, 147, 148, 246, 247, 261, 263, 280 pronominal reference  121, 122, 123, 125, 126, 133 R Radical Construction Grammar (RCG), 195 reference point  25–50, 124, 125, 134, 154, 245, 248, 249, 253–258 relational profile  53, 54, 59, 61, 65, 75, 79, 80, 84 relative clause  134–145, 153, 155, 156, 162, 171, 180, 195–207, 253, 269, 274 S saliency effects  53, 54, 62 schematicity  261



Subject index  search domain  25, 26, 29, 31, 32, 33, 34, 36, 37, 48 sequential scanning  14, 20–22, 87, 96–100, 127, 227, 234, 236, 261, 263–272, 276, 278 sequentiality  16, 22, 87 serial verb construction  11, 103, 119 serializability  103, 104, 111, 115, 117, 119 sociolinguistic variable  53, 54 spatial terms  25, 26 stage-management  53, 84 stressed pronoun  121, 122, 124, 129, 130, 134 subject  4, 15, 16, 18, 25, 26, 27, 28, 29, 31, 38–46, 50, 53, 54, 59–66, 71, 72, 75, 80, 81, 104, 110, 111, 124–128, 138, 146, 163, 166–188, 197, 209, 210, 213–245, 249, 252, 256–258, 264–265 subject clause  75, 227, 231, 241 subject conceptualizing constructions  53, 59 subject control  209, 215, 218, 219, 220, 221, 223, 235 subject of perception  26, 27, 28, 29, 38, 39, 46 subjectification  51–54, 86, 227, 237, 242

subjective construal  26, 28, 39, 43, 227, 237, 239 subordination  60, 87, 88–102, 135 -174, 180, 191–197, 207, 243, 262 summary scanning  39, 41, 43, 114, 118, 227, 234, 261, 263, 266–269, 280 symmetric-event description  121, 122, 126–131, 133 syntactic variation  53 tertium comparationis  261 T Thai  103, 104, 109, 111, 113, 118, 119 though-raising construction  227, 240 topic marker  91, 121, 122, 130–134 Toqabaqita  173–180, 182, 183, 190 trajector/landmark  137, 139 translation  53, 55, 79, 80, 199 transparency (isomorphism), 195, 196, 205 typology  23, 102, 151, 170, 171, 172, 193, 196, 207, 208, 224, 225, 261

U undesiderative marker  190 uninflected infinitival construction  227, 232, 234 unmarkedness  53, 65 V valence relation  103, 104, 115–119 vantage point  53, 54, 60–64, 72, 81, 143, 237 verb  3–11, 14, 16–23, 33, 38–50, 55, 57, 62, 65, 67, 73–77, 85, 89, 101–119, 147, 155, 157–159, 163, 165, 167–200, 208, 214, 216, 218, 228–234, 237–243, 246, 249, 251–256, 261–270, 273–280 verb coordination  3 verb serialization  22, 103, 119 viewing arrangement  25–30, 35–39, 43, 45, 48–50, 53, 54, 59, 60, 64–66, 69, 71, 73, 80, 83–85, 227, 228, 236, 239 volition verbs  227, 232, 238, 239 W Walbiri  162 word order  50, 90, 153, 200, 208, 227, 232, 241, 266, 269

In the series Converging Evidence in Language and Communication Research the following titles have been published thus far or are scheduled for publication: 12 Zlatev, Jordan, Timothy P. Racine, Christopher Sinha and Esa Itkonen (eds.): The Shared Mind. Perspectives on intersubjectivity. xiii, 378 pp. + index. Expected July 2008 11 Lewandowska-Tomaszczyk, Barbara (ed.): Asymmetric Events. 2008. vii, 287 pp. 10 Steen, Gerard J.: Finding Metaphor in Grammar and Usage. A methodological analysis of theory and research. 2007. xvi, 430 pp. 9 Lascaratou, Chryssoula: The Language of Pain. Expression or description? 2007. xii, 238 pp. 8 Plümacher, Martina and Peter Holz (eds.): Speaking of Colors and Odors. 2007. vi, 244 pp. 7 Sharifian, Farzad and Gary B. Palmer (eds.): Applied Cultural Linguistics. Implications for second language learning and intercultural communication. 2007. xiv, 170 pp. 6 Deignan, Alice: Metaphor and Corpus Linguistics. 2005. x, 236 pp. 5 Johansson, Sverker: Origins of Language. Constraints on hypotheses. 2005. xii, 346 pp. 4 Kertész, András: Cognitive Semantics and Scientific Knowledge. Case studies in the cognitive science of science. 2004. viii, 261 pp. 3 Louwerse, Max and Willie van Peer (eds.): Thematics. Interdisciplinary Studies. 2002. x, 448 pp. 2 Albertazzi, Liliana (ed.): Meaning and Cognition. A multidisciplinary approach. 2000. vi, 270 pp. 1 Horie, Kaoru (ed.): Complementation. Cognitive and functional perspectives. 2000. vi, 242 pp.