Linguistics of Vietnamese: An International Survey 9783110289411, 9783110289220

Citation preview

Daniel Hole and Elisabeth Löbel (Eds.) Linguistics of Vietnamese

Trends in Linguistics Studies and Monographs

Editor Volker Gast Founding Editor Werner Winter Editorial Board Walter Bisang Hans Henrich Hock Natalia Levshina Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume Walter Bisang

Volume 253

Linguistics of Vietnamese An International Survey

Edited by Daniel Hole and Elisabeth Löbel

ISBN 978-3-11-028922-0 e-ISBN 978-3-11-028941-1 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. 6 2013 Walter de Gruyter GmbH, Berlin/Boston Typesetting: RoyalStandard, Hong Kong Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Preface The present collection of chapters grew out of a workshop on “Linguistics of Vietnamese” at the University of Stuttgart in July 2009. The chapters that were accepted for inclusion in this volume have passed at least one internal and one external reviewing cycle, where “internal” refers to reviews written by authors who have themselves contributed to the volume. The external reviewing cycle fulfilled the standards of a double-blind peer review. The following external reviewers have agreed to be acknowledged for sharing their highly valued expertise with us: Walter Bisang, Katie Drager, Mary Erbaugh, Zhuo Jing-Schmidt, Yen-hui Audrey Li, Edgar Onea, Stavros Skopeteas, Thomas Stolz, Rolf Thieroff, Tue Trinh and Henk van Riemsdijk. In addition to these editorial measures, an anonymous reviewer for the whole volume appointed by the publishing house has instigated further changes. We would like to express our gratitude to the persons and institutions which have had a share in rendering the 2009 workshop and this publication possible. Klaus von Heusinger has certainly been our most encouraging background supporter – thank you, Klaus! Moreover, we gratefully acknowledge Julia Jürgens’ help with the formatting of the contributions, as well as Waltraud Ott’s accounting assistance. Finally, we would like to thank the Fritz Thyssen Foundation for the Promotion of Science, which has supported the workshop with substantial funding. Without it, this volume could no have been realized, either. The work of the editors is almost done. What remains to be done is to say that it has been a great pleasure for us to prepare this volume. We would like to thank the series editor Walter Bisang for his helpful supervision of the content aspects of the editing process. All authors, and no less Birgit Sievert, Julie Miess and Wolfgang Konwitschny from the publishing house, have had their share in establishing the highly collaborative spirit in which the enterprise “Linguistics of Vietnamese – an international survey” was carried out. Berlin/Cologne, April 2013 Daniel Hole Elisabeth Löbel

Table of contents Preface Introduction

v 1

I

Phonology/Phonetics

1

Marc Brunelle and Stefanie Jannedy The cross-dialectal perception of Vietnamese tones: Indexicality and convergence 9

2

Hạ Kiều Phương Prosodic means in repair initiation as an activity in Northern Vietnamese conversation 35

II

Noun phrase syntax

3

Nguyễn Hùng Tưởng The Vietnamese noun phrase

57

4

Jennie Tran Vietnamese classifier phrases from the perspective of how children acquire them 87

III

Clausal and verb phrase syntax

5

Nigel Duffield Head-First: On the head-initiality of Vietnamese clauses

6

Andrew Simpson and Hồ Hảo Tâm Vietnamese and the typology of passive constructions

7

127

155

Theresa Hanske Serial verbs and caused change of location constructions in Vietnamese 185

viii

Table of contents

IV

Pronouns and minor word classes

8

Tran Thuan and Benjamin Bruening Wh-phrases as indefinites: A Vietnamese perspective

9

Marie-Claude Paris and Lê Thị Xuyến On conjunction and comitativity in Vietnamese

Daniel Hole 10 Focus particles and related entities in Vietnamese Index

305

217

243

265

Introduction Linguistic lore has it that Vietnamese (Austro-Asiatic, Việt-Mường) constitutes the paradigm case of an isolating language with lexical tones. Moreover, Vietnamese with its roughly 80 million native speakers worldwide is one of the major languages of East and South East Asia. Nonetheless, detailed knowledge about Vietnamese is somewhat scarce among linguists, and those researchers or students who wish to enhance their knowledge about this language will often find that reliable information is scattered across the most varied publication sources, or is hard to come by for other reasons. To give just one example: One of the most detailed grammars of Vietnamese available in a western language (Trương Văn Chình 1970) was written in French and has never been translated into any other language; cf. Thompson 1965 for a comprehensive grammar in English). For this reason, it is generally ignored by the majority of linguists without literacy in French. Given these circumstances, the editors have taken special care to cater to the needs of a readership which should be as broad as possible. Each chapter in this volume is self-contained in the sense that no other chapter, let alone a grammar of Vietnamese, is necessary to make sense of it. Another guiding principle for the editing process has been to take an inclusive stance as far as the commitment to different frameworks and research methodologies is concerned. This decision implies that no attempt was made to level out differences in terminology. The following survey of the contents of this volume is organized in the following way. We first single out four chapters and discuss them in some detail before the other contributions are briefly summarized. We have chosen this mode of presentation because some of the chapters are not just valuable additions to our knowledge of Vietnamese grammar, they may simultaneously serve to illustrate in a good way the wide array of methodological stances taken by individual contributions to this volume. Brunelle and Jannedy’s study is a contribution to the cutting-edge area of experimental sociophonetics. It demonstrates how the observer’s paradox can be rendered fruitful in the domain of Vietnamese tones. The authors show among other things that dialectal traits in the speech of the experimenter influence hearers’ perception of stimuli. The methodological position taken in the paper challenges simplistic views of the separation between competence and performance. It emphasizes the life-long (and frequency-dependent) plasticity of speakers’ grammars instead. This position is put to work with the help of meticulous experimental design and refined statistical analysis. If we were to attempt a characterization of Brunelle and Jannedy’s position in a slogan, “a

2

Introduction

revision of the competence-performance divide with experimental back-up” would be the result. The chapter by Nguyễn defends a different, highly reductionist and universalist position. In accordance with generative assumptions, Nguyễn develops an account of Vietnamese noun phrase, or DP, syntax which derives all possible word orders of noun phrases in all languages from a single universal cascade of functional categories inside the DP. The linearization found with Vietnamese noun phrases is analyzed as one pattern predicted by this universal cascade. The surface linearizations of Vietnamese noun phrases result from the application of movements which are constrained by general principles. It is worth pointing out that the reductionist methodology of Nguyễn’s contribution allows one to clearly state the conditions which would falsify the proposal – a highly desirable feature of the analysis. Given the cascade of categories and the movements assumed by the author, certain word orders are predicted not to exist in the languages of the world. If they turn out to be attested in the end, one will be able to discard the proposal. Not all linguistic frameworks have falsifiability conditions which are as clear – at least in theory. All in all, Nguyễn may be said to subscribe to “syntax-driven mainstream generativism”. The authors Paris and Lê chose a classic topic of research into isolating languages, viz. the syntactic and semantic construal of conjunction and comitativity. Their chapter illustrates the usefulness of a modernized structuralist approach with a strong taxonomic underpinning. By relying mostly on clear distributional diagnostics, Paris and Lê carve out the detailed generalizations in this area of Vietnamese grammar for the first time. Và is established as a true ‘and’-type conjunction, whereas với is polysemous between a conjunctive use and a use in which it heads a comitative adjunct. Throughout the chapter, the situation found in Vietnamese is compared with that in Mandarin Chinese and French. “Advanced taxonomic structuralism informed by typology and semantics” could maybe describe the methodological position of the authors. Hạ’s contribution, by contrast, illustrates a clearly functionalist and conversation-analytic standpoint. The author reports findings about the prosody of repair initiations in Vietnamese telephone calls. These findings are paired with an autosegmental analysis which accounts for different interaction patterns of lexical tones with repair-initializing high boundary tones. By tying the presence of high boundary tones to the function of the conversation-analytic category of repair initiation (as opposed to, say, the signaling of discourse incompleteness), Hạ subscribes to a research paradigm which prefers intuitively grounded basic notions over highly abstract principles for which language users lack intuitions: “autosegmentally-informed conversation analysis”, in a slogan.

Introduction

Phonology/ Phonetics

Noun phrase syntax

Clausal and verb phrase syntax

Pronouns and minor word classes

3

Author(s)

Title

Further specification of content

Marc Brunelle/ Stefanie Jannedy

The Cross-dialectal Perception of Vietnamese Tones: Indexicality and Convergence

Sociophonetics, Tonal Phonology

Hạ Kiều Phương

Prosodic means in repair initiation as an activity in Northern Vietnamese conversation

Conversation Analysis, Autosegmental Phonology

Nguyễn Hùng Tưởng

The Vietnamese noun phrase

(Areal) Typology, Noun Phrase Structure, Classifiers

Jennie Tran

Acquisition, Noun Phrase Vietnamese classifier phrases from the perspective Structure, Classifiers of how children acquire them

Nigel Duffield

Head-First: On the headinitiality of Vietnamese clauses

(Areal) Typology, Functional Clause Structure

Andrew Simpson/ Hồ Hảo Tâm

Vietnamese and the typology of passive constructions

(Areal) Typology, Voice, Causatives

Theresa Hanske

Serial verbs and caused change of location constructions in Vietnamese

Aspectuality, Serial Verb Constructions

Tran Thuan/ Benjamin Bruening

Wh-phrases as indefinites: a Vietnamese perspective

Alternative Semantics, Indefinites

Marie-Claude Paris/ Lê Thị Xuyến

On conjunction and comitativity in Vietnamese

(Areal) Typology, Conjunction, Comitativity

Daniel Hole

Focus particles and related entities in Vietnamese

(Areal) Typology, Information Structure

Table 1: Contributions to this volume

If one assesses the qualitative vs. quantitative stance in methodology of the four chapters just mentioned, then qualitative approaches certainly prevail (qualitative: Nguyễn, Paris/Lê, Hạ; quantitative: Brunelle/Jannedy). This bias is representative of the whole volume. If formalist vs. functionalist explanatory patterns are taken as a dimension of classification, formalist thinking is found to characterize the majority of chapters (formalist: Nguyễn, Paris/Lê; functionalist: Hạ, Brunelle/Jannedy). Again, this is representative of the whole collection of chapters.

4

Introduction

The classifier structures which are characteristic of many Vietnamese nominals are the primary object of study in Tran’s acquisition study. It is the first such study on Vietnamese, and the author presents findings gained from both longitudinal and cross-sectional child data. The first structures produced by children are classifier+noun, classifier+demonstrative and classifier+wh-word beginning from around the age of 2. Moreover, the data reveals more omission errors than found in acquisition studies on other classifier languages. Duffield’s contribution focuses on clausal syntax. It makes a strong case for strictly left-headed structures in Vietnamese even in those domains that would seem to involve right-headed projections. Clause-final complementizers are Duffield’s main concern. In the further course of the chapter, modals and demonstratives enter the picture, and parallels to Mandarin Chinese are drawn. Simpson and Hồ direct their attention to passives and related constructions in Vietnamese, mainly in comparison with Mandarin Chinese. Apart from analyzing the neatly elaborated system of beneficiary and adversative passives found in Vietnamese, they deal with the important issue of whether the passive can be recognized as a clear-cut category in Vietnamese (and beyond) in the light of control structures and causative structures which make use of the same coding devices as translational equivalents of passives. In Hanske’s contribution, a detailed analysis of Vietnamese change-oflocation constructions is developed which incorporates the insight that aspectual (or aktionsart) differences of the lexical items involved play an important role in the architecture of the resulting serial verb constructions. Most importantly, dynamic second verbs yield action readings while stative second verbs yield a perfect-of-result reading. The chapter by Tran and Bruening on Vietnamese wh‑indefinites, i.e. whwords with (mostly indefinite) uses in non-question sentences, approaches its empirical domain from a (formal) syntax-and-semantics perspective. Non-veridical operators are identified as licensing non-interrogative uses of wh‑words, and existential closure or choice functions lead to the wh-readings which may be characterized as indefinite. Hole, finally, provides a first survey of Vietnamese lexical items and syntactic patterns which are used to convey focus-semantic meanings like ‘also’, ‘even’ or ‘only’. The extraordinarily rich system found in Vietnamese outranks the partly similar system of Mandarin Chinese in complexity and calls for further research in this domain. A strong undercurrent of the majority of chapters is the comparison of the Vietnamese facts with analogous phenomena in Mandarin Chinese, at least among other languages (Duffield, Nguyễn, Tran, Simpson/Hồ, Hanske, Tran/ Bruening, Paris/Lê, Hole). It may seem objectionable to some readers to promote a more visible ‘Linguistics of Vietnamese’ with such a strong bias towards

Introduction

5

comparison with Chinese. However, we would like to emphasize that, in contradistinction to Vietnamese, Chinese is a very well-studied language, and has been in close contact with Vietnamese for millennia. Moreover it is structurally similar to Vietnamese in some respect (classifiers, verb serialization, focus particles, to name just a few areas). Thus we submit that it is not a drawback if many chapters of this volume study Vietnamese against the background of the rich knowledge that we have about Mandarin Chinese.

References Thompson, Laurence C. 1987. A Vietnamese Reference Grammar. Honolulu: University of Hawai‘i Press. Trương Văn Chình. 1970. Structure de la langue vietnamienne. Paris: Geuthner.

I Phonology/Phonetics

Marc Brunelle and Stefanie Jannedy

1 The cross-dialectal perception of Vietnamese tones: Indexicality and convergence* Previous research has shown that speech production and speech perception can be influenced by social factors such as inferred attributes of speakers (Strand and Johnson 1996; Niedzielski 1999; Hay and Drager 2010) and experimenters (Hay, Drager and Warren 2009; Hay, Warren and Drager 2010). This paper reports an experiment testing if the rate of misidentification of tones in Vietnamese is influenced a) by the dialect of stimuli, b) the dialect of the experimenter to which the hearer is exposed and c) the amount of exposure to the speech of an experimenter. Our results indicate that listeners do adjust their interpretation of some tonal stimuli in accordance to the dialect of the person administering the experiment, strongly suggesting that perceptual cues contained in the signal, inferred social factors and amount of exposure all play a role in the categorization of tones in Vietnamese. Keywords: Vietnamese, cross-dialectal tonal perception, perceptual divergence, experimenter effect

1 Introduction Usage-based or experience-based models of language (for an overview, see Barlow and Kemmer 2000) operate under the assumption that there is a very close relationship between linguistic structures and instances of use. In these models, linguistic competence is not cognitively separated from performance: they are both continually shaped over time through usage events (Bod, Hay and Jannedy 2003). The role of experience and exposure is essential in these models as the frequency of occurrence of a linguistic event is both a result and * We wish to thank Hạ Kiều Phương and Nguyễn Thị Minh Tâm for conducting the experiment in Hồ Chí Minh City, Phạm Thị Lê Uyên for recruiting subjects, Joe Roy for help with the statistical analysis, the editors and anonymous reviewers of this paper, and Bernd Pompino-Marschall (HU Berlin) for supporting this work. The authors are listed in alphabetical order. A 4-page preliminary version of this paper of this paper was published in the Proceedings of the 15th International Congress of Phonetic Sciences. Stefanie Jannedy acknowledges the support by BMBF grant 010G0711.

10

Marc Brunelle and Stefanie Jannedy

a shaping force of the system. Learning is construed as taking place over the entire life-span of the speaker and their linguistic system is updated continuously through new experiences. Speakers thus display changes in their production and perception over time. In short, the traditional sharp distinction between competence and performance is weakened if not dissolved by the fact that comprehension and production are seen as integral to the linguistic system: performance becomes a part of a speaker’s competence. The absence of a distinction between performance and competence can be captured by models which provide a way of accounting for the production, representation and perception of variation. In exemplar accounts, for instance, the acoustically rich signal is stored in the lexicon together with linguistic, contextual and social information (Johnson 1997; Pierrehumbert 2002). These signals have a lexical value, which means they have semantic content. However, these signals also have an indexical value which encodes the social characteristics of the talker and the speech situation, that is, where, how and through whom this signal was encountered before. The storage of these acoustically rich signals results in exemplar ‘clouds’ (phonetic density distributions), which are active in speech perception and speech production. Speech production then involves averaging over relevant parts of such distributions while speech perception involves matching an acoustic input with the exemplar clouds it most resembles. Such models can account for the apparent sensitivity of listeners to social information such as age (Drager 2011), social class (Hay et al., 2006), gender (Strand and Johnson 1996; Strand 1999) and dialect (Niedzielski 1999; Hay, Drager and Warren 2009; Hay and Drager 2010). The larger question is how speakers and hearers’ productions and categorizations are altered through immediate experience and interactions and what are the implications of these effects for phonetic variation and phonological representations. Through accommodation and convergence (Goldinger 1998; Pickering and Garrod 2004), interacting talkers become more similar-sounding and listeners adapt to speakers perceptually as part of a subconscious socio-cognitive learning process. While such convergent behavior seems to be a fundamental aspect of human behavior in general. (i.e. interlocutors accommodate each other as part of a perceptual-behavioral link), there is also evidence for divergence effects whereby speakers maximize the linguistic contrast to socially dissociate themselves from their interlocutors. While some researchers maintain the view that imitation and convergence of interlocutors is mechanistic and an automatic consequence of cognitive reflexes (Goldinger 1998; Pickering and Garrod 2004) just as yawning is (Bargh and Williams 2006), others maintain the view that convergence in verbal behavior also involves choices, even if these are subconsciously made, and that these choices are mediated by social factors (Giles

The cross-dialectal perception of Vietnamese tones

11

and Coupland 1991; Pardo 2006; Babel 2009). Currently, it is not fully understood which aspects of the acoustic signal lend themselves to convergence: while Gregory et al. (2001) found no convergence in f0 when the signal was high-pass filtered, both Goldinger (1997) and Gregory et al. (2001) showed experimentally that speakers in conversational tasks are converging (imitating each other) to some extend in terms of f0 when the signal was unaltered or low-pass filtered (see Babel 2009 for a more detailed review of that literature).

1.1 Evidence from production Throughout adulthood, the phonetic detail of speech production continues to be subtly influenced by the speech of those around us and causes change over time (for example Harrington 2006). It has been found that speakers can, and do, shift their pronunciations towards their addressees, overhearers and even referees (Bell 1984; Rickford and McNair-Knox 1994; Hay, Jannedy and Mendoza-Denton 2010). Harrington (2006) for example showed that over time, the lax vowel /ɩ/ as in happy in the speech of Elizabeth II (in her annual Christmas addresses) displayed tensing of this vowel: it became less RP-like and more like the speech of the general population. Hay et al. (2010) found that the US talk-show host Oprah Winfrey shifts her pronunciations towards the referee, the person she is talking about. Furthermore, it has also been found that shifting towards the model of the addressee or away from the model of the addressee is dependent on the speaker’s estimation of who he believes the addressee is. Hay, Warren and Drager (2010) report on two experiments specifically designed to test the influence of prior exposure to different dialects on the responses in a lexical access task in New Zealand English. They find that there is an effect of prior exposure, probably due to accommodation and adaptation. The authors suggest that an explanation for such a finding is found along the lines of exemplar models of speech production and speech perception. The authors argue that exposure to an experimenter with a different dialect than the subjects in either perception or production tasks can result in a higher activation level of distinct forms of the same word in production and perception: it simultaneously helps speakers to generate greater contrasts in minimal pairs and hampers their identification of such word pairs. (It may be worth noting, that other researchers have different explanations for such evidence: Pardo (2006) attributes phonetic convergence to entrainment while Namy et al. (2002) assume a perceptual reflex underlying phonetic convergence.) Hay et al. (2009) also found an effect of experimenter identity on the production of minimal pairs currently undergoing a merger in New Zealand English.

12

Marc Brunelle and Stefanie Jannedy

They conducted several minimal-pair reading tasks of words containing the diphthongs /iǝ/ and /eǝ/ as in beer and bare involved in a merger in progress in New Zealand English. While the distinction is maintained in production in the different dialects, perceptually, it is merged for most listeners. They interpret their results in two ways. First, participants that received their instructions from an US experimenter were more likely to retain a contrast in production between these two vowels than participants who met with an experimenter from New Zealand. Second the participants who interacted with the US experimenter and did not produce a contrast in their own speech nonetheless self-reported that they preserved that contrast. The authors suggest that this is evidence for accommodation to the speech of the US experimenter, in whose speech the distinction between the diphthongs is retained. However, Trudgill (1981) showed that he as the interviewer and native of Norwich (UK) accommodated to the speech of an interviewee from Norwich with regard to social salient markers of identity. Thus, there is evidence that accommodation occurs in both directions.

1.2 Evidence from perception An effect of perceptual learning and an interpretation of the acoustic stimuli in light of the hearers’ social categories has also been found in perception. Strand and Johnson (1996) and Strand (1999) found a gender effect in the categorization of re-synthesized acoustic stimuli: women generally produce non-periodic noise for fricatives like /s/ and /ʃ/ at higher frequencies than men. To evaluate listeners’ sensitivity to this fine phonetic detail, Strand and Johnson (1996) tested listeners on prototypically male and prototypically female, as well as non-prototypically male and non-prototypically female sounding voices. They combined each of these voices with a synthesized 9-step fricative continuum from /s/ to /ʃ/ and played the resulting composite stimuli to listeners. Their results show that the prototypically male and female sounding voices were rated differently, with female voices pushing up the perceptual boundary towards higher frequency /s/, while the male voices shifted down these boundaries towards /ʃ/. When matching non-prototypical voices with pictures of male and female faces, listeners were more likely to perceive /ʃ/ with the co-presentation of a male than a female face (and vice versa). This Face-Gender-Effect is a strong indication that audio-visual integration and the expectations that we have about the speaker strongly influence the perception and interpretation of identical acoustic signals. Niedzielski (1999) showed that listeners’ responses to a perceptual vowel matching task were biased by the supposed origin of the speaker: speakers

The cross-dialectal perception of Vietnamese tones

13

from Detroit, Michigan, produce Canadian Raising, or raised variants of the diphthong /aʊ/, just like their Canadian neighbors. However, speakers from Detroit do not seem to be aware of this raising in their own speech and associate it with Canadian speakers. In her study, Niedzielski asked Detroit listeners to match natural speech tokens in words containing /aʊ/ recorded from a Detroit speaker to tokens synthesized along a vowel continuum. There were two experimental conditions: in one condition, the word ‘Canadian’ was written on top of the answer sheet, while in the second condition, the word ‘Michigan’ appeared on it. Her results show a perceptual divergence effect whereby the Detroit listeners perceived less Canadian raising when they inferred from the answer sheet that the speaker was from Detroit like themselves. She concludes that the expectations that listeners have about speakers – in this case the speakers’ supposed dialect, but expectation could be based on other markers – can influence the categorization of acoustic stimuli even though the two dialects produce rather similar variants. Hay et al. (2006) conducted a follow-up to the Niedzielski study, using the same kind of methodology and exploiting differences in the pronunciation of the front vowel /ɩ/ between Australian and New Zealand English. The /ɩ/ is more raised in Australian English and more centralized in New Zealand English. In one experimental condition, the word ‘Australian’ appeared on the answer sheet and in the second condition, the word ‘New Zealander’ was written on it. While nearly all the New Zealand listeners reported that they knew that they were listening to a speaker from New Zealand, they nevertheless selected more fronted versions of the synthesized vowel to be matched with the recording of the New Zealand speaker when ‘Australian’ was written on their response sheet. It seems that listeners display a perceptual divergence effect even when they are aware of the geographic identity of the speaker. Hay et al. (2009) investigated the impact of social stereotyping on the identification of the two vowels /iǝ/ (as in near) and /eǝ/ (as in square) that are currently undergoing a merger in New Zealand English. Production data showed that the distinction between the diphthongs /iǝ/ and /eǝ/ in this variety of English is more likely to be retained when the speaker is older and not from a lower social class background. In the experiment, the perceived age and the perceived social class of the speaker was manipulated by showing listeners photographs of people of different ages and pictures of identical individuals wearing clothes typical for white- and blue-collar workers in New Zealand. It was found that listeners who self-reportedly maintain a distinction between these two vowels in their own speech had a higher identification accuracy in natural speech tokens when a picture of a younger rather than older person was co-represented. Results further indicate that the vowel merger between

14

Marc Brunelle and Stefanie Jannedy

near /iǝ/ and square /eǝ/, which is more advanced in the working class, was perceived more systematically when a picture of the supposed speaker in working class attire was shown in contrast to the same person wearing formal clothes. They conclude that the social characteristics that are being attributed to speakers have an influence on the identification of sounds undergoing change. In a recent experiment, Hay and Drager (2010) were able to show that priming of listeners with objects denoting national identity also has an effect on categorization. They exposed speakers of New Zealand English to experimental primes as subtle as stuffed toys located in the experimental room (kangaroos and koalas for Australia and kiwis for New Zealand). Just as in the previous study, speakers of New Zealand English were asked to match naturally produced vowel tokens from a speaker of New Zealand English to vowels from a synthesized vowel continuum ranging from a raised and fronted /i/ to a more New Zealand-like centralized version of this vowel. Their results indicate that the categorization shifts as a function of which stuffed toy category was seen by the listener. All these experiments support the view that speech perception is influenced by inferences made from non-linguistic observations. Evidence is mounting that perception cannot be separated from interpretation and subconscious classification. Thus, the relationship between speech perception and speech production and the degree to which hearers generalize over and weigh heard forms based on actual previous encounters or simple stereotypes all have an impact on the interpretation and categorization of acoustic signals. Controlled perception studies show that social information such as the speakers’ supposed regional dialect (and therefore inferred cultural background), gender, socio-economic status or nationality biases the perception and rating of naïve listeners (Strand and Johnson 1996; Niedzielski 1999; Strand 1999; Brunelle and Jannedy 2007; Hay, Drager and Warren 2009; Hay and Drager 2010; Hay, Warren and Drager 2010). These factors ought to be taken into account when explaining differential patterns in the categorization of identical stimuli. Perceptual divergence effects demonstrate a great sensitivity of listeners to social stereotypes about inferred social or linguistic groups. However, all studies (Niedzielski 1999; Hay, Drager and Warren 2009; Hay, Walker and Drager 2010; Hay, Warren and Drager 2010) these effects have been conducted on varieties of English or with carefully constructed acoustic continua. We are not aware of work showing convergence or divergence effects in the perception of lexical tones in tone languages, a phenomenon that would be expected in light of the fact that f0 imitation and convergence has been found to be a form of social status accommodation in English (Goldinger 1997; Gregory, Green, Carrothers and Dagan 2001). In this paper, we are therefore testing for divergence or convergence effects in Vietnamese, a language that, just like English, has dialectal

The cross-dialectal perception of Vietnamese tones

15

diversity associated with a strong sense of regional identity: North-South political division is a recurrent theme of Vietnamese history and the current rivalry between Hà Nội, the political center, and Hồ Chí Minh City, the industrial and economic metropolis, coincides with significant dialectal differences. However, rather than using synthesized acoustic continua, we have used a minimal set of natural speech tokens recorded from actual speakers because Vietnamese tonal contours are too multi-dimensional (pitch height, contour and voice quality) for step-wise modifications. As we will see, the internal patterns of tonal contrasts of Vietnamese dialects can be affected by indexical factors introduced in the experimental set-up, just like in the studies discussed above.

1.3 Vietnamese tones Vietnamese dialects have complex tone systems combining pitch, contour and voice quality distinctions. Dialectal differences in tone inventory are important. Dialects range from four to six tones in open syllables and syllables closed by sonorants. Checked syllables, i.e. syllables closed by voiceless stops, bear one of two additional tones, which are sometimes considered allotones of tones B1 and B2 (they will not be addressed here). The pitch height and contour of tones are sometimes totally different in distinct dialects. Moreover, while glottalization and breathiness are common in Northern and Central varieties, they are mostly lost in Southern ones. In this paper, we will focus on the two regional dialects that have the most speakers: Northern Vietnamese (NVN) and Southern Vietnamese (SVN). Northern Vietnamese (NVN) has 6 tones that combine complex pitch contours with voice quality distinctions (Han 1969; Earle 1975; Vũ 1981; 1982; Seitz 1986; Nguyễn and Edmondson 1997; Michaud 2004; Michaud and Vũ 2004; Vũ, d’Alessandro and Michaud 2005; Michaud, Vũ, Amelot and Roubleau 2006; Brunelle 2009b; Kirby 2010). It is the standard variety of the language and benefits from a widespread diffusion in the national media. It is prestigious even outside the North and a certain admixture of NVN features by speakers of other varieties is common in formal contexts. Nonetheless, NVN is not perfectly homogeneous. For instance, standard NVN reflects a somewhat conservative variety of Hà Nội Vietnamese, while colloquial NVN dialects, especially the varieties spoken in Hà Nội, exhibit a large amount of variation (in age, gender and social class) that affects some of their tones. In this paper, our NVN speaker and experimenter both use an urban middle-class variety in which two of the tones diverge from standard NVN and

16

Marc Brunelle and Stefanie Jannedy

to which speakers of other dialects are not frequently exposed. To avoid confusion between the two types of NVN speech, we will refer to them as standard NVN and colloquial NVN, respectively. The second regional dialect discussed in this paper is Southern Vietnamese (SVN). SVN exhibits less tonal variation than NVN. It is also less standardized than NVN, but the variety spoken in Hồ Chí Minh City, the largest urban agglomeration in the country, has unofficially become a regional standard that is now used on radio and television alongside with standard NVN. It has 5 tones that rely exclusively on pitch height and contours (Gsell 1980; Vũ 1981; 1982; Brunelle 2009b; Kirby 2010). Figures 1 and 2 illustrate the mean f0 values of the five utterances of each tone used as stimuli for this experiment as produced by a female (colloquial NVN) and by a male (SVN) speaker on the syllable /ma/ (f0 was measured on the vowel /a/).

Figure 1: The tone system of the northern speaker (NVN)

The tones of our NVN speaker are representative of the Hà Nội middle-class dialect. The three NVN tones that have a laryngealized voice quality are represented with dotted lines in Figure 1. Laryngealization is realized as a strong final glottalization in B2 and as a salient medial creak in tone C2. It is produced as word-final breathiness, harshness or weak glottalization in tone C1 (Michaud

The cross-dialectal perception of Vietnamese tones

17

Figure 2: The tone system of the southern speaker (SVN)

2004). There are two significant departures between the tones of our speaker and standard NVN. First, colloquial tone B1 is low-rising rather than high-rising. This innovative variant of tone B1 started out among young female Northerners, but is now spreading to other groups as well. Interestingly, native speakers do not seem aware of this change in progress. The other important difference is that tone C1, which is a falling-rising tone in standard NVN, has lost its final rise in colloquial NVN. The Southern Vietnamese subjects who took part in the experiment described below are unlikely to be aware of these two colloquial variants. The tone system of our SVN speaker is representative of the southern regional standard. The most important differences between SVN and NVN are that laryngealization has been lost altogether and that tones C1 and C2 have long been fully merged into a single tone in SVN, leaving it with five tones instead of six as in NVN. SVN tones B2 and C1-C2 typically end on a slight rise, but this final rise is not visible in Figure 2 because it was averaged out by two non-rising tokens of tone B2 and one non-rising token of tone C1-C2 respectively. Note that because of the very low pitch range of our SVN speaker, his tones A2, B2 and C1-C2 sometimes break into a non-contrastive creak. Not surprisingly given these acoustic differences, the exact acoustic cues used in tone perception are not identical in NVN and SVN (Vũ 1981; Michaud

18

Marc Brunelle and Stefanie Jannedy

2004; Brunelle 2009b; Kirby 2010). Glottalization and direction of contour are the dominant cues in NVN, while pitch height and contour complexity are the main cues in SVN. Because of these differences, listeners tend to misidentify tones from other dialects more often than tones from their own dialect. More interestingly, it seems that despite dramatic differences in the acoustic realization of the tones across dialects, listeners are able to adjust their perception to other dialects to a certain degree. This is either due to exposure to other varieties through the media or internal migration, or to an ability to quickly adapt and accommodate unfamiliar accents. The aim of the present study is to evaluate the effect of the following factors on tone identification: 1. Effect of dialect: We expect that listeners will perform better at identifying tones from their own dialect than from a less familiar variety. 2. Effect of the experimenter: We hypothesize that the dialect of the experimenter present during the procedure will cause a perception shift. Following previous studies (Niedzielski 1999; Hay, Drager and Warren 2009; Hay and Drager 2010; Hay, Warren and Drager 2010), we expect listeners to be biased towards the speech of the experimenter. 3. Learning effect (adaptation): If listeners can rapidly adjust their perception to the tones of an unfamiliar variety, we expect that the amount of time spent with an experimenter speaking this unfamiliar variety should improve performance on the identification of its tones. In order to test these effects, we conducted an experiment in which we asked SVN listeners to identify a minimal set produced by a colloquial NVN speaker and a SVN speaker. The words are all composed of the segments /ma/, but distinguished by their tones. A NVN experimenter conducted the experiment with half of the listeners, while a SVN experimenter conducted the experiment with the other half. With each experimenter, half of the subjects first listened to the NVN tones and then to the SVN tones, while the other half followed the opposite order. The experimental design is described in more detail in section 2. An important difference between current work and previous studies is that rather than using subtle phonetic continua ranging between two variants as stimuli (Strand and Johnson 1996; Niedzielski 1999; Hay, Nolan and Drager 2006; Hay, Drager and Warren 2009), we use naturally produced tones, assuming that the effects on perception will be strong enough to override some phonemic distinctions and cause misidentifications. Since some tones are not markedly different in NVN and SVN, we are not expecting the same amount of perceptual confusability with all the tones. The phonetic properties of the tones of the two varieties under study allow us to make predictions about expected errors. Previous studies (Brunelle 2009b; Kirby 2010) reveal that SVN listeners have no difficulty identifying standard NVN tones

The cross-dialectal perception of Vietnamese tones

19

A1, A2, B1 and B2. The problems arise with tones C1 and C2, which are merged into a tone bearing little phonetic resemblance with either of them in SVN. In Brunelle (2009b), SVN speakers mostly identified standard NVN tone C1 as B2 (52 times out of 60 times), which seems to be attributable to its fairly robust laryngealization, which they interpret as the strong final glottalization NVN B2. Standard NVN C2, on the other hand, was often identified as B1 (16 times out of 60), which seems to be due to a failure to perceive the glottalization on this rising tone, and as C1 (5 times out of 60), the tone with which it is merged in SVN. In the same vein, Kirby (2010) finds that the pairs of colloquial NVN tones B2-C1 and C1-C2 are difficult to discriminate for SVN listeners. However, the pair B1-C2 does not seem as problematic in discrimination, suggesting that SVN listeners are sensitive to glottalization, but have a hard time using it for lexical retrieval. An inspection of the tonal system of our NVN speaker suggests that our listeners are likely to meet the same difficulties (see Fig. 1). His tones A1, A2, B1 are similar to their SVN counterparts and should not pose any problem to SVN listeners. Tone B2, on the other hand, should be easy to identify for a different reason: it has a characteristic strong final glottalization that SVN listeners are able to use as a reliable identification cue due to their exposure to standard NVN (Brunelle 2009b). On the other hand, colloquial NVN tone C1, which is high, slightly falling and laryngealized, should be unfamiliar to SVN listeners as it greatly differs from its standard NVN counterpart. If SVN listeners perceive its laryngealization, it should be confused with B2. On the other hand, if they fail to perceive it, it should be misidentified as tones A1 (high) or A2 (falling), which have resembling contours or height. Similarly, tone C2 is likely to be misidentified as B1 if SVN listeners fail to perceive its laryngealization, as in Brunelle (2009b). We also expect that SVN listeners will make some identification errors in their native dialect. This is due to the very dense lower pitch range of our SVN speaker, added to the fact that some tokens of tones B2 and C1-C2 used in the experiment do not have the typical final rise (2/5 and 1/5, respectively). Tones A2, B2 and C1-C2, which are all low and falling, should therefore be occasionally confused with one another. We will now describe our methodology and experimental results and will come back to our research questions and to the predictions we just made in the discussion.

2 Methods Forty native listeners of SVN were asked to identify words with natural tones produced by native speakers of NVN and SVN (one speaker of each dialect). A

20

Marc Brunelle and Stefanie Jannedy

NVN experimenter administered the test to half of the subjects while a SVN experimenter administered the test to the other half.

2.1 Stimuli Recordings of the Vietnamese stimuli were made with two speakers of NVN and SVN. The NVN speaker is a 25 year old woman who had been living in Hà Nội for eight years at the time of recording, while the SVN speaker is a 23 year old man from Tiền Giang province who had been living in Hồ Chí Minh City for 4 years. Their speech is representative of colloquial NVN and SVN, respectively. Ten recordings of each tone were made with a Marantz PMD-680 recorder and a AKG-C5900 microphone in Hà Nội and Hồ Chí Minh City. The speakers were asked to read the syllable /ma/ with the six tones (five for SVN) in a frame sentence where the target word was preceded and followed by words with the level tone A1 so as to control for tonal coarticulation effects (Han and Kim 1974; Brunelle 2003; 2009a). (1) Tôi chào sư __ xem ông ấy có nhớ tôi không. [toj caw sɨ __ xɛm oŋ ɛj kɔ ɲɤ toj xoŋ] “I greet monk __ to see if he remembers me” ma: spirit, ghost mà: but má: cheek (also “mother” in SVN) mả: tomb mã: password, code (also “horse” in Sino-Vietnamese, or “appearance”) mạ: rice seedling Five utterances of each tone were selected. The target words /ma/ in these sentences (5  6 tones = 30 for NVN; 5  5 tones = 25 for SVN) were extracted to serve as stimuli for the identification task. Although the six syllables used as stimuli correspond to real words that have different lexical frequencies, frequency effects were not controlled for due to lack of a reliable frequency database.

2.2 Subjects The experiment was administered to 20 male and 20 female native speakers of SVN in Hồ Chí Minh City. They were all born and raised in southern Vietnam (from Quảng Nam province to the southern tip) and none of them had ever spent

The cross-dialectal perception of Vietnamese tones

21

more than one year in northern Vietnam at the time of the experiment. All these subjects have regular exposure to standard NVN through radio, TV and music. Some of them might also be familiar with some varieties of colloquial NVN because of significant NVN migration to Hồ Chí Minh City.

2.3 Procedure An identification task was administered in Praat 4.4.16 on a laptop computer with Sennheiser headphones. Each of the 30 natural stimuli recorded from the NVN speaker were played five times in isolation and in randomized order, for a total of 150 tokens. The subjects had to click on one of six boxes containing the six possible answers (in Vietnamese spelling) to identify the word they had just heard. The next stimulus was then played. This same procedure was conducted with five repetitions of the 25 SVN stimuli, for a total of 125 tokens. Two linguistically-aware female experimenters conducted the experiment independently with 20 subjects each. The first experimenter is a 26 year old native speaker of Hà Nội Vietnamese while the second one is a 31 year old speaker, born in Quảng Ngãi and raised in Tây Nguyên, who spoke the SVN standard during the experiment. With each experimenter, 10 subjects carried out the task with the NVN stimuli, followed by the SVN stimuli, while 10 subjects were presented with the two blocks in the opposite order. Before the experiment, the experimenters greeted the subjects and explained the procedure for a few minutes in their dialect. The subjects took a five-minute pause between the two blocks during which they talked to the experimenter. Thus, four variables were controlled for in the procedure: 1. Dialect of stimulus (SVN/NVN) 2. Dialect of the experimenter (SVN/NVN) 3. Order of presentation of the two blocks of stimuli (SVN first/NVN first) 4. Tone of stimulus

2.4 Statistical analysis Identification results were statistically analyzed in a binary logistic regression model1, implemented in R 2.10.1 with the glm function. The analysis was run on 1 This model and its link function are better fitted to the data than the model used in the preliminary report published in Brunelle and Jannedy (2007). Overall results are similar, but with differences for individual factors and interactions in some tones.

22

Marc Brunelle and Stefanie Jannedy

each tone independently. The dependent variable is a binary variable (0 = incorrect, 1 = correct). The fixed factors are the dialect of the speaker, the order of presentation of the two dialect blocks and the identity of the experimenter (assumed to reflect the experimenter’s dialect). All interactions between fixed factors were also included. Due to multiple comparisons across the six tones, the p-level normally assumed for the single comparison (p < 0.05) was adjusted to p < 0.01 (a literal Bonferroni adjustment would yield p < 0.008333, but no p-value falls between 0.008 and 0.01 – see the appendix). This increases the chance of obtaining false negatives, but we prefer to err on the conservative side than to report spurious effects. Our statistical model cannot be perfectly balanced because of the merger of NVN tones C1 and C2 in SVN. For this reason, three tonal categories will not be dissociated from their dialect (NVN C1, NVN C2 and SVN C1-C2) in the statistical analysis: the dialect of the speaker will not be included as a fixed factor for these categories. In order to allow a comparison of errors and an interpretation of the binary logistic regression results, the number of errors made for each independent factor and interaction were corrected by means of estimated marginal means. These corrected errors are an evaluation of the number of errors made by listeners for each factor or interaction after filtering out the effect of all other factors and interactions.

3 Results Before looking at the results of the statistical analysis, a breakdown of the raw data is necessary. Table 1 gives error rates for the various factors. It suggests a strong effect of some of the four factors on the overall distribution of perception errors. First, there is a much larger proportion of errors when our subjects listen to stimuli from the NVN (21.4%) than the SVN (7.4%) dialect. This is not surprising as listeners are all native speakers of SVN. The raw differences caused by the experimenter and the order of presentation of the dialectal blocks are very small, but we will see that these factors do have an effect on perception for some individual tones. Finally, some tones cause more errors than others, especially tone C1 (74.7%) and, to a lesser extent, tone C2 (13.3%). This is not unexpected as these tones are merged in SVN and as the resulting merged tone is different from its corresponding NVN counterparts.

23

The cross-dialectal perception of Vietnamese tones

Factor

Level

Number

%

Dialect

NVN SVN

1281/6000 369/5000

21.4% 7.4%

Experimenter

NVN SVN

775/5500 875/5500

14.1% 15.9%

Order

1st 2nd

846/5500 804/5500

15.4% 14.6%

Tone of stimulus

A1 A2 B1 B2 C1 C2

137/2000 188/2000 140/2000 172/2000 747/1000 266/2000

6.9% 9.4% 7.0% 8.6% 74.7% 13.3%

Table 1: Number and proportion of errors for main factors

A closer look at the differences between dialects and between tones reveals some interesting generalizations. Table 2 is a confusion matrix for the NVN stimuli, where the largest number of responses for each category is marked in boldface. As predicted in section 1.3, NVN tones A1, A2, B1 and B2 (stimulus rows) are well-identified overall (Tone B1 is identified as C1 71/1000 times, corresponding to 7.1%, probably because its low rising contour is reminiscent of the final rise of standard NVN C1). Most misidentifications fall in the C1 and C2 columns. This probably reflects an awareness by subjects that NVN C1 and C2 are phonetically very different from SVN C1-C2: these categories become default answers when listeners are faced with unusual stimuli. Response Stimulus A1 A2 B1 B2 C1 C2

A1

A2

B1

B2

C1

C2

Total

978 4 5 22 209 0

0 938 1 13 401 1

0 1 868 1 2 110

0 14 8 936 67 2

22 40 71 17 253 141

0 3 47 11 68 746

1000 1000 1000 1000 1000 1000

Table 2: Response matrix for Northern Vietnamese tones (errors reaching 10% or more of the stimuli are italicized)

Interestingly, tone C1, which is low-falling and mildly laryngealized in NVN, is mostly identified as A2, a low-falling tone in both dialects. This response pattern reflects the lack of familiarity of SVN listeners with NVN contrastive laryngealization (Brunelle 2009b; Kirby 2010). For tone C2, 83/110 of the B1

24

Marc Brunelle and Stefanie Jannedy

responses were elicited by a single token of C2 (recall that there are five tokens of each tone). Once again laryngealization plays a role. SVN listeners seem to be able to perceive the laryngealization of C2 when it is salient, but the misidentified token is less laryngealized than normal and, as a result, is treated as a simple rising tone. Lastly, the large number of NVN C2 misidentified as C1 can be attributed to the fact that C1 and C2 are merged into a single tone in SVN and therefore, are not clearly distinct for the SVN subjects. As predicted in section 1.3, Table 3 reveals that for some tones, SVN subjects make more mistakes when listening to stimuli in their own dialect than in NVN. This may be attributable to the absence of voice quality cues in SVN listeners that makes some tones less distinct: two tones with similar shapes and heights are more likely to be confused if they have no other contrastive acoustic property. More specifically, two significant misidentification patterns must be reported. First, 92/1000 tokens of tone A2 are misidentified as B2. This is due to the very similar contours of the two tones: they are both low-falling in SVN, their main difference being a slight final rise at the end of B2. Second, tone C1-C2 is identified as either C1 and C2, two answers that are equally accurate as the two tone marks are pronounced the same way in SVN (the bias for C2 is unexpected, but could simply be due to the visual layout of the answer boxes on the screen). In the rest of the paper, C1 and C2 are both counted as instances of correct identification. Response Stimulus

A1

A2

B1

B2

C1

C2

Total

A1

885

18

45

5

41

6

1000

A2

3

874

0

92

6

25

1000

B1

2

5

992

0

0

1

1000

B2

0

46

1

892

24

37

1000

C1-C2

1

2

1

8

373

615

1000

Table 3: Response matrix for Southern Vietnamese tones

Raw results are misleading as they can hide variation going in opposite directions in different subcategories or aggregate mistakes stemming from different factors or interactions of factors. For instance, the apparently higher error rates for the tone stimulus A1 in Table 2 than Table 3 and for the tone stimulus B1 in Table 3 than in Table 2 are not significantly attributable to dialect as they are a sum of relatively small and non-significant errors (which are attributable to the factor experimenter and to the interactions dialect*experimenter and dialect* experimenter*order). It is therefore crucial to look at the more sophisticated results obtained from the binary logistic regression. We see in Table 4 that all factors and interactions except order and dialect*order are responsible for variation in misidentification patterns. Note that tones A1 and B1 are not affected by

The cross-dialectal perception of Vietnamese tones

25

the various factors, which is expected since they have the same perceptual cues in the two dialects (Brunelle 2009b): A1 is modal and neither rising nor low, while B1 is rising and modal. Factor(s) dialect

A1

A2 **

B1

B2

NVN C1

NVN C2

**

*

SVN C1-C2

**

experimenter order dialect*exp.

**

*

dialect*order exp.*order dialect*exp.*ord.

** **†

Table 4: Factors and interactions affecting identifications, by tone class **p < 0.001, *p < 0.01, †low p-value due to outliers (explanation below), full coefficient tables in Appendix 1.

In order to look at the significant factors and interactions in more detail, their corrected means (estimated marginal means) are reported in Table 5. The corrected means represent the percentage of incorrect answers, so a high number means a low accuracy. Effect of individual variables The effect of dialect on tones A2 and B2 is fairly easy to explain. The higher proportion of misidentifications in SVN tones A2 and B2 than in their NVN counterparts can be attributed to the more crowded lower pitch range of SVN, which leads to more confusion. The effect of experimenter on NVN tone C1 is also relatively easy to explain. As this tone is only found in NVN, the effect of the experimenter for tone C1 can be compared to the effect of dialect*experimenter in other tones: it shows the differences in the perception of NVN tone C1 when the experimenter is NVN as opposed to SVN. It seems that the presence of the NVN-speaking experimenter creates a reference that can be used by SVN listeners to identify NVN C1. When the experimenter speaks SVN, the error rate is at chance level (85.3%, which roughly corresponds to 1/6 chances of a coincidental correct response). However, performance improves significantly (error rate: 67.4%) when the experimenter speaks NVN because her tones C1 can be used as a reference point. It appears that listeners learn to pay attention to new perceptual cues from the speech of the experimenter.

26

Marc Brunelle and Stefanie Jannedy

Tone

Factor(s)

p value

Estimated Marginal Means (EMM)

A2

dialect

p < 0.001

NVN 4.6% SVN 10.9%

dialect * exp.

p < 0.001

dial. NVN * exp. NVN 8.9% dial. NVN * exp. SVN 2.4% dial. SVN * exp. NVN 8.6% dial. SVN * exp. SVN 13.7%

dialect * exp. * order

p < 0.001

dial. NVN * exp. NVN * 1st block 5.2% dial. NVN * exp. SVN * 1st block 2.8% dial. SVN * exp. NVN * 1st block 14.0% dial. SVN * exp. SVN * 1st block 7.6% dial. NVN * exp. NVN * 2nd block 14.8% dial. NVN * exp. SVN * 2nd block 2.0% dial. SVN * exp. NVN * 2nd block 5.2% dial. SVN * exp. SVN * 2nd block 23.6%

dialect

p < 0.001

NVN 5.6% SVN 10.8%

dialect * exp.

p < 0.01

dial. NVN * exp. NVN 3.5% dial. NVN * exp. SVN 8.7% dial. SVN * exp. NVN 11.2% dial. SVN * exp. SVN 10.4%

experimenter

p < 0.001

NVN 67.4% SVN 85.3%

exp. * order

p < 0.001

exp. NVN * 1st block 80.8% exp. NVN * 2nd block 50.4% exp. SVN * 1st block 76.4% exp. SVN * 2nd block 91.2%

experimenter

p < 0.01

NVN 29.4% SVN 20.8%

B2

NVN C1

NVN C2

Table 5: Corrected means for statistically significant factors and interactions

Surprisingly, the opposite effect is found in NVN C2: SVN listeners perceive it slightly better when the experimenter is SVN than when she is NVN. The milder glottalization in the fast speech of the experimenter than in the elicitation forms used as stimuli might account for this. When SVN listeners hear the NVN stimuli in presence of the SVN experimenter, they use their knowledge of the well-enunciated NVN C2 that they hear in the media as a reference and do relatively well. However, the weak glottalization in the fast colloquial speech of the NVN experimenter departs from the formal realization of C2 and leads the subjects to confuse NVN C2 with tones that have similar contours, but no glottalization, like B1 or C1.

The cross-dialectal perception of Vietnamese tones

27

Interactions of two variables As the effect of the interaction between dialect and experimenter is more transparent for tone B2 than for A2, we will discuss it first. The effect of the interaction seems to be based on the salient glottalization of NVN tone B2 (Brunelle 2009b). When the strong final glottalization of the NVN tone B2 can be matched with its presence in the speech of a NVN experimenter, the error rate is very low (3.5%). When stimuli are NVN, but the experimenter is SVN, the glottalization is still a salient cue, but its perceptual relevance must be inferred from previous exposure to NVN (without additional confirmation from the speech of the experimenter) leading to a higher proportion of errors (8.7%). Finally, the error rate goes up when tone B2 is produced in SVN (10.4% with a SVN experimenter, 11.2% with a NVN experimenter) because glottalization is not available as a cue and because B2 is realized in a crowded tonal space in this dialect. In the case of tone A2, we are unable to explain why the congruent combination of SVN listeners rating SVN stimuli in the context of an SVN experimenter yields the highest error rate. However, it is clear that the difficulty of identifying tone A2 in the crowded lower pitch range of SVN already mentioned interacts with the experimenter. Note that the error rates with the NVN experimenter are exaggerated by two listeners who display idiosyncratic patterns of responses. However, the interaction is still significant after excluding them from the statistical analysis (p < 0.001, E.M. Means: dial. NVN*exp. NVN 4.3%; dial. NVN*exp. SVN 2.6%; dial. SVN*exp. NVN 5.1%; dial. SVN*exp. SVN 15.3%). The effect of experimenter and order is only found in tone C1, for which stimuli are all NVN. While performance is better in the second block than in the first one when the experimenter is NVN, it is actually worse in the second than in the first block when the experimenter is SVN. This interaction seems to illustrate a learning effect: the performance of SVN subjects exposed to an unfamiliar NVN tone improves with the amount of exposure to the speech of the NVN experimenter. When the experimenter is SVN, subjects do better if they have to identify the NVN stimuli right in the first block. If they spend the first half of the experiment session hearing SVN stimuli and interacting with a SVN experimenter, they get entrenched in their “SVN mode” and fare worse on the NVN stimuli in the second block. Three-way interaction The only apparent significant three-way interaction is between dialect, experimenter and order in tone A2. However, this effect turns out to be due to two listeners who behave in a way markedly different from the other 8 listeners when the dialects of the stimuli and of the experimenter are both SVN. While most subjects occasionally confuse tone A2 with tone B2, one of these two

28

Marc Brunelle and Stefanie Jannedy

subjects confuses it with C1 20 times out of 25, and the other always confuses it with C2. If these two outliers are excluded, the three-way interaction is no longer significant (p = 0.421). Note that this exclusion does not affect the statistical significance of other factors and interactions affecting A2.

4 Discussion and conclusion Results show that the tones whose perception is affected by dialectal differences are those that vary across dialects (B2, C1, C2) or that are produced in dense areas of the tonal space (A2, B2, C1). Tones A1 and B1, which have distinctive contours and tend to be produced in the less crowded upper tone range, are not significantly affected by the factors included in this study. The effect of dialect depends on the acoustic and perceptual properties of each tone and cannot be mechanically attributed to a greater familiarity with one’s native dialect. Unfamiliarity with NVN tone C2 does explain why listeners misidentify it more often than its SVN counterpart C1-C2, but in the case of tones A2 and B2, it seems that the more crowded lower pitch range of SVN triggers more errors despite the fact that SVN is the native dialect of the listeners. In the case of tone B2, the better identification of the NVN stimuli could also be attributed to its very salient final glottalization, with which SVN listeners are already familiar through exposure to the standard NVN variety used in the media (as already noted in Brunelle 2009b). Experimenter only seems to affect perception when it interacts with the dialect of the stimuli. It seems that for tones that are very different in the two dialects, performance in the unfamiliar NVN variety is better if the experimenter is also NVN (C1 and B2). In other words, when having to identify an unfamiliar stimulus, subjects rely in part on their immediate experience of the speech of the experimenter to carry out the task. This suggests that listeners are able to rapidly adapt to and exploit the surrounding linguistic environment. This effect is another instance of the shift observed in studies in which listeners are biased towards particular perceptual thresholds via linguistic – the experimenter – (Hay, Drager and Warren 2009; Hay, Warren and Drager 2010) or even nonlinguistic referents (Niedzielski 1999; Hay and Drager 2010). NVN tones C1 and B2 are perceptually similar to tones A1 and A2, but are crucially distinguished from them by their laryngealization. When a NVN experimenter is present, listeners are induced to pay more attention to laryngealization, leading to a better performance than when the experimenter is SVN and does not make use of laryngealization in her patterns of tonal contrasts.

The cross-dialectal perception of Vietnamese tones

29

The interaction of the experimenter and the order of presentation of the dialectal blocks in tone C1 further supports this view. Not only does the presence of a NVN experimenter result in better identification of NVN tone C1, but a longer exposure to her speech leads to further improvement, suggesting that there is learning and rapid adaptation. Interestingly, there also seems to be a corresponding negative effect. SVN subjects fare the worst on NVN tone C1 (91.2% errors) when the experimenter is SVN and the NVN stimuli are presented in the second block. This is probably due an entrenchment in SVN mode during the first half of the experiment, which creates confusion when the unexpected NVN stimuli are presented. The experiment also yields one partly unexpected effect: the interaction between the dialect of the stimuli and the experimenter for A2 shows the highest error rate when the listeners are primed with SVN and hear SVN stimuli. Part of this effect seems to be attributable to the experimenter given that the error rate dramatically drops when the experimenter is NVN (all other things remaining constant), but it cannot be easily interpreted. It is difficult to determine if our results are due to the idiosyncratic characteristics of the experimenters and/or the speakers from whom the stimuli were recorded or if the effects can be generalized to larger dialectal influences. However, if we view dialects as speech varieties spoken by individuals who share linguistic and regional properties, this question becomes circular. If our speakers and experimenters are not representative of their “dialects”, the magnitude and direction of the effects we found might shift, but the mechanisms that underlie them will remain the same.

References Babel, Molly E. 2009. Phonetic and Social Selectivity in Phonetic Accommodation. Unpublished PhD Dissertation. University of California at Berkeley. Bargh, John A. and Williams, Erin L. 2006. The automaticity of social life. Current Directions in Psychological Science 15. 1–4. Barlow, Michael and Kemmer, Suzanne. 2000. Usage Based Models of Language. Stanford: CSLI Publications. Bell, Allan. 1984. Language Style as Audience Design. Language in Society 13. 145–204. Bod, Rens, Hay, Jennifer and Jannedy, Stefanie. 2003. Probabilistic Linguistics. Cambridge: MIT Press. Brunelle, Marc. 2003. Tone Coarticulation in Northern Vietnamese. Proceedings of the 15th International Congress of Phonetic Sciences. 2673–2676. Brunelle, Marc. 2009a. Northern And Southern Vietnamese Tone Coarticulation: A Comparative Case Study. Journal of Southeast Asian Linguistics 1. 49–62.

30

Marc Brunelle and Stefanie Jannedy

Brunelle, Marc. 2009b. Tone perception in Northern and Southern Vietnamese. Journal of Phonetics 37. 79–96. Brunelle, Marc and Jannedy, Stefanie. 2007. Social Effects on the Perception of Vietnamese Tones. Proceedings of the 16th International Congress of Phonetic Sciences. 1461–1464. Drager, Katie. 2011. Speaker, Age and Vowel Perception. Language and Speech. Earle, Michael Allan. 1975. An Acoustic Phonetic Study of Northern Vietnamese Tones. Santa Barbara: Speech Communications Research Laboratory Inc. Giles, Howard and Coupland, Nikolas. 1991. Language: Contexts and Consequences. Keynes: Open University Press. Goldinger, Stephen D. 1997. Perception and production in an episodic lexicon. In: Keith Johnson and John W. Mullennix (ed.). Talker Variability in Speech Processing. 33–66. San Diego: Academic Press. Goldinger, Stephen D. 1998. Echoes of echoes? An episodic theory of lexical access. Psychological Review 105. 251–279. Gregory, Stanford W., Green, Brian E., Carrothers, Robert M. and Dagan, Kelly A. 2001. Verifying the primacy of voice fundamental frequency in social status accommodation. Language and Communication 21. 37–60. Gsell, René. 1980. Remarques sur la structure de l’espace tonal en Vietnamien du Sud (parler de Saigon). Cahiers d’Études Vietnamiennes (4). 1–26. Han, Mieko. 1969. Studies in the Phonology of Asian Languages VIII: Vietnamese Tones. Los Angeles: University of Southern California. Han, Mieko and Kim, Kong-On. 1974. Phonetic variation of Vietnamese tones in disyllabic utterances. Journal of Phonetics 2. 223–232. Harrington, Jonathan. 2006. An acoustic analysis of ‘happy-tensing’ in the Queen’s Christmas broadcasts. Journal of Phonetics 34(4). 439–457. Hay, Jennifer and Drager, Katie. 2010. Stuffed toys and speech perception. Linguistics 48(4). 865–892. Hay, Jennifer, Drager, Katie and Warren, Paul. 2009. Careful who you talk to: An effect of experimenter identity in the production of the NEAR/SQUARE merger in New Zealand English. Australian Journal of Linguistics 29(2). 269–285. Hay, Jennifer, Jannedy, Stefanie and Mendoza-Denton, Norma. 2010. Oprah and /ay/: Lexical Frequency, Referee Design and Style. In: Miriam Meyerhoff and Erik Schleef (ed.). The Routledge Sociolinguistic Reader. 53–58. London: Routledge. Hay, Jennifer, Nolan, Aaron and Drager, Katie. 2006. From fush to feesh: Exemplar priming in speech perception. Linguistic Review 23. 351–379. Hay, Jennifer, Walker, Abby and Drager, Katie. 2010. Listening and Speaking with Attitude, ms. Hay, Jennifer, Warren, Paul and Drager, Katie. 2010. Short-term exposure to one dialect affects processing of another. Language and Speech 53(4), 447–471. Johnson, Keith. 1997. Speech Perception without Speaker Normalization: An Exemplar Model In: Keith Johnson and John Mullenix (ed.). Talker Variability in Speech Processing. 145– 165: Academic Press. Kirby, James. 2010. Dialect experience in Vietnamese tone perception. Journal of the Acoustical Society of America 127(6). Michaud, Alexis. 2004. Final Consonants and Glottalization: New Perspectives from Hanoi Vietnamese. Phonetica 61. 119–146. Michaud, Alexis and Vũ, Ngọc Tuấn. 2004. Glottalized and Nonglottalized Tones under Emphasis: Open Quotient Curves Remain Stable, F0 Curve is Modified. Proceedings of Speech Prosody 2004, Nara, Japan. 745–748.

The cross-dialectal perception of Vietnamese tones

31

Michaud, Alexis, Vũ, Ngọc Tuấn, Amelot, Angélique and Roubleau, Bernard. 2006. Nasal realease, nasal finals and tonal contrasts in Hanoi Vietnamese: an aerodynamic experiment. Mon-Khmer Studies 36, 121–137. Namy, Laura L., Nygaard, Lynne C. and Sauterberg, Denise. 2002. Gender differences in vocal accommodation: the role of perception. Journal of Language and Social Psychology 21. 422–432. Nguyễn, Văn Lợi and Edmondson, Jerold. 1997. Tones and voice quality in modern northern Vietnamese: Instrumental case studies. Mon-Khmer Studies 28. 1–18. Niedzielski, Nancy. 1999. The effect of social information on the perception of sociolinguistic variables. Journal of Language and Social Psychology 18(1). 1–18. Pardo, Jennifer S. 2006. On phonetic convergence during conversational interaction. Journal of the Acoustical Society of America 119. 2382–2393. Pickering, Martin J. and Garrod, Simon. 2004. Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences 27. 169–226. Pierrehumbert, Janet. 2002. Word-specific Phonetics. In: Carlos Gussenhoven and Natasha Warner (ed.). Laboratory Phonology VII. 101–140. Berlin: Mouton de Gruyter. Rickford, John and McNair-Knox, Faye. 1994. Addresse- and topic-influences style shift: A quantitative sociolinguistic study. In: Douglas Biber and Edward Finegan (ed.). Sociolinguistic Perspectives on Register. 235–276. New York: Oxford University Press. Seitz, Philip. 1986. Relationship between tones and segments in Vietnamese. Ph.D., Linguistics, University of Pennsylvania. Strand, Elizabeth A. 1999. Uncovering the Role of Gender Stereotypes in Speech Production. Journal of Language and Social Psychology 18. 86–99. Strand, Elizabeth A. and Johnson, Keith. 1996. Gradient and visual speaker normalization in the perception of fricatives. In: Dafydd Gibbon (ed.). Natural Language Processing and Speech Technology. 14–26. Berlin: Mouton. Trudgill, Peter. 1981. Linguistic accommodation: Sociolinguistic observations on a sociopsychological theory. In: Roberta Hendrick, Carrie Masek and Mary Frances Miller (eds.). Papers from the Parasession on Language and Behavior. 218–237. Chicago: Chicago Linguistics Society. Vũ, Ngọc Tuấn, d’Alessandro, Christophe and Michaud, Alexis. 2005. Using open quotient for the characterization of Vietnamese glottalised tones. Interspeech, Lisbon. Vũ, Thanh Phương. 1981. The Acoustic and Perceptual Nature of Tone in Vietnamese. Unpublished PhD Dissertation. Australian National University at Canberra. Vũ, Thanh Phương. 1982. Phonetic Properties of Vietnamese Tones across dialects. In: David Bradley (ed.). Papers in Southeast Asian Linguistics. 55–75. Sydney: Australian National University.

32

Marc Brunelle and Stefanie Jannedy

Appendix 1: Statistical analyses (R formulas and coefficients) TONE A1 Call: glm(formula = I(Correctness == "Wrong") ~ Dialect * Order * Experimenter, family = binomial("logit")) Coefficients: (Intercept) Dialect(NVN) Order(1) Experimenter (NVN) Dialect(NVN):Order(1) Dialect(NVN):Experimenter(NVN) Order(1):Experimenter(NVN) Dialect(NVN):Order(1):Exp.(NVN) Signif. codes: 0 ‘**’ 0.001 ‘*’ 0.01

Estimate –6.80587 –4.70435 3.93545 0.41516 4.12039 0.06304 0.22872 0.24948

Std. Error 120.23356 120.23356 120.23356 120.23356 120.23356 120.23356 120.23356 120.23356

z value –0.057 –0.039 0.033 0.003 0.034 0.001 0.002 0.002

Pr(>|z|) 0.955 0.969 0.974 0.997 0.973 1.000 0.998 0.998

TONE A2 Call: glm(formula = I(Correctness == "Wrong") ~ Dialect * Order * Experimenter, family = binomial("logit")) Coefficients: (Intercept) Dialect(NVN) Order(1) Experimenter(NVN) Dialect(NVN):Order(1) Dialect(NVN):Exp.(NVN) Order(1):Exp.(NVN) Dial.(NVN):Order(1):Exp.(NVN) Signif. codes: 0 ‘**’ 0.001 ‘*’ 0.01

Estimate –2.56045 –0.46267 –0.13044 0.21748 –0.07158 0.47890 0.11421 –0.48856

Std. Error 0.10135 0.10135 0.10135 0.10135 0.10135 0.10135 0.10135 0.10135

z value –25.263 –4.565 –1.287 2.146 –0.706 4.725 1.127 –4.820

Pr(>|z|) |z|) 0.947 0.972 0.991 0.988 0.981 0.975 0.979 0.983

TONE B2 Call: glm(formula = I(Correctness == "Wrong") ~ Dialect * Order * Experimenter, family = binomial("logit")) Coefficients: (Intercept) Dialect(NVN) Order(1) Experimenter(NVN) Dialect(NVN):Order(1) Dialect(NVN):Experimenter(NVN) Order(1):Experimenter(NVN) Dialect(NVN):Order(1):Exp.(NVN) Signif. codes: 0 ‘**’ 0.001 ‘*’ 0.01

Estimate –2.47208 –0.35845 0.04968 –0.21654 –0.01207 –0.25719 0.19117 0.17237

Std. Error 0.09052 0.09052 0.09052 0.09052 0.09052 0.09052 0.09052 0.09052

z value –27.309 –3.960 0.549 –2.392 –0.133 –2.841 2.112 1.904

Pr(>|z|) |z|) |z|) |z|)

investigated line in the transcript Final pitch movement ? high rising . (deep) falling Shifted pitch register

low pitch register Loudness and speech tempo

forte, loud

allegro, fast

II Noun phrase syntax

Nguyễn Hùng Tưởng

3 The Vietnamese noun phrase* This chapter provides an analysis of the Vietnamese noun phrase from a generative perspective. Given the empirical evidence that Vietnamese has a class of lexical articles, the Vietnamese noun phrase is analyzed as a DP (determiner phrase), which contains multiple functional projections, including a classifier (CL), a numeral (Num), and a demonstrative (Dem) phrase. Crosslinguistic evidence supports the proposed analysis for Vietnamese. Specifically, five different nominal word orders across Austro-Asiatic, Austro-Tai and Sino-Tibetan language families are demonstrated to be derived by applications of syntactic movement from a single underlying DP structure. Keywords: classifier, measure word, determiner, demonstrative reinforcer, specifier, count, mass, word order, typology, phrasal movement

As a typical classifier language, Vietnamese presents itself as an interesting test case for the study of the internal structure of noun phrases. The Vietnamese noun phrase has not been investigated to any great extent in the syntactic literature. This chapter examines the Vietnamese noun phrase from a generative perspective (Chomsky 1981, 1986), incorporating further developments in the analysis of functional categories (Abney 1987; Pollock 1989; Grimshaw 1991; Kayne’s Antisymmetry approach 1994; Chomsky’s Minimalist Program 1993, 1995). The aim is to see how Vietnamese fits in crosslinguistically in terms of noun phrase structure. Controversial issues such as the existence of lexical articles in Vietnamese, the ‘extra’ CÁI, and the distinction between classifiers and measure words will be discussed. Moreover, the study provides an account of word order possibilities within the noun phrase across several language families of Southeast Asia, with a focus on Vietnamese. It is suggested here that the five nominal word orders attested in 30 different languages are derived from a single underlying * This research was originally presented at the “Linguistics of Vietnamese” Workshop at the University of Stuttgart in 2009. The present chapter summarizes findings from my doctoral dissertation (2004) at Boston University, with some further developments. I thank the audience at the Workshop for a lively discussion of the topic and two anonymous reviewers for helpful comments on a previous version of the chapter. Naturally, any remaining errors are mine. For a detailed description and analysis of the Vietnamese noun phrase, a literature review, and relevant references, see Nguyễn, Hùng Tưởng (2008) The Structure of the Vietnamese Noun Phrase.

58

Nguyễn Hùng Tưởng

structure. A unified analysis of the typology of DPs involving demonstratives and demonstrative reinforcers across languages is also proposed.

1 Underlying structure of the Vietnamese noun phrase Empirical considerations of word order facts provide evidence that the Vietnamese noun phrase consists of the lexical projection of NP, dominated by various functional projections including: Classifier (CL), Numeral (Num), Demonstrative (Dem), and Determiner (D). It is proposed that the Vietnamese noun phrase has the underlying structure as shown in Figure 1.

Figure 1: Underlying structure of the Vietnamese DP

The Vietnamese noun phrase

59

The surface structure of the Vietnamese DP is derived by an obligatory movement of the entire NumP to the specifier of the DemP. Analyses of each of the component phrases follow. Since any nominal projection in Vietnamese is interpreted as being either definite or indefinite (to be discussed in 8.1), it will be assumed that the DP is always projected.

2 Nouns in Vietnamese In Vietnamese, nouns are morphologically unmarked for number, as shown in (1): (1) a.

mèo cat

b.

‘(a/the) cat(s)’

đường sugar ‘(the) sugar’

Thus, a bare noun can refer to one or more than one entity (1a), or to mass substances (1b). In Vietnamese, there is no lexical distinction between count and mass nouns. Vietnamese nouns typically behave like English mass nouns (e.g., sugar or furniture), although they may refer to discrete entities. Consistent with their apparent mass-like properties, Vietnamese nouns such as mèo and đường cannot be directly counted without the help of a classifier (2c) or measure phrase (2d): (2)

a. *hai mèo two cat

b. *hai đường two sugar

‘two cats’ c. hai con mèo two CL cat ‘two cats’

‘two kilograms of sugar’ d.

hai kí đường two kilogram sugar ‘two kilograms of sugar’

In other words, nouns in Vietnamese are non-individuated and thus they need to be individuated via classifiers or measure phrases before they can be counted or measured.1

1 It should be noted that there is a lexically specified class of nouns that do not appear with classifiers in numeration. These include nouns denoting time units (as in (i)), or geographical and administrative units (as in (ii)), or (Sino-Vietnamese) disyllable compound nouns (as in

60

Nguyễn Hùng Tưởng

3 Structure of the Vietnamese classifier phrase 3.1 Distinction between classifiers and measure words Classifiers2 and measure words (Nmeasure) are often treated as belonging to a single category in prior literature3 because linearly they take the slot between the number and the head noun and they do not co-occur, as shown in (3) and illustrated in (4). (3)

Num {CL , Nmeasure} N

(4) a.

Num – CL – N hai cái bát

b. Num – Nmeasure – N hai bát cơm

two CL bowl

two

bowl

rice

‘two bowls’

‘two bowlfuls of rice’

However, there are significant differences between classifiers and measure words that support the idea that they are not of the same grammatical category. The following table sums up their differences.

(iii)). These are often called ‘non-classified nouns’ (see Emeneau 1951 and Nguyễn, Đình-Hòa 1957, 1997, for further descriptions). (i) hai ngày two day ‘two days’

(ii) ba tỉnh three province ‘three provinces’

(iii) năm sinh viên five student person ‘five students’

2 See Nguyễn, Hùng Tưởng (2008) for a list of Vietnamese classifiers according to their semantic types (such as unit classifiers, kind classifiers, and event classifiers). Also, see Allan (1977) and Aikhenvald (2000) for general discussions of classifiers. 3 In fact, the term classifiers is often used to cover measure words as well as classifiers proper (see, e.g., Emeneau 1951; Nguyễn Đ. H. 1957; Jones 1970, among others). Lyons (1977) refers to classifiers as “sortal classifiers”, and measure words as “mensural classifiers”, although he does not consider their different distributions. Li and Thompson (1981: 106) state that “any measure word can be a classifier.” Cheng and Sybesma (1999) make no distinction between classifiers and measure words in their discussion of Mandarin and Cantonese. Tai and Wang (1990) propose that “while a classifier ‘categorizes’ an object, a measure word simply ‘measures’ an object”, but they do not go beyond the different functional distinctions of classifiers and measure words.

61

The Vietnamese noun phrase

Differences

MW

CL

Examples

Can be modified

+

–

(5 & 6)

Phonological stress

+

–

(7)

Selectional restriction

–

+

(8 & 9)

Quantity reading

+

–

(10)

Membership

open

closed

Table 1: Distinction between Measure Words and Classifiers

(5)

a. Measure sequence ba muỗng đường three spoon sugar ‘three spoonfuls of sugar’ b. ba [muỗng vun] đường (cf. * ba muỗng [vun đường]) 4 three spoon heaping sugar ‘three heaping spoonfuls of sugar’

(6)

a.

Classifier sequence hai cuốn [sách cũ] two CL book old ‘two old books’

b. *hai [cuốn cũ] (7)

sách

a. Measure sequence một BÁT cơm (or một BÁT CƠM) one bowl rice ‘a bowlful of rice’ b. Classifier sequence một cái BÁT (cf. *một CÁI BÁT) one CL bowl ‘a bowl’

4 Note that prosodically this string should be uttered as ba |muỗng vun| đường, but not as *ba muỗng |vun đường|.

62

Nguyễn Hùng Tưởng

(8) a.

cái nhà mới CL:inanimate house new ‘a new house’

b.

(9)

*con nhà mới CL:animate house new

a. một quả cam one CL:round orange ‘an orange’ b. một quả banh one CL:round ball ‘a ball’

(10)

a. Nouns used as measures (not preceded by a CL) Hai bát cơm này ngon. two bowl rice this delicious ‘These two bowlfuls of rice are delicious.’ (Not ‘These two rice-bowls. . .’) b. Nouns not used as measures (preceded by a CL) Nó làm vỡ hai cái bát. he make break two CL bowl ‘He smashed two rice-bowls.’ (Not ‘. . . two bowlfuls’)

3.2 Structure of the Vietnamese classifier phrase Evidence based on their distributional, phonological, lexical, and semantic properties just presented suggests that classifiers and measure phrases belong to two distinct categories. Indeed, classifiers are head-like and measures are phrasal in nature, and thus they should be base-generated in different syntactic positions within the same phrase. Specifically, it is proposed that a classifier heads the classifier phrase (CLP), while a measure phrase is base-generated as a DP in Spec,CLP, as shown in Figure 2.

The Vietnamese noun phrase

63

Figure 2: Structure of the Vietnamese Classifier Phrase

Note that, as shown in (c) above, a measure phrase (DP) is in complementary distribution with an overt classifier head within the Classifier Phrase. Empirical facts such as (a–c) above indicate that a noun can be individuated with either a classifier (a) or a measure phrase (b), but not with both (c). In other words, once a noun has been individuated by a classifier, it is not compatible with a measure phrase.5

4 Structure of the Vietnamese numerated noun phrase Once head nouns have been individuated by a classifier or a measure word, the resulting phrase [CL-N] or [measure-N] is countable and can be quantified numerically. It is proposed here that numerals project a separate Num head, selecting a CLP complement. Thus, NumP (Numeral Phrase) is structurally higher than the CLP, containing the classifier and the measure DP, as shown in Figure 3.

5 For a detailed analysis of the internal structure of the measure DP, see Nguyễn, Hùng Tưởng (2008, Chapter 3).

64

Nguyễn Hùng Tưởng

Figure 3: Structure of the Vietnamese Numerated Noun Phrase

Evidence in support of this proposal includes the following. First, unlike in other languages, numerals and classifiers in Vietnamese are distinct morphemes, with distinct semantic functions. Second, ellipsis data also support the idea that numerals are higher than the CLP (see note 6): (11) Nó có năm con trâu, còn anh nó chỉ có một con trâu. he have five CL buffalo but brother he only have one CL buffalo ‘He has five buffaloes, but his brother has only one.’ Third, prosodic evidence in Vietnamese appears to support the constituent structure [Num [CL-N]], rather than [[Num CL] N]. For example, a noun phrase such as ba con bò ‘three cows’ is uttered with a prosodic break between the numeral and the [CL-N] phrase, as shown in (12a), rather than between the Num-CL sequence and the N, as shown in (12b). (12) a.

Ba | con bò. three CL cow

b.

?*Ba con | bò

The Vietnamese noun phrase

65

Finally, numerals and classifiers can be separated by the focus marker CÁI (which will be addressed in the next section) in Vietnamese, thus suggesting that numerals are outside the CLP.6 The existence of CÁI provides strong motivation for projecting two independent projections NumP and CLP.

5 Analysis of the ‘extra’ CÁI 5.1 The ‘extra’ CÁI versus the classifier cái There is another special function word in the Vietnamese noun phrase that can intervene between a numeral (if there is one) and the classifier, as illustrated in (13). That is the particle CÁI,7 which is homonymous with the classifier cái, but has a function different from that (or any other) classifier. (13) (Num) – CÁI – CL – N – modifier CÁI con chó NÀY chỉ biết sủa suốt ngày. CÁI CL dog this only know bark all day ‘THIS VERY dog is barking all the time.’ Arguments that this CÁI is not a classifier include the following. First, when the particle CÁI is used in a noun phrase, the presence of a classifier is obligatory, as shown in (14a). However, the particle CÁI cannot be used before the homonymous classifier, as shown in (14b):8 (14)

a. Hai CÁI *(con) mèo này hay ăn vụng. two CÁI CL cat this often eat stealthily ‘THESE VERY TWO cats always eat on the sly.’ b. *Tôi muốn mua hai CÁI cái bàn này. I want buy two CÁI CL table this ‘I want to buy THESE VERY two tables.’

6 This is counter to Tang’s (1990) proposal that Num and CL are base-generated in a single head CL because they are inseparable in Chinese. Another piece of evidence that indicates that Num and CL should occur in distinct positions is that some languages allow an adjective to be inserted between a numeral and a classifier (see, e.g., Simpson 2005). 7 As will be discussed later, this particle normally bears phonological stress. For this reason, CÁI, together with its focal elements, will hereafter be printed in capitals. In contrast, the classifier cái is never phonologically stressed. Therefore, it will not be capitalized. 8 One possible way to account for the non-occurrence of the sequence CÁI-cái is haplology, a morphological process whereby co-occurrence of two homophonous morphemes in a sequence is often avoided in languages.

66

Nguyễn Hùng Tưởng

Given that classifiers cannot co-occur, examples such as (14) demonstrate that CÁI in such constructions is not a classifier. Another piece of evidence that the particle CÁI and classifiers differ is stress patterns. The particle CÁI is always phonologically stressed (as indicated by capital letters) while classifiers are never stressed (as indicated by lower-case letters): (15)

a. Cuốn sách này hay thật. (cf. *CUỐN sách này . . .) CL book this good real ‘This book is really good.’ b. CÁI cuốn sách này hay thật. CÁI CL book this good real ‘THIS VERY book is really good.’

In addition, CÁI is not a classifier because, unlike classifiers, the particle CÁI can occur with any kind of nouns, whether classified nouns (15b), unclassified nouns (16a), mass-denoting nouns (16b), or measure nouns (16c): (16) a.

CÁI ngày ấy CÁI day that ‘THAT VERY day’

c.

b. CÁI thịt ít mỡ CÁI meat little fat ‘the LEAN meat’

Uống thử CÁI ấm trà này coi có ngon không. drink try CÁI pot tea this see QUES delicious QUES ‘Try THIS VERY potful of tea to see if it’s delicious.’

The evidence above demonstrates that CÁI is different in distribution and function from the homonymous classifier with which it cannot co-occur.

5.2 Structure of the ‘extra’ CÁI What, then, is the function of the ‘extra’ CÁI in the structure of the Vietnamese noun phrase? What is its position? When CÁI occurs in a noun phrase, it explicitly draws attention to a focal element to be identified in the noun phrase, as shown in (17): (17)

a. Tôi thích CÁI con ngựa ĐEN. I like CÁI CL horse black ‘I like the BLACK horse (but not the brown one).’

The Vietnamese noun phrase

67

b. Tôi thích CÁI con NGỰA đen. I like CÁI CL horse black ‘I like the black HORSE (but not the black cow).’ It is proposed here that the particle CÁI is a focus marker in the Vietnamese noun phrase. It serves as a formal device to signal that there is a focus in the NP. Examples such as (14a) and (17) indicate that CÁI is structurally higher in the tree than the classifier. Conceptually, CÁI is an adjunct, optionally inserted to signal a focus in the noun phrase. Thus the focus marker CÁI is analyzed as being left-adjoined to the CLP, as shown in Figure 4.

Figure 4: The Structure of the Vietnamese Focus Phrase

6 Analysis of demonstratives Given that Vietnamese demonstratives are free morphemes and always occur following the head noun (18a) and following any other post-nominal modifiers as the rightmost constituent in the noun phrase (18b), it is proposed that demonstratives are heads of the DemP, complement of D, and that the extended NP (i.e., NumP) crosses over them.9 Under this analysis, the surface structure of

9 Many researchers (e.g., Mallen 1989; Tang 1990; Li 1998, 1999; among others) have adopted Abney’s (1987) DP-analysis to account for demonstratives. This proposal was based on the observation that articles and demonstratives are in complementary distribution in languages such as English. However, it cannot account for the fact that demonstratives can co-occur with

68

Nguyễn Hùng Tưởng

the nominal word order involving demonstratives in Vietnamese is derived by an obligatory leftward movement of the functional phrase NumP containing the head noun to Spec,DemP. This is shown in Figure 5. (18)

a. cuốn sách này CL book this ‘this book’ b. hai cuốn sách vàng này two CL book yellow this ‘these two yellow books’

Figure 5: Structure of the Vietnamese DemP

articles in many other languages (including Irish, Javanese, Vietnamese, Hungarian, Romanian and Spanish), suggesting that demonstratives are not in D. Giusti (1993, 1997) proposes that demonstratives are base-generated in a lower specifier of a functional projection and may move to Spec,DP at LF. It is not clear how Giusti’s (1993, 1997) head-movement analysis can straightforwardly account for the Vietnamese fact that both the head noun and its post-nominal modifiers precede the demonstrative (see Nguyễn, Hùng Tưởng (2008, Chapter 2) for a review of relevant analyses). Simpson (2005) also analyzes Dem as occupying D to account for noun phrases in Southeast Asia. One purpose of this study is to argue that it is XP movement inside DP, rather than head movement, that accounts for nominal word order typology. This idea is supported by Cinque (2005, 2008).

The Vietnamese noun phrase

69

One advantage of projecting DemP is that it provides an extra landing site for XP movement. This proposal can unify the analysis of the distribution of demonstratives relative to the noun phrase in other Southeast Asian languages, which will be discussed in section 10. Second, it would be difficult to see how noun movement alone would derive the postnominal position of adjectives as well as the DP-final position of demonstratives in Vietnamese. This analysis also straightforwardly accounts for the Vietnamese nominal word order facts, presented in (18) where both adjectives and demonstratives surface postnominally, and also in (19) where articles and demonstratives can co-occur. (19) [ DP những / các [ DemP [cuốn sách] này]] ART ART CL book this ‘these books’ Third, given that demonstratives belong to a closed class and never seem to be modifiable, it is not unreasonable to posit that they are heads. Finally, analyzing demonstratives as base-generated to the left of the head noun is consistent with the Vietnamese nominal word order distributions: articles, numerals, focus marker CÁI, classifiers, as well as demonstratives themselves precede the head noun and modifiers such as adjectives and relative clauses follow it.

7 Demonstrative reinforcers The Demonstrative-as-head account proposed here also provides a unified analysis for the demonstrative-reinforcer constructions across languages.10 Crosslinguistically, there are three attested nominal word orders involving demonstratives and their reinforcers, as shown in (20): (20)

a. Swedish:

Dem Reinforcer NP den här mannen the here man-the ‘this man’

10 Note that Bernstein (1997) adopts and develops Giusti’s (1993) demonstrative-in-specifier analysis to account for demonstrative reinforcement constructions in Romance and Germanic languages. Syntactically, her proposal to move an XP (i.e., Dem) to an X position (i.e., D0) is rather unwelcome with respect to the structure-preservation principle. However, it is possible to extend her NP-movement analysis to account for the DP-final position of demonstrative and reinforcer in Vietnamese (see Nguyễn, Hùng Tưởng (2008, Chapter 2) for a review).

70

Nguyễn Hùng Tưởng

b. French:

Dem NP Reinforcer cette femme-ci this woman-here ‘this woman’

c. Vietnamese:

NP Dem Reinforcer cuốn sách này đây CL book this here ‘this book here’

It is claimed that these three language groups share the same underlying word order corresponding to the configuration in Figure 6. Derivations of the surface word orders can be accounted for in terms of occurrence or nonoccurrence of two XP-movements, namely NumP to Spec,FP and NumP to Spec, DemP. The first possibility is NumP raising to Spec,FP, deriving the Dem-NPReinforcer order of French. The second possibility involves NumP raising to Spec,FP and then up to Spec,DemP, resulting in the NP-Dem-Reinforcer order of Vietnamese. The third possibility, involving no movement, patterns with the Dem-Reinforcer-NP order of Swedish.

Figure 6: Structure of Demonstrative-reinforcer constructions

71

The Vietnamese noun phrase

8 Structure of the Vietnamese DP 8.1 Does Vietnamese have a set of lexical articles? Empirical considerations from Vietnamese provide evidence that there is a set of lexical articles in Vietnamese, contrary to what has generally been assumed.11 Specifically, the articles are shown in (21): (21) một: những: các:

[−definite; −plural] [−definite; +plural] [+definite; +plural]

Evidence in favor of treating một and những as indefinite articles is summarized in Table 2.12 Indefinite environments

một

những

1. Existential sentences

Z

Z

2. Initial mentions of referents

Z

Z

3. Complement of là ‘be’, trở thành ‘become’

Z

Z

4. Articles do not bear phonological stress

Z

Z

Table 2: Indefinite environments where một and những occur

As shown above, một and những can occur in environments that require indefinite noun phrases. Conversely, as indefinite articles, một and những cannot occur in definite environments, as shown in the following table. Definite environments

một

những

1. Unique reference (i) Superlatives (ii) đầu tiên ‘the first’ (iii) duy nhất ‘the only’ 2. Second and subsequent mentions of the referent

No No No

No No No

No

No

Table 3: Definite environments where một and những cannot occur

11 Vietnamese những and các have often been called plural markers or pluralizers (Thompson 1965; Dương 1971; Nguyễn Đ. H. 1997; among others). 12 For reasons of space, see Nguyễn, Hùng Tưởng (2008, Chapter 1) for further explanations and illustrated examples. Also, see (23) below for some typical examples.

72

Nguyễn Hùng Tưởng

Evidence in favor of treating các as a plural definite article includes: (22)

1. 2.

3.

các contributes an inclusive interpretation (i.e., denoting a maximal set of entities), which is associated with definiteness.13 các is not used in contexts that are associated with indefiniteness (e.g., existential sentences, first mentions, and as complement of là ‘be’, trở thành ‘become’, etc.). các only occurs as part of plural, definite DPs.

8.2 The surface structure of the Vietnamese DP As markers of indefiniteness, một and những are argued to occur only in indefinite environments, while các is used only in contexts that are associated with definiteness. The Vietnamese data show that the articles một, những, and các occur in the same position in D by Spell-Out, although indefinite một may originate in a lower position for numerals and raises up during the course of the syntactic derivation.14 To illustrate, a Vietnamese DP such as các con mèo này ‘these cats’ has the structural representation in Figure 7. The surface structure is derived by obligatory movement of the extended nominal phrase (NumP) to Spec, DemP. The possible interpretations of nominal constructions involving a noun (such as bò ‘cow’) in Vietnamese can be summarized as in Table 4 and illustrated in (23). The definiteness/indefiniteness of a bare noun [N], as in (23a), is normally pragmatically constrained and there is no way to use a determiner to unambiguously mark [+singular; +definite] in Vietnamese. When there are no overt markers of plurality, a [CL-N] phrase, as in (23b), is invariably interpreted as

13 Treating các as a definite article is consistent with a criterion that uses ‘inclusiveness’ to define definiteness. Hawkins (1978) associates the English definite article the with this ‘inclusiveness’ hypothesis of definiteness: This property of the definite article to refer to all objects or all the mass in the pragmatically limited domain of quantification, whereupon the sentence as a whole makes some claim about these objects, I shall refer to as ‘inclusiveness’. This term is intended to capture the fact that the reference is all-inclusive, i.e. all the objects in the shared set satisfying the descriptive predicate are being referred to, and none are being excluded. (Hawkins 1978: 161) 14 Given that the indefinite article một may well have developed from the homonymous numeral, it is plausible to propose that một moves from its base-generated position in Num to D. Simpson (2005) also proposes a similar analysis for numeral ‘one’ in some Southeast Asian languages.

The Vietnamese noun phrase

73

Figure 7: The surface structure of the Vietnamese DP

singular,15 but it is compatible with either definite or indefinite reference.16 If the determiner một is present, as in (23c), then only an indefinite interpretation is possible. Noun phrases with các and những have a plural reading, with [các-CLN] being definite, as in (23d), and [những-CL-N] being indefinite, as in (23e).

15 The same is also true for other classifier languages, such as Thai (Hundias and Konver 1983), and Mandarin and Cantonese (Cheng and Sybesma 1999), among others. 16 With respect to definiteness, Cheng and Sybesma (1999) treat classifiers as the equivalent of the definite article, which does not seem semantically and syntactically plausible for Vietnamese, given that there is a class of lexical articles in Vietnamese. Since the definiteness of a nominal projection is encoded on D, it is possible to postulate that the CL head moves to D to result in a definite reading, as proposed by Tang (1990), Li (1998, 1999), and Simpson (2005).

74

Nguyễn Hùng Tưởng

Noun phrases

SG

PL

Indef

Def

Examples

N

+

CL

+

+

+

+

(23a)

–

+

+

(23b)

một-CL-N các-CL-N

+

–

+

–

(23c)

–

+

–

+

những-CL-N

(23d)

–

+

+

–

(23e)

Table 4: NP interpretational possibilities in Vietnamese

(23)

a. Bò đang ăn lúa kìa! cow PROG eat paddy over there. ‘Look! (A/The) cow(s) is/are eating your paddy!’ b. Con bò đang ăn lúa kìa! CL cow PROG eat paddy over there ‘Look! A/The cow is eating your paddy!’ c. Một con bò đang ăn lúa kìa! ART CL cow PROG eat paddy over there ‘Look! A cow is eating your paddy.’ d.

Các con bò đang ăn lúa kìa! ART CL cow PROG eat paddy over there ‘Look! The cows are eating your paddy.’

e.

Những con bò đang ăn lúa kìa! ART CL cow PROG eat paddy over there ‘Look! Some of the cows are eating your paddy.’

9 Analysis of postnominal modifiers Although lengthy clusters of modifiers are not commonly used in Vietnamese, the head noun can be post-modified by an adjective phrase (AP), a prepositional phrase (PP), a relative clause (RC), and a possessive phrase (PossP).17 The linearization of these postnominal modifiers is typically as follows: (24) Head > AP > PP > RC > PossP

17 See Nguyễn, Hùng Tưởng (2008, Chapter 1) for a detailed discussion of postnominal modification.

The Vietnamese noun phrase

75

These postnominal modifiers are analyzed as phrases adjoined to an N’, with the simplified structure in Figure 8 below:18

Figure 8: Structure of Postnominal Modifiers in the Vietnamese Numerated Noun Phrase

If there is a complement to the head noun, it appears right after the head noun and before any modifiers, as shown in example (25), and represented in Figure 9 below:

18 This is a tentative analysis of postnominal modifiers in the Vietnamese noun phrase. It appears to work for other Southeast Asian languages under study as well as for Vietnamese. Given various approaches, namely adjunction analysis, head-analysis, and Spec-analysis, to adjective syntax in the literature (Abney 1987; Valois 1991; Bernstein 1991; Cinque 1994, 2008; Giusti 1991, 1997; Laenzlinger 2000), there remains much to be explored, and I leave this for future research.

76 (25)

Nguyễn Hùng Tưởng

a. Một người giáo viên toán tận tâm one CL teacher math devoted ‘a devoted math teacher’ b. *Một người giáo viên tận tâm toán

Figure 9: Numerated Noun Phrase with Complement NP

10 Crosslinguistic evidence Crosslinguistic evidence lends support to the analysis proposed in this study. It will be shown that empirical evidence suggests that nominal word order differences across three language families of Southeast Asia, namely Austro-Asiatic, Austro-Tai, and Sino-Tibetan, can be derived from the same underlying structure.

10.1 Nominal word order facts in Southeast Asian languages Southeast Asia is a typical Sprachbund, a linguistic area where representatives of various language families share quite a few typological features because of

The Vietnamese noun phrase

77

close geographical proximity. It is, therefore, not surprising that many of the language families in the area exhibit similar nominal structures. There are five nominal word order patterns attested in the three language families under study, as shown in (26):19 (26)

Type 1 Type 2 Type 3 Type 4 Type 5

Dem Dem N Num N

Num N A CL A

CL A Dem N Num

N Num Num A CL

A CL CL Dem Dem

(Yao-type) (Burmese-type) (Lisu-type) (Vietnamese-type) (Thai-type)

Type 1 word order sequence is found in Yao. Type 2 language groups include Burmese, Lolo, Maru, Lahu, Rawang, and Ancient Chinese. Type 3 sequence is attested in Lisu.20 Type 4 patterns closely with Vietnamese, Malay, Indonesian, Nung, Miao (also called Hmong or Meo), White Tai, Black Tai, Sedang, Sre, Katu, Cham, and Brôu. Type 5 is characteristic of Thai, Khmer, Lao, Javanese, Khmu, Shan, Palaung, and Karen.

10.2 Derivation of the surface structure of the five language types It is possible to give a unified analysis of the typology of DPs across languages if one posits that the five language groups described above have the same basic underlying word order shown in (27): (27) [Dem [Num [CL [[N] A] ]]] This surface word order corresponds to the configuration in Figure 10:

19 Since evidence is lacking about the existence of lexical articles in these language groups, this comparative analysis will focus on noun phrases involving classifiers, numerals, demonstratives, adjectives and nouns. For further discussion of lexical articles in the languages of Southeast Asia and China, see Nguyễn, Hùng Tưởng (2008, Chapter 4). 20 I am grateful to an anonymous reviewer for raising the theoretical possibility of this nominal word order. It turns out that this order is also attested in Jones (1970). As predicted, the derivation of this pattern is in line with the present proposal, as will be shown in 10.2.3.

78

Nguyễn Hùng Tưởng

Figure 10: Underlying nominal structure of Southeast Asian languages

The five attested nominal word orders found in Southeast Asian can be derived in terms of application or non-application of the two phrasal movement rules shown in (28):21 (28)

i.

NP to Spec,NumP (and then to Spec,DemP)

ii.

NumP to Spec,DemP

It is possible to suggest that there is a restructuring of nominal constituents across these language families, such that one, the other, both, or neither of the two movement processes may occur, yielding the five attested word order possibilities.

21 It should be noted that the analysis proposed here focuses on the attested word orders found in Jones (1970), thus leaving other possible orders for future research. For present purposes, I will not consider the prenominal adjective pattern [Dem Num CL A N] either, which is found in various Chinese dialects.

The Vietnamese noun phrase

79

10.2.1 Derivation of the Yao-type languages Nominal elements in this language group have the word order [Dem-Num-CL-NA], as illustrated in the Yao example below (Jones 1970): (29)

naiqteij pyei taub juq kia’q this four CL dog black ‘these four black dogs’

The surface word order in this language group corresponds to underlying order of elements in the DP, as illustrated in Figure 10. Thus, no movement is involved.

10.2.2 The Burmese-type languages The Burmese-type nominal word order is characterized as [Dem-N-A-Num-CL], illustrated in the examples below (Jones 1970): (30)

a.

Burmese dí lú ìl θôun yaw this person big three CL ‘these three big people’

b.

Lolo eé ts’ò sêizé seū joù this man fat three CL ‘these three fat men’

c.

Maru maw myaw oyu pit tau that horse white four CL ‘those four white horses’

The surface word order [Dem-N-A-Num-CL] is derived by an obligatory NP movement to Spec,NumP.22 This is illustrated in Figure 11:

22 Evidence from the Burmese-type languages supports NP-movement, rather than N-movement: when the head noun moves up, the adjectival modifier cannot be left stranded.

80

Nguyễn Hùng Tưởng

Figure 11: Derivation of the Burmese-type surface structure

(i) Burmese a. dí [lú ìl ] θôun yaw this person big three CL ‘these three big people’ b. *dí lú

θôun yaw

ìl

The same applies to other language groups: (ii) Vietnamese a. hai con [trâu trắng] này two CL buffalo white this ‘these two white buffaloes’ b. *hai con trâu này trắng (grammatical on the reading where this is a sentence, meaning ‘These two buffaloes are white.’) (iii) Thai a. [máa dam] soon tua nan dog black two CL that ‘those two black dogs’ b. *máa soon tua

nan dam

The Vietnamese noun phrase

81

10.2.3 Lisu-type languages Nominal word order in this language type has the pattern [N-A-Dem-Num-CL], as illustrated in the Lisu example below (Jones 1970):23 (31) la 5htsaw 4 da 5ma 4 htê hti 5 raw person big this one CL ‘this big person’ In this language group, the NP raises to Spec,NumP and then to Spec,DemP, deriving the surface word order [N-A-Dem-Num-CL], as shown in Figure 12.

Figure 12: Derivation of the Lisu-type surface structure

10.2.4 Vietnamese-type languages Nominal phrases in this language group follow the word order sequence [NumCL-N-A-Dem]. Examples are provided below ((32a & b) from Jones 1970):

23 The Lisu example shown in (31) is a combination of two examples in Jones (1970: 5): (i) ná 2yi 3 da 5ma 4 river big

(ii) la 5 htsaw 4 htê hti 5 raw person this one CL

82 (32)

Nguyễn Hùng Tưởng

a. Malay dua lai kmeja putela ini two CL shirt white this ‘these two white shirts’ b. Nung slam ắn tẳng slay này three CL chair small this ‘these three small chairs’

(33)

Vietnamese hai con trâu trắng này two CL buffalo white this ‘these two white buffaloes’

The Dem-final word order in this language group results from an obligatory movement of NumP (containing the head noun) to Spec,DemP, deriving [NumCL-N-A-Dem], as shown in Figure 13.

Figure 13: Derivation of the Vietnamese-type surface structure

The Vietnamese noun phrase

83

10.2.5 The Thai-type languages Nominal phrases in this language group have the word order pattern [N-A-NumCL-Dem]. Examples are given below (Jones 1970): (34) Thai máa dam soon tua nan dog black two CL that ‘those two black dogs’ (35)

Khmer serəy lo o bəy nĕaq nuh girl pretty three CL that ‘those three pretty women’

(36)

Javanese kertas shang sepuluh lambar kuwi paper red ten CL that ‘those ten sheets of paper’

The Dem-final word order in this language group is derived by the same two obligatory XP movements. First, the NP raises to Spec,NumP, and then the whole NumP raises to Spec,DemP, producing the surface order [N-A-Num-CL-Dem]. This is represented in Figure 14:

Figure 14: Derivation of the Thai-type surface structure

84

Nguyễn Hùng Tưởng

11 Conclusion This research has provided empirical evidence that there is a class of lexical articles in Vietnamese and thus supports analyzing the Vietnamese noun phrase as a DP. A distinction between classifiers and measure words has been established and an analysis of this distinction has been proposed. The ‘extra’ CÁI has been reevaluated and hypothesized to be a focus marker. Crosslinguistic findings from this study show that there is only one underlying word order available for the three language families of Southeast Asia and that word order variations are the result of the occurrence or non-occurrence of relevant movement processes. The surface nominal word orders are derived either via leftward phrasal movements containing the head noun or via roll-up movements. The proposal made here suggests that head movement inside DP cannot account for nominal word order typology and that phrasal movement not containing the lexical head is not possible.24 An analysis of demonstrative reinforcers across languages has also been proposed. It is hoped that this study provides insights into the understanding of the internal structure of noun phrases, both in Vietnamese and crosslinguistically.

References Abney, Steve. 1987. The English noun phrase in its sentential aspect. Ph.D. dissertation, MIT, Cambridge, MA. Aikhenvald, Alexandra. 2000. Classifiers: a Typology of Noun Categorization Devices. New York: Oxford University Press. Allan, Keith. 1977. Classifiers. Language 53: 284–310. Bernstein, Judy. 1991. DPs in French and Walloon: Evidence for Parametric Variation in Nominal Head Movement. Probus 3(2): 101–126. Bernstein, Judy. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Cheng, Lisa and Rint Sybesma. 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry 30: 509–542. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press.

24 Recently, Cinque (2005, 2008) has examined a large array of word orders, including noun phrase elements, attributive adjectives, adverbs, speech-act modifying adverbs, and circumstantial PPs, and proposes that there is only one word order or structure available for all languages and the attested orders are derivable via independent conditions of phrasal movement.

The Vietnamese noun phrase

85

Chomsky, Noam. 1993. A minimalist program for linguistic theory. In: K. Hale & S. J. Keyser (eds.), The View from Building 20: 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP. In: G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, and R. Zanuttini (eds.), Paths towards Universal Grammar: Studies in honor of Richard S. Kayne, 85–110. Georgetown University Press. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36: 315–332. Cinque, Guglielmo. 2008. The fundamental left-right asymmetry of natural languages. In: S. Scalise, E. Magni, and A. Bisetto (eds.), Universals of Language Today, 165–184. Berlin: Springer. Dương, Thanh Bình. 1971. A Tagmemic Comparison of the Structure of English and Vietnamese Sentences. The Hague: Mouton. Emeneau, Murray Barnson. 1951. Studies in Vietnamese (Annamese) Grammar. Berkeley: University of California. Giusti, Giuliana. 1991. The categorial status of quantified nominals. Linguistische Berichte 136: 438–454. Giusti, Giuliana. 1997. The categorical status of determiners. In: L. Haegeman (ed.), The New Comparative Syntax, 95–123. Harlow: Addison Wesley Longman. Gorald, D. R. 1978. Numeral classifier systems: A Southeast Asian cross-linguistic Analysis. Linguistics of the Tibeto-Burman Area 4: 1–72. Greenberg, Joseph. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In: J. H. Greenberg (ed.), Universals of language, 73–113. Cambridge, MA: MIT Press. Grimshaw, Jane. 1991. Extended projection. Unpublished manuscript, Brande is University. Hawkins, John. 1983. Word Order Universals. New York: Academic Press. Hundias, H. and Ulrike Kolver. 1983. Syntax and semantics of numeral classifiers in Thai. Studies in Language 7: 165–214. Jones, Robert. 1970. Classifier constructions in Southeast Asia. Journal of the American Oriental Society 90: 1–12. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Laenzlinger, Christopher. 2000. French Adjective Ordering: Perspectives on DP-internal Movement Types. Generative Grammar in Geneva 1: 55–104. Li, Charles and Sandra Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley, CA: University of California Press. Li, Yen-hui Audrey. 1998. Argument determiner phrases and number phrases. Linguistic Inquiry 29: 693–702. Li, Yen-hui Audrey. 1999. Plurality in a classifier language. Journal of East Asian Linguistics 8: 75–99. Lyons, John. 1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. Mallen, Enrique. 1989. The internal structure of determiner phrases. Ph.D. dissertation, Cornell University. Nguyễn, Tài Cẩn. 1975. Từ Loại Danh Từ trong Tiếng Việt Hiện Đại [The word class of nouns in modern Vietnamese]. Hanoi: Khoa Hoc Xa Hoi. Nguyễn, Đình-Hòa. 1957. Classifiers in Vietnamese. Word 13: 124–152.

86

Nguyễn Hùng Tưởng

Nguyễn, Đình-Hòa. 1997. Vietnamese: Tiếng Việt Không Son Phấn [Vietnamese without Veneer]. Philadelphia: John Benjamins B.V. Nguyễn, Phú Phong. 1995. Questions de Linguistique Vietnamienne. Les Classificateurs et les Déictiques. Paris: Presses de L’Ecole Française D’Extrême-Orient (Monographies, No. 180). Nguyễn, Hùng Tưởng. 2004. The Structure of the Vietnamese Noun Phrase. Ph.D. dissertation, Boston University. Nguyễn, Hùng Tưởng. 2008. The Structure of the Vietnamese Noun Phrase. Saarbrücken, Germany: VDM Verlag Dr. Müller Aktiengesellschaft & Co. KG. Pollock, Jean-Yves. 1989. Verb movement, UG, and the structure of IP. Linguistic Inquiry 20: 365–424. Simpson, Andrew. 2005. Classifiers and DP structure in Southeast Asia. In: G. Cinque and R. Kayne (eds.), The Oxford Handbook of Comparative Syntax, 806–838. Oxford: Oxford University Press. Tai, James and Lianqing Wang. 1990. A semantic study of the classifier Tiao. Journal of the Chinese Language Teachers Association, Volume XXV, No. 1: 35–56. Tang, Jane. 1990. Chinese phrase structure and the extended X’-theory. Ph.D. dissertation, Cornell University. Thompson, Lawrence. 1965. A Vietnamese Grammar. Seattle: University of Washington Press. Valois, Daniel. 1991. The internal syntax of DP. Unpublished Ph.D. dissertation, UCLA.

Jennie Tran

4 Vietnamese classifier phrases from the perspective of how children acquire them This study investigates the structure of numeral and non-numeral classifier phrases in Vietnamese by examining how such phrases and their syntactic properties develop in very young Vietnamese children between the age of two to five, employing both longitudinal and cross-sectional child data. The results show that Vietnamese children demonstrate early knowledge of the classifier slot in a noun phrase. As early as age two, they can produce an obligatory classifier, not only with a noun, but also with a demonstrative and an interrogative. They combine classifiers with a demonstrative before they combine it with a number, and produce the classifier with a noun rather than without. They tend to combine the number with a classifier before they combine it with a head noun. Grammatical omission of the head noun precedes the mastery of a full classifier phrase. The first classifier phrases that Vietnamese children build at age two or earlier are all non-numeral and consist of two elements, namely classifier + noun, classifier + demonstrative, classifier + wh-word. Several differences are found when the results are compared with previous studies of other numeral classifier languages. The largest differences are the higher percentage of omission errors and classifier + noun sequences. Keywords: classifiers, classifier phrases, noun phrases, Vietnamese, child Vietnamese, first language acquisition, child language, child language acquisition

1 Introduction Vietnamese is an isolating language with lexical tones and monosyllabic word structure. Verbs have no tense or person/number agreement markings. Nouns do not have any morphological inflections for case, gender, person or number. Compared to a language like German, which has tense, person/number agreement markings on verbs, and case, gender, person, and number markings on nouns, Vietnamese seems easier for foreigners to learn and perhaps simpler for children to acquire. However, this is not necessarily true because unlike German and other European languages, Vietnamese is one of several Asian languages with a complex numeral classifier system. This system is not easy for foreigners

88

Jennie Tran

to grasp and not easy for children to acquire. This is due to the large number of classifiers (over 200), the complex semantic nature of classifiers, and the fixed order of elements in classifier phrases. Classifiers may be particularly difficult to learn in the context of Vietnamese noun phrases, which are not exclusively head-final, but have both postnominal and prenominal modifying elements: Demonstratives, wh-words, adjectives, possessive pronouns and relative clauses follow the noun, whereas numerals and classifiers precede the noun. The word order of a full five-element classifier phrase is [numeral/quantifier + classifier + noun + adjective + demonstrative/wh-word/possessive/relative clause]. (Cf. Nguyễn’s contribution to this volume for an analysis of such expressions in terms of constituency.) From a semantic perspective, classifiers are unbound function words that categorize the head noun based on inherent or salient features of the noun’s referent, such as animacy, shape, length, dimension, function, or material. From a syntactic perspective, numeral classifiers can be defined based on three criteria: First, they are obligatory in noun phrases containing a numeral, a demonstrative, an interrogative, or a combination of these elements, with or without an overt head noun, as shown in (1).1 Second, they can be repeated, as is typical in Thai and Burmese (2), or can be reduplicated, as is unique to Cantonese (3). (1) hai con (chó ) này (Vietnamese) two CL (dog) this ‘these two dogs’ (2)

thaleesàap saam thaleesàap (Thai – data from Hundius and Kölver 1983: 164) lake three (CL: lake) ‘three lakes’

(3)

go3 go3 hok6saang1 (Cantonese – data from Wong 1998: 16) CL CL student ‘every student’

Third, the numeral and the classifier are in all cases adjacent, that is, they are an inseparable pair. An adjective cannot separate the numeral + classifier constituent. The word order of the elements in the numeral classifier phrase can vary depending on the language.

1 Exceptions exist in Vietnamese and Thai, where the presence of the demonstrative does not require an obligatory classifier (as can be seen in Table 1).

Vietnamese classifier phrases from the perspective of how children acquire them

89

‘this dog’

‘these two dogs’

‘two black dogs’

Japanese

Dem-N

Dem-Num-CL-N

Adj-N-Num-CL

Cantonese

Dem-CL-N

Dem-Num-CL-N

Num-CL-Adj-N

Mandarin

Dem-CL-N

Dem-Num-CL-N

Num-CL-Adj-N

Thai

N-(CL)-Dem

N-Num-CL-Dem

N-Adj-Num-CL

Vietnamese

(CL)-N-Dem

Num-CL-N-Dem

Num-CL-N-Adj

Table 1: Order of elements in NPs in various numeral classifier languages

Since Vietnamese has a different order of classifier phrase elements than that of other languages studied so far, an investigation of the developmental pattern of Vietnamese classifier phrases will add cross-linguistic depth to the classifier acquisition literature. This study represents the first work conducted on the acquisition of the numeral classifier system in Vietnamese. The focus is on investigating how children acquire classifier phrases both in earlier and later stages, as well as on examining the emergence in child speech of the first two-, three- and four-element classifier phrases. Although the study investigates the syntactic as well as the semantic development of classifiers in young Vietnamesespeaking children, the content of the present chapter covers only the development of the syntactic aspect of classifiers. The semantic aspect of classifier development will be dealt with in a separate, future paper.

2 Previous research on the syntactic development of classifier phrases In numeral classifier languages, classifiers play an important role in the development of noun phrase structures. So far, there have been studies on numeral classifier development in six languages (Cantonese: Mak 1991; Poon 1981; Szeto 1996; Wong 1998; Japanese: Matsumoto 1985a, 1985b, 1987; Muraishi 1983; Sanches 1977; Uchida and Imai 1996, 1999; Yamamoto 2000; Korean: Lee 1994, Lee and Lee 2005; Mandarin: Erbaugh 1982, 1986; Fang 1985; Hsu 1987; Hu 1993; Loke 1991; Loke and Harrison 1986; Ng 1991; Tse et al. 1991; Thai: Carpenter 1987, 1991; Gandour et al. 1984, and Malay: Salehuddin and Winskel 2009.) Among these, the only studies that examined the syntactic properties of classifier development are those by Erbaugh (1982) and Hu (1993) on Mandarin, Carpenter (1987) on Thai, and Wong (1998) on Cantonese. Erbaugh’s and Wong’s studies are the only two longitudinal studies examining naturalistic data from very young children in the age range 1;10 to 3;10 and 1;09 to 2;09 respectively.

90

Jennie Tran

Carpenter’s is a seminal experimental study on how children between the ages 2;0 to 11;0 acquire the semantic system of Thai classifiers. Erbaugh’s (1982), Hu’s (1993), and Carpenter’s (1987) studies examined the syntactic aspect of classifier phrases rather peripherally. Only Wong’s (1998) study had a primary focus on the syntactic development of noun phrases, including classifier phrases. With respect to the earliest strategies, previous studies found essentially similar syntagmatic intralinguistic patterns. From very early on, at around age 1;9, children are sensitive to the word order requirements for classifier constructions. They show knowledge of the position of the classifier in the noun phrase. The very first strategy is the ‘blank attempt’, reported by Gandour et al. (1984), in which children do not yet produce a classifier but make a hesitation after the numeral and a pause to mark the classifier slot. As young as age two, they proceed with the next strategy, filling the classifier slot with the most general classifier, which they use as a default. The general classifier thus serves as a placeholder for the grammatical position of the classifier. Thai acquisition of classifiers manifests two additional strategies: an across-the-board usage of one particular classifier, regardless of the head noun, at around age 3;0, along with an overuse of ‘repeaters’ across all age groups, where children use the head noun as its own classifier (a strategy referred to as ‘overspecification’) (Carpenter 1987). These early strategies are important stepping stones into the system as they indicate that children understand the syntactic properties of classifier phrases very early. Children know that (a) the classifier is an obligatory element in the classifier phrase, and (b) classifiers constitute a closed class. Interestingly, the results of Carpenter’s (1987) elicitations show that the child subjects never responded with a word that did not conventionally belong in the classifier position, suggesting that children are very aware of the constraints on which words may be classifiers. The studies by Erbaugh (1982) on Mandarin found that early on, Mandarinspeaking children produce the general classifier together with the demonstrative ‘this’ and the number ‘one’, but in the case of specific classifiers, they produce them later, first with the numbers ‘two’ and above, and then with a demonstrative. Erbaugh further found in her child data that classifiers occur with a noun rather than without. The results in Wong’s (1998) study on Cantonese, by contrast, indicate that the grammatical omission of the head noun precedes the mastery of a full classifier phrase. Children tend to combine the classifier with a number before they combine it with a head noun. Their speech in the early multi-word stage exhibits two-element noun phrase structures, namely a very large number of demonstrative + classifier structures at first, then numeral + classifier, followed by classifier + noun. At around age 2;6, their speech exhibits three-element noun phrase structures: demonstrative + classifier + noun first,

Vietnamese classifier phrases from the perspective of how children acquire them

91

followed by numeral + classifier + noun. In addition, the classifier phrase occurs earlier and more frequently in object than in subject position. The only instances of ungrammatical patterns are the incorrect use of double classifiers and the omission of an obligatory classifier in the presence of a numeral. These errors are, however, very sporadic (below 0.6%). Similarly, Erbaugh’s (1982) investigation of longitudinal naturalistic data comprising over 64 hours of recordings reports a total of only six omissions of obligatory classifiers in the 44,158 utterances produced by a single child. As can be seen from these findings, children perform better in non-numeral than in numeral constructions. Carpenter’s (1987) study on Thai shows that twoyear-olds could produce the semantically appropriate classifier in combination with a demonstrative or an adjective, but failed to respond or responded with a single, inappropriate classifier in combination with a numeral. The combination of classifiers with demonstratives is easier because children’s understanding of deixis precedes that of counting. This pattern suggests an important relation between cognitive and linguistic development.

3 Classifier phrases in Vietnamese The three main syntactic properties to be observed when constructing a Vietnamese classifier phrase are the following. First, the classifier is obligatory in the presence of a numeral, both ordinal and cardinal numbers (4a–b), some quantifiers such as mỗi ‘every’ (4c), một vài ‘a few’ (4d), and the plural markers những, các, mấy (4e). Classifiers are not obligatory with the quantifier nhiều ‘many, much’ (4f). (4a) (4b)

ordinal number + CL CL + cardinal number

một con (chó) con (chó) thứ nhất

‘one CL (dog)’ ‘CL (dog) first’

(4c)

every + CL

mỗi con (chó)

‘every CL (dog)’

(4d)

a few + CL

một vài con (chó)

‘a few CL (dog)’

(4e)

plural + CL

những/các con (chó)

‘PL CL (dog)’

(4f)

many/much + (CL) + noun

nhiều con nhiều chó

‘many CL’ ‘many dog’

The classifier is also obligatory in the presence of demonstratives in direct deixis (4g). (4g)

CL + demonstrative

con (chó) này con (chó) đó con (chó) kia

‘CL (dog) this’ ‘CL (dog) that’ ‘CL (dog) that over there’

92

Jennie Tran

Classifiers are also required with the wh-words gì ‘what’ and nào ‘which’, when the noun referred to is specific (or definite) (4h), or with the question words bao nhiêu or mấy ‘how many’ that require a numeral response (4i). (4h)

CL + wh-word

‘CL (dog) what?’ ‘CL (dog) which?’

con (chó) gì? con (chó) nào?

(4i) how many + CL bao nhiêu/mấy con (chó)? 2 ‘how many CL (dog)?’ The classifier is not obligatory in direct deixis with demonstratives and whwords when the noun is non-specific. Piriyawiboon (2009) notes that in Thai, the noun in a [N-Dem] phrase, where the classifier is absent, can have both an object and a sub-kind reading, whereas the noun in the [N-CL-Dem] phrase, where the classifier is present, has only an object reading. Vietnamese does not pattern like Thai. In (5a), no ambiguity exists in the reading of chó ‘dog’; it has a sub-kind reading (non-specific). In (5b), the presence of the classifier unambiguously gives chó ‘dog’ a specific object reading (specific). (5a)

chó này khôn lắm dog this smart very (i) *‘this particular dog is smart’, (ii) ‘this kind of dog is smart’

(5b)

con chó này khôn lắm CL dog this smart very (i) ‘this particular dog is smart’,

(ii) *‘this kind of dog is smart’

Another condition for the classifier not being obligatory in direct deixis is the grammaticality and acceptability of the non-occurrence of the classifier with the demonstratives and wh-words in less precise speech (Nguyễn 1957). The absence of the classifier is more common in spoken discourse. (6a)

Hai anh muốn ngồi ø bàn nào? two you want sit table which ‘At which table do you want to sit?’

(6b)

Anh ngồi xuống ø ghế nầy. you sit down chair this ‘Sit down on this chair.’ (examples from Emeneau 1951: 99)

2 Generally, mấy is used for numbers below 10 and bao nhiêu for numbers above 10.

Vietnamese classifier phrases from the perspective of how children acquire them

93

The CL + Dem sequence in (4g) carries the meaning of singularity and definiteness, whereas in (5a), (6a–b) it is definite, but does not indicate the singular/ plural contrast. Second, the classifier can be used anaphorically, that is, it can co-occur with a numeral (7a), a demonstrative (7b), a wh-word (7c), an adjective (7d), a possessive expression (7e), or a relative clause (7f) without the head noun, if the head noun has been sufficiently identified by the previous context. The omitted head noun in (7a)–(e) may, for example, be chó ‘dog’. (7a)

Num + CL

hai con

two CL

‘two’

(7b)

CL + Dem

con này

CL this

‘this one’

(7c)

CL + wh-word

con nào?

CL which

‘which one’

(7d)

CL + Adj

con nhỏ

CL small

‘the small one’

(7e)

CL + Poss

con của Mi

CL of Mi

‘Mi’s’

(7f)

CL + Rel Clause

cái con (voi)

(mà)

anh thấy lúc nảy

CL CL (elephant) (which) you see a while ago ‘the elephant (which) you saw a while ago’ (Nguyễn 1957: 130) The classifier + relative clause construction in (7f ) is a special characteristic of the Vietnamese classifier phrase as it allows an ‘extra classifier’ to precede the actual classifier, even though the noun is already classified (Goral 1978; Nguyễn 1957). This ‘extra classifier’ is always the general classifier cái, but it is not permitted when the actual classifier is cái itself, that is, a sequence of cái cái is impossible. Third, the numeral and the classifier are in all cases adjacent. An adjective, for example, cannot separate the Num-CL constituent, as illustrated (6a) and (6b). (8a)

Num-CL is inseparable Num-CL-Adj

(8b) Num-CL is inseparable *Num-Adj-CL

một

quả

to

one

CL:fruit;big-round

big

*một to

quả

one big CL:fruit;big-round

‘one big one’ (referring to an apple) Unlike in Thai and Burmese, in Vietnamese, numeral classifiers cannot be repeated. There are at least 24 possible two- to four-element classifier phrases in Vietnamese. Table 2 illustrates possible combinations of elements.

94

Jennie Tran

2 elements

Examples

Morphological gloss

Gloss

CL-N

con chó quả trứng cái này con này cái đó cái gì? cái nào? con nào? con đen cái của con hai cái ba con

CL:animal dog CL:fruit;big-round egg CL:general this CL:animal this CL:general that CL:general what CL:general which CL:animal which CL:animal black CL:general of I two CL:general three CL:animal

‘the dog’ ‘the egg’ ‘this one’ ‘this one’ ‘that one’ ‘what?’ ‘which one?’ ‘which one?’ ‘the black one’ ‘mine’ ‘two’ ‘three’

CL-Dem a4

CL-wh-word a

CL-Adj a CL-Poss a Num-CL a 3 elements

Examples

Morphological gloss

Gloss

CL-N-Dem

trái5 banh này cái nồi này chiếc xe nào? cái bát nào? cái xe củ quả bóng vàng cái đàn của con con mèo của em hai cây kem ba quả bóng hai cái này mấy cái này hai cái gì đây? hai cái nào? nhiều quả to một ngón nhỏ hai cái của con

CL:fruit;big-round ball this CL:general pot this CL:vehicle car which CL:general bowl which CL:general car old CL:fruit;big-round ball yellow CL:general guitar of I CL:animal cat of I two CL:long-straight ice cream three CL:fruit;big-round ball two CL:general this plural CL:general this two CL:general what here two CL:general which many CL:fruit;big-round big one CL:finger,toe small two CL:general of I

‘this ball’ ‘this pot’ ‘which car?’ ‘which bowl?’ ‘the old car’ ‘the yellow ball’ ‘my guitar’ ‘my cat’ ‘two ice creams’ ‘three balls’ ‘these two’ ‘these’ ‘what are these two?’ ‘which two?’ ‘many big ones’ ‘one small one’ ‘(the) two of mine’

4 elements

Examples

Morphological gloss

CL-N-Adj-Dem CL-N-Adj-whword CL-N-Adj-Poss

quả trứng bự này CL:fruit;big-round egg big this chiếc xe nhỏ nào? CL:vehicle car small which

‘this big egg’ ‘which small car?’

quả bóng màu CL:fruit;big-round ball yellow of I vàng của con hai cục pin này two CL:small-roundish battery this hai trái banh nào? two CL:fruit;big-round ball which hai chiếc xe nhỏ two CL:vehicle car small one CL:one of pair sock of L một cái tất của plural CL:general racket Linh mấy cái vợt of Daddy của ba

‘my yellow ball’

CL-N-wh-word CL-N-Adj CL-N-Poss Num-CL-N Num-CL-Dem a Num-CL-wh-wor a Num-CL-Adj a Num-CL-Poss a

Num-CL-N-Dem Num-CL-N-whword Num-CL-N-Adj Num-CL-N-Poss

Gloss

‘these two batteries’ ‘which two balls?’ ‘two small cars’ ‘Linh’s one sock’ ‘Daddy’s rackets’

Table 2: Possible 2- to 4-element 3 classifier NPs in Vietnamese with examples 3 By ‘element’ is meant the grammatical category and not the number of morphemes needed to construct a classifier phrase. For example, the construction CL-Poss consists of two elements, the classifier (CL) and the possessive construction (Poss), but in an actual phrase, CL-Poss consists of three morphemes, e.g. cái của con, in which cái is CL and của con is Poss.

Vietnamese classifier phrases from the perspective of how children acquire them

95

It is essential to differentiate between numeral and non-numeral classifier phrases because, as mentioned above, children perform better in non-numeral than numeral contexts. Table 3 lists this differentiation. Non-numeral

Numeral

CL-N

Num-CL a

CL-Dem a

Num-CL-N

CL-wh-word a

Num-CL-Dem a

CL-Adj a

Num-CL-wh-word a

CL-Poss a

Num-CL-Adj a

CL-N-Dem

Num-CL-Poss a

CL-N-wh-word

Num-CL-N-Dem

CL-N-Adj

Num-CL-N-wh-word

CL-N-Poss

Num-CL-N-Adj

CL-N-Adj-Dem

Num-CL-N-Poss

CL-N-Adj-wh-word CL-N-Adj-Poss Table 3: Non-numeral vs. numeral classifier phrases

4 Methodology 4.1 Subjects All child subjects for the current study, a total of 42 participants, were children from monolingual families living in Vietnam. The four subjects for the longitudinal study were from four different families of the same economic and educational level. The other 38 subjects for cross-sectional study were from a daycare center in Ho Chi Minh City, Vietnam.

4.2 Research design The study employed two different types of data collection: longitudinal naturalistic data and cross-sectional experimental elicited-production data.

4 The phrases marked with the superscript a symbol are anaphoric uses of the classifier, where the head noun is grammatically omitted. All the other phrases, where the head noun is not omitted, are referred to as full classifier phrases.

96

Jennie Tran

4.2.1 Longitudinal The longitudinal naturalistic data was collected from four children, ages 1;9, 1;11, 2;4, and 2;5, twice a month over a period of six to nine months. Table 4 provides information about these children. Name

Gender

Age range

Duration of recording

Minh Hà Mi Liêm Giang

boy girl boy girl

1;9–2;3 1;11–2;5 2;4–2;10 2;5–3;2

6 months 6 months 6 months 9 months

Table 4: The children in the longitudinal study

Each recording session covered about one hour of interaction between the child and his/her parents or caregiver in the child’s home. All sessions were both audio-taped and video-recorded. The recording scenarios were playtime, mealtime, bedtime, bath time, and T.V. time. For each session, the investigator brought along various types of toys, props, puppets, pictures and picture books to encourage parents/caregivers and children to use various types of nouns, since the targeted syntactic constructions are classifier phrases involving nouns. For uncontrolled elicitation, picture book and toys were used. In addition to the recorded child speech, a questionnaire about the child’s noun repertoire was given to the parents to fill out prior to the first recording and then once every month. This questionnaire was designed following the MacArthur Communicative Development Inventory (CDI) for infants and toddlers. It served to reveal any discrepancy between the child’s performance during the recorded sessions and his/her actual knowledge of the nouns. The six to nine-month-stretch of data collection from these four children of different ages is representative of a one and a half-year range of development from 1;9 to 3;2. The age 1;9 is the youngest suitable age for such a study, as it is theoretically two months after the vocabulary spurt and four months before the two-word stage, in which we predict the very first appearance of a classifier, whether in an adult-like form or a phonologically reduced form. On the other hand, the age 3;2 is the oldest appropriate age for such a longitudinal study based on the finding in a pilot study 5 that children between 3;0 and 3;5 were 5 A pilot study was conducted two years prior to the actual data collection for the present study at a daycare in Ho Chi Minh City, Vietnam with 50 children between the ages 2;6 and 6;5 to find out how many and what types of classifiers Vietnamese children produce, and at what age they can produce which classifiers.

Vietnamese classifier phrases from the perspective of how children acquire them

97

able to produce numeral + classifier + noun sequences, thus indicating that they had acquired the syntactically adult-like full numeral classifier phrase. This 1.5year age range is therefore crucial to investigate longitudinally as it covers transitions from bare nouns to two-element and three-element noun phrases. It is during this time that children acquire the grammatical slot of the classifier, the general classifier, and a number of different types of classifier phrases, as reported in previous studies.

4.2.2 Cross-sectional The cross-sectional experimental elicited-production data was collected every day for a period of six weeks from 38 children between the ages 2;10 and 5;7 in a daycare center in Ho Chi Minh City. The 38 children were divided into three groups. More details can be seen in Table 5. Group #

Group name

Gender

Age range

I II III

Mầm (youngest) Chồi (mid) Lá (oldest)

6 girls, 5 boys 7 girls, 5 boys 8 girls, 7 boys

2;10–3;7 3;8–4;4 4;7–5;7

Table 5: The children in the cross-sectional study

The experimental design consisted of two tasks. Each task was conducted with each child individually. The tasks were administered on different days to avoid getting false results due to overtesting and fatigue. The aim of these production tasks was to elicit the obligatoriness of classifiers in constructions with numerals and the use of classifiers in clauses with demonstratives. The children’s responses were audio-recorded. Prior to the start of the first elicitation task, each participating child was asked to count a set of eight objects to determine whether the child has mastered the concept of numbers. 114 stimuli were used, such as pictures, toys, and real objects. Task I elicits the obligatory classifier in numeral constructions: The child is shown pictures of multiple objects and entities on picture cards and asked to say how many of the same objects s/he sees. The quantity of the objects ranges from 2 to 10. There are 58 different pictures and objects, which attempt to elicit 17 different classifiers. The targeted syntactic structure is numeral + classifier + noun. Task II elicits a classifier phrase with a demonstrative: The child is shown two items that differ only in one feature, for example, two plastic apples – one red and the other green. The child is asked to describe how the two apples are different. There are 43 different pictures and objects. The targeted syntactic structure is classifier + (noun) +

98

Jennie Tran

demonstrative. The test items used were the same for both tasks, except that task I showed more than two of the same object, whereas task II showed only two of the same object. For all types of data collection described above, both familiar and unfamiliar objects and entities were used in a 1:1 ratio. For the longitudinal data, the familiar items are those listed by the parents in the noun inventory questionnaire. For the cross-sectional data, the familiar items are those available at the daycare center. The unfamiliar items are those brought along by the investigator.

4.3 Transcription and coding All children’s speech was transcribed and coded following the CHAT format (MacWhinney 2000), the common transcription format for child language. Each adult and child utterance in all 56 hours of longitudinal data was transcribed with a %mor tier and English gloss. Each single child utterance was coded as spontaneous, imitation, or repetition. Besides the conventional coding for syntactic categories, this study also contains classifier-specific coding and error coding. The analysis was run using the CLAN program (Computerized Language Analysis).

5 Results Results are primarily derived from the longitudinal data. In the cross-sectional data, only two structures, numeral + classifier + noun and classifier + (noun) + demonstrative, were examined, whereas in the longitudinal data, any classifier phrase produced by any child was extracted.

5.1 General versus specific classifiers The general classifier in Vietnamese is cái. It is used to classify a large group of nouns, covering all kinds of concrete, immobile objects, including furniture, most household items, kitchenware, appliances, tools, clothing items, body parts, very small insects or living things, and abstract nouns. The occurrence of the general classifier is widespread, whereas that of specific classifiers is restricted. The reason for this is that the general classifier also serves as a default classifier. Speakers resort to this classifier when they choose not to use or cannot think of the specific classifier. On the other hand, a (semantically)

Vietnamese classifier phrases from the perspective of how children acquire them

99

specific classifier can be applied only with a particular group of nouns whose referents share some specific semantic features and common characteristics associated with the classifier. Using another specific or the general classifier inappropriately with a noun that requires its own semantically appropriate classifier(s) would lead to an unacceptable, ungrammatical overgeneralization on the grounds of semantics and/or pragmatics. Here is a list of the classifiers that occurred in the children’s speech on a frequent basis, in the order of frequency, beginning with those that occurred more often and ending with those that occurred less often. General classifier: cái general classifier, inamimate Specific classifiers: con animal, animate quả/trái 6 fruit, big-round chiếc secondary general classifier, vehicle, one in a pair cây long-straight-rigid củ root vegetables cục small-roundish, undefined đôi pair sợi long-thin-malleable viên small-round bài lesson; song; speech ngôi house, building; temple; grave chùm bunch, bundle ngón finger, toe hột small-roundish, seed-like

5.2 Findings Because the focus of the present chapter is on the syntactic development of classifier phrases, the analysis of the classifier phrases listed in Table 6 was geared toward only the structure of the classifier phrase, not the semantics of the classifier. This means that the results for each classifier phrase type include both those produced with the general and those produced with the specific classifiers.

6 Both quả and trái are the classifiers for fruits and big, round shapes. Quả is used in the Northern dialect and trái in the Southern dialect.

100

Jennie Tran

#

1

Minh 1;09–2;03 (6 months)

Hà Mi 1;11–2;05 (6 months)

Liêm 2;04–2;10 (6 months)

Giang 2;05–3;2 (9 months)

CL-N

762 (82%) 95 (10%) 20 (2%) 2 (0.2%)

584 (48%) 285 (24%) 74 (6%) 4 (0.3%)

1638 (59%) 238 (9%) 337 (12%) 15 (1%) 1 (0.04%) 92 (3%)

64 (5%) 14 (1%) 26 (2%) 32 (3%) 11 (1%) 6 (0.5%)

3 (0.1%) 9 (0.3%) 31 (1%) 43 (2%) 287 (10%) 22 (1%) 2 (0.2%) 2 (0.2%) 1 (0.04%)

1 (0.1%) 3 (0.2%) 2 (0.2%)

1 (0.04%)

2

CL-Dem

3

CL-Wh

4

CL-Adj

5

CL-Poss

6

Num-CL

5 (1%)

591 (77%) 64 (8%) 41 (5%) 4 (1%) 3 (0.4%) 11 (1%)

CL-N-Dem

8 (1%)

6 (1%)

8 (2%) 8 (1%) 3 (0.3%)

1 (0.1%) 30 (4%) 3 (0.4%) 1 (0.1%)

7

2-element

CL Phrase Type

3-element

8

CL-N-Wh

9

CL-N-Adj

10

CL-N-Poss

11

Num-CL-N

12

Num-CL-Dem

13

Num-CL-Wh

14

Num-CL-Adj

15

Num-CL-Poss

16

4-element

Num-CL-N-Dem

18

Num-CL-N-Adj

19

Num-CL-N-Poss

1

Errors

3 (0.4%)

CL-N-Adj-Poss

17

32 (3%)

7 (0.3%) 2 (0.2%)

*Num-N

1

3

29

65

TOTAL

925

770

1206

2799

Table 6: Total number of tokens of classifier phrases produced by each child in all sessions, including impermissible NPs (errors)

Vietnamese classifier phrases from the perspective of how children acquire them

101

Table 6 shows that, in the collected longitudinal data of the four children, classifiers occurred in 19 grammatical and three ungrammatical syntactic positions. Of the 19 grammatical syntactic positions, 11 were produced by all four children; the other eight were produced only by the two older children (2;4– 3;2). All four children made ungrammatical syntactic constructions (errors). (See Appendix A for explication of each child’s errors.) The most frequent position in which the classifier occurs is the one preceding the noun, i.e., classifier + noun (CL-N). All four children overwhelmingly prefer this pre-noun position, with usage rates ranging from 59% of all classifier constructions for the oldest child to 82% for the youngest child. The second most frequent position for the classifier is just before the determiner, i.e., classifier + demonstrative (CL-Dem). Compared to the pre-noun position, however, this construction occurs much less frequently, ranging only from 8% to 24% of total usage. The third most frequent position is that preceding the wh-word, i.e., classifier + wh-word (CL-Wh), which also occurs much less often than the pre-noun position at only 2% to 12%. Grammatical omission of the head noun precedes the mastery of a full classifier phrase. Compare #1–6 with #7–19 in Table 6. The classifier phrases with determiners and wh-words that display head noun omission, CL-Dem and CL-Wh, are more frequent, and also first acquired by all children. The fullfledged three-element classifier phrases with a determiner or wh-word are produced much less frequently than the two-element classifier phrases with the determiner or wh-word. Vietnamese children combine the classifier with a demonstrative before they combine it with a number. Compare #1 and #6 in Table 6. This pattern also prevails in three-element combinations. The three younger children prefer to combine the demonstrative with CL-N rather than with Num-CL. Compare #7 and #12 in Table 6. Vietnamese children tend to combine numbers with a classifier alone before they combine them with the classifier + head noun, as can be seen in Figure 1. Numeral + classifier emerges at around 2;01, whereas numeral + classifier + noun emerges more than half a year later, at around 2;09. An important finding is the order of acquisition of the two-element classifier phrases as it can be compared to previous findings in other numeral classifier languages. This study’s findings show that Vietnamese children produce classifier + noun first, classifier + demonstrative second, and numeral + classifier third, as shown in (7a)–(7c).

102

Jennie Tran

Figure 1: Age of acquisition of Numeral-Classifier and Numeral-Classifier-Noun

(7a)

cái ly CL glass ‘the glass’

(7b)

cái này CL this ‘this (one)’

(7c) hai cái two CL ‘two’

5.3 Number of classifier phrase types Information on how many classifier phrase types each of the children could produce over the duration of the study is provided in Table 7.

103

Vietnamese classifier phrases from the perspective of how children acquire them

CL phrase types

Minh 1;09–2;03

Mi 1;11–2;05

Liêm 2;04–2;10

Giang 2;05–3;02

Possible number of CLP types

9

12

16

18

22

Table 7: Number of CL phrase types produced by each child

The children all display diversity of structure and a neat developmental trajectory. The number of classifier phrase types increases over time: Each older child produces a greater number of classifier phrase types than the next youngest child. Minh, the youngest child, produces nine; Hà Mi, the next youngest, produces 12; Liêm, the second oldest child, produces 16; and Giang, the oldest child, produces 18 different types of classifier phrases. More three-element classifier phrases and the more complex four-element ones appear in the speech of the two older children. The eight most frequently occurring noun phrase and classifier phrase types, arranged in rough order of frequency across all four children, are shown in Table 8. NP Type

Minh 1;09–2;03

Hà Mi 1;11–2;05

Liêm 2;04–2;10

Giang 2;05–3;02

N CL-N CL-Dem CL-Wh N-Poss N-Dem N-Adj Num-CL

1751 762 95 20 71 66 10 5

857 591 64 41 124 19 12 11

1212 584 285 74 103 84 48 32

2139 1638 238 337 232 28 49 92

Table 8: The eight most frequent NP types in tokens

Based on the frequency order presented in Table 8, the most often-used NP types are the one- and two-element ones, whether or not the noun phrase contains a classifier.

5.4 Child data examples Selected actual child utterances from the data illustrating the children’s production of some of the classifier phrases listed in Table 6 are presented below. The target utterances are underlined. Note that due to lack of space, the morphological gloss and the English gloss was included only for the target utterances

104

Jennie Tran

produced by the child; for the other utterances, only one of the glosses was included. (Key of tagging code: sfp = sentence final particle, EMP = emphatic particle, IMP = imperative marker). CL-N Minh and auntie Tam were conversing. Suddenly, Minh pointed to an eraser and said: *CHI:

cái CL:general

cục CL:small roundish

cái cục. CL:small roundish

*TAM: cục gì? CL:small roundish what? *CHI:

cục gôm (Minh, Session 2, Age: 1;10) CL eraser

CL-Adj Giang and her aunt were looking at picture books. Giang pointed at a picture and asked: *CHI:

a cái gì vậy Cô_Linh? ‘Oh what’s that, auntie Linh?’

*AUN: đố Giang? ‘I ask you.’ *CHI: cái . . . cái a umbrella nè. CL . . . CL oh umbrella sfp *AUN: giỏi. ‘Good.’ *CHI: cái dù umbrella nè. CL umbrella umbrella sfp *CHI: cái nhỏ nè. cái to nè. (Giang, Session 16, Age: 3;0) CL small sfp. CL big sfp. CL-Poss Hà Mi and her Mom were looking at pictures on a poster. Her Mom asked her to name the different things she saw on the poster.

Vietnamese classifier phrases from the perspective of how children acquire them

*MOT: cái này? ‘This?’ *CHI: cái của Mi. (Hà Mi, Session 5, Age: 2;1) CL of Mi *MOT: hử ? huh? *CHI: cái của Mi nè. CL of Mi sfp *MOT: à, cái của Mi hả? oh, CL of Mi sfp ‘Oh, it’s yours?’ CL-N-Dem Liêm and his Mom were playing with cars. *CHI: cái xe này bể rồi. (Liêm, Session 1, Age: 2;4) CL car this break PERF *MOT: à, xe này bể rồi. ‘Oh yes, this car is broken.’ *CHI: xe mới đi. car new IMP *MOT: xe mới đi hả? ‘Oh you want a new car?’ CL-N-Wh Liêm and his Mom were playing with cars. *CHI: con chỉ mẹ. con chơi tiếp. ‘I show you. I continue playing.’ *CHI: con chỉ mẹ chiếc xe nào nha. (Liêm, Session 7, Age: 2;7) I show you CL car which sfp ‘I show you which car.’ *MOT: rồi con chỉ mẹ đi. ‘Okay show me.’

105

106

Jennie Tran

CL-N-Adj *CHI: tập đi ba. notebook IMP Daddy *CHI: cái tập đen. (Minh, Session 6, Age: 2;0) CL notebook black *DAD: rồi ra đây, nè. ‘Ok, come out, here.’ *CHI: đâu rồi ba? tập đâu rồi ba? tập đâu rồi? ‘Where, Daddy? Where’s the notebook, Daddy? Where’s the notebook?’ *DAD: đây nè. ‘Here.’ Num-CL-N Minh was playing tearing pictures out of a picture book with Dad and Mom. When he found the picture with two balls, he tore it out. He called this game ‘cutting’. *CHI: ba cắt cho. trái banh đâu mẹ? trái banh đâu? Daddy cut sfp. CL ball where Mommy? CL ball where? *MOT: con kiếm đi. ‘Look for it.’ *CHI: trái banh đâu mẹ? CL ball where Mommy *MOT: hả? huh? *CHI: hai trái banh. (Minh, Session 6, Age: 2;0) two CL ball ‘Two balls.’ *CHI: cắt. giụt. xong rồi. ‘Cut. Throw away. Finished.’ Num-CL-Dem Liêm and his Dad were sitting at the table eating lunch. Liêm asked: đây? (Liêm, Session 8, Age: 2;8) *CHI: mấy cái này gì PL CL this what here ‘What are these?’

Vietnamese classifier phrases from the perspective of how children acquire them

107

*DAD: rau. ‘Greens’ *CHI: rau để ăn hả? ‘Greens for eating?’ *DAD: ừm. ‘Yes.’ *CHI: bỏ vô miệng ăn ba nha. put in mouth eat Daddy sfp *DAD: ừm, để đó ba lấy ba ăn. ‘Ok, leave it there, I’ll take and eat it.’ Num-CL-Wh Giang and her aunt were looking at picture books. Auntie pointed at a picture. *AUN: Cô Linh chỉ nè. ‘I point here.’ *CHI: hai cái gì vậy cô Linh? (Giang, Session 16, Age: 3;0) two CL what sfp auntie Linh ‘What are those two?’ *AUN: Hư? ‘Huh?’ *CHI: chiếc thuyền hả? CL boat sfp ‘Is it a boat?’ *AUN: đây mà cái thuyền a hả? ‘What? This is a boat?’ Num-CL-Adj Auntie asked Giang to tell what she saw in the picture. *AUN: Cô Linh chỉ cái hình này là Giang nói nè. Gì đây nè? Đây nè. ‘I show you this picture then you will tell me. What is this? This.’ *CHI: Cô Linh ơi nhiều quả to Cô Linh ơi nhiều quả. auntie Linh ex many CL big auntie Linh ex many CL ‘Auntie Linh, there are many big ones, auntie Linh, many.’ (referring to balloons) (Giang, Session 15, Age: 3;0)

108

Jennie Tran

5.5 Emergence order This study employs Wong’s (1998) criteria for emergence order: If the child uses a certain type of structure continuously across three consecutive sessions, the child is deemed able to use that structure productively and the first session of that series is regarded as the onset of this development. Figure 2 below illustrates this emergence order.

Figure 2: Emergence Order

Vietnamese classifier phrases from the perspective of how children acquire them

109

The classifier phrase that emerges first is CL-N. Evidence for this is provided by Minh, the youngest child, who could produce such phrases already at 1;9; additionally, CL-N phrases appear in the other three children’s speech from the time of the first recording. The two-element non-numeral possessive construction, N-Poss, is acquired early as well, before the age of 1;9. Two-element noun phrases also emerge early at around 1;10. Children begin using combinations of the classifier (or the noun) with the determiner and with the interrogative (CLDem, N-Dem, CL-Wh) between 1;11 and 2;1. Children build their first non-numeral three-element classifier phrases at around age 2 by adding a third element to the CL-N sequence. These are CLN-Dem and CL-N-Poss. The general pattern of expanding a two-element to a three-element classifier phrase involves adding an element to the fixed CL-N sequence. CL-N + Dem Poss Adj CL-N-Poss is the first three-element NP that emerges, at around 2;0. CL-NDem emerges at around the same time. Before 2;0, there is no productive use of three-element phrases. Results from Task II of the cross-sectional data reconfirm this development as they show a 100% correct production of the CL-N-Dem construction in the children aged 2;10 to 5;5. The first numeral classifier phrase that children build consists of two elements, namely Num-CL. This appears later than non-numeral classifier phrases, at around age 2;2. The first numeral three-element classifier phrases are built by adding a third element to the numeral-classifier sequence. These are Num-CL-N, Num-CL-Dem, Num-CL-Poss. They appear later, at around age 2;8. Num-CL + N Dem Poss Classifier omission errors (*Num-N) also appear at around age 2;8. CL-N-Adj emerges between 2;7 and 2;8, about the same time as Num-CL-N. CL-N-Adj emerges earlier than CL-Adj. This is because CL-Adj occurs rarely in adult speech. Shortly after the three-element classifier phrases have developed, at around 2;9, the first four-element classifier phrases emerge, including both nonnumeral and numeral constructions.

110

Jennie Tran

Other constructions that occurred in small numbers are the three-element CL-N-Wh, the four-element CL-N-Adj-Poss, Num-CL-N-Dem, Num-CL-N-Adj, and Num-CL-N-Poss. Only the two older children could produce these more sophisticated four-element classifier phrases; they emerged between ages 2;8 and 3;2.

5.6 Classifier omission errors Classifier omission errors (*Num-N) involve the ungrammatical omission of the obligatory classifier in a classifier structure containing a numeral (Num-CL-N). Here are some actual examples from the longitudinal data. (8a)

*CHI: ba lấy cho Minh một bóng đi. (Minh 2;3) you get for Minh one ball IMP ‘Daddy, get a ball for me.’

(8b) *CHI: có hai chổi. (Hà Mi 2;2) have two broom ‘There are two brooms.’ (8c)

*CHI: hai mèo luôn. (Hà Mi 2;3) two cat EMP ‘There are two cats!’

(8d)

*CHI: hai banh đây. (Liêm 2;9) two ball here ‘Here are two balls.’

(8e) *CHI: một tất nè. (Giang 3;1) one sock here ‘Here is one sock.’ (8f)

*CHI: hai chuối. (Ngoc Tien 3;4, cross-sectional) two banana ‘Two bananas.’

(8g)

*CHI: hai kẹo. (Gia Thinh 3;9, cross-sectional) two candy ‘Two candies.’

In all of these ungrammatical utterances (*Num-N), the obligatory classifier is missing. In (8a), (8d) and (8f), the classifier for fruits and big/round objects,

Vietnamese classifier phrases from the perspective of how children acquire them

111

trái or quả, is missing. In (8b), the classifier cây for long, thin objects is missing. In (8c), the classifier for animals, con, is missing. In (8e), the classifier chiếc for one of a pair is missing. In (8g), the classifier viên for small, roundish objects is missing. The most classifier omission errors occurred in the speech of the two older children, at around 2;8. The two younger children produced very few classifier structures involving a numeral (Num-CL-N) (only four tokens in the case of Minh and six tokens in the case of Mi) and almost half of them were classifier omission errors (*Num-N). Table 9 shows a comparison among the main numeral classifier phrases: numeral + classifier (Num-CL), numeral + classifier + noun (Num + CL + N), and the erroneous *numeral + noun (*Num + N). This illustrates which of these constructions emerges first and at what ages the children make classifier omission errors. It can be inferred from this comparison table that the first numeral classifier phrase to emerge in the speech of Vietnamese children is the combination of the numeral + classifier (Num-CL), at around 2;2. The three-element numeral + classifier + noun (Num-CL-N) starts to be used productively at around 2;8. The period before that, between 2;2 and 2;8, is the period of trial and error: Children produce the correct forms while also making classifier omission errors. These errors do not stop when they learn to use the three-element numeral + classifier + noun productively. Rather, such errors continue to occur concurrently with the correct forms, but to a lesser extent, as can be seen in the data from the oldest child, Giang. The rates of classifier omission errors are shown in Table 10. The two younger children, Minh and Hà Mi (1;9–2;5) displayed classifier omission error rates of 25% and 50%. Surprisingly, the third child, Liêm (2;4– 2;10), had a high error rate of 73%. The oldest child, Giang (2;5–3;2) displayed an omission rate of 19%. Despite this inconsistent progression, however, the general trend is for a positive correlation between age and frequency of use of numeral classifier constructions. A child like Liêm with a high rate of omission is passing through the period of trial and error. The results from Task I of the cross-sectional data further show that after around 3;2, classifier omission errors start to decrease. Table 11 illustrates the cross-sectional omission rates. The youngest group (2;10–3;7) had an error rate of 17.5%, the mid group (3;8–4;4), 13%, and the oldest group (4;7–5;7), 4%. The youngest group of the cross-sectional study (2;10–3;7) displayed on average a slightly lower error rate (17.5%) than the oldest child (2;5–3;2) in the longitudinal study (19%). The period in which children’s error rates begin to decrease can therefore be roughly estimated to fall between 3;2 and 3;7.

112

Jennie Tran

Minh

Hà Mi

Session

Age

1 2 3 4 5 6 7 8 9 10 11 12 13

1;09;20 1;10;06 1;10;20 1;11;04 1;11;19 2;00;03 2;00;16 2;01;01 2;01;15 2;01;29 2;02;13 2;02;27 2;03;11 TOTAL

NumCL

NumCL-N

*NumN

1

1 3

1

1

1

1

5

3

1

Liêm

Session

Age

1 2 3 4 5 6 7 8 9 10 11 12

1;11;10 1;11;27 2;00;11 2;00;29 2;01;27 2;02;11 2;02;27 2;03;16 2;04;00 2;04;14 2;04;28 2;05;12

NumCL

4 2 2

NumCL-N

3

*NumN

1 1 1

3

TOTAL

11

3

3

Session

Age

NumCL

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

2;05;01 2;05;18 2;06;10 2;06;23 2;07;08 2;07;22 2;08;01 2;08;10 2;09;01 2;09;18 2;10;00 2;10;14 2;11;00 2;11;17 3;00;03 3;00;17 3;01;00 3;01;14 3;01;27 3;02;13

1 4 2 5 5 4 5 18 6 2 12 18 3

9 7 7 10 24 44 95 42 39

2 7 6 3 2 5 7 15 10 6

TOTAL

92

287

67

Giang

Session

Age

1 2 3 4 5 6 7 8 9 10 11

2;04;09 2;04;19 2;05;22 2;06;09 2;06;20 2;07;07 2;07;22 2;08;11 2;08;25 2;09;08 2;09;29 TOTAL

NumCL 1 5 4 1 4 5 5 6 1 32

NumCL-N

*NumN

1 1 1

1 3 2 1 3 11

3

5 20 30

Table 9: Comparing the main numeral classifier phrases

NumCL-N

*NumN

1 1 6

2 3 4 2 1

1

Vietnamese classifier phrases from the perspective of how children acquire them

113

Child

Age

total CL omission tokens

total expected Num-CL-N tokens

total % classifier omission

Minh Hà Mi Liêm Giang

1;9–2;3 1;11–2;5 2;4–2;10 2;5–3;2

1 3 30 67

4 6 41 354

25.0% 50.0% 73.0% 19.0%

Table 10: Classifier omission errors in longitudinal study

Age

# of subjects

total CL omission tokens

total elicited tokens

total % classifier omission

2;10–3;7 3;8–4;4 4;7–5;7

n = 11 n = 12 n = 15

109 89 34

623 687 847

17.5% 13.0% 4.0%

Table 11: Classifier omission errors by children in cross-sectional study

5.7 General classifier as a placeholder To find out whether young Vietnamese children use the general classifier in the classifier slot as a placeholder, it is necessary to examine data of the two younger children in the longitudinal study. This is because their speech shows which classifiers children use to build their first classifier phrases. The results reveal that these two youngest Vietnamese children do use the general classifier as a placeholder to build their first two-element classifier phrases, as shown in Table 12. For a detailed chart, refer to Appendix B. Child Age

CL phrase

general CL tokens

Minh 1;9–2;3

CL-Dem CL-Wh Num-CL Total

93 21 5 119

Child Age

CL phrase

general CL tokens

specific CL tokens

Hà Mi 1;11–2;5

CL-Dem CL-Wh Num-CL Total

63 32 12 107

4 11 9 24

Table 12: Use of the general classifier by the two youngest children

specific CL tokens 4 3 1 8

114

Jennie Tran

As can be seen in this table, the use of the general classifier far exceeds the use of specific classifiers in these two children’s early two-element classifier constructions, classifier + demonstrative (CL-Dem), classifier + wh-word (CL-Wh), and numeral + classifier (Num-CL). The results in Appendix B also show that very young children before the age of two and a half can mainly produce only four specific classifiers, which are con, trái, chiếc, and cây, out of the very large inventory of about 200 Vietnamese classifiers. These same classifiers also occurred with the highest rate of frequency in the speech of the children’s parents or caretakers. A detailed frequency analysis of individual classifiers in the input was conducted but not provided here due to lack of space.

6 Discussion 6.1 Observance of syntactic properties Vietnamese children acquire non-numeral classifier phrases early and make no errors with the structure of these phrases. At the time of their emergence (between 1;9 or earlier and 2;8), all two-to-three element classifier phrases are well-formed syntactically when used with the noun, demonstrative, wh-word, possessive, and adjective. When the classifier phrase with a numeral first emerges (at around 2;2), children can produce well-formed two-element numeral + classifier phrases. They observe two of the three syntactic properties in the construction of the Vietnamese classifier phrase: (1) they use the classifier anaphorically and (2) they position the numeral and classifier adjacent to each other. However, children have difficulty observing the third syntactic property, which mandates obligatory classifiers in the presence of numerals. For this reason, they make classifier omission errors when producing the three-element numeral + classifier + noun phrase. Other three-element classifier phrases without nouns, numeral + classifier + demonstrative and numeral + classifier + possessive, which emerge in limited usage at around the same time, are well-formed. Results from the cross-sectional data show that classifier omission errors decrease only after age 3;2, and begin amore rapid decline after 4;7. Young Vietnamese children cannot yet use ordinal numbers and the quantifiers ‘every’ or ‘a few’ with the classifier.

6.2 Order of emergence Most non-numeral two-element noun phrases, except for classifier + adjective, are acquired before 2;0. Between 2;0 and 3;2, five of the nine possible threeelement classifier phrases are used productively. These results are quite similar

Vietnamese classifier phrases from the perspective of how children acquire them

115

to Wong’s (1998). Vietnamese children tend to use classifier phrases in more diverse ways after the age of two and a half. The first noun phrase structure involving a numeral that emerges is numeral + classifier, which appears at around age 2;2. The three-element numeral + classifier + noun appears alongside classifier omission errors (*numeral + noun) at age 2;8. Between 2;8 and 3;2, children produce both grammatical and ungrammatical numeral classifier structures. During this same period of time, more three-element and the more complex four-element classifier phrases appear. Not until late in their third year do children produce grammatical numeral classifier phrases.

6.3 Comparisons with previous studies These results show that Vietnamese children demonstrate early knowledge of the classifier slot in noun phrases in non-numeral contexts. They know early on that the slot for the classifier precedes the noun. They do not make any word order errors with classifier phrases. As early as age 1;9, they can produce an obligatory classifier, not only with a noun, but also with a demonstrative and an interrogative. This is consistent with results of previous studies in other Asian languages investigating the development of classifier phrases (Erbaugh 1982, Hu 1993 on Mandarin, Carpenter 1987 on Thai, Wong 1998 on Cantonese). As for Vietnamese children’s use of the ‘blank attempt’ strategy, there is no clear evidence in the current study. A more in-depth phonetic-prosodic examination of the data may reveal that children can rely on phonological or tonal cues to identify the syntactic position of classifiers. As for the ‘general classifier as placeholder’ strategy, clear supporting evidence is found in the current study. Vietnamese children use the general classifier to fill the grammatical position of the classifier. A previous classifier acquisition study by Erbaugh (1982) on Mandarin found that classifiers occur with both a demonstrative and a number first. The results of the present study on Vietnamese show that classifiers occur with a demonstrative before they occur with a number. (Refer to Table 6 and Figure 2.) Erbaugh further found that classifiers occur with a head noun rather than without. By contrast, a previous classifier acquisition study by Wong (1998) on Cantonese found that grammatical omission of the head noun precedes the mastery of the full classifier phrase. Wong’s finding is consistent with the Vietnamese data. When constructing a classifier phrase, Vietnamese children produce the classifier without a head noun rather than with it. The classifier phrases without the head noun, classifier + possessive, classifier + demonstrative, classifier + wh-

116

Jennie Tran

word and numeral + classifier, do precede their full-fledged counterparts in the Vietnamese children’s speech. (Refer to Figure 2.) Wong’s study further found that children tend to combine the classifier with a number before they combine it with a head noun. She explained that, among two-element noun phrase structures, children produce demonstrative + classifier first, then numeral + classifier, then classifier + noun. The results of the present study on Vietnamese differ from Wong’s findings. The emergence order of two-element noun phrase structures for Vietnamese children is classifier + noun first, classifier + demonstrative second, and numeral + classifier third. (Refer to Figure 2 and examples (7a–c).) The present results clearly show that very young Vietnamese children employ a very high number of classifier-noun constructions. The youngest child alone produced 58 tokens at the first session at age 1;09. (Refer to Appendix A.) The children in this study either learned all nouns together with their appropriate classifiers as noun chunks, or they learned only certain nouns with the classifier, and only used these nouns in their speech in their beginning sessions. In order to determine this, an in-depth analysis of the nouns used by each child in each session is necessary. Wong (1998) further found that at around age 2;6, children’s speech begins to exhibit three-element noun phrase structures, starting with demonstrative + classifier + noun and then growing to include numeral + classifier + noun. This is similar to the emergence order in Vietnamese. Classifier + noun + demonstrative is acquired first at around 2;0 and numeral + classifier + noun develops later at around 2;8. (Refer to Figure 2.) Previous findings have shown that children perform better when using nonnumeral constructions than when using numeral constructions. The results of the present study on Vietnamese fully support this finding. However, previous studies did not report any significant rate of classifier omission errors. These previous studies report that by age three, children make very few structural errors such as omission of the classifier from numerical constructions (below 0.6%). On the contrary, the Vietnamese two, three and four year-old children in this study made a higher percentage of omission errors than those found in previous studies in other languages. (Refer to Tables 10 and 11.) I speculate that this is because the semantic knowledge that children would need to select the appropriate classifier for each noun is not well developed; thus, young children prefer to leave out the classifier. In other words, something in the system of Vietnamese classifier phrases makes children be more conservative in their acquisition of the classifier system of this language than child speakers of other languages with similar classifier systems.

Vietnamese classifier phrases from the perspective of how children acquire them

117

Vietnamese children choose to omit the obligatory classifier rather than deliver the wrong classifier; they prefer to make mistakes in syntax for the sake of avoiding mistakes in semantics.

References Carpenter, Kathie Lou. 1987. How children learn to classify nouns in Thai. Ph.D. dissertation, Stanford University. Carpenter, Kathie Lou. 1991. Later rather than sooner: extralinguistic categories in the acquisition of Thai classifiers. Journal of Child Language 18: 93–113. Emeneau Murray. B. 1951. Studies in Vietnamese (Annamese) Grammar. Berkeley, Los Angeles: University of California Publications in Linguistics. Erbaugh, Mary. 1982. Coming to order: Natural selection and the origin of syntax in the Mandarin speaking children. Ph.D. dissertation, University of California at Berkeley. Erbaugh, Mary. 1986. Taking stock: The development of Chinese noun classifiers historically and in young children. In: C. G. Craig (ed.), Noun Classes and Categorization, 399– 436. Amsterdam/Philadelphia: John Benjamins. Fang, Fuxi. 1985. An experiment of the use of classifier by 4 to 6 year olds. Acta Psychologica Sinica 17(4): 384–392. Hsu, Joseph H. 1987. A study of the various stages of development and acquisition of Mandarin Chinese by children in Chinese milieu. National Science Council Research Report, College of Foreign Languages, Fu Jen Catholic University. Hu, Qian. 1993. The acquisition of classifiers by young Mandarin-speaking children. Ph.D. dissertation, Boston University. Gandour, Jack, Soranee Holasuit Petty, Rochana Dardarananda, Sumalee Dechongkit and Sunee Mukngoen. 1984. The acquisition of numeral classifiers in Thai. Linguistics 22: 455–479. Goral, Donald R. 1978. Numeral Classifier Systems: A Southeast Asian cross-linguistic analysis. Linguistics of the Tibeto-Burman Area 4: 1–72. Hundius, Harald and Ulrike Kölver. 1983. Syntax and semantics of numeral classifiers in Thai. Studies in Language 7: 165–214. Lee, Kwee-Ock. 1994. Acquisition of Korean classifiers: Syntactic and semantic factors. Paper presented at the Symposium on Linguistics Theory and the Acquisition of Korean Semantics and Syntax. Lee, Kwee-Ock and Sun-Yong Lee. 2005. The acquisition of Korean numeral classifiers. Unpublished manuscript. Kyungsung University and Kyung Hee University. Loke, Kit Ken. 1991. A semantic analysis of young children’s use of Mandarin shape classifiers. In: A. Kwan-Terry (ed.), Child Language Development in Singapore and Malaysia, 98–116. Singapore: Singapore University Press. Loke, Kit Ken and G. Harrison. 1986. Young children’s use of Chinese (Cantonese and Mandarin) sortal classifiers. In: Linguistics, Psychology, and the Chinese Language. Hong Kong: Centre of Asian Studies, University of Hong Kong. MacWhinney, Brian. 2000. The CHILDES Project: Tools for Analyzing Talk. 3rd edition. Mahwah, NJ: Lawrence Erlbaum Associates. Mak, David Lai-Woon. 1991. The acquisition of classifiers in Cantonese. Ph.D. dissertation, University of Reading.

118

Jennie Tran

Matsumoto, Yo. 1985a. Japanese numeral classifiers: Their structure and acquisition. M.A. Thesis, Sophia University. Matsumoto, Yo. 1985b. Acquisition of some Japanese numeral classifiers: The search for convention. Stanford University Papers and Reports in Child Language Development 24: 89–96. Matsumoto, Yo. 1987. Order of acquisition in the lexicon: Implications from Japanese numeral classifiers. In: K. Nelson and A. van Kleck (eds.), Children’s Language, Volume 6, 229–60. Hillsdale, NJ: Lawrence Erlbaum Associates. Muraishi, Shozo. 1983. Josuushi tesuto [classifier tests]. In: [The National Language Research Institute] (ed.), [Conceptual Development and Language in Children]. Tokyo Shoseki: Tokyo. Ng, Bee Chin. 1991. Word meaning acquisition and numeral classifiers. La Trobe University Working Papers in Linguistics 4: 73–83. Nguyễn, Đinh Hoa. 1957. Classifiers in Vietnamese. Word 13: 124–52. Piriyawiboon, Nattaya. 2009. The role of classifiers in N + Dem in Thai. Paper presented the 19th annual meeting of the Southeast Asian Linguistics Society (SEALS). Poon, Emma Yuen-wai. 1980. Some aspects of the ontological development of nominal classifiers in Cantonese. M.A. Thesis, University of Hong Kong. Salehuddin, Khazriyati and Heather Winskel. 2007. The role of input and cognitive development in Malay numeral classifier acquisition. Paper presented at the 15th Biennial Conference of the Australian Human Development Association. Sanches, Mary. 1977. Language acquisition and language change: Japanese numeral classifiers. In: B. Blount and M. Chanse (eds.), Sociocultural Dimensions of Language Change. New York: Academic Press. Szeto, Kitty K. 1996. Classifiers in Cantonese-speaking children from 1 year 5 months to 3 years 8 months. In: T. Lee, C. Wong, S. Leung, P. Man, A. Cheung, K. Szeto and C. Wong (eds.), The Development of Grammatical Competence in Cantonese-speaking Children – Report of a project funded by RGC earmarked grant CUHK 1991–94, 175–202. Hong Kong. Tse, John Kwok-ping, Tang Ting-chi, Shie Yu-huei and Cherry Y. Li. 1991. Chinese children’s language acquisition and development. National Science Council Research Report, Department of English, National Taiwan Normal University. Uchida, Nobuko and Mutsumi Imai. 1996. A study on the acquisition of numeral classifiers among young children: The development of human-animal categories and generation of the rule of classifiers applying. Japanese Journal of Educational Psychology 44(2): 126– 135. Uchida, Nobuko and Mutsumi Imai. 1999. Heuristics in learning classifiers: The acquisition of the classifier system and its implications for the nature of lexical acquisition. Japanese Psychological Research 41(1): 50–69. Wong, Cathy Sin Ping. 1998. The acquisition of Cantonese noun phrases. Ph.D. dissertation, University of Hawaii. Yamamoto, Kasumi. 2000. The acquisition of Japanese numeral classifiers. Ph.D. dissertation, Departement?, Cornell University.

95

TOTAL

762

5 3 10 5 17 6 21 6 9 9

3 1

1;09;20 58 1;10;06 46 1;10;20 50 1;11;04 33 1;11;19 79 2;00;03 30 2;00;16 99 2;01;01 37 2;01;15 23 2;01;29 57 2;02;13 30 2;02;27 36 2;03;11 184

CL + Dem

1 2 3 4 5 6 7 8 9 10 11 12 13

CL + N

Age

20

2 1 3 11

1

1 1

CL + Wh

2

1

1

0

2 8

1

5

2 3

1

CL + N + Dem

1 3

CL + Num + Poss CL

CL + Adj

0

CL + N + Wh

2

8

2

3

1 1

1

CL + N + Adj

3

1

6

5

4

3-element classifier phrases

2-element classifier phrases

Classifier Phrase Type # 1 2 3

S

MINH

Classifier NPs

APPENDIX A

8

4

1 2

1

CL + N + Poss

4

MINH

3

1

1

1

Num + CL + N

5

7

8

9

1

2

3

4

0

0

0

0

1 1

1

1

6

3

1

2

7

2

1 1

1

2

926

65 48 53 38 85 43 105 58 34 90 39 48 219

Num + Num + Num + Num + CL + Dem CL + Wh CL + Adj CL + Poss *Num + N *CL *CL + CL *CL + V TOTAL

6

Errors/ungrammatical

64

591

TOTAL

41

4

1

4 8 1 2 2

12 6

1

1

4 9 10 12

1 1

3

2

1

11

3

4 2 2

CL + Num + Poss CL

CL + Adj

2 4 8 9

CL + Wh

2 3 5 1

1;11;10 33 1;11;27 82 2;00;11 82 2;00;29 102 2;01;27 18 2;02;11 36 2;02;27 88 2;03;16 70 2;04;00 27 2;04;14 13 2;04;28 12 2;05;12 28

CL + Dem

1 2 3 4 5 6 7 8 9 10 11 12

CL + N

Age

6

1

1 1

2 1

CL + N + Dem

0

CL + N + Wh

2

1

1

CL + N + Adj

3

1

6

5

Classifier Phrase Type # 1 2 3

4

3-element classifier phrases

30

3 1 3 2 2 3 1 4 6 5

CL + N + Poss

4

HÀ MI

2-element classifier phrases

S

Hà Mi

Classifier NPs

3

3

Num + CL + N

5

7

8

9

1

2

3

4

1

1 0

0

3

1 2

3

1

1 1

2

2

4

2

1 1

8

1

2 2 3

773

37 91 102 115 23 54 110 94 48 18 34 46

Num + Num + Num + Num + CL + Dem CL + Wh CL + Adj CL + Poss *Num + N *CL *CL + CL *CL + V TOTAL

6

Errors/ungrammatical

584

TOTAL

285

17 22 24 6 33 24 29 28 32 5 65

CL + Dem

74

4 5 2 2 15 3 6 20 9 8

CL + Wh

3

2

1

0

32

5 6 1

1 4 5

1 5 4

64

6 7 4 18 1 5

1 12 8 2

CL + Num CL + N + Poss CL Dem

CL + Adj

14

5

8

1

CL + N + Wh

4-element classifier phrases: 1 CL-N-Adj-Poss in Session 9 3 Num-CL-N-Dem in Session 9,10,11 2 Num-CL-N-Adj in Session 9,11

46 45 45 28 46 49 102 26 35 27 135

2;04;09 2;04;19 2;05;22 2;06;09 2;06;20 2;07;07 2;07;22 2;08;11 2;08;25 2;09;08 2;09;29

1 2 3 4 5 6 7 8 9 10 11

CL + N

Age

2

26

8 2 9

1 2

2 2

CL + N + Adj

3

1

6

5

Classifier Phrase Type # 1 2 3

4

3-element classifier phrases

2-element classifier phrases

S

Liêm

Classifier NPs

32

1

1 5 2 1 2 9 6 1 4

CL + N + Poss

4

LIÊM

11

2 1 3

1 3

1

Num + CL + N

5

7

8

9

6

2

2 1

1

0

0

0

Num + Num + Num + Num + CL + Dem CL + Wh CL + Adj CL + Poss

6

29

5 20

2

1 1

*Num + N

1

Errors/ungrammatical

7

1

1 2 1 1

1

*CL

2

4

51

37

3

2 1 1 6

1

15

7

1 2 2 1

2

1235

71 98 88 41 86 114 168 69 137 51 281

*CL + CL *CL + V TOTAL

3

1638

TOTAL

238

3 8 1 1 5 9 58 10 25 10 21 13 14 49 9

2

CL + Dem

337

2 1 3 16 5 3 13 8 13 22 19 33 10 27 10 47 15 55 8 27

CL + Wh

15

2 5

1 1 2

1 3

6

3

2

1

92

1 4 2 5 5 4 5 18 6 2 12 18 3

1 6

CL + Num + Poss CL

CL + Adj

3

1

1

1

CL + N + Dem

4-element classifier phrases: 1 CL-N-Adj-Poss in Session 9 7 Num-CL-N-Adj: 1 in Session 16, 2 in Session17, 4 in Session18 2 Num-CL-N-Poss in Session 18,19

64 13 29 40 17 37 101 118 109 122 46 87 55 46 39 142 55 181 84 253

2;05;01 2;05;18 2;06;10 2;06;23 2;07;08 2;07;22 2;08;01 2;08;10 2;09;01 2;09;18 2;10;00 2;10;14 2;11;00 2;11;17 3;00;03 3;00;17 3;01;00 3;01;14 3;01;27 3;02;13

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

CL + N

Age

9

5

2

1 1

CL + N + Wh

2

31

1 6 4 11

1

2

1 1 2 1 1

CL + N + Adj

3

1

6

5

Classifier Phrase Type # 1 2 3

4

3-element classifier phrases

43

3 3 6 1 4 6

4 2 2 5 1 4

1

1

CL + N + Poss

4

GIANG

2-element classifier phrases

S

GIANG

Classifier NPs

287

9 7 7 10 24 44 95 42 39

3 4 2 1

Num + CL + N

5

7

8

9

1

2

3

4

22

1

1 8 4 5 1 2

2

2

2

1

1

1

1

65

2 7 6 3 2 5 7 15 8 6

1

2

1

1

1

21

1

1

5 2 6 1 1

2

1

1

46

5

2

12 5 3 6 2 2 1 1 1

3 1

2

2864

70 15 37 58 32 44 144 142 137 168 85 214 96 127 95 259 132 374 216 349

Num + Num + Num + Num + CL + Dem CL + Wh CL + Adj CL + Poss *Num + N *CL *CL + CL *CL + V TOTAL

6

Errors/ungrammatical

Vietnamese classifier phrases from the perspective of how children acquire them

123

APPENDIX B Closer analysis of CL-Dem and CL-Wh (to find out which classifiers children use in CL-Dem and CL-Wh phrases) Child

CL-Dem

Minh

CL cái cái con chiếc cái cái cái con trái chiếc

Ha Mi

Dem này đó này này kia

at at at at at

1;9 or before 1;11 1;11 2:1 2;3

71 (occurs at each session) 2 (1;11), 7 (2;1), 12 (2;2) 1 (1;11), 1 (2;0), 1 (2;2) 1 (2;1) 1 (2;3) 26 total

này đó này này này

at 1;11 or before none at 1;11 at 2:0 at 2;4

63 (occurs at each session) 1 (1;11) 1 (2;0), 1 (2;2) 1 (2;4) 4 total number of occurrence

Wh gì gì gì

at 1;10 at 1;10 at 2;03

21 (more at 2;2, 2;3) 1 (1;10), 1 (1;11) 1 (2;03) 3 total

cái con

gì gì

1;11 or before at 1;11

trái trái cái

gì nào nào

at 2;00 at 2;00 at 2;00

29 (each session) 1 (1;11), 2 (2;0), 3 (2;4), 2 (2;5) 2 (2;00) 1 (2;00) 2 (2;00), 1 (2;2) 13 total

CL-Wh

Minh

CL cái con trái

Child

Num-CL

Minh

Num 1 1,2,3 2,3 1,2 2 1,2 2 2 3

Ha Mi

number of occurrence

time of first occurrence

Child

Ha Mi

time of first occurrence

time of first occurrence

number of occurrence

CL chiếc cái

2;1 2;2

3 (2;2), 2 (2;3)

cái trái cây chùm chiếc món con

2;2 2;2 2;2 2;2 2;3 2;3 2;5

12 2 1 3 1 1 1

III Clausal and verb phrase syntax

Nigel Duffield

5 Head-First: On the head-initiality of Vietnamese clauses* This paper provides a detailed analysis of complementizer-like elements on the left and right periphery of Vietnamese clauses, comparing these with their counterparts in other East Asian varieties, as well as in more familiar Western European languages. The focus of the study is on the derivation of Yes-No questions in Vietnamese, which – in contrast to constituent questions – exhibit a number of theoretically significant co-occurrence constraints, including a ban on negative, topicalized and future tense questions. Re-interpreting a more traditional, descriptive treatment (Nguyễn 1997), this paper explains these distributional constraints through a derivational analysis involving predicate-raising around sentencemedial functional heads (cf. Kayne 1994). Implications of this analysis for other apparent rightward heads in Vietnamese are also considered. Keywords: Complementizers; CP Structure; Negation; Tense-Aspect Interactions; Yes-No Questions; Topicalization; Head-Initial vs. Head-Final

1 Introduction This paper is concerned with the analysis of functional morphemes in Vietnamese that surface on the right edge of interrogative clauses, specifically, the negative morphemes không (‘not’) and chưa (‘not yet’) illustrated in (1a) and (1b), respectively.1,2 * I am very grateful to the editors and to two anonymous reviewers of an earlier draft of this paper for their constructive and insightful comments, especially to the first reviewer for directing my attention to relevant work by Theresa Biberauer and her colleagues in respect of extra-sentential final particles, and to the second for helping to better situate the work in a broader historical context, and for probing questions about the analysis (only some of which I have been able to address). I should also like to thank Trang Phan and Tue Trinh for their invaluable judgments and suggestions. Naturally, I am responsible for all remaining errors and weaknesses. Finally, I wish to acknowledge the financial support provided by the UK Arts and Humanities Research Council, through the Research Leave Extension Award Scheme, in the initial stages of the research reported here. 1 The following abbreviations are used in the glosses and trees: ANT = anterior; ASR = assertion; ASP = aspectual; CLS = classifier; COMP = complementizer; CONJ = conjunction; BE . LOC = locative be; DEM = demonstrative (1,2,3 = proximal, non-proximal, distal); EXCL = exclamative;

128

Nigel Duffield

(1) a.

Chị có PRN ASR

mua cái nhà không? buy CL house NEG

‘Did you (elder sister) buy (the) house?’ b.

Con đã PRN ANT

uống thuốc chưa? 3 drink medicine not.yet

‘Have you (child) taken your medicine yet?’ These elements, which have been treated previously in generative work (see for example, Duffield 1999, 2007; Trinh 2005), attract attention not so much in virtue of their unexpected syntactic distribution – apparently head-final elements in an otherwise rigidly head-initial language – but because of how they interact with other functional categories, in ways that suggest a more complex derivational history. Among the surprising interactions that this paper considers are the impossibility of negative Yes-No questions, exemplified in (2a), and the incompatibility of final không/chưa with preverbal topic and (future) tense markers, illustrated in (3a) and (4a), respectively. (2)

a. *Anh ấy

không đến không? come NEG

PRN DEM NEG

‘Isn’t he coming?’ b. Anh ấy

không đến. come

PRN DEM NEG

‘He isn’t coming.’ c. *Con chưa uống thuốc chưa? PRN not.yet drink medicine not.yet ‘Haven’t you [child] taken your medicine yet?’ d.

Con chưa uống thuốc. PRN not.yet drink medicine ‘You [child] haven’t yet taken your medicine.’

FUT = future tense; NEG = negation; POL = politeness marker; POSS = possessive; PRN = pronominal (including kinship terms); Q = Question Morpheme; REFL = reflexive; TM = topic marker. 2 The original conference talk treated final elements occurring in both Yes-No and Whquestions (không/chưa and thế, respectively). Here, I focus on the former set: for a more detailed discussion of wh-questions in Vietnamese, see Bruening and Tran (2006a); for a critique of that analysis, Duffield (2009, in prep.). 3 The main difference between the two morphemes không and chưa is that questions with final không simply ask about the truth of the proposition, whereas those with chưa ask whether the event or state of affairs denoted by the proposition has been realized.

Head-First: On the head-initiality of Vietnamese clauses

(3)

129

a. *Xã bên thì ruộng tốt không? village side TM rice field good NEG ‘(As for) the neighboring village, are its rice-fields good (fertile)?’ b. Xã bên thì ruộng tốt. village side TM rice field good ‘(As for) the neighboring village, its rice-fields are good (fertile).’

(4) a. *Vợ anh sẽ (có) làm việc ỏ Paris không? wife PRN FUT ASR work be.loc Paris NEG ‘Will your wife work in Paris?’ b.

Vợ anh sẽ làm việc ỏ Paris. wife PRN FUT work be.LOC Paris ‘Your wife will (indeed) work in Paris.’

There are three points to observe immediately concerning these examples. The first is the rather obvious fact that the final interrogative morpheme không in the (a) examples is formally identical to the sentential negation morpheme found in corresponding declarative clauses – except for its distribution: sentential negation in Vietnamese is strictly pre-verbal, occurring in a fixed position between the clausal subject and the predicate, such that final placement of không/chưa in declarative clauses to express negation is ungrammatical. This is shown by the contrasts in (5) and (6): (5)

a. Chị (đã) không mua cái nhà. PRN ANT NEG buy CL house ‘You [elder sister] did not buy (the) house.’ b. Con (đã) chưa uống thuốc. PRN ANT not.yet drink medicine ‘You [child] have not taken your medicine.’

(6)

a. Anh khỏe không? PRN well NEG cannot mean ‘You [older brother] are unwell.’ b. Con uống thuốc chưa. PRN drink medicine not.yet cannot mean ‘You [child] have not taken your medicine.’

130

Nigel Duffield

The second point of these examples is that Yes-No questions marked by simple không are generally restricted to unselected root clauses (unless the complement clause is introduced by the leftward complementizer liệu to be discussed presently). The third point to observe is the contrast in co-occurrence restrictions between the future/irrealis morpheme sẽ – illustrated in (4) above, which is incompatible with final không – vs. the “anterior” morpheme đã exemplified in (1b), and (8) below, which is fully compatible with final chưa.4 The standard structuralist assumption has been that Yes-No questions are formed by means of two parallel “brace constructions” in which the pre-verbal morphemes and có and đã – the former glossed ASR (for assertion) in (1a), following Duffield (2007, to appear) – are paired with final không and chưa, respectively: according to this brace analysis, schematized in (9), simple Yes-No questions involving only final không/chưa – that is, with no medial element – are derived through the optional omission/deletion of the pre-verbal element, in a manner apparently very similar to the omission of preverbal ne in negative sentences in most varieties of spoken French. (7)

a.

[(có) . . . VP . . . không]

b. [(đã) . . . VP . . . chưa] The discrepancy between the two pre-verbal morphemes is unexpected – and largely inexplicable – if đã and sẽ are both exponents of tense, as has sometimes been claimed in the descriptive literature (see, for example Nguyễn 1997; Lo Cicero 2001; amongst others); there would seem to be no reason why a construction may be used to ask past tense Yes-No questions, but not future tense ones. On the other hand, the contrast can be explained if one assumes that these two morphemes are initially projected in distinct structural positions, as will be argued presently; see also Duffield and Phan (2010). The aim of the present study, therefore, is to offer a syntactic analysis of Vietnamese Yes-No questions that derives these effects, and which reconciles the surface position of these sentence-final question markers with the strongly head-initial character of Vietnamese phrase-structure more generally. The empirical claim will be that Vietnamese is in fact head-initial throughout (up to and including the CP level), and that the surface distributions in (1) as well as the constraints illustrated in (2)–(6) reflect a series of complex derivational steps

4 See Duffield and Phan (2010) for further justification of this label (i.e., anterior).

Head-First: On the head-initiality of Vietnamese clauses

131

involving predicate-raising around leftward heads.5 It will also be shown how the syntactic and interpretive behaviour of the anterior morpheme đã provides additional support for the projection of aspectual morphemes in a position below negation but above the maximal vP, that is to say, the position labelled “Outer Aspect” in Travis (1991, 2010). The functional phrase structure implied by this proposal is diagrammed in (8), which is very similar that of Travis (2010), the only modifications being the addition of a Topic Phrase projection to accommodate the topic-prominent nature of Vietnamese syntax – see also Cao (1991), Duffield (1999) – and the relabeling of E(vent) P(hrase) as AsrP, in order to be consistent with Duffield (2007). Notice in this phrase-markers, Vietnamese morphemes are realized at the point of initial merger: (8)

Our starting point for discussion will be the null hypothesis in any analysis of non-thematic elements, namely, that the surface position of such elements directly reflects their underlying position; in other words, that no syntactic 5 Notice that the argument developed here only goes through if it is assumed that there exist intrinsically head-final languages – Japanese being a canonical example – in which heads are initially merged to the right of their phrasal complements; in other words, that the type of “roll up movement” proposed here is not a generalized solution to the issue of head-finality, but rather a way of accounting for sporadic exceptions in head-initial languages. Hence, in this paper I adopt a Kaynian approach to predicate raising, without adopting the more radical assumption that all languages are underlyingly head-initial.

132

Nigel Duffield

movement of any kind is involved. If this were correct, then one could directly translate to Vietnamese the analysis of final Q-markers in Mandarin Chinese proposed by Cheng (1997). As the Mandarin examples in (9) show, these Yes-No questions exbibit near identical surface word-orders. Adopting Cheng’s (1997) analysis, không/chưa would be treated as occupying a rightward (head-final) Comp position, as schematized in (10): (9)

a. Nĭ xĭhuan zhè bĕn shū ma? you like this CL book Q ‘Do you like this book?’ b. Qiáofēng măi-le shénme ne? Qiaofeng buy-perf what QWH ‘What did Qiaofeng buy?’

(10)

The theoretical problems with such a translation, however convenient, should be obvious. In the first place, it offends against the head-initial character of Vietnamese phrase-structure; more importantly, it offers no possibility of accounting for the contrasts just introduced. Most significantly perhaps, there is a considerable amount of direct empirical evidence showing that all other plausible complementizers in Vietnamese appear to the left of their complements, which speaks against this analysis. The argument to be developed runs as follows. First, it will be established that Vietnamese has leftward complementizers in other clausal contexts, including interrogative complementizers in indirect questions. Next, it will be demonstrated that these final elements are not tags, or other extra-sentential markers: Vietnamese does have productive tag questions, which are superficially similar to Yes-No questions; however, tag questions display none of the constraints illustrated above. These apparently contradictory findings – that Vietnamese has leftward complementizers, and that final Q-elements are nevertheless integrated functional heads – are reconciled by the ‘Kaynian’ conclusion that head-final word-order in Vietnamese interrogatives is derived through predicate-raising (Kayne 1994, and subsequent work): the analysis proposed in the final section analyzes the observed word order facts in this fashion, and by doing so directly accounting for the interpretive and morphological constraints. This analysis is

Head-First: On the head-initiality of Vietnamese clauses

133

also shown to make the correct cut between the two pre-verbal morphemes sẽ and đã, a distinction that is independently motivated by the semantic and interpretive properties of đã (see Trinh 2006; Duffield and Phan 2010).

2 Leftward complementizers One of the most basic objections to the idea that these sentence-final elements occupy a rightward Comp position is the fact that Vietnamese is otherwise a radically head-initial language. The examples in (11) and (12) below illustrate various ways in which Vietnamese is head-initial in its canonical word order, not simply with respect to head-complement order, but also with regard to most other head-modifier relations: the examples in (11) show that transitive verbs, prepositions, (complement-taking) nouns and adjectives, and most Trelated elements precede their phrasal complements; those in (12) demonstrate that attributive adjectives, (restrictive and non-restrictive) relative clauses, adjunct clauses, and prepositional phrases all invariably follow the head of the phrase they modify: (11) a.

Tôi chỉ [ V P mua hai trái cam thôi]. I only buy two CL oranges at.all ‘I only bought two oranges.’

b.

Bếp nằm [ P P ngay sau lểu ngủ]. kitchen directly behind tent sleep ‘The kitchen is directly behind the sleeping tent.’

c.

[ N P Giả thuyết [trái đất phẳng]] (thì) dễ bác bỏ. belief earth flat TM easy disprove ‘The belief that the earth is flat is easy to disprove.’

d.

về con gái mình] Người đàn ông [ AP rất tự hào person man very self proud about CL girl own ‘a man very proud of his daughter’

e.

chúng tôi trở về nhà. Mẹ tôi [ T P đã mua] và mother I ANT buy CONJ plural I return house ‘My mother bought (it) and we returned home.’

(12) a.

Một người rất thông mình one person very intelligent ‘A very intelligent person’

134

Nigel Duffield

b.

Quyển sách [mà anh thích nhất] thì bán chạy. CL book RM PRN like most TM sell run ‘The book that you like most is selling well.’

c.

Người phụ nữ hài lòng [vì đã quyết định sáng suốt]. person woman happy because ANT decide clearly ‘The woman was happy because the right decision had been made.’

d.

mẹ của tôi mother POSS I ‘my mother’

Given this overwhelming head-initiality,6 it would be surprising if regular clausal complements in Vietnamese were not also head-initial, and even if Vietnamese were like Mandarin Chinese in possessing no overt complementizers.7 As it happens though, it does.

2.1 Complementizers in declarative contexts In embedded declarative clauses, the two most plausible candidates as complementizers are the clause-initial elements rằng and là, the former largely restricted to formal written registers, the latter more commonly found in colloquial registers, and homophonous with the identificational copula. The sentences in (13) illustrate typical examples of verbal sentential complements. In all of these cases, either rằng or là may introduce the embedded clause, depending on the register:8

6 The main exceptions to this “cross-category harmony” (Hawkins 1983) are two functional elements that unexpectedly appear on the right periphery of their phrasal complement, namely, the abilitative/epistemic modal được (on the right edge of the verb-phrase), and demonstrative modifiers, which appear on the right edge of the noun-phrase. See below for further discussion: for detailed discussion of the peculiarities of dược – arguably the most recalcitrant element in Vietnamese grammar – see Duffield (1999, 2001), Simpson (1998, 2001); see also Nguyễn (this volume) for an analysis of demonstrative placement. 7 A reviewer observes that Mandarin Chinese and other varieties do have emerging complementizers (citing shuō in Mandarin as such a case). 8 In this paper, là/rằng are presented as stylistic variants of one another. This is not strictly accurate: as discussed in Duffield (in prep.), there are contexts from which one or other of

Head-First: On the head-initiality of Vietnamese clauses

(13) a.

135

Ông ta nói rằng/là công việc không thích hợp với ông ta. say COMP work NEG suitable with PRN

PRN

‘The man said that the job was unsuitable for him.’ b.

Tôi mong-ước là/rằng mình có thể có ngọn đèn như thế. COMP self can possess lamp like so I wish ‘I wish that I had a lamp like that.’

c.

Tôi nghỉ [rằng các bạn sẽ nói [rằng tôi dược hưởng I think COMP PL friend FUT say COMP I can inherit một nềngiáo dục tốt đẹp nhất [mà tiền bạc có thể mua đuợc.]]] one education good beautiful most RM money can buy can ‘I guess that you could say that I had the best education that money can buy.’

Example (13c) provides evidence of the fact that Vietnamese permits multipleembedding with either element, such recursion being a further indication that these are complementizers of the familiar sort.

2.2 Interrogative Complementizers Turning now to interrogative complement clauses, the expectation is that Vietnamese should have leftward complementizers here too: if this is the case, then it would seem to militate strongly against the idea that không and chưa are in C. The examples in (14) below feature the clause-initial element liệu, which introduces indirect questions in Vietnamese, and is normally translated as “if, whether” – clear prima facie evidence of its status as a complementizer. Notice that, as mentioned earlier, embedded Yes-No questions are normally signaled by a (subtly) different device, namely, hay không (‘or not’), which exhibits a distinct set of co-occurrence restrictions to final không in root clauses, as do the other tag questions discussed below. (14)

a. Cô gái hỏi [liệu cô có thể đi đến bữa tiệc được không]. PRN girl ask LIEU PRN ASR possibility go arrive party can NEG ‘The girl asked if she could go to the party.’

these elements is excluded, as well as sentences such as those in (i) below, in which the two co-occur: (i) Phải nói rằng là thế hệ trẻ của chúng ta rất tài năng. modal say COMP COMP generation young of PL PRN very talented ‘(I) have to say that our young generation is very talented.’ Space constraints preclude further elaboration of these contrasts, which are in any case tangential to the main argument.

136

Nigel Duffield

b. Cô giáo hỏi các em sinh viên có tham gia chuyến đi hay không. PRN teacher ask PL PRN student ASR intention trip go or NEG ‘The teacher asked whether the students were going on the trip.’ c.

Người đàn ông tự hỏi [liệu cô bồ có ở lại với person man self ask LIEU PRN friend ASR be.loc stay with ông ấy PRN DEM

(hay không)]. or NEG

‘The man wondered whether (or not) his girlfriend would stay with him.’ Notice also in passing that it is quite possible to embed a Yes-No question within a Yes-No question, yielding a double occurrence of final không, as exemplified in (21) below (though as noted earlier, it is preferable to signal the embedded question by means of a preceding hay ‘or’): (15)

a. Nó có PRN ASR

hỏi mày có về nhà (hay) không không? ask PRN ASR return home or NEG NEG

‘Did he ask whether you went home (or not)?’ Summarizing this section, there seems to be robust distributional evidence to support the view that Vietnamese CP projections are head-initial in both declarative and interrogative complement clauses, consonant with the headinitial character of the other constituent phrases that were illustrated in (11) and (12) above. In the light of this, the strong implication must be that rightward không and chưa occupy some other phrase-structural position.

3 Analyzing the right periphery If the evidence of the previous section implies that không and chưa are not in C, the question then arises as to their correct analysis. The next subsection considers, and rejects, two straightforward analytic possibilities: first, that final không and chưa are extra-sentential elements; second – more plausibly perhaps – that they are sentential tags.

3.1.1 Final không as extra-sentential discourse marker The notion that these elements might be extra-sentential (possibly extragrammatical) morphemes, on a par with the final discourse particles exemplified

Head-First: On the head-initiality of Vietnamese clauses

137

in (16) below, is schematized in (17). Were this the case, their right-peripheral position would present no challenge to the assertion that Vietnamese phrasestructure is consistently head-initial; but neither would it be a theoretically interesting conclusion. (16) a.

thế à? fact Q ‘Is that so?’

b.

Anh đang làm gì thế ạ? PRN PROG do what Q POL ‘What are you doing?’

c.

Tôi đã bảo là tôi đi câu cơ mà! 9 I ANT say COMP I go fishing EXCL ‘I already told you that I’m going fishing!’

(17)

[ CP . . . . . . . . . . . . ] không?

The idea that morphemes such as those in (16) – and their correlates in other languages – should be analyzed as extra-syntactic may be controversial (see note below), but it is by no means unprecedented in formal analysis.10 Especially relevant to the present paper is recent work by Theresa Biberauer

9 [http://hahien.wordpress.com/2009/05/08/th%E1%BA%BF-ma-toi-c%E1%BB%A9-t%C6% B0%E1%BB%9Fng-la-ong-di-cau/] accessed 1/07/09. 10 A reviewer of a previous version of this paper expressed shocked incredulity (“[The] statement left me flabbergasted”) at the claim that “many languages have lexical elements that are extra-metrical in this sense – present in utterances but not in sentences”. However, I know of no formal analysis that treats elements such as final ‘alright’, ‘ok’, ‘yeah’, etc – as in (i) below – as structurally integrated into English clauses, even though these elements also are functionally clause-typing, signaling a (rhetorical) question: (i) a. I’m coming, alright?! b. She’s my sister, yeah?! c. I know what I’m doing, ok?! Nor are these the only elements that can be viewed as extra-grammatical, but nonetheless linguistic objects: from affective noises (Brrr!, Whoosh, Zoom-Zoom) to the ubiquitous, conversational like, to paralinguistic gestures, including nods, head-shakes, turn-taking uh-huhs etc., natural language utterances are populated with morphemes – unique conventionalized pairings of sound and meaning – that show no signs of grammatical integration. Some of these elements are of course treated in theories of pragmatics or in cognitive theories of communication – see, for example, Kita and Ide (2007) – but not as syntactic objects.

138

Nigel Duffield

and her colleagues (Biberauer, Holmberg and Roberts 2007, 2008), whose FinalOver-Final Constraint (FOFC)11 specifically requires final ‘force particles’ in headinitial languages to be treated as extra-metrical with respect to Kayne’s LCA, in virtue of not projecting categorial features. Interestingly, Biberauer, Holmberg and Roberts (2007) specifically cite Mandarin Chinese as a key example of a head-initial language with extra-metrical final particles. That said, such an analysis for Vietnamese không is unpromising. For one thing, it fails to explain why these elements should be homophonous with morphemes that are clearly syntactically integrated, and which have a clear context-free semantics of their own; by any intuitive understanding of the definition, không projects categorial features. Second, it predicts incorrectly that these elements should be restricted to the right edge of matrix clauses, and also implies that they should be separated from sentential material by some type of intonational break (though it does not require this, of course). The examples in (18), however, show that không/chưa are both integrated into some kind of subordinate clause such that they can appear within that constituent in fronting contexts.12 This mobility contrasts sharply with that of the elements in (16) above, all of which are restricted to utterance-final positions. (Notice further that where không co-occurs with any such discourse particles, the latter elements always occur to the right of không). (18)

a. [Lan có gặp Thơ không]j, Tân biết t j (Bruening and Tran 2006a) Lan ASR meet Tho NEG Tan know ‘Tan knows whether Lan met Tho.’ b. *[Lan có gặp Thơ]j, Tân biết t j không. Lan ASR meet Tho Tan know NEG ‘*Tan knows whether Lan met Tho.’ c.

Tân biết [Lan có gặp Thơ ?* (hay) không]13 Tan know Lan ASR meet Tho or NEG ‘Tan knows whether Lan met Tho.’

Finally, as noted at the outset, both không and chưa are ‘syntactically bound’ to the pre-verbal morphemes có and đã. The syntax of these two latter 11 Originally due to Holmberg (2000). 12 The reason for the circumspect phraseology is that I am not convinced, pace Bruening and Tran (2006), that strings such as those in (18) involve either fronting or true complement clauses: if this were the case, then one would expect the unmoved, unfronted alternant (18c) to be grammatical without hay, contrary to the judgments of most speakers consulted. 13 In this context, hay is obligatory for speakers consulted.

Head-First: On the head-initiality of Vietnamese clauses

139

elements is discussed in greater detail in Duffield (2007) and Duffield and Phan (2010), respectively. Here, the crucial point to observe is that it is these (optionally realized) pre-verbal elements, rather than không or chưa, that actually determine the force of the clause – the “clausal type”, in Cheng’s (1997) sense. So, for instance, whereas sentence (19a) is ambiguous between a preferred direct and indirect reading, those in (19b–c) are unambiguous: where có precedes the embedded verb, as in (19b), only the indirect reading is possible; conversely, where có precedes the matrix verb (19c), only the direct reading is available. This contrast is indicated by the final bracketing to the left and right of không, respectively. In other words, it is có rather than không – correspondingly, đã rather than chưa – that functions as the scope marker in each instance. (19) a.

Tân biết Lan gặp Thơ không? Tan know Lan meet Tho NEG ‘Does Tan know whether Lam met Tho?’ marginally: ‘Tan knows whether Lan met Tho.’

b.

Tân biết [Lan có gặp Thơ hay không]? Tan knows Lan ASR meet Tho or NEG Cannot mean: ‘Does Tan know whether Lan met Tho?’ ‘Tan knows whether (or not) Lan met Tho.’

c.

Tân có biết [Lan gặp Thơ] không? Tan ASR knows Lan meet Tho NEG ‘Does Tan know whether Lan met Tho?’ Cannot mean: ‘Tan knows whether (or not) Lan met Tho.’

The main implications of such alternations are two-fold: first, that không cannot be treated as an extra-grammatical discourse marker, otherwise its dependence on (interrogative) có would be inexplicable; second, as argued previously in Duffield (2007), that in Vietnamese [wh]-features are not inherent properties of C-related heads, but are associated with a much lower functional category. See also Duffield (to appear). Other Vietnamese facts discussed in that paper provide further support for this proposal, which recalls earlier analyses of Noonan (1989), and especially Rizzi (1996). The following quote from Rizzi (1996) is germane: . . . It is natural to assume that such a position (for +wh features) can be the main inflection (or one of the main inflectional heads, if some version of the Split Infl hypothesis is adopted, as in Pollock (1989)), the head that also contains the independent tense specification of the whole sentence. I would like to propose that among the other autonomously licensed specifications, the main inflection can also be specified as [+wh] . . . [italics mine; N. D.] Rizzi (1996: 66).

140

Nigel Duffield

The assumption that [wh]-features are located lower than C-related projections also immediately accounts for the distribution of Q-morphemes in some other languages, for example, the question marker kĕ in Mandarin Chinese, discussed in Schaffar (n.d.) and exemplified in (20); also, the pre-verbal positioning of the Q-morpheme puas in Mong Leng, an areally and typologically related language, described by Bruhn (2007) and illustrated in (21) below: (20)

a.

Nĭ kĕ xĭhuan zhè bĕn shū? you Q like this book ‘Do you like this book?’

(21) a.

Lauj puas tau pum tug us. Lao Q PERF see CL duck ‘Has Lao seen the duck?’

b.

Lauj nug kuv saib Maab puas nyam nwg. Lao ask 1 SG whether Mang Q like 3 SG ‘Lao asked me whether Mang liked him.’

All of these distributional facts then tend towards the idea, originally articulated in Aoun and Li (1993), that [wh]-features in East Asian languages more generally are located in a position below C-related projections.

3.1.2 Final không as tag question Though the facts illustrated in (18) and (19) above clearly militate against a treatment of final không as a extra-grammatical element, they do not preclude a different possibility, namely, that final không should be analysed as a polarity tag-question. The most direct analogy would thus be to the English negative tags illustrated in (22): these also exhibit a close dependency with pre-verbal morphology, as signalled by person and number agreement as well as by the ‘inverse polarity’ effects (-+, +-, *++, *--): (22)

a.

He should have done that, shouldn’t he?

b. She has finally realised what she’s doing, hasn’t she? c. *She has realized what she’s doing, isn’t she? Culicover (1998) analyzes English tag questions as involving a covert proIP, and one might wonder whether a similar analysis could be applied to

Head-First: On the head-initiality of Vietnamese clauses

141

Vietnamese final không. What makes this especially plausible is the fact that Vietnamese does have a very productive tag-question strategy, employing some of the same morphological elements. This is shown in (23): (23)

a. Anh ấy

không đến, (có) phải không? come ASR right NEG

PRN DEM 2 NEG

‘He isn’t coming, is he?’ b. Bạn chưa xem phim này, (có) phải không? friend not.yet see film DEM 1 ASR right NEG ‘You haven’t seen the film yet, have you?’ c.

Mày không có xu nào, (có) phải không? have money which ASR right NEG

PRN NEG

‘You haven’t got any money, have you?’ However, simple Yes-No questions contrast with these tag questions in several crucial ways that show that final không requires a separate treatment (whatever the correct analysis of tags may be). The most obvious difference between the two is that tag questions in both English and Vietnamese are strictly root phenomena: the interpretation of the examples in (24) shows that phải không can only modify matrix predicates, while the contrast between the sentences in (25) and those in (18) above demonstrate that phải không, unlike không, cannot appear within a fronted subordinate clause. Thus, the two constructions clearly diverge, interpretively and distributionally. (24) a.

Chị nói anh ấy không đến, phải không? say PRN DEM 2 NEG come right NEG

PRN

‘She said he isn’t coming, didn’t she/*is he?’ b. Chị nói bạn chưa xem phim này, phải không? PRN say friend not.yet see film DEM right NEG ‘She said you haven’t seen the film, didn’t she/*have you?’ (25)

a. *[Lan có gặp Thơ phải không]j, Tân biết t j. Lan ASR meet Tho right NEG Tan know ‘*Lan met Tho, didn’t he, Tan knows.’ b.

?[Lan có gặp Thơ], Tân biết, (có) phải không? Lan ASR meet Tho Tan know ASR right NEG ‘*That Lan met Tho, Tan knows, didn’t he?’

142

Nigel Duffield

c.

*Tân biết Lan có gặp Thơ hay phải không. Tan know Lan ASR meet Tho or right NEG ‘Tan knows whether Lan met Tho.’ (cf. 19b)

The second notable difference between final không and phải không is that tags are not licensed in formal registers by preverbal có: in fact, as the examples in (26) demonstrate, phải không is incompatible with preverbal interrogative có.14 (26)

a. Tân (*có) biết Lan gặp Thơ, phải không? Tan ASR knows Lan meet Tho right NEG ‘Does Tan know whether Lan met Tho?’ b.

Anh (*có) học tiếng Việt, phải không? study lang. Viet. right NEG

PRN ASR

‘He studies Vietnamese, doesn’t he?’ Finally, phải không shows none of the co-occurrence restrictions on final không flagged in the introduction, and which – as will be shown directly – offer the key evidence for a transformational analysis of the latter element. On the contrary, as shown by the examples in (27)–(28) phải không is fully compatible with preceding topic markers and tense markers, as well as with negative assertions: (27)

a. Xã bên thì ruộng tốt, (có) phải không?15 village side TM rice field good ASR right NEG ‘(As for) the neighboring village, its rice-fields are good (fertile), aren’t they?’ b.

Vợ anh sẽ làm việc ỏ Paris, (có) phải không? wife PRN FUT do work be-LOC Paris ASR right NEG ‘Your wife will work in Paris, won’t she?’

(28)

a. Anh ấy

không đến, (có) phải không? PRN DEM NEG come ASR right NEG ‘Isn’t he coming?’

14 Notice that phải không is compatible with pre-verbal có where có functions as an emphatic marker (without interrogative function), or as a lexical main verb meaning ‘have’: see Duffield (2007, to appear) for further discussion. 15 Adapted from Cao (1991: 145).

Head-First: On the head-initiality of Vietnamese clauses

143

b. Bạn chưa xem phim này (có) phải không? friend not.yet see film DEM 1 ASR right NEG ‘Haven’t you seen the film yet?’ c.

Mày không có xu nào, (có) phải không? have money which ASR right NEG

PRN NEG

‘Haven’t you got any money?’ In short, the evidence against the idea that final không should be treated as a reduced tag is as varied and robust as against the notion that final không is an extra-sentential discourse marker. Hence, it must be concluded that final không is an integrated functional head. Yet, as the first section showed, this is not a C-related head, since Vietnamese CPs are demonstrably head-initial, even in Yes-No questions. This apparent impasse calls for a more radical solution.

3.2 A derivational approach to yes-no questions Interestingly, one possible way through this impasse – which might be termed a “radical ellipsis” analysis – is already offered by Nguyễn (1997). Nguyễn proposes that final không questions are derived from more complex bi-clausal representations in which không (chưa) occupies the same underlying position in both clauses, as a pre-verbal marker of sentential negation. Nguyễn writes: . . . Like có . . . không? the discontinuous expression đã . . . chưa asks whether the action has taken place yet, and the question . . . Con đã uống thuốc hay (là) chưa uống thuốc? child ANT drink medicine or COP not.yet drink medicine ‘Have you taken your medicine (or not) yet?’ . . . undergoes successive deletions, until đã itself can be left out: > Con đã uống thuốc hay là chưa? > Con đã uống thuốc chưa? > Con uống thuóc chưa?. . . (1997: 152–153):’ This insightful analysis seems quite plausible both as a diachronic explanation of the origins of final không/chưa, also as a synchronic analysis of the hay không construction, discussed earlier. However, given the contrasts obtaining between bare không/chưa questions and phải không tag question just discussed,

144

Nigel Duffield

it is less credible as a synchronic analysis. It is certainly not a Minimal(ist) solution.16

3.2.1 Final không is not final (Kaynian manoeuvre) Consider then how we might retain the intuition behind Nguyễn’s (1997) suggestion within a more economical derivational approach. Given the structure presented in (8) above, the proposed analysis of Yes-No questions is schematized in (29): (29)

On this treatment, wh-features of không, projected as the head of NegP, force movement of the immediately subjacent constituent phrase to its specifier. At a stroke, this analysis explains not only the common form and semantics of preverbal sentential negation and final không, but also the impossibility of negative Yes-No questions mentioned at the outset, and further illustrated in (30); these examples may be compared directly with the tag question examples in (27) above. (30)

a. *Anh ấy PRN

DEM 2

không đến không? NEG come NEG

‘Isn’t he coming?’

16 This is not so say that one could not imagine a Minimalist implementation of this solution, perhaps in terms of Sideward Movement (Nunes 2004): see Bruening and Tran (2006b) for an analysis of bare conditional structures in Vietnamese and Chinese along these lines. Whatever the merits of such an analysis for bare conditionals may be, it seems implausible that it should be applied in the present case: for one thing, it would not account for the absence of negative Yes-No questions, nor for the other constraints discussed immediately below.

Head-First: On the head-initiality of Vietnamese clauses

b.

145

*Bạn chưa xem phim này không? friend NOT.YET see film DEM NEG ‘Haven’t you seen the film yet?’

c.

*Mày không có xu nào không? PRN NEG have money which NEG ‘Haven’t you got any money?’

The reason that such sentences are excluded should be clear: a head is unable to trigger movement of a phrase containing itself around itself without violating some fundamental derivational constraint (e.g., Structure Preservation, Chain Uniformity, Extension Condition, etc.): (31) *Không [ TP anh ấy [ NegP không [ vP anh ấy NEG

PRN DEM

NEG

PRN DEM

đến]] arrive

3.2.2 Immediate consequences In addition to excluding negative Yes-No questions, this analysis also derives the fact that, in contrast to normal declaratives and tag questions, Yes-No questions may only involve functional categories that are initially merged lower than the projection headed by không. Specifically, it predicts correctly that Yes-No questions cannot contain the future tense marker sẽ or the topic marker thì; I temporarily set aside the anterior marker đã. The contrasts discussed earlier – and recapitulated in (32) and (33) below – show that both of these predictions are borne out: in each case, tense and topic markers that are fully acceptable in tag questions – the (a) examples – are excluded from simple Yes-No questions, even though they are readily interpretable.17 (32)

a. Vợ anh sẽ làm việc ỏ Paris, (có) phải không? wife PRN FUT do work be-LOC Paris ASR right NEG ‘Your wife will work in Paris, won’t she?’

17 The contrast between (33b) and (33c) shows that topic information may still be left-adjoined – in apposition to the Yes-No question: it just cannot appear in the integrated topic position ({Spec, TopP}).

146

Nigel Duffield

b. *Vợ anh sẽ (có) làm việc ỏ Paris không? wife PRN FUT ASR do work be-LOC Paris NEG ‘Will your wife work in Paris?’ c.

Vợ anh có làm việc ỏ Paris không? wife PRN ASR do work be-LOC Paris NEG ‘Does/Will your wife work in Paris?’ [Tense unspecified]

(33)

a. Xã bên (thì) ruộng tốt. village side TM rice-field good ‘(As for) the neighboring village, its rice-fields are good (fertile).’ b. *Xã bên thì ruộng tốt không? village side TM rice-field good NEG ‘(As for) the neighboring village, are its rice-fields good (fertile)?’ c.

Xã bên kia, ruộng có tốt không? village side DEM rice field ASR good NEG ‘That neighboring village over there, are its rice-fields good (fertile)?’

Notice now that the analysis has immediate consequences for the treatment of two other grammatical morphemes that can appear in Yes-No questions: viz., the interrogative/emphatic marker có and the anterior marker đã, shown in (1) above, and elsewhere, repeated here for convenience: it requires that these elements must originate within the phrasal complement of không. (1) a.

Chị có

mua cái nhà không? PRN ASR buy CL house NEG ‘Did you [elder sister] buy (the) house?’

b.

Con đã PRN ANT

uống thuốc chưa? drink medicine not.yet

‘Have you [child] taken your medicine yet?’ Taking có first, it was claimed in Duffield (2007) that in Yes-No questions the negation morpheme không was adjoined to the same functional projection (AsrP) that was headed by có: this head was assumed to combine formal features associated with polarity (±N EG ), clausal type (±WH ), and emphasis (±A SR ) under one node, as schematized in (34):

Head-First: On the head-initiality of Vietnamese clauses

(34)

147

(cf. Duffield 2007: 790)

Whatever advantages this analysis may have, it is clearly incompatible with the present account. Instead, in order to derive questions such as those in (1a), có must be projected independently below the functional projection associated with không: moreover, the thematic subject must undergo raising out of the verb-phrase into the specifier of this projection prior to generalized raising of this derived complement to không. A possible derivation is of có questions is diagrammed in (35): (35)

Similar considerations apply to the derivation of Yes-No questions involving đã as in (1b). If we assume that the negative morpheme chưa occupies the same syntactic position as (the default negative) không, it follows that đã too must be projected below the NegP projection: in other words, that anterior đã – in contrast to future sẽ – is not a tense marker underlyingly, but is instead associated with a lower aspectual projection (Outer Aspect: OAsP). This is shown in (36): (36)

148

Nigel Duffield

As it turns out, there is independent evidence that these two pre-verbal markers – anterior đã vs. future sẽ – have a distinct syntax and semantics: see Duffield (2009), Duffield and Phan (2010), for details. Nor is this assumption novel: Trinh (2005) also proposes on independent grounds that đã undergoes head-raising to T (in contrast to future sẽ, which is assumed to be base-generated in T). The central piece of interpretive evidence that directly supports such an analysis has to do with the interpretation of anterior đã in different sentential contexts. Trinh (2005) observes that whereas đã is generally ambiguous between a perfective and preterite interpretation (37a), only the latter reading is available in negative assertions (37b). (37)

a. Nó đã PRN ANT

dọc sách. read book

[examples from Trinh 2005]

‘He has read books.’ (perfect reading) or ‘He did read books.’ (preterite reading) b. Nó đã

không dọc sách. PRN PAST NEG read book

‘He did not read books.’ (Only preterite reading allowed) This interpretive difference can be made to follow directly from the assumption that generally đã is projected below NegP. Assume that – in the absence of sentential negation – đã is inserted under OAsp and raises via head movement to T (accounting both for the ambiguity of đã, and for the incompatibility of aspectual đã with future sẽ.)18 In Yes-No questions, đã continues to be inserted under OAsp, and interpreted aspectually, but in negative declaratives, when Neg is projected, head movement to T is blocked: this forces đã to be merged late directly under T, and resulting in a purely temporal interpretation (compare Pollock 1989, and much subsequent work on verb raising). Finally, this analysis of Yes-No questions correctly predicts the incompatibility of (aspectual) đã with (eventive) có in interrogative structures, even though these same elements are not mutually exclusive in declarative clauses: compare (38a) and (38b) below. On the present analysis, this contrast follows from the impossibility of combining the derivation in (35) with that in (36):

18 This incompatibility cannot be excluded on semantic grounds: as discussed in Duffield and Phan (2010), in non-interrogrative clauses đã is perfectly compatible with future perfect readings.

Head-First: On the head-initiality of Vietnamese clauses

(38) a. *Anh ấy

đã

có

PRN DEM ANT ASR

149

trở về không? return NEG

‘Did he come back?’ b.

Anh ấy

đã

có

PRN DEM ANT ASR

trở về. return

‘He did come back.’ Summarizing, I have argued in this section that the various morphological, distributional and interpretive constraints on simple Yes-No questions that distinguish these from tag questions and negative declaratives are best explained in terms of a Kaynian turn, by which không/chưa is taken occupiy a constant head-initial, sentence medial position underlyingly, its final position in interrogative structures derived through predicate raising of the immediately subjacent phrase to a higher specifier position. Before considering some further implications of this analysis, it is worthwhile addressing the formal motivation for the predicate raising postulated here. A number of analytic options are available, but essentially these boil down to two: feature-driven movement on the one hand, and “scope evasion” on the other. As for the first, one might imagine that this obligatory predicateraising satisfies the checking of [+wh] features of không, along the lines of Rizzi’s Wh-criterion (Rizzi 1996), the difference being that such checking takes place in the original (medial) position where [+wh] features are initially inserted – see quote above – rather than in C. However, though this might provide an adequate formal account of the movement, it is as stipulative as any appeal to movement having no clear interpretive effects: it also seems anomalous that Vietnamese should require overt movement for Yes-No questions, but not for constituent wh-questions. An alternative explanation is suggested by the obligatory movement exhibited in Vietnamese by wh-words interpreted as universal quantifiers, discussed in Duffield (2007). As the examples in (39)–(41) show, object and adjunct phrases containing the underspecified ‘wh-words’ ( gì ‘what’, ai ‘who’, nào ‘which’, etc.) are interpreted as universal quantifiers just in case they are displaced from their canonical postverbal position to a position to the left – beyond the scope of – the operator cũng; cf. Hole (this volume) for a focus-semantic-based interpretation of this kind of displacement. The alternations between the OSV orders in (39) and SOV orders in (40) demonstrate that the leftward landing site of such movement is irrelevant (suggesting that feature-checking is not the driving force); what is crucial is that the phrases are displaced to the left of cũng, as proven by the strict ungrammaticality of the in situ examples in (41).

150 (39)

Nigel Duffield

a. [Từ nào]i anh ấy cũng nhớ ti . word which PRN DEM also remember ‘He remembers every word.’ b. [Ai]i cô ấy cũng quen ti. ai PRN DEM also know ‘She knows everybody.’ c.

ti . [Ngày nào]i tôi cũng tập thể thao day which I also practise exercise ‘I do exercises every day.’

(40)

a. Anh ấy PRN DEM

[từ nào]i cũng nhớ ti . word which also remember

‘He remembers every word.’ ấy

[ai]i cũng quen ti . PRN DEM ai also know

b. Cô

‘She knows everybody.’ (41)

a. *Anh ấy PRN DEM 2

b. *Cô

ấy

PRN DEM 2

cũng nhớ từ nào. also miss thing which cũng quen ai. also know who

c. *Tôi cũng tập thể thao ngà nào. PRN also practice sport day which The present proposal is that parallel considerations apply to predicateraising around không. Suppose that like cũng (‘also’), không is ambiguous with respect to the (interpretable) features that can be associated with it: in principle, it can bear either negative features or wh-features, but not both simultaneously. The default interpretation of không is obviously as a negation morpheme, negating the constituent that appears overtly to its right (within its linear scope). It may be assumed then that in order to evade this negative interpretation, and to signal the presence of wh-features on không, the predicate is forced to raise outside of its scope, just as underspecified wh-constituents are forced to raise to avoid (default) wh-question interpretations. Thus, the best metaphor for raising here is not so much Greed (Chomsky 1989), or even Enlightened Self-Interest (Lasnik 1999), but a more functionalist notion: Survival (of interpretation).

Head-First: On the head-initiality of Vietnamese clauses

151

Notice finally that this analysis of Yes-No questions has theoretical and analytic implications beyond Vietnamese, in particular, for Cheng’s (1997) analysis of Mandarin Yes-No questions (introduced earlier) and for the C LAUSAL-T YPING H YPOTHESIS (CTH) that this analysis supports:19 (42) Clausal Typing Hypothesis Every clause needs to be typed. In the case of typing a wh-question, either a wh-particle in C is used or else fronting of a wh-word to the Spec of C is used, thereby typing a clause through C by Spec-Head Agreement (Cheng 1997: 22).

The fundamental empirical contrast at the heart of the CTH is between English and (Mandarin) Chinese questions, illustrated above (repeated here for convenience) in which sentence-final particles in Mandarin questions are assumed to be exponents of a head-final CP: (9)

a. Nĭ xĭhuan zhè bĕn shū ma? you like this book Q ‘Do you like this book?’ b. Qiáofēng măi- le shénme ne? Qiaofeng buy PERF what Q WH ‘What did Qiaofeng buy?’

The analysis of Yes-No questions presented in this paper offers several distinct challenges to this hypothesis. First – restricting attention to Vietnamese – the evidence that Vietnamese is consistently head-initial at all hierarchical levels up to CP shows that final Q-markers in Yes-No questions are not in C. If so, then Vietnamese stands as a counter-example to the CTH in having neither whparticles in C, nor overt wh-movement. Second, the analysis of Vietnamese Yes-No questions presented here, as well as the related evidence from other East Asian varieties (e.g. Mandarin kĕ and Mong Leng puas mentioned above) provides both theoretical and empirical evidence for the idea that wh-features are inherent properties of a much lower, clause-medial, projection, which again argues against the idea that clausal typing is uniquely determined by elements originating in C.

19 This is by no means the first paper to address the CTH: aside from authors mentioned above, see Hoge (1998), Bruening (2007), Stevens (2008); also Ultan (1978), amongst many others. It should also be noted that the CTH includes substantive claims that go beyond the presence or absence of (overt) wh-movement: in particular, Cheng draws attention to a correlation between multiple wh-fronting and the possibility of forming wh-words from the morphological base of indefinites: see again Bruening (2007), Hoge (1998), for discussion.

152

Nigel Duffield

Taking a broader view, it is not inconceivable that the core Chinese data that provided the initial motivation for the CTH might be reanalyzed in a similar way to that proposed for Vietnamese. On the face of it, the distribution and discourse functions of elements on the right periphery in wh-questions are strikingly similar in both languages: if the existence of leftward complementizers licenses the inference that these final elements are not in C in Vietnamese, it raises the suspicion of an alternative analysis for Chinese as well: see Huang (2008) for one promising proposal.

4 Conclusion In this paper, I have provided a detailed exposition and analysis of the left and right periphery of Vietnamese clauses. Various arguments have been presented in support of the claim that Vietnamese is consistently head-initial up to CP. In place of a rightward C position, the appearance of the negative morphemes không/chưa in sentence-final position in Yes-No questions has been explained in terms of a Kaynian turn, in which the immediately subjacent complement of không is raised to a higher specifier position, leaving không/chưa on the right periphery. This analysis has been shown to explain a number of significant restrictions on Yes-No questions in Vietnamese, including the absence of negative Yes-No questions, and the exclusion of tense and topic markers (all of which are freely available in superficially similar tag questions, and in wh-questions). Finally, I have noted how this analysis fits with previous treatments of other anomalous elements in Vietnamese and areally-related languages, and considered the implications of this analysis for Mandarin Chinese – and for the Clausal Typing Hypothesis more generally.

References Aoun, Joseph and Yen-Hui Audrey Li. 1993. WH-elements in situ: syntax or LF. Linguistic Inquiry 24: 199–238. Biberauer, Theresa, Anders Holmberg and Ian Roberts. 2007. Disharmonic word-order systems and the Final-over-Final-Constraint (FOFC). In: Antonietta Bisetto and Francesca Barbieri (eds.), Proceedings of XXXIII Incontro di Grammatica Generativa, 86–105. Biberauer, Theresa, Anders Holmberg and Ian Roberts. 2008. Structure and linearization in disharmonic word orders. In: Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics, 96–104. Somerville/Massachusetts: Cascadilla Proceedings Project. Bruening, Benjamin. 2007. Wh-in-situ does not correlate with wh-indefinites or question particles. Linguistic Inquiry 38: 139–166.

Head-First: On the head-initiality of Vietnamese clauses

153

Bruening, Benjamin and Thuan Tran. 2006a. Wh-questions in Vietnamese. Journal of East Asian Linguistics 15: 319–341. Bruening, Benjamin and Thuan Tran 2006b Wh-conditionals in Vietnamese and Chinese: against unselective binding. Conference Presentation. To appear in Berkeley Linguistics Society (BLS) 32. Bruhn, Daniel. 2007. LF Wh-movement in Mong Leng. Term Paper, UC Berkeley, Department of Linguistics. Cao, Xuan Hạo. 1991. Some preliminaries to the syntactic analysis of the Vietnamese sentence. Mon-Khmer Studies 20: 137–151. Cheng, Lisa L.-S. 1997. On the typology of WH-Questions. New York: Garland Publishing. Chomsky, Noam. 1989. Some Notes on Economy of Derivation and Representation. In: Itziar Laka and Anoop Mahajan (eds.), MIT Working Papers in Linguistics 10, Cambridge/MA: MIT. Culicover, Peter. 1992. English tag questions in Universal Grammar. Lingua 88: 21–54. Duffield, Nigel. 1995. Particles and Projections in Irish Syntax. Dordrecht: Kluwer. Duffield, Nigel. 1996. On Structural Invariance and Lexical Diversity in VSO Languages: Arguments from Irish Noun-Phrases. In: Robert Borsley and Ian Roberts (eds.), The Syntax of the Celtic Languages, 314–340. Cambridge: Cambridge University Press. Duffield, Nigel. 1998. Auxiliary placement and interpretation in Vietnamese. In: M. Catherine Gruber, Derrick Higgins, Kenneth E. Olson and Tamra Wysocki (eds.), Proceedings of the 34th Regional Meeting of the Chicago Linguistic Society, 95–109. Chicago: CLS. Duffield, Nigel. 1999. Final modals, adverbs and antisymmetry in Vietnamese. Revue québécoise de linguistique 27: 92–129. Duffield, Nigel. 2001. On certain head-final effects in Vietnamese. In: Karen Megerdoomian and L. A. Bar-el (eds.), Proceedings of West Coast Conference on Formal Linguistics 20, 101– 114. Somerville, MA: Cascadilla Press. Duffield, Nigel. 2007. Aspects of Vietnamese clause structure: separating tense from assertion. Linguistics 45: 765–814. Duffield, Nigel. 2009. Head-first: on the head initiality of Vietnamese phrase structure. Paper presented at Linguistics of Vietnamese Workshop. University of Stuttgart, July 2009. Duffield, Nigel. in prep. Close to Prefect: Particles and Projections in Vietnamese Syntax. Ms., University of Sheffield. Duffield, Nigel. to appear. Low assertion as low modality in Vietnamese and English. Lingua. Duffield, Nigel and Trang Phan. 2010. Aspect Exposed: On the Syntactic Representation of Tense and Aspect in Vietnamese. Paper presented at South East Asian Linguistics Society XX, Zurich, June 2010. Hawkins, John A. 1983. Word Order Universals. New York: Academic Press. Hoge, Kerstin. 1998. Review of Cheng (1997). Linguist List http://www.linguistlist.org/issues/9/ 9-1725.html. Holmberg, Anders. 2000. Deriving OV order in Finnish. In: Peter Svenonius (ed.), The derivation of VO and OV, 123–152. Amsterdam: John Benjamins Publishing Company. Huang, C.-T. James. 1994. More on Chinese word order and parametric theory. In: Barbara Lust, M. Suñer and John Whitman (eds.), Syntactic Theory and First Language Acquisition: crosslinguistic perspectives. (Volume 1 Heads, Projections and Learnability), 15–35. Hillsdale/New Jersey: Lawrence Erlbaum Associates. Huang, Xiao-You Kevin. 2007. Initialness of Sentence-final Particles in Mandarin Chinese. In The 21st Pacific Asia Conference on Language Information and Computation Proceedings, 182– 191.

154

Nigel Duffield

Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge/Massachusetts: MIT Press. Kita, Sotaro and Sachiko Ide. 2007. Nodding, aizuchi, and final particles in Japanese conversation: How conversation reflects the ideology of communication and social relationships. Journal of Pragmatics 39: 1242–1254. Lasnik, Howard. 1999. Minimalist Analysis. Oxford: Blackwell. Nguyễn, Dinh Hoa. 1997. Vietnamese. (Volume 9: London Oriental and African Language Library). Amsterdam/Philadelphia: John Benjamins Publishing Company. Noonan, Maire. 1989. Operator Licensing and the Case of French Interrogatives. In: E. Jane Fee and Kathryn Hunt (eds.), Proceedings of the Eighth West Coast Conference of Formal Linguistics, 373–383. Stanford, CA: Stanford Linguistics Association. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge/ MA: MIT Press. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20: 365–424. Rizzi, Luigi. 1996. Residual verb-second and the Wh-Criterion. In: Adriana Belletti and Luigi Rizzi (eds.), Parameters and Functional Heads, 63–90. New York/Oxford: Oxford University Press. Schaffar, Wolfram. n.d. Typology of Yes-No questions in Chinese and Tai languages, Ms., University of Tübingen. Simpson, Andrew. 1998. VP-final modals and Pied Piping in S.E. Asian. In: Pius Tamanji and Kiyomi Kusumoto (eds.) Proceedings of North East Linguistics Society 28, 437–451. University of Massachusetts, Amherst: GLSA Publications, Simpson, Andrew. 2001. Focus, Presupposition and Light Predicate Raising in East and South East Asia. Journal of East Asian Linguistics 10: 89–128. Stevens, Jeff. 2008. Grammar, Performance and the Wh-Question Typology. Ms., University of Washington. Trinh, Tue. 2005. Aspects of Clause Structure in Vietnamese, Masters thesis, Humboldt University Berlin. Travis, Lisa. 2010. Inner Aspect. The Articulation of VP. Berlin: Springer. Ultan, Russell. 1978. Some general characteristics of interrogative systems. In: Joseph H. Greenberg, Charles A. Ferguson and Edith A. Moravcsik (eds.), Universals of Human Language, Volume 4: Syntax, 211–248. Stanford: Stanford University Press.

Andrew Simpson and Hồ Hảo Tâm

6 Vietnamese and the typology of passive constructions* Passive-type constructions formed with bị in Vietnamese are compared to the bèi passive of Mandarin Chinese and it is shown that there are both strong syntactic similarities between bị and bèi sentences and also significant differences. The existence of subject gap “passives” with bị in Vietnamese, in particular, is shown to impact attempts to arrive at a cross-linguistic definition of passive in terms of a set of minimal shared properties, and calls into question whether “the passive” indeed exists as a meaningful and definable linguistic construction. The chapter also considers how to account for the syntactic licensing of subject gap passive forms in Vietnamese but their exclusion from Chinese when it seems that appropriate pragmatic-semantic conditions for their occurrence are regularly met within both languages in bèi/bị forms. Keywords: passive, adversity passive, Chinese, Control, null operator constructions

1 Introduction This chapter examines passive-type constructions in Vietnamese in comparison with similar structures found in Mandarin Chinese and explores the significance of Vietnamese for a general typology of passive, both from formal and functional perspectives. Much interesting research has been carried out on the syntactic structure of Mandarin passive constructions in recent years, with significant results described in Ting (1998), Huang (1999), and Tang (2001). Huang (1999), in particular, places modern Mandarin bèi-constructions in a broad comparative perspective, incorporating insights from the diachronic development of bèi

* An early version of this chapter was presented at the 20th North American Conference on Chinese Linguistics held in 2008 in Columbus, Ohio, and appears in the on-line proceedings of that conference. The contents of the chapter was subsequently developed further and presented at the UCLA Conference on Southeast Asian Languages in 2009, and the Conference on the Linguistics of Vietnamese held at the University of Stuttgart in 2009. Many thanks to the audiences at all three conferences for their suggestions and comments on aspects of the chapter as it has evolved over time.

156

Andrew Simpson and Hồ Hảo Tâm

passives, and the synchronic realization of passive in non-Mandarin varieties of Chinese (Cantonese and Southern Min) as well as other East Asian languages such as Japanese and Korean. Passive-type structures in Vietnamese show much obvious similarity to those in Mandarin, and seem to be closer to Chinese in surface structure than the passive in Japanese and Korean are. In order to position the passive-type structure found in Vietnamese in a broader, crosslinguistic perspective, and chart the properties of this construction present in Vietnamese, the chapter probes how passive-type constructions in Vietnamese are both similar and also different from its apparent ‘closest cousin’ − modern Mandarin bèi constructions, and discusses how the existence of certain forms in Vietnamese requires a reassessment of the typology of passive structures presented in recent works focusing on East Asian languages. A consideration of the variety of passive-related structures in Vietnamese, in particular those involving subject-to-subject dependencies, raises questions concerning the limits of ‘passive’ as a definable construction. The chapter is organized as follows. Section 2 reviews a number of important defining characteristics of the bèi passive constructions in Mandarin Chinese. Section 3 then shows how bị structures in Vietnamese show many clear similarities to Mandarin bèi passives, suggesting that the analysis of bị and bèi passives should be fundamentally the same, despite a difference in the degree to which indirect passives appear to be available in the two languages. Section 4 focuses more squarely on ways in which Vietnamese and Chinese passive constructions show further surface differences and highlights both the use of different passive ‘auxiliary’ verbs and the occurrence of intransitive passives in Vietnamese. This leads on to a re-consideration of properties that may be taken to be universal to the passive in section 5, and how the patterns in Vietnamese impact on crosslinguistic characterizations of the passive. Section 5 also considers what syntactic factors may be responsible for the parametric variation between Chinese and Vietnamese, and why the range of forms found in Vietnamese are not all permitted to occur in Chinese. A summary of the chapter and its cross-linguistic consequences is provided in section 6.

2 Passive in Chinese The Mandarin bèi construction has been well described in a number of works in recent years, for example Shi (1997), Ting (1998), Huang (1999), and Tang (2001). Huang (1999) identifies a number of important syntactic properties of sentences such as (1) which support a bi-clausal analysis of Chinese passives, in which bèi occurs as a predicate embedding a second clause.

Vietnamese and the typology of passive constructions

157

(1) Zhāngsān bèi [Lĭsì dă-le]. Zhangsan BEI Lisi hit-ASP ‘Zhangsan was hit by Lisi.’ First, it is noted that a subject-oriented adverb such as gùyì ‘deliberately’ can occur preceding bèi and be construed as referring to the action of the initial NP in the sentence (Zhāngsān in (1)), identifying this NP as an Agent. This is taken to suggest that the initial NP may be base-generated as the Agent subject of a higher clause, rather than being raised to this position from a lower object position, where it would receive a Patient theta role. Movement between two independent theta positions is assumed to be unavailable due to restrictions imposed by the Theta Criterion. (2)

Zhāngsān shì gùyì bèi Lĭsì dă-de. Zhangsan BE deliberately BÈI Lisi hit-DE ‘Zhangsan deliberately got hit by Lisi.’

Second, it is observed that either the NP preceding bèi or the NP following bèi can bind the subject-oriented anaphor zìjĭ in sentences such as (3).1 The interpretations available in (3) therefore suggest that both the NPs Zhāngsān and Lĭsì are in subject positions, and hence that (3) contains two clauses, each with its own subject, as schematized in (4) (3)

Zhāngsāni bèi Lĭsìk guān zài zìjĭ i/k-de jiā-lĭ. Zhangsan BEI Lisi shut in self’s house-in ‘Zhangsan was locked up by Lisi in his/her own house.’

(4)

This leads to an analysis in Huang (1999) in which the ‘gap’ position present in examples such as (1–3) results from movement of an empty operator basegenerated in the object-of-verb position to a clause-initial position, where it converts the subordinate clause into a secondary predicate construed as referring to

1 For discussion of the subject-oriented nature of the Chinese anaphor zìjĭ, see Huang, Li and Li 2009: 337, fn. 8), and Cole, Hermon and Sung (1990), Pan (2001).

158

Andrew Simpson and Hồ Hảo Tâm

the subject of bèi, through co-indexation of this NP and the empty operator, as in (5): (5)

a.

Zhāngsāni bèi [Opi Lĭsì dă-le t i ]

b. NP1 BEI [Opi NP2 V t i ]

The operator-trace dependency posited in passive sentences such as (1–3) is argued to be potentially unbounded and able to span multiple clauses, as illustrated in (6). It is also constrained by syntactic islands, as shown in (7). Both of these observations support the view that passive sentences may involve A’-movement – for Huang (1999) the A’-movement of an empty operator (cf. Browning 1987; Chomsky 1977). (6)

Zhāngsān bèi Lĭsì pài jĭngchá zhuā-zŏu-le. Zhangsan BEI Lisi send police grab-away-ASP ‘Lisi sent the police to seize Zhangsan and take him away.’

(7) *Zhāngsān bèi wŏ tóngzhì Lĭsì bă [[ RC zànmĕi _ de] shū] dōu măi-zŏu-le. Zhangsan BEI I inform Lisi BA praise DE book all buy-off-ASP ‘I told Lisi to buy up all the books that praised Zhangsan.’ Such a conclusion receives further support from two other patterns. First, the particle suŏ, which otherwise only occurs in relative clauses (and hence is associated with A’-operator movement), may occur in bèi sentences of the form considered so far, where bèi is followed by an overt NP agent. This is illustrated in (8). (8) Zhèxiē shìqing bù néng bèi tā suŏ liăojiĕ. these thing not can BEI he SUO understand ‘These things cannot be understood by him.’ Second, it is possible for a resumptive pronoun to occur in the position of the object gap, when a frequency adverbial also appears, as shown in (9). The potential occurrence of resumptive pronouns is a property which is crosslinguistically associated with instances of A’-movement rather than A-movement (Hornstein 2001).

Vietnamese and the typology of passive constructions

(9)

159

Zhāngsān bèi Lĭsì dă-le tā yí-xià. Zhangsan BEI Lisi hit-ASP him one-time ‘Zhangsan was hit by Lisi once.’

The above-noted patterns all characterize bèi sentences in which an agent subject of the main descriptive verb is overtly present in the sentence. In addition to such forms, Mandarin also allows for there to be no overt realization of the agent of the main verb, as illustrated in (10): (10)

Zhāngsān bèi dă-le. Zhangsan BEI hit-ASP ‘Zhangsan was hit.’

Interestingly, such agentless passives, which Huang (1999) refers to as the ‘short passive’ form, have certain different syntactic properties from the ‘long passive’, where an agent is present. These differences, observed in Huang (1999), are summarized in (11), and are argued to call for a somewhat different analysis from that of the long passive: (11) P ROPERTIES OF THE M ANDARIN S HORT PASSIVE (Huang 1999) a.

No resumptive pronouns (even when frequency phrases appear).

b.

No particle suŏ possible.

c.

No unbounded dependencies possible.

Because subject-oriented agentive adverbs are possible in the short passive, as in the long passive, Huang concludes that the pre-bèi NP is base-generated as a subject in a higher clause and related to the gap position by an occurrence of A-movement (hence no resumptive pronouns, suŏ, or unbounded dependencies). In the short passive, bèi is suggested to select for a VP construed as a secondary predicate of the pre-bèi NP through co-indexation of a PRO which undergoes movement from the object gap position to SpecVP, as indicated in (12a) and (12b) (from Huang, Li and Li 2009: 134):2

2 The moved element is posited to be PRO rather than pro due to the obligatory control property that its reference can only be controlled by the subject of bèi.

160

Andrew Simpson and Hồ Hảo Tâm

(12) a.

Zhāngsāni bèi [VP PROi dă-le t i ]

b.

Both long and short passive constructions are consequently analyzed as having bi-clausal structures, with a simple difference in the size of the constituent that occurs as the secondary predicate combined with bèi – either a full clause with an overt subject and the occurrence of A’-operator movement, or a VP with A-movement of a PRO.

3 Passive-type structures in Vietnamese Turning now to consider Vietnamese, sentences with a passive meaning similar to the Chinese examples in section 2 are in many cases constructed with the morpheme bị, which is likely to have been borrowed from Chinese bèi. As in Mandarin, there are both ‘long’ and ‘short’ passive patterns in Vietnamese, and the appearance of the agent NP associated with the main verb is quite optional: (13) Nam bị Nga đánh. Nam BI Nga hit ‘Nam was hit by Nga.’ (14)

Nam bị đánh. Nam BI hit ‘Nam was hit.’

Similar to Chinese (as pointed out for Mandarin by Huang 1999), the passive morpheme and the following agent NP cannot undergo any repositioning as

Vietnamese and the typology of passive constructions

161

a sequence (example 15), hence do not pattern like a PP constituent, unlike English passive ‘by-phrases’. (15) *Bị Nga Nam đánh. BI Nga Nam hit Int.: ‘Nam was hit by Nga.’ Combined with the observation that the NP following bị is able to bind an anaphor, as shown in (16), this would seem to favor a bi-clausal analysis of bịsentences in which bị embeds a subordinate clause (at least in cases of overtagent long passive structures). Anaphors such as mình are regularly only bound by subjects, as illustrated in (17). The post-bị NP in passive sentences like (16) therefore patterns like a subject, similar to the post-bèi NP in Mandarin: (16) Nami bị Ngak nhốt trong phong ngủ của mìnhi/k . Nam BI Nga lock in room sleep of self ‘Nam was locked by Nga in his/her own room.’ (17)

Ngak nhốt Nami trong phong ngủ của mìnhk/*i . Nga lock Nam in room sleep of self ‘Nga locked Nam in her own (Nga’s) room.’

Long passive sentences in Vietnamese are also characterized by restrictions on the embedding of a Patient gap position which are typical of A’-dependencies, as in Mandarin. Long-distance dependencies similar to those in the bèi-passive are possible, but only in long-passive structures (i.e. where the Agent is overt as in (18)), and never into island constituents (not illustrated here). (18)

Nam bị *(Nga) bảo cảnh sát đến bắt. Nam BI Nga call police come arrest ‘Nga called the police to come and arrest Nga.’

With regard to a range of passive-like sentence forms, Vietnamese therefore shows patterns which clearly parallel those found in Chinese. This seems to suggest that the analysis of passive phenomena in Vietnamese and Chinese should be similar, and a bi-clausal treatment of both Vietnamese and Chinese appears to be warranted, at least in the instance of overt agent long passive structures.3 3 With regard to the possible occurrence of agentive, subject-oriented adverbs such as ‘deliberately’ in the higher clause of bị sentences, Vietnamese seems to permit this as long as a modal element such as muốn also occurs. At this point it is not clear whether muốn should

162

Andrew Simpson and Hồ Hảo Tâm

Parallels between Vietnamese and Chinese also extend further, with the occurrence of ‘indirect passive’ sentences in both languages. The term ‘indirect passive’ is commonly used to refer to instances of passive in which the ‘passivized’ surface subject does not correspond to any direct argument NP of the main descriptive verb such as the direct object, or indirect object. In both Chinese (19–21) and Vietnamese (22–24), it is found that the subject of bèi/bị may co-refer with the possessor of the object of the main verb, when the action of the verb clearly affects the possessor through action being applied to the object, which is frequently a body-part or some item closely associated with the subject: (19) Lăo Zhāng bèi dă-diào-le yáchĭ. old Zhang BEI hit-lose-ASP teeth ‘Zhang had his teeth knocked out.’ (Shi 1997) (20)

Tā bèi jĭngchá mòshōu-le zhí-zhào. he BEI police confiscate-ASP driving-license ‘He had his driving license confiscated by the police.’ (Shi 1997)

(21) Zhāngsān bèi tŭfĕi dă-sĭ-le fùqīn. Zhangsan BEI bandit hit-dead-ASP father ‘Zhangsan’s father was killed.’ (Huang 1999) (22)

Tôi bị Nga làm gẫy một ngón tay. finger I BI Nga make snap 1 ‘Nga broke one of my fingers.’

(23)

Nga bị Nam giật tóc. Nga BI Nam pull hair ‘Nga had her hair pulled by Nam.’

(24) Nam bị cảnh sát tịch thu ra đô của Nam. confiscate radio of Nam Nam BI police ‘Nam had his radio confiscated by the police.’

(Le 1976)

be analyzed as a raising modal auxiliary or as a control verb, a difference which may affect the way that the patterning here can be interpreted. (i) Nam cồ ỳ muốn bị cảnh sát bắt. Nam deliberately want BI police arrest ‘Nam deliberately wanted to be arrested by the police.’

Vietnamese and the typology of passive constructions

163

The full distribution of indirect passives in Vietnamese is, however, rather more restricted than that in Mandarin (and Taiwanese; Huang 1999), in two distinct ways. First, where the object of the main verb is a kin term and refers to a relative of the subject (e.g. ‘son’, ‘father’ etc), an indirect passive structure is licensed in Chinese (ex. 21) but not in Vietnamese (even with a resumptive possessor): (25) *Nga bị một người găng xtọ giết ba (của Nga). Nga BI 1 CL gangster kill father (of Nga) Intended: ‘A gangster killed Nga’s father.’ (26) *Nga bị Ông thầy giáo phạt con trai (của Nga). (of Nga) Nga BI Ong teacher criticize son Intended: ‘Nga’s son was criticized by teacher Ong.’ Second, Chinese permits the occurrence of certain ‘adversity passives’ in which the subject of bèi does not appear to correspond to any obvious argument or possessor gap position in the clause following bèi, as for example in: (27)

Lĭsì yòu bèi Wángwŭ jìchū-le yì-zhī quánlĕidă. Lisi again BEI Wangwu hit-ASP 1-CL home-run ‘Lisi again had Zhangsan hit a home run on him.’

(28)

(Huang 1999)

Wŏ bèi tā zhème yí zuò, jiù shénme dōu kànbujiàn-le. BEI he thus one sit then everything all can.not.see-ASP I ‘As soon as he sat this way on me, I couldn’t see anything at all.’ (Huang 1999)

This kind of passive structure licensed purely by the adverse effect of the action on the subject does not seem to be possible in Vietnamese: (29) *Cảnh sát bị tên sát nhân trốn thoát. police BI murderer escape Int.: ‘The murdered escaped from the police (and this adversely affected the police).’ In section 5, we will return to consider how such differences might be accounted for in an extension of the analysis of indirect passives proposed in Huang (1999). First, though, we will present two other sets of differences

164

Andrew Simpson and Hồ Hảo Tâm

between Vietnamese and Chinese passive forms, one which is primarily lexical and can be simply accommodated in existing treatments of East Asian passives, and another which is syntactic and has more serious consequences for characterizations of the passive as a cross-linguistic construction.

4 Lexical and syntactic variation in Chinese and Vietnamese passives 4.1 Negative and positive effect passives in Vietnamese An interesting lexical difference between Vietnamese and Chinese is that Vietnamese regularly makes use of two different functional morphemes in its ‘passive’ structures. In addition to the morpheme bị, present in all of the Vietnamese examples thus far, a second verbal element được also frequently occurs in fully parallel sentence forms. The key semantic difference between bị and được is as follows: (30)

a.

bị is used in sentences where the event depicted by the main verb is understood as affecting the subject in a generally negative way.

b.

được occurs in parallel sentence forms where the event depicted by the main verb is understood to affect the subject in a generally positive way.

Ðược itself appears to be cognate with Chinese 得 ‘to get’ (pronounced as dé in modern Mandarin, and as dak in Cantonese), and has a main verb use with the meaning ‘to get/receive’, as well as a post-verbal use as a modal with the meaning ‘to be able to’ (very similar to modern Cantonese dak, as described in Simpson 2001). Example (32) illustrates the use of được in a passive frame parallel to bị. (31) Nam bị thầy giáo phạt. Nam BI teacher punish ‘Nam was punished by the teacher.’ (32)

Nam được thầy giáo khen. Nam DUOC teacher praise ‘Nam was praised by the teacher.’

Vietnamese and the typology of passive constructions

165

Structurally, được ‘passives’ correspond fully to bị passives and allow for the same kinds of syntactic patterns.4 Bị and được therefore seem to simply be two (semantically different) values of the same functional verb type used to encode passive in Vietnamese. Example (33) shows how được can occur in an indirect passive-type use (with beneficial effect), similar to the use of bị in (24): (33)

Tôi được Nga đọc lá thư của tôi. I DUOC Nga read letter of I ‘I had Nga read my letter.’

In terms of meaning and patterns of use, bị most commonly occurs with verbs which encode an obviously unpleasant action on their objects, hence verbs such as ‘criticize’, ‘hit’ etc., rather than verbs indicating a positive effect on their objects, e.g. ‘praise’, which naturally occur with được. However, verbs such as ‘praise’ may in fact occur with bị if the effect of the action of the verb is contextually understood as being negative (e.g. creating embarrassment for the subject), and verbs such as ‘punish’ may occur with được if the action of ‘punishing’ is somehow contextually understood to be positive for the subject: (34) Nam bị thầy giáo khen. Nam BI teacher praise ‘Nam was praised by the teacher.’ (35)

Nam được thầy giáo phạt. Nam DUOC teacher punish ‘Nam was punished by the teacher.’

Consequently, interpretations of the subject being negatively or positively affected by the action of the verb in the Vietnamese passive are primarily a function of

4 In this way, Vietnamese appears to be different from another Southeast Asian language with similar patterns – Thai. In Thai, the corresponding ‘positive effect’ passive verb combination day-rap (in which the day component shows clear signs of being related to Chinese dé/dak and Vietnamese được – Simpson 2001) does not permit a long passive form with an overt Agent, unlike the negative effect passive formed with thuuk (Iwasaki and Ingkaphirom 2005): (ii) phom day-rap (*khruu) chom I PASS teacher praise ‘I was praised (by the teacher).’

166

Andrew Simpson and Hồ Hảo Tâm

the choice of bị and được, and not principally dictated by the content of the main descriptive verb.

4.2 Passives of intransitive verbs A second, particularly striking syntactic property of Vietnamese bị passives, which distinguishes them from Chinese bèi sentences and passives in most other languages, is the occurrence of intransitive verbs in the bị passive frame. This is a frequent property of intransitive verbs referring to unpleasant states or actions, as illustrated in examples (36) and (37) referring to sickness: (37)

Nga bị ốm/bệnh. Nga BI sick/ill ‘Nga got sick.’

(38) Nga bị bệnh ung thư. Nga BI ill cancer ‘Nga got cancer.’ Verbs of this type often occur with bị, but they also can occur without bị in non-passive clauses: (39)

Tôi nghe nói là Nam ốm/bệnh lắm. I hear say C Nam sick/ill much ‘I heard that Nam is very ill.’

(40)

Nam đang ốm/bệnh (lắm). Nam PROG ill/sick much ‘Nam is very sick.’

Examples (41–45) provide further illustration of intransitive passives referring to bodily conditions and actions which are viewed as negative. Both new, longterm states such as ‘blindness’ and ‘becoming crippled’ as well as short-term physical experiences such as ‘coughing’ and ‘vomiting’ occur naturally in these passive-of-intransitive verb structures, and terminal negative events such as ‘drowning’ may also be represented with a passive structure:

Vietnamese and the typology of passive constructions

(41)

167

Nam bị mù. Nam BI blind. ‘Nam is/became blind.’

(42) Nam bị tàn tật. Nam BI crippled ‘Nam is/became crippled.’ (43) Nam bị ho. Nam BI cough ‘Nam coughed.’ (44)

Nam bị ói. Nam BI vomit ‘Nam vomited.’

(45)

Nam bị chết đuối. Nam BI drown ‘Nam drowned.’

This kind of passive structure embedding intransitive verbs is not at all possible in Chinese, as illustrated in (46) and (47), and represents a very clear difference between Chinese and Vietnamese: (46)

*Tā bèi bìng-le. he BEI sick-ASP

(47)

*Ta bèi késòu-le he BEI cough-ASP

Presently, it will be seen that the occurrence of apparent intransitive passive forms in Vietnamese also has significant consequences for any characterization of ‘passive’ in terms of universal, cross-linguistic properties.

5 Significance of the patterns for functional and theoretical approaches to passive The Vietnamese patterns presented above, and particularly those in section 4.2, are significant for both formal and functional analyses of the passive as a construction having clearly definable, cross-linguistically stable properties. Func-

168

Andrew Simpson and Hồ Hảo Tâm

tional descriptions of the passive frequently claim that passive constructions exist to fulfill either one or both of the following manipulations of perspective/ viewpoint (Givón 1984; Shibatani 1985, 1988):

AGENT DEMOTION – removal of the Agent from prominent subject position and demotion to a less salient role in the syntactic structure (or full elimination of the Agent from the sentence) PATIENT

PROMOTION

– promotion of the Patient from object to subject

position In the Vietnamese passive of intransitive verbs, however, there is neither any agent demotion, nor any patient promotion, and the prominence of the single argument of the verb is not changed by the use of a passive structure. The function of the use of passive morpheme bị in such sentences is to signal and emphasize the negative impact of the event on the subject of the verb. The extension of passive bị to such intransitive verbs thus poses a clear challenge to current, heavily restrictive classifications of passive morphology and syntactic structure in terms of their functional use. With regard to the formal, generative modeling of the passive within Government and Binding Theory and various of its ‘Principles and Parameters’ successors, the surface syntactic properties of passive sentences in European languages such as Italian, English and German have been suggested to be due to two common underlying features of passive (Burzio 1986; Haegeman 1991): (48)

a.

Passivization eliminates the accusative case assigning potential of the verb

b. Passivization eliminates the external theta role of the verb The interaction of (48a) and (48b) is suggested to cause the Patient/Object argument of a passivized verb to undergo movement to the subject position of a finite clause to be assigned/check case. As pointed out by Huang (1999), however, the patterning of passive constructions in Chinese and other East Asian languages necessitates a re-assessment of (48a/b) when considered as putative cross-linguistic properties of the passive.5 In Chinese-type passives, there is no evidence that any accusative case assigning potential of the verb is lost, and overt NP objects may still occur in canonical post-verbal positions in passive

5 See also Simpson (1990) for similar conclusions based on Thai.

Vietnamese and the typology of passive constructions

169

sentences. This is illustrated in the ‘retained object’ indirect passive examples (19–21). In Chinese long passives, the Agent argument of the verb is also not eliminated, and may surface overtly, as in examples (1–3, 6, 8, 9). Neither of the core properties of passive identified on the basis of Romance and Germanic languages seem to be relevant for languages such as Chinese. While East Asian languages therefore clearly question the validity of (48a/b) as potentially definitive, cross-linguistic properties of passive structures, Huang (1999: 67) suggests that it may still be possible to identify certain basic shared features of passive constructions across typologically diverse languages: (49) ‘. . . there is nevertheless a universal notion of passivization that can be maintained, namely that all passives involve intransitivization and a dependency relation between the surface subject and underlying object position . . .’ Such a revised perspective on the passive requires a little further explanation before we consider the relevance of the patterns found in Vietnamese. Specifically, with regard to indirect passives, where the direct object of the verb is overtly present and not directly linked to the surface subject position (examples 19–21), Huang (1999) suggests that the surface subject is actually linked to an ‘outer object’ position in the embedded clause. It is proposed that an empty operator originates in a higher object position within VP, raises to a clauseperiphery position as in other instances of long passive, and binds a pro in the possessor position of the direct object/Patient NP, as illustrated in (50): (50)

Zhāngsān bèi [Op i tŭfĕi [VP t i dă-sĭ-le [pro i fùqīn]]. Zhangsan BEI bandit hit-die-ASP father

The possible occurrence of such structures is suggested to be due to an ability of Chinese to case-license the outer object base position of the empty operator. This availability of case for outer objects in Chinese is itself argued to be manifested in a second significant pattern and occur in the ‘ba construction’, where overt NPs are introduced and licensed by the element ba in outer object positions:6 6 Similar suggestions that the occurrence of indirect passives in Chinese may in some way be related to the bă-construction are given in Shi (1997), who points to clear parallelisms between many ‘retained object’ indirect passives and bă-forms: (i) Huā bèi wŏ jiāo-le shuĭ. flower BEI I add-ASP water ‘The flowers were watered by me.’

(ii) Wŏ bă huā jiāo-le shuĭ. I BA flower add-ASP water ‘I watered the flowers.’

170

Andrew Simpson and Hồ Hảo Tâm

(51)

Tŭfĕi bă Zhāngsān dă-sĭ-le fùqīn. bandit BA Zhangsan hit-die-ASP father ‘The bandits killed Zhangsan’s father.’

Considering relevant patterns found in other languages, Huang notes that in Korean similar outer objects are clearly case-marked (with Accusative case), and only possible where the outer object is affected by the action of the verb, as in Chinese: (52)

Mary-ka John-ul tali-lul cha-ess-ta/*po-ess-ta. Mary-NOM John-ACC leg-ACC kick-PAST- DEC / SEE - PAST- DEC ‘Mary kicked/*saw John in the leg.’

Concerning the Chinese forms referred to as adversity passives (examples 27 and 28), Huang (1999) suggests that these are similarly derived by the movement of an empty operator from a higher outer object position. NPs base-generated in such a position are suggested to receive a theta role with the meaning of ‘entity adversely affected by the action of the verb’.7 Such an analysis of indirect and adversity passives has two immediate consequences, both of which seem to be positive. First, Huang is able to maintain that passive sentences in Chinese uniformly incorporate a dependency between the surface subject position and some underlying object position – either the direct object position, or one of the two outer object positions. This subsequently allows for the statement of (49) as a putatively general property of passive both in Chinese and other languages. Second, the case-theoretic approach to indirect passives allows for a principled way to describe and possibly even predict crosslinguistic variation in the occurrence of such structures. Earlier it was noted that ‘non-gap’ adversity passives do not occur in Vietnamese, unlike Chinese. This difference between Chinese and Vietnamese may be attributed to differences in the availability of abstract case in the two languages. The objective case which is suggested to license higher outer objects in adversity passives in Chinese may be (iii) Nà kuài dì bèi tāmen zhòng-le guā. (iv) Tā bă nà kuài dì zhòng-le guā that CL land BEI they plant-ASP melon he BA that CL land plant-ASP melon ‘They planted melons in that bit of land.’ ‘They planted melons in that bit of land.’ 7 Due to this theta-related restriction on interpretation, indirect passives with no clear meaning of adversity implied by the predicate are not acceptable: (i) *Zhāngsān bèi Lĭsì păo huí jiā qù le. Zhangsan BEI Lisi run return home go ASP Intended: ‘Zhangsan suffered from Lisi running back home.’

Vietnamese and the typology of passive constructions

171

suggested to be unavailable in Vietnamese, accounting for the unacceptability of forms such as (27) and (28) in Vietnamese.8 Having clarified the status of indirect ‘retained object’ passives, we are now in a position to reflect on how Vietnamese and certain of its passive structures may impact on (49). This redefined, cross-linguistic characterization of passive as minimally and necessarily involving a dependency between a surface subject and an underlying object position, inspired by differences between languages such as Chinese and English, Italian etc., would seem to require further reconsideration as a result of the Vietnamese data presented in section 4.2. Whereas much of what is found in Vietnamese can be satisfactorily modeled with the basic analysis of passive in Chinese presented in Huang (1999) (also Ting 1998; Tang 2001), there are certain patterns which cannot be straightforwardly accounted for in the analysis as it stands, and Vietnamese significantly seems to extend the use of passive structures from the canonical linking of a subject with an underlying object position to other dependencies which connect a surface subject and a second underlying subject position. This highly distinctive use of the passive was illustrated in section 4.2, where it was shown that the subjects of intransitive verbs may participate in bị constructions in a way parallel to the objects of transitive verbs. The occurrence of such patterns therefore calls into question whether a restriction can be placed on characterizations of the passive limiting it to cases where the surface subject of a passive structure is connected (by movement or operator-mediated secondary predication) only to underlying object positions. Rather, it would seem that the possible boundaries of what is commonly referred to as passive may need to be recognized as less narrowly defined, and may in theory also connect a surface subject to other syntactic/argument positions located in the same clause or alternatively an embedded clause in various East Asian languages.9 8 The fact that Vietnamese permits indirect passives with retained objects that are possessed body-parts but not kin terms, unlike Chinese, may however call for a finer understanding of the hypothesized case-licensing of outer objects. It may be that ‘kin term’ retained object passives are licensed in the same way as adversity passives, both as higher outer objects, hence the availability of the former may be linked to that of the latter: both licensed in Chinese, neither possible for speakers of Vietnamese. 9 A reviewer points out that the approach to passive developed in Keenan (1980, 1985) and Keenan and Timberlake (1985) may allow for the occurrence of a wider array of passive structures than the more classical GB view (and its extensions) described here. According to Keenan, passivization is an operation that existentially binds the highest argument of a predicate that it applies to. In many languages this may be limited to being an external argument, but in some languages it is not, giving rise to the occurrence of impersonal passives of unaccusatives. However, the reviewer suggests that such a broader conceptualization of passive may still not capture the Vietnamese patterns, as there may be no existential closure of an argument in such structures.

172

Andrew Simpson and Hồ Hảo Tâm

The ‘subject passive’ patterns found with bị and intransitive verbs denoting an unwelcome outcome/experience for their subjects can additionally be noted to extend further in Vietnamese, in two directions. First, there are instances where the surface subject of a bị sentence can form a dependency with the subject gap position of a transitive verb, when the latter describes an action that is obviously unpleasant and which may involve suffering on the part of the subject, as illustrated in (53) and (54): (53)

Nam bị xem một phim kinh dị. Nam BI watch one film horror ‘Nam watched a horror film (and this was unpleasant for Nam).’

(54)

Sắp bị lọt vào miệng con quái vật thì . . . ASP BI fall enter mouth animal odd creature then ‘He was about to fall into the monster’s mouth when . . .’ (Daley 1998: 92)

Second, the ‘positive experience passive’ verb counterpart to bị in Vietnamese – được – also regularly occurs with its subject linked to a lower subject position: (55)

Nam được đi Pari. Nam DUOC go Paris ‘Nam went/got to go to Paris’.

Therefore with both bị and được both subject-to-object and subject-tosubject dependencies are possible in structures which are built with these ‘negative/positive passive’ morphemes. This is schematized in (56), where square brackets around an NP in the embedded clause indicate the gap position linked to the subject of the sentence. (56)

subject-to-object dependencies NPi bị/được [NPk verb [NPi ]]

‘transitive passive’

subject-to-subject dependencies NPi bị/được [[NPi ] verb]

‘intransitive passive’10

NPi bị/được [[NPi ] verb NPk ]

‘transitive passive’

10 The set of intransitive verbs which occur with bị are non-agentive undergoers of an unpleasant state. In this sense, they correspond to verbs of the unaccusative grouping, which are often non-agentive cross-linguistically. Clear, independent, syntactic tests for diagnosing unaccusativity and distinguishing unaccusative from unergative verbs in Vietnamese have thus far not been identified.

Vietnamese and the typology of passive constructions

173

Vietnamese thus makes use of an extensive array of linking options in the projection of passive-related structures, challenging assumptions about the necessary limits of such forms and raising new questions for both the formal and functional modeling of the passive. Previously, initial cross-linguistic characterizations of the passive based on patterns in European languages were modified by the observation of non-prototypical (though robust) forms such as the passive of unergative intransitives in German, illustrated in (57), where agent demotion occurs but no patient promotion (the pronoun es ‘it’ in (57) is expletive): (57)

Es wurde getanzt. it became danced Lit: ‘It was danced.’

East Asian languages, such as Chinese (also Japanese, Korean, Thai) then forced a further re-consideration of universal properties of the passive, as noted above and discussed at length in Huang (1999). Vietnamese with its bi-clausal subject dependency passives now indicates an additional limit of variation which needs to be factored into and acknowledged in global descriptions of the ‘passive’. Given what is observed in Vietnamese, a universal set of ‘minimal required properties’ of the passive may need to acknowledge that prototypical passive constructions may be stretched to incorporate (and be licensed by) dependencies between two subject positions, bringing the passive syntactically close, in this instance, to the constructions commonly referred to as ‘Control structures’, where the reference value of the covert subject of an embedded clause is provided (‘controlled’) by the subject of a higher clause, as in (58): (58) John i wanted [PRO i to leave]. The potential similarity of subject dependency passives such as (53–55) to Control structures raises interesting questions about the syntactic distinctions between passive and Control structures in particular, and, in general, whether the ‘passive’ as a syntactic construction can still be isolated from other constructions in a meaningful way through the identification of a distinct set of ‘passive’ properties. Increasingly it may seem that the boundaries between what can be labeled ‘passive’ and the properties of various other constructions are not so fully clear and significant.11 11 With regard to the passive/Control similarities referred to briefly here, it is interesting to note that recent analyses of Control structures have argued for a movement approach and against earlier ideas that ‘Control’ sentences are formed with a base-generated PRO element in

174

Andrew Simpson and Hồ Hảo Tâm

Faced with such issues, a number of theoretical positions might be considered. One initial reaction to the Vietnamese and Chinese patterns presented here and in Huang (1999) might be to suggest that the term ‘passive’ should in fact simply be discarded, being too broad in what it appears to include across languages to allow for a distinct and exclusive set of properties to be defined for such a construction. Such a position would concede that there would be no real theoretical value in the continued use of the label passive as a term with intended universal, cross-linguistic application. A second kind of stance that might perhaps be adopted would be to suggest that the term ‘passive’ still be maintained, but in a narrow, more restrictive way, to identify just those archetypal passive forms found in languages such as English, Italian etc, with common case and theta-theoretic properties, and that ‘passive’ not be applied as a term to refer to constructions with similar meanings (and a partially-overlapping syntax) in East and Southeast Asian languages (e.g. Chinese, Japanese, Vietnamese, Hmong, Thai etc). Such a position would potentially allow for the term ‘passive’ to be maintained with a clearly-defined theoretical identity, but might also seem to be imposing boundary conditions on membership in the passive in a rather proscriptive and even arbitrary way (for example, if loss of accusative case is presented as a necessary property of passive verbs, this will exclude unergative passives in languages such as German which otherwise look very much as if they should be referred to as passives). It is not clear that setting highly strict entry definitions of what may be considered a form of the passive is a step that actually makes any real progress in understanding cross-linguistic correspondences and the ways that structures may occur in different languages. A third reaction, similar in direction to the position just outlined, might be to propose that subject-to-subject dependencies in bị and được sentences simply be ignored as potential instances of ‘passive’, so as to maintain a broad (but not too broad) notion of passive still inclusive of Chinese-type (object-gap) passives. However, such a move would again seem to be proscriptive and potentially imposed merely to safeguard established ideas about the passive as a construction. It would seem to disregard the fact that canonical passive structures in Vietnamese clearly do spill over into the occurrence of subject gap ‘passive’ patterns without the presence of any obvious linguistic boundary that might justify their treatment as a different construction type. Given that the same morphemes bị and được occur in both ‘regular’ affected object passives and affected subject structhe subject position of the embedded clause (e.g. Hornstein 1999). The hypothesis that Control structures are the result of movement from the embedded clause subject position brings such forms closer still to ‘passive’ sentences resulting from movement of the subject of the verb in the embedded clause.

Vietnamese and the typology of passive constructions

175

tures, and would seem to be readily recognized as the same morphemes by speakers of Vietnamese, there may be no non-arbitrary reasons to exclude the latter from consideration as non-prototypical members set of a larger passive paradigm with fewer necessarily shared syntactic properties. Despite such arguments against the elimination of bị sentences from consideration as part of a more extended passive paradigm, a reviewer of the chapter in fact advocates adopting either of the latter two positions mentioned above as a way to avoid the conclusion that the boundaries of the passive construction may be less clear than previously imagined, suggesting: ‘Why shouldn’t we just declare that the Vietnamese construction (and probably the Chinese one as well) is not a passive, but is something else.’. The reviewer suggests that instead of being a passive morpheme, bị might be viewed as a general inchoative verb meaning ‘Subject comes to be in the state of Complement’ hence that examples such as (31) might be analyzed as ‘Nam comes to be in the state of being punished by the teacher’. The reviewer adds that such sentences would then not count as passive sentences (though it is perhaps not really evident why they would not count as passive, given the passive meaning embedded under the inchoative verb, as seen in the English gloss provided by the reviewer). The reviewer continues that even if one were to consider examples like (31) with embedded transitive verbs as being passive, one might still think that bị sentences with intransitive verbs are not passive forms, and it might be possible that there are two (or more) bị elements in Vietnamese, one a passive morpheme, the other(s) not (the latter occurring with intransitive verbs). However, the possibility of some accidental polysemy in Vietnamese resulting in the chance occurrence of two (or more) independent morphemes with the same form bị can be fairly swiftly discounted, given the common distinctive meaning associated with bị in its regular occurrences with both transitive and intransitive verbs. As noted earlier, bị is not just some general inchoative verb with the meaning ‘BECOME’, but consistently functions to signal that the subject of the sentence is affected in a negative, adversative way by the event depicted in the predicate/ clause following bị. This highly specialized semantic function of bị occurs not only when bị combines with transitive verbs but also with intransitives (and with subject gap transitives such as (53) and (54) too). It is therefore clear that bị is the same morpheme in all its occurrences and a uniform analysis is consequently desirable. As a single uniform analysis for all bi forms furthermore can be assumed (see below), developing and supplementing Huang’s analysis of passive forms in Chinese, by Occam’s Razor it seems unwarranted to attempt to forcibly declare that the combination of bị with transitive and intransitive verbs be categorized as constructions of different types with different labels.

176

Andrew Simpson and Hồ Hảo Tâm

Quite generally, then, the occurrence of bị (and được) subject gap dependencies can be shown to lead to challenging general questions about the way that linguistic terms are used and correspond to clearly definable constructs in natural language. Similar issues have previously been raised with regard to the use of the linguistic label ‘subject’, and whether this term picks out a unitary cross-linguistic phenomenon which is definable by a set of universal properties, or whether ‘subjects’ are best viewed as NPs sharing a sub-set of prototypical properties (Keenan 1976). However such terminological disputes may ultimately be settled (or continue to cause disagreement)12, empirically there still remains the puzzle of a set of forms that interestingly occur in one language – the bị (and được) subject gap dependencies in Vietnamese) – but unexpectedly do not get repeated in other languages, even though appropriate conditions for their occurrence might seem to be present. In other words, it remains to be explained why forms such as (37–38), (41–45) and (53–54) are syntactically well-formed in Vietnamese but not in Chinese and other languages with similar bi-clausal passive structures formed with morphemes entailing negative effects (i.e. Mandarin bèi, Thai thuuk etc.). In trying to identify what makes Vietnamese ‘special’ in allowing for these kinds of subject-gap dependencies built with the ‘passive’ morpheme, one might wonder whether this could be related to another way in which Vietnamese bị seems to be different from Chinese bèi – its selection of a range of different complement types. Although less frequent in occurrence than the combination of bị with a clause/verbal predicate, ‘passive’ bị may actually be combined with a variety of non-verbal constituents, such as nouns/NPs, adverbs, and adjectives.13 This is illustrated in (59–62) below: (59)

bị + adverb bị chậm ‘be delayed’ chậm ‘slowly (adv.)’ (đi chậm ‘go slowly’)

(60)

bị + noun bị hoang tưởng ‘be paranoid’

hoang tưởng ‘delirium (N)’

bị virus ‘get a virus’ bị nạn lụt ‘be flooded’

virus (N) nạn lụt ‘flood disaster (N)’

12 Croft (2001), for example, takes the extreme position in his Radical Construction Grammar that grammatical categories may actually be language-specific or even construction-specific, and so cannot be compared cross-linguistically. 13 Though it might perhaps be argued that there is strictly no separate category of adjectives in Vietnamese, only ‘stative verbs’, and that adjectival elements are also verbal in nature. Thanks to a reviewer for bringing up this point.

Vietnamese and the typology of passive constructions

(61) a.

Computer của anh ấy bị virus. compute of he BI virus ‘His computer got a virus.’14

(62)

bị + adjective bị nghèo đi ‘be impoverished’ bị hư ‘be damaged’

b.

177

Mỹ bị nạn lụt. USA BI flood ‘The US was flooded.’

nghèo ‘poor (Adj)’ hự ‘damaged (Adj)’

While this ability of bị to combine with a range of complement types further distinguishes it from Mandarin bèi, it is actually not clear that it could be responsible for the Vietnamese-Mandarin difference with regard to subject-gap dependencies with bị /bèi. What seems to be minimally necessary for such a dependency to occur is the presence of a constituent that can contain a subject NP, hence a vP/VP or a TP, and it has been argued in Huang (1999) that Mandarin bèi can actually select either vP or TP as its complement (for short and long passives). Consequently, Mandarin would not seem to be lacking in any relevant complement type that Vietnamese permits in a special way for the occurrence of subject gap ‘passives’. The more one considers the Vietnamese-Chinese difference in subject gap ‘passives’, the more it may seem that it is actually Chinese which is in need of some special explanation rather than Vietnamese, and the question should more naturally be: ‘Why aren’t more languages like Vietnamese in allowing for subject gaps in clauses embedded by morphemes such as bị/bèi etc?’ The meaning contribution of bị and bèi (and also Thai thuuk) is essentially the same in the sentences they introduce, and communicates that some unwelcome mental/ physical experience occurs for the subject of the sentence, and this referent is negatively affected as a consequence.15 Frequently, this will result in the bị/bèi

14 Concerning (61a), it can be noted that the word ‘virus’ can be modified by a quantifier and by adjectives or PPs, which confirms its nominal status: (i) Computer của anh ấy bị nhiều/mấi loại virus nặng lắm/từ nước Đức. computer of he BI many/several type virus serious very from Germany ‘His computer got many/several types of very serious virus from Germany.’ 15 This is indeed the most common interpretation in bèi/bị passives, and the regular situation with NP subjects which are animate. However, both Mandarin and Vietnamese have extended the use of passive structures to allow for inanimate NPs to occur as subjects, in certain situations and registers. In various instances, this does not cause an interpretation in which any discourse referent is negatively affected, and may be the result of the influence from passive structures in languages such as English, where no negative impact interpretation is necessary with animate subjects and inanimate subjects are common in the passive.

178

Andrew Simpson and Hồ Hảo Tâm

subject corresponding to the patient argument of the embedded clause, hence an object gap, but there is also a natural expectation that it could alternatively be interpreted as the subject of the embedded clause, if this NP is understood to suffer an unwelcome state/event, as with the subjects of negative impact predicates such as verbs of sickness, psychological fear etc. Consequently, the ungrammaticality of examples such as (46) and (47) would not seem to be attributable to any reasons of semantic ill-formedness, and might well be expected to occur in Mandarin, Thai, and other languages with similar passive structures. The fact that they do not occur, despite the presence of appropriate semantic licensing conditions, would seem to clearly suggest that they are disallowed for syntactic reasons.16 The generalization that needs to be accounted for is that, in the analysis of operator-movement assumed for Chinese bèi passives (and Vietnamese bị sentences), extraction and operator movement can occur from object positions, but (in Chinese) not from subject positions. This suggests that what may be syntactically relevant in accounting for the difference between Vietnamese and Chinese is a principle regulating the extraction of subject NPs, operating successfully in Vietnamese bị sentences, but failing to license subject extraction in Mandarin. Cross-linguistically, there are actually many instances of constructions in which subject/object asymmetries in extraction are found, where bans on the extraction of subjects exist and may (sometimes) be overcome only with the use of some additional licensing mechanism. For example, tough constructions in English permit the extraction of an empty operator from object position, but not from subject positions: (63) John is easy [Op k PRO to please t k ]. (64) *John is easy [Op k t k to be happy]. (65) *John is easy [Op k t k to please Mary].

16 It also seem implausible to suggest that the difference between Vietnamese and Mandarin with regard to subject gap ‘passives’ can be dismissed as an issue of simple pragmatics, claiming that Vietnamese is just more permissive than Chinese, therefore allowing for subjectgaps in bị sentences. Elsewhere it has been seen that Vietnamese is actually less permissive than Mandarin in its bị passives, and does not allow for pure adversity-type passives or indirect passives involving possessed kinship terms in the object of verb position (examples 25, 26, 29). This latter difference was given a syntactic explanation in terms of case availability for outer objects. An explanation of subject gap passives in terms of some parametrized syntactic principle would therefore seem to be desirable, licensing such passive forms in Vietnamese but not in Chinese.

Vietnamese and the typology of passive constructions

179

The factor often used to account for the extraction asymmetry in tough constructions is the presence/absence of case in the position of extraction. A’-operator movement is critically assumed to be possible only from case-marked positions, hence not from the SpecTP position of non-finite clauses. Though a case-based account of operator-movement will correctly characterize instances of tough-movement, as in (63–65), it would not seem to allow for an explanation of the differences between Vietnamese and Chinese under consideration here. The simple reason for this is that case appears to be readily available in the subject position of clauses embedded by bèi in Mandarin, licensing overt subjects in instances of the long passive. This presence of case might consequently be expected to allow for subject operator movement if this were to be the relevant licensing factor, yet subject gap passives clearly do not occur in Chinese. Looking elsewhere for useful clues, there are further instances of asymmetries in the extraction of subjects and objects in certain languages which are not attributable to case and which are therefore more informative for the analysis of the Chinese/Vietnamese difference. Two well-documented examples of this in the literature are the that-trace paradigm in English, and qui/que alternations in French relativization.17 In both constructions, it has been argued in Rizzi (1990) that licensing of the extraction-site of empty operator movement is regularly achieved through (proper) government of the subject position by an appropriate C, significantly failing to occur when ‘that’ occupies the C position in English and when qui occurs in C in French. Quite generally, extraction from 17 The ‘that trace effect’ refers to the observation that extraction of a subject from a position immediately following the complementizer that is regularly judged to be deviant and unacceptable, as in (i) (Haegeman 1991: 362): (i) Who i do you think that t i left?

In qui/que alternations (Rizzi 1990: 56), the regular complementizer que occurs in a special form qui when there is subject extraction from a position immediately following the complementizer, as in the relativization of a null operator in cases such as (ii). Where there is relativization of a null operator from other, non-subject positions, as in (iii), the regular form of the complementizer que occurs: (ii) la personne [Oi qui t i est venu] the person C is come ‘the person who came’ vue t i ] (iii) la personne [O i que j’ai the person C I-have seen ‘the person I saw’

180

Andrew Simpson and Hồ Hảo Tâm

subject (and object) position is characterized as requiring a formal licensing by some adjacent c-commanding head, and may not be possible if an appropriate head is either lexically not available or contextually not supplied. Such a line of analysis seems to be much more promising as a precedent and parallel for attempting to explain the Vietnamese/Chinese differences with regard to subject gaps in bị and bèi sentences. Adopting the approach in Rizzi (1990), it can be suggested that what underlies the difference between Vietnamese and Chinese in bị/bèi constructions is the nature of the head which potentially licenses extraction from SpecTP in the embedded clause of bị and bèi sentences. In the analysis of Huang (1999), the clause embedded in long passive sentences is actually argued to be IP/TP rather than CP (with the empty operator moving to adjoin to IP/TP). Consequently, the head position immediately c-commanding SpecTP in the embedded clause, and hence the extraction-site of the subject, will in fact be the higher clause predicate position instantiated by bèi. It can therefore be suggested that the difference between Vietnamese and Mandarin lies directly in the lexical difference between bèi and bị and their role in licensing the extraction of a following subject. Bèi, like English complementizer that, can be suggested to lack the ability to license extraction of an operator from SpecTP, with the result that subject gap passive forms cannot be created in Mandarin. Bị, on the other hand, can be suggested to pattern like other lexical elements such as verbs and retain the ability to license extraction from an appropriate, adjacent, c-commanded position – in the instance under discussion, the SpecTP position of the clause embedded by bị. Subject-to-subject dependencies in bị sentences formed by extraction of an operator from SpecTP are therefore licensed to occur and syntactically made available (in principle) in bị sentences unlike in Mandarin bèi constructions.18 Such a relatively simple, lexically-parametrized difference between bị and bèi will effectively suffice to account for the robust non-occurrence of subject gap passives in Mandarin (and also Thai), which otherwise remain unaccounted for, though are (arguably) predicted to occur given the similar semantic content of bị and bèi (and Thai thuuk). Quite possibly, the syntactic difference assumed to exist between bị and bèi might also be hypothesized to be part of a broader difference in diachronic development exhibited by the two elements. Specifically, if the general process of grammaticalization, where lexical elements become more functional over time, involves an incremental loss of properties typically associated with lexical 18 The full range of occurrence of subject gap ‘passives’ in bị sentences may, of course, turn out to be further constrained by certain other (as yet unidentified) pragmatic, semantic or even syntactic factors.

Vietnamese and the typology of passive constructions

181

elements (Roberts and Roussou 2003), it may be speculated that the inability of bèi to license subject gaps may be a manifestation of a more advanced process of grammaticalization having occurred with bèi than with bị, the latter element retaining more of its source properties as a verb, which cross-linguistically are more commonly licensors of movement and extraction.19 Finally, such a hypothetically less grammaticalized status of bị might also link up naturally with the observation that, unlike bèi, bị allows for a range of lexical complements in addition to clauses (examples 59–62), hence patterns more globally in a less purely auxiliary-type way than Mandarin bèi. A general, principled view of the intriguing differences between bị and bèi sentences is thus possible, and can be grounded in a plausible, well-established analysis of cross-linguistic restrictions on subject extraction. With this as our present conjecture about bị and bèi and the syntax of bi-clausal ‘passive’ constructions, we will close the chapter now and summarize a number of the key properties of this challenging and revealing area of Vietnamese syntax.

6 Summary of similarities and differences in Chinese and Vietnamese passives As noted in the introduction, the primary concern of this chapter has been how passive-type constructions in Vietnamese compare with similar structures found in other languages, in particular Mandarin Chinese, and how patterns found in Vietnamese have a significance for the establishment and defense of a general cross-linguistic typology of passive, within both formal and functional approaches to linguistics. In terms of relevant similarities between Vietnamese and Chinese, it was shown that both languages present evidence for a bi-clausal structure in passives, and the existence of a formally distinct long and short passive with different syntactic and structural properties. In long passive structures in Vietnamese and Chinese, the evidence available suggests the occurrence of A’-movement in the lower clause creating a gap position linked in its interpretation to the NP in main clause subject position, preceding bị and bèi. Vietnamese and Chinese were also seen to both exhibit retained object ‘possessor-passives’ in addition to simple object-gap passives, where the main clause subject is interpreted as 19 Vietnamese bị, like Mandarin bèi, shows certain patterns which distinguish it from simple lexical verbs, however, for example the ability to occur as a short answer form in yes/no questions. This patterning is described and commented on in Appendix 2.

182

Andrew Simpson and Hồ Hảo Tâm

co-referential with the implicit possessor of the object of the embedded clause, when this is a body-part or certain other kinds of entity. Such clear similarities between Vietnamese and Chinese were taken to indicate that the analysis of bị and bèi sentences in the two languages should essentially be parallel, at least in primary, fundamental aspects of their underlying syntax, and an extension of Huang’s (1999) analysis was shown to be able to account for much core data in Vietnamese. Vietnamese and Mandarin bị and bèi sentences were then shown to be interestingly different in various ways. First it was noted that Chinese permits pure ‘adversity passives’ with no gap corresponding to either the object of the verb or the possessor of the object, whereas Vietnamese does not appear to allow for such structures with bị. Second, the chapter highlighted the fact that Vietnamese makes productive use of two different morphemes in passive type structures, one for events with negative impacts/outcomes (bị), the other for positive impact events (được), while Mandarin lacks an equivalent for the latter được-based forms. Third, it was seen that Vietnamese bị can combine with a range of complement types with a similar passive-like meaning of being negatively affected, again showing a difference to Mandarin bèi. Finally, the chapter discussed the occurrence of subject-gap ‘passive’ structures, both with intransitive verbs and transitive verbs, which are present in Vietnamese, but not in Mandarin. It is these differences between bị and bèi sentences which turn out to be the most intriguing aspect of a comparison of Vietnamese and Mandarin, giving rise to interesting questions and challenges for the typology of passive, and the general question of whether passive indeed exists as a meaningful and definable linguistic construction. Last of all, and independently of how the broader theoretical issues are ultimately resolved, the presence of ‘subject gap’ forms such as (37– 38), (41–45) and (53–55) in Vietnamese but not Mandarin poses the interesting challenge of how to account for the exclusion of such forms from one language but their presence in the other when it seems that appropriate pragmatic-semantic conditions for their occurrence are regularly met within both languages in bèi/bị forms.

References Browning, Marguerite. 1987. Null operator constructions. Ph.D. dissertation, Department of Linguistics and Philosophy, Massachusetts Institute of Technology. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. Chomsky, Noam. 1977. On WH-movement. In: Peter Culicover, Thomas Wasow and Adrian Akmajian (eds.), Formal Syntax, 71–132. New York: Academic Press.

Vietnamese and the typology of passive constructions

183

Cole, Peter, Gabriella Hermon and Li-May Sung. 1990. Principles and parameters of long distance reflexives. Linguistic Inquiry 21: 1–22. Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Daley, Karen. 1998. Vietnamese Classifiers in Narrative Texts. Dallas: Summer Institute of Linguistics. Givón, Talmy. 1984. Syntax: a Functional-Typological Introduction, Vol. 1. Amsterdam: John Benjamins. Haegeman, Liliane. 1991. Introduction to Government and Binding Theory. Oxford: Blackwell. Hornstein, Norbert. 1999. Movement and Control. Linguistic Inquiry 30: 69–96. Hornstein, Norbert. 2001. Move! A Minimalist Theory of Construal. Oxford: Blackwell Publishers. Huang, C.-T. James. 1999. Chinese passives in comparative perspective. Tsing Hua Journal of Chinese Studies 29: 423–509. Huang, C.-T. James, Audrey Li and Yafei Li. 2009. The Syntax of Chinese. Cambridge: Cambridge University Press. Iwasaki, Shoichi and Preeya Ingkaphirom. 2005. A Reference Grammar of Thai. Cambridge: Cambridge University Press. Iwashita, Mami. 2007. Being Affected: the Meanings and Functions of Japanese Passive Constructions. Munich: Lincom. Keenan, Edward. 1976. Towards a universal definition of ‘subject of ’. In: Charles Li (ed.), Subject and Topic, 303–333. New York: Academic Press. Keenan, Edward. 1980. Passive is phrasal (not sentential or lexical). In: Teun Hoekstra, Harry van der Hulst and Michael Moortgat (eds.), Lexical Grammar, 181–214. Dordrecht: Foris. Keenan, Edward. 1985. Passive in the world’s languages. In: Timothy Shopen (ed.), Language Typology and Syntactic Description, Volume 1: Clause Structure, 243–281. Cambridge: Cambridge. University Press. Keenan, Edward and Alan Timberlake. 1985. Predicate formation rules in Universal Grammar. In: Jeffrey Goldberg, Susannah MacKaye and Michael Wescoat (eds.), Proceedings of the 4th West Coast Conference on Formal Linguistics, 123–138. Stanford: CSLI. Le, T.D. 1976. Vietnamese passives. Chicago Linguistics Society 12: 438–449. Pan, Haihua. 1997. Constraints on reflexivization in Chinese. New York: Garland. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Massachusetts: MIT Press. Roberts, Ian and Anna Roussou. 2003. Syntactic Change: a Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Shi, Dingxu. 1997. Issues on Chinese passive. Journal of Chinese Linguistics 25(1): 41–69. Shibatani, Masayoshi. 1985. Passives and related constructions: a prototype analysis. Language 61: 821–848. Shibatani, Masayoshi. 1988. Passive and Voice. Amsterdam: John Benjamins. Simpson, Andrew. 1990. The passive in Thai. Manuscript. University of London. Simpson, Andrew. 2001. Focus, presupposition and light predicate raising in Southeast Asian languages and Chinese. Journal of East Asian Linguistics 10(2): 89–128. Tang, Sze-wing. 2001. A complementation approach to Chinese passives and its consequences. Linguistics 39: 257–295. Ting, Jen. 1998. Deriving the bei-construction in Mandarin Chinese. Journal of East Asian Linguistics 7(4): 319–354.

Theresa Hanske

7 Serial verbs and change of location constructions in Vietnamese* This chapter deals with a specific type of serial verb construction (SVC) in Vietnamese that encodes caused change of location. The construction is of particular theoretical interest because it exhibits an alternation that is triggered by aspectuality, subject to the dynamic vs. stative character of the verb that occupies the second verb (V2 ) position within the SVC. Caused change of location constructions with dynamic V2 s refer to events as ongoing or completed actions. The constructions with stative V2 s, however, denote the result that is brought about by the event to which the first verb in the series (V1 ) refers, yielding a perfect of result reading. The aspectual alternation is explained by a difference in inherent configurations of phases and boundaries. While the dynamic V2 construction type may or may not imply a result, the result is made explicit and enforced on the constructional level in the stative V2 construction type. In the stative V2 construction, both the stative character of the V2 and the lexical presence of a resultative component of the verbs in V1 position are crucial for the perfect of result reading. Keywords: serial verb constructions, aspectuality, situation type, viewpoint, resultativity

1 Introduction This chapter discusses two distinct subtypes of serial verb constructions expressing caused change of location and proposes that the alternation between the two types should be attributed to an aspectual rather than a spatial distinction. I will show that an analysis in terms of aspectually relevant configurations at the constructional level is needed in order to account for this alternation. Additionally, I will show that aspectual distinctions account for the co-occurrence restrictions of certain verb classes within caused change of location constructions.

* I would like to thank Leila Behrens, Pamela Perniss, Tưởng Hùng Nguyễn and two anonymous reviewers for their helpful comments on an earlier version of this chapter. All errors are mine.

186

Theresa Hanske

The two construction types under consideration are exemplified in (1) and (2).1,2 (1) Chị ấy đặt cái cốc lên trên bàn. she put CL cup ascend RN table

Agent V1 Theme V2

dynamic

Goal

(i) ‘She is/was putting the cup on the table.’ (ii) ‘She put the cup on the table.’ (2)

Chị ấy đặt tách trà ở trên bàn. she put tea-cup be.at RN table

Agent V1 Theme V2

stative

Goal

‘She has put the tea-cup on the table.’ Caused change of location constructions refer to events in which an Agent causes a Theme to move to a Goal. In Vietnamese, caused change of location events may be expressed by serial verb constructions (SVCs) which consist of two verbs and their respective arguments (other non-serializing caused change of location constructions exist, but will not be considered here). They have a fixed word order following the pattern NP V1 NP V2 ( RN ) NP, whereby pragmatic non-realization is allowed for the NPs, i.e. any of the NPs can be a zero-argument. The linguistic status of elements such as lên ‘ascend’ (in example (1)) and ở ‘be at’ (in example (2)) is controversially debated in the literature. They have been variously analyzed as verbs, prepositions, or coverbs (Clark 1978; Kuhn 1990; Bisang 1992). The term “coverb” has become common for cases in which they appear with prepositional function with a verb for which main verb status is assumed. However, I analyze these elements as verbs because they also can occur as single verbs of clauses and are thus equivalent to main verbs (Slobin 2004). When they co-occur with another verb in a serial verb construction, I refer to them according to their position within the SVC, i.e. as verbs in second verb position (V2-position) or V2 s, for short. The verbs which denote the subevents causing the change of location, such as đặt ‘put’ in the examples above, are referred to as verbs in first verb position (V1-position or V1 s). I analyze elements

1 If not indicated otherwise, the examples in this chapter are from my own data collected in Hanoi and Köln (spoken narratives and elicitations for which I used stimulus materials such as “The Topological Relations Picture Series” (Melissa Bowerman), “Mouse films” (WDR, Sendung mit der Maus) and “Put films” designed and compiled by the Max-Planck-Institute for Psycholinguistics, Nijmegen). Additionally, examples are taken from my corpus of written texts (CWT). 2 The following abbreviations are used in this chapter: CL : (nominal) classifier; DEM : demonstrative; FUT : future; IMPFV : imperfective viewpoint; PFV : perfective viewpoint; PROG : progressive viewpoint; RN : relator noun; RV : resultative verb; SVC : serial verb construction; TOP : topic marker.

Serial verbs and change of location constructions in Vietnamese

187

such as trên ‘upper part’ in (1) and (2) as relator nouns (RN ), i.e. nouns which function as spatial relators (Thompson 1987).3 The first verb (V1 ) in the caused change of location SVC refers to the action that causes the Theme’s movement, cf. đặt ‘put’ in (1) and (2). In (1), the Theme’s direction and its topological relation to the Goal is expressed by the second verb (V2 ) lên ‘ascend’ and a relator noun (RN ), trên ‘upper part’ respectively. In (2), the V2 ở ‘be at’ indicates a stationary location of the Theme which is further specified as one of contact with a horizontal surface of the Goal bàn ‘table’ by the relator noun trên ‘upper part’. The relator nouns in both subtypes of these serial verb constructions are optional. The exact function of relator nouns occurring alone or in addition to V2 s in caused change of location constructions is complex in itself, but lies outside the present scope of investigation. As such, I exclude relator nouns from my account of caused change of location constructions. The difference between the dynamic (in (1)) vs. stative (in (2)) character of V2 verbs within caused change of location SVCs does not correspond to a semantic difference in respect to the expression of spatial topological relations (e.g. ON, IN relations). That is, both types of verbs can be used to express the same spatial configuration (cf. section 2). However, the use of the dynamic vs. stative V2 verb exhibits a systematic difference in aspectual interpretation, as can be seen in their English translations. With dynamic verbs in V2-position (i.e. the motion verb lên ‘ascend’ in (1)), caused change of location SVCs express the event either as an ongoing action or as a completed action (see (1)). In example (2) the locational stative verb ở ‘be at’ denotes a result state of the action, yielding a perfect of result reading. The sentences are not marked for tense, and accordingly, are open to different interpretations in respect to the relation of the events to the moment of speaking.4 The different aspectual readings of the two construction types will be presented in detail in section 1.2. Section 3 will lay out the theoretical framework applied in my analysis of the alternation.

3 In the literature two possible analyses of these elements are discussed, namely, relator nouns as nouns or as verbs. See Emeneau (1951), Kuhn (1990)) for an analysis as verbs and Clark (1978), Nguyễn (1997), Thompson (1987) for an analysis as nouns. 4 From now on, I will make use of the following conventions for translations of Vietnamese examples: In isolated sentences I distinguish the processual meaning and the completed meaning, where relevant, by use of progressive forms (usually the present progressive) and simple past forms in the English translations. The perfect of result reading is usually conveyed by the English present perfect (if the present perfect is acceptable in the respective English sentence). In examples from narrative texts, the narrative time that is clear from the wider context will be accounted for by the respective English tenses. In examples taken from other sources, I used the translations provided.

188

Theresa Hanske

Section 4 will address the contribution of dynamic and stative V2 s to the aspectuality of caused change of location SVCs and the compositionality of caused change of location phrases. The remaining section relates the phenomenon in Vietnamese to more general issues discussed in the theoretical literature on aspectuality. Like English put, its Vietnamese translational equivalents đặt and để occur in prototypical instances of caused change of location constructions. Apart from placement verbs, verbs from other semantic classes also appear in caused change of location constructions, such as ném ‘throw’ (caused motion verb), dựng ‘lean’ (attachment verb), and xiên ‘pierce’ (impalement verb). The verbs group into two sets according to their distribution in caused change of location SVCs. One set of verbs, which I call put verbs, encompasses placement verbs and attachment verbs. Impalement verbs are subsumed under caused motion verbs, due to their behavior in caused change of location SVCs. While both put verbs and caused motion verbs co-occur with dynamic V2 s, put verbs also occur in the caused change of location SVCs with stative V2 s. As will be shown in section 4.2, membership in the put verb class correlates with occurrence in the stative locative construction with a stationary Theme-subject. In contrast, when the verbs of the other set appear with a Theme-subject, i.e. in a motion construction, they are interpreted as processes.5 In (3) and (4) we see the relevant two classes grouped by constructional subtype. Examples (5)–(8) illustrate their co-occurrence restrictions with the subtypes of caused change of location SVCs. (3)

Alternating verbs (V1 s): verbs combined either with dynamic V2 s or stative V2 in caused change of location constructions Put verbs:6 buộc ‘tie’, dán ‘stick, (make) adhere’, dựng ‘lean’, đặt ‘put’, để ‘put’, treo ‘hang’

5 Note that I treat the SVCs consisting of V1 and V2 (and their respective arguments) as referring to the three-participant events which constitute caused change of location constructions. The two sets in (3) and (4) do not represent a classification of V1-lexemes according to argument structure. 6 There is a further placement verb cho ‘put’ which does not participate in the alternation. The fact that cho ‘put’ behaves differently from đặt ‘put’ and để ‘put’ may be related to its incorporation of a purposive meaning which is not compatible with stative V2s in caused change of location constructions. That is, in addition to its lexical meaning ‘put, give’ and its function as recipient/beneficary marker, cho has assumed the function of a purposive or indirect causative marker.

Serial verbs and change of location constructions in Vietnamese

189

(4) Non-alternating verbs (V1 s): verbs only combined with dynamic V2 s in caused change of location constructions Caused motion verbs: cắm ‘pierce’, đánh ‘hit’, giơ ‘raise, lift’, hất ‘throw’, lăn ‘roll’, ném ‘throw’, nhấc ‘lift’, tung ‘throw’, vứt ‘cast off, throw away’, xiên ‘pierce’, xuyên ‘(make) go through’ (5)

a. Tôi để quả táo vào cái bát. I put CL apple enter CL bowl ‘I am putting/put the apple in the bowl.’ b. Tôi để quả táo ở trong cái bát. I put CL apple be.at RN CL bowl ‘I have put the apple in the bowl.’

(6)

a. Cô dán cái tem vào phong bì. you stick CL stamp enter envelope ‘You are sticking/stuck the stamp on the envelope.’ b. Cô dán cái tem ở phong bì. you stick CL stamp be.at envelope ‘You have stuck the stamp on the envelope.’

(7)

a. Chị ấy ném sách lên giường. she throw book ascend bed ‘She is throwing/threw the book onto the bed.’ b. *Chị ấy ném sách ở giường. she throw book be.at bed intended: ‘She has thrown the book on the bed.’

(8) a.

Chị ấy xiên quả táo qua cái que. she pierce CL apple cross CL stick ‘She is piercing/pierced the apple with a stick.’

b.

*Chị ấy xiên quả táo ở cái que. she pierce CL apple be.at CL stick intended: ‘She has pierced the apple with a stick.’

190

Theresa Hanske

1.1 Compatibility with the progressive marker đang Usually, the progressive test is used to discriminate between states and momentary events on the one hand, and processes on the other (Dowty 1979). As a compatibility test, it delimits states (*Mary is knowing French) from activities (Mary is working) and accomplishments (The child is painting a picture). Achievement predicates exhibit special behavior. Some of them, the so called achievements with a prelude, are compatible with progressive marking (Bill is dying), while momentary achievements are not (*John is finding his key). Compatibility with the progressive, however, does not delimit states unambiguously. It is well known that some state predicates do allow progressive marking, such as be a hero in John is being a hero. With the progressive marking, this subclass of predicates is typically interpreted as denoting temporary states (Dowty 1979; Behrens 1998: 293, EN 33). In Vietnamese, the optional progressive marker đang is likewise sensitive to the inherent aspectuality of predicates and can be used to delineate states and processes, as the opposition in (9) exemplifies. (9)

a. *Quyển sách đang ở trên kệ sách. CL book PROG be.at RN shelf intended: ‘The book is on the shelf.’ b. Chị ấy đang cắt bánh mì trên đĩa. she PROG cut bread RN plate ‘She is cutting the bread on a plate.’

In certain contexts, stative predicates may be used with đang, but in these cases they always have a temporary state meaning. In (10a) this reading is salient because the subject is animate and volitional and therefore may control its locational state. A reading by which a temporary locational state is obtained may be less readily accessible for an inanimate Theme such as ‘the ball’ in (10b). Some speakers accept (10b) under the temporary state reading (while others do not allow for đang in these types of sentences at all, as (9a) demonstrates). With non-statives, đang yields a clear processual reading (compare (9b) and (10)). (10)

a. Chị ấy đang ở nhà. she PROG be.at house ‘She is staying at home.’

Serial verbs and change of location constructions in Vietnamese

191

b. Quả bong đang ở dưới gầm chiếc ghế. CL ball PROG be.at RN space CL chair ‘The ball is lying under the chair.’ The progressive marker đang also discriminates between the two caused change of location construction types. It may be employed with caused change of location SVCs with dynamic V2 s, as (11) demonstrates. Yet, the progressive is blocked from caused change of location SVCs with the stative V2 , shown in (12). When đang is compatible with the contruction type, compatibility holds irrespective of whether the verb in V1-position belongs to the alternating or nonalternating verb class (see (3) and (4)). That đang is blocked from the construction with stative V2 can only be shown for put verbs, since only put verbs occur in this alternant, as (12) illustrates. (11) a.

Nó đang đặt cái hộp vào trong túi. He PROG put CL box enter RN bag ‘He is putting the box into the bag.’

b.

Nó đang dựng cái thang vào bức tường. he PROG lean CL ladder enter CL wall ‘He is leaning the ladder against the wall.’

c.

Nó đang tung bong lên mái nhà. he PROG throw ball ascend roof ‘He is throwing the ball up onto the roof.’

d.

Nó đang xiên quá táo vào cái que. he PROG pierce CL apple enter CL stick ‘He is piercing the apple with a stick.’

(12) a. *Nó đang đặt cái hộp ở trong túi he PROG put CL box be.at RN bag intended: ‘He has been putting the box into the bag.’ b. *Nó đang dựng cái thang ở bức tường. he PROG lean CL ladder be.at CL wall intended: ‘He has been leaning the ladder against the wall.’ Although caused change of location SVCs with stative V2 as well as states cannot get a processual reading, it will become clear in the course of this chapter

192

Theresa Hanske

that the former differ from states. It will be argued that caused change of location SVCs with the stative V2 have a perfect of result reading and that this is responsible for the blocking of đang. I will come back to this issue in greater detail in section 3.3.

1.2 Aspectual readings 1.2.1 Caused change of location SVCs with dynamic V2 s: atelic and telic interpretations on the sentence level As has been shown above, caused change of location SVCs with dynamic V2 s in isolated sentences can have an atelic and a telic reading. This means that they can be used to present an event either as an ongoing action or as a completed action. This ambiguity in respect to aspectuality is reflected in their English translation equivalents. They may be translated with the (present or past) progressive or with simple past verb forms. The examples in (11) demonstrate that the construction with a dynamic V2 is compatible with the progressive marker đang. The combination with đang yields a processual reading on the sentence level. The event is presented as an ongoing action without any reference to its inherent endpoints. Thus, with the progressive, caused change of location SVCs with dynamic V2 s have an atelic reading on the sentence level. In Vietnamese, time adverbials such as trong hai phút ‘lit. in two minutes’, are ambiguous between a durative time reading (‘for x time’) or a punctual time reading (‘in x time’). Due to their lexical ambiguity, they cannot aptly be used as a compatibility test to delimit inherently telic predicates from inherently nontelic predicates. However, it can be shown that caused change of location SVCs with dynamic V2 s pattern with inherently telic predicates by their possible interpretations with trong hai phút ‘in/for two minutes’.7 They have a telic but not an atelic reading ((13i) and (14i) as opposed to (13iii) and (14iii)). Example (13) has yet another interpretation, namely (ii), by which the time adverbial modifies the implicit result. This reading is not available for example (14), as (ii) shows.

7 I have not taken into consideration possible inactual meanings, for example the iterative interpretation, or a prospective meaning corresponding to the adverbial’s third reading as ‘after two minutes’.

Serial verbs and change of location constructions in Vietnamese

193

(13) Chị đặt quả táo lên bàn trong hai phút. she put CL apple ascend table in two minute

(14)

(i)

‘She put the apple on the table in two minutes.’ (telic sense: ‘It took her two minutes to put the apple onto the table.’)

(ii)

‘She put the apple for two minutes on the table.’ (resultative sense: the resultant state of the apple obtains for two minutes.)

(iii)

*‘She put the apple on the table for two minutes.’ (*atelic sense: ‘She spent two minutes putting the apple on the table.’)

Chị ném cành cây xuống đất trong hai phút. she throw branch descend ground in two minute (i) ‘She threw the branch down to the ground in two minutes.’ (telic sense: ‘It took her two minutes to throw the branch to the ground.’ (ii) *‘She threw the branch for two minutes down to the ground.’ (*resultative sense: the resultant state of the branch holds for two minutes.) (iii) *‘She threw the branch down to the ground for two minutes.’ (*atelic sense: ‘She spent two minutes throwing the branch down to the ground.’)

The applicability of the time adverbial trong hai phút ‘in/for two minutes’ confirms the inherent telicity of caused change of location SVCs with dynamic V2 s. However, as noted at the beginning of the section, this does not mean that they must be interpreted as telic on the sentence level. Moreover, as has been shown, their inherent telicity can be neutralized by the progressive marker đang. In discourse, events are presented in relation to each other. Two events may form a sequence of actions (sequential), occur simultaneously (coincidence), or one event may intersect with another ongoing event (incidence). These kinds of relations between events are usually called taxis.8 Taxis configurations trigger particular readings of predicates, namely as referring to a completed, temporallybounded, or ongoing event. In narrative discourse, the dynamic V2 construction type is conventionalized to express events within sequential taxis. The stative V2 8 In languages with a grammaticalized aspect system, the perfective/imperfective opposition is highly sensitive to taxis relations. This is why taxis configurations may be used in these languages as diagnostic contexts for identifying perfective vs. imperfective aspect markers.

194

Theresa Hanske

caused change of location SVC is not employed in these contexts (its use in discourse will be discussed in section 1.2.2). The following examples illustrate the use of dynamic V2 caused change of location SVCs in a sequence of actions, where they refer to realized and completed events. In (15), the caused change of location SVC is completed and then followed by another event. In (16), two caused change of location SVCs with dynamic V2 s are in sequential relation. Here the conjunction rồi ‘and then’ links the successive events. Note that in (16) the stative verb ở ‘be at’ introduces a relative clause; the serial verb of the caused change of location SVC is vào ‘enter’. (15)

Bà mẹ đặt nhẹ thanh củi xuống nền nhà, tiến lại mother put gently stick wood descend floor house advance come.back bên đầu giường, sững sờ nhìn ngắm. side head bed stand.motionless feel see gaze ‘The mother gently set her makeshift club down on the floor and approached the head of the bed, staring in shock.’ (CWT)

(16) rồi tung lên đầu rồi cuối cùng là nó để miếng trứng ở then throw ascend head then end be it put piece egg be.at trên đầu đấy vào chảo. head there enter pan

RN

‘then [it] threw [the egg] up on its head, then in the end it put the piece of egg that was on its head into the pan.’ Sentence (17b) illustrates an incidence configuration. The event expressed by the dynamic V2 caused change of location SVC is the foregrounded event that intersects with the backgrounded event, which in turn is marked by the progressive marker đang. The foregrounded event is again understood to reach its final endpoint, i.e. it is interpreted as completed. This holds irrespective of future time reference marked at the beginning of the paragraph by the future marker sẽ (see (17a)). (17)

a. Chị sẽ gào lên điều đó rất nhiều lần. [. . .] you FUT scream ascend sentence DEM very many times b. rồi ném cái ghế vào chân chị đang bỏ chạy. and then throw CL bench enter leg you PROG leave run ‘You will cry out this many times. [. . .] and [then he’ll] throw it at your leg as you run away.’ (CWT)

Serial verbs and change of location constructions in Vietnamese

195

In (18), the verb đặt ‘put’ from the alternating class occurs in a caused change of location SVC with the dynamic V2 sang ‘pass’. Example (18) has two potential readings: First, the caused change of location SVC refers to an intersecting event within an incidence taxis configuration; second, the first event is anterior to the second, whereas its result is simultaneous to the second event. (18)

Clive đặt quyển sách sang một bên trong khi anh đang ăn cơm. Clive put CL book pass one side while he PROG eat rice ‘Clive put the book aside while he ate his meal.’ (linkdict.com)

The interpretation of caused change of location SVCs with dynamic V2 s as completed events can be neutralized by a second clause which explicitly denies that the inherent endpoint has been reached. In Vietnamese this is possible if a caused motion verb occurs in V1-position. (19) Nó ném quả bong vào lưới; nhưng trượt. he throw CL ball enter net but miss ‘He aimed the ball into the net; but missed.’ (linkdict.com)

1.2.2 Caused change of location SVCs with stative V2: the perfect of result reading The caused change of location SVCs with stative V2 do not have processual readings on the sentence level. The progressive marker đang is blocked with this construction type (see (12)). The construction has a perfect of result reading. It presents the result as a relevant postphasal state which is ascribed to the Theme argument but is still linked to the action which brought about the result. Stative V2 caused change of location SVCs occur in contexts in which the resultant state is especially relevant. In discourse, the construction with a stative V2 may express backgrounded or scene setting events, as in the following fragment from a renarration. (20)

Con chuột thì đang nằm ngủ ở trên giường CL mouse TOP PROG lie sleep be.at RN bed còn con voi thì buộc một chiếc khăn ở trên đầu như tỏ remain CL elephant TOP tie one CL cloth be.at RN head like vẻ rất mệt và ốm yếu. seem very tired and sick.

196

Theresa Hanske

Nó đi lại xung quanh phòng. . . it go come.back surrounding area room ‘The mouse lay in bed in order to sleep, and the elephant had tied a piece of cloth around its head and seemed very tired and sick. It walked around in the room. . .’ In contrast to the caused change of location SVC with a dynamic V2 , the construction with the stative V2 is not action-oriented. In narrative discourse, this construction type is not the conventional means to express completed actions and typically does not appear in sequential taxis configurations: (21) *Bà mẹ đặt nhẹ thanh củi ở nền nhà, rồi. . . mother put gently stick wood be.at floor house and then. . . ‘The mother gently put the club on the floor, and then. . .’ The caused change of location SVC with a stative V2 is not acceptable in the incidence taxis configuration either. The construction is odd here, even if the result is understood as being simultaneous to the ongoing event. (22)

*Clive đặt quyển sách ở một bên trong khi anh đang ăn cơm. Clive put CL book be.at one side while he PROG eat rice intended: ‘Clive had put the book aside while he ate his meal.’

Given that the construction is result-oriented rather than action-oriented, it is not surprising that the stative V2 caused change of location SVC is underrepresented in narrative discourse. In instructions, however, caused change of location SVCs with the stative V2 are more commonly used. In (23), it is the result that ought to be ensured before proceeding to the next step. In contrast to the sequential taxis depicting single actions in sequence, the result here is a precondition. As such, it is especially relevant. Similarly in (24), the stative V2 caused change of location SVC is used to prevent the result (i.e. the altar being in a passageway) rather than the action (i.e. putting the altar somewhere). (23)

Bước 2: Đặt embé ở vị trí an toàn, chẳng hạn trong một chiếc step 2 put baby be.at position safe for example RN one CL nôi. Sau đó bạn có thể bắt đầu chuẩn bị tã. cradle after this you can begin prepare.to.get.ready diaper ‘Step 2: Put the baby in a safe position, e.g. in a cradle. Afterwards you can begin to change the diaper.’ (CWT)

Serial verbs and change of location constructions in Vietnamese

197

(24) Không đặt bàn thờ ở chỗ có đường đi lại NEG put table worship be.at place there.is road go come.back ‘Don’t put the altar in a place which is a passageway’. (CWT)

2 The traditional analysis: coverbs and spatial relators In the literature on Vietnamese, the dynamic and stative V2 s under consideration in this chapter are commonly subsumed under the notion of coverb. Their contribution to the inherent aspectuality of phrases has received much less attention than their contribution to argument structure. Some authors regard coverb constructions as a prime example of grammaticalization with serial verb constructions. According to this assumption, coverbs are serial verbs which grammaticalize into elements which introduce arguments or mark semantic roles (Bisang 1992; Durie 1997; Aikhenvald 2006; Clark 1978; Kuhn 1990). Functional equivalence between coverbs and prepositions is frequently assumed in the literature on serial verb constructions, suggesting an account of their semantics primarily in terms of spatial relationships.9 In caused change of location SVCs, dynamic V2 s do indeed express a change in spatial configuration, as shown in example (25). (25)

lên ‘ascend’

UP/ ON , AT

a. Anh ấy lăn cái thùng lên núi. he roll CL barrel ascend hill ‘He rolled the barrel up the hill.’ b. Chị ấy treo bức tranh lên tường. she hang CL picture ascend wall ‘She hung the picture on the wall.’

9 However, a growing interest in the aspectual contribution of prepositions in non-serializing languages can be observed within the theoretical literature (Jackendoff 1991; Zwarts 2008). This may also initiate a shift of focus toward the contribution of coverbs to aspectuality. Wechsler (2003), for example, argues for a parallelism between the coverb khâw ‘enter’ in Thai and the preposition to in English. He points out that both are relevant for aspectual class membership of predicates.

198

Theresa Hanske

vào ‘enter’ c.

IN , AT, AROUND

Anh ấy lăn cái thùng vào phòng. he roll CL barrel enter room ‘He rolled the barrel into the room.’

d.

Chị ấy buộc băng vào trán. she tie band enter forehead ‘She tied a band around her forehead.’

xuống ‘descend’ e.

DOWN

Anh ấy lăn cái thùng xuống núi. he roll CL barrel descend hill ‘He rolled the barrel down the hill.’

f. Nó đặt chiếc máy nhỏ xuống cái đĩa. he put CL machine little descend CL disk ‘He put the arm of the player down to the record.’ In terms of spatial semantics, the stative V2 ở ‘be at’ expresses a general location whose lexical meaning is neutral in respect to spatial configuration. Because of its generality, the V2 ở ‘be at’ may be adequately used to depict the same spatial configurations as the respective dynamic V2 s, which give more specific spatial information (compare the minimal pairs in (5)–(6)). Spatial underspecification can be compensated by relator nouns, such as trong ‘inside’ in (5b), but the relator noun is not obligatory. Despite its generality with regard to spatial information, the stative V2 cannot replace dynamic V2 s in caused change of location SVCs in discourse (see section 1.2). This is because the alternation between the dynamic V2 construction and the stative V2 construction does not depend on spatial, but rather on aspectual semantics. The inchoative alternation provides evidence, independent from caused change of location SVCs, that coverbs are relevant for aspectuality. In this alternation, a quality predicate alternates with an SVC containing the quality verb in V1-position and a dynamic V2 . The SVC has an inchoative meaning.10 The two constructions differ systematically in aspectual class by either belonging to the state class or to the accomplishment class.11 10 For an account of this type of SVCs in terms of metaphorical extensions of the spatial meanings of dynamic V2s, see Bisang (1992: 301–305). 11 For Mandarin Chinese a similar alternation, namely that between state predicates and inceptive predicates has been pointed out by Li (1991), see also section 5.

Serial verbs and change of location constructions in Vietnamese

(26) a rẻ ‘be cheap’

199

a 0 . rẻ xuống ‘become cheaper’

b. béo ‘be fat’

b 0 . béo ra ‘get fat, to gain weight’

c. khỏe ‘be strong’

c 0 . khỏe ra ‘become strong’

d. đẹp ‘be beautiful’

d 0 . đẹp lên ‘become (more) beautiful’

e. nhanh ‘be fast’

e 0 . nhanh lên ‘get faster’

f. to ‘be big’

f 0 . to lên ‘get big’

3 Aspectuality 3.1 General background In the first part of this chapter, I focused on the readings which the two types of caused change of location SVCs may have when instantiated in sentences, both with and without the progressive marker. I mentioned the aspect class of a predicate or construction is relevant for its aspectual interpretation on the sentence level. For example, it was pointed out that aspect markers, such as the progressive marker đang, cause different readings on the sentence level depending on the aspect classes of the predicates. Up to now, this has been subsumed under the term “aspectuality” or “aspectual”. Yet, it is well-known that different levels of representation (or dimensions, as some authors would have it) are, in fact, involved here. They do interact within the domain of aspectuality, but there are good reasons for keeping them separate (Bertinetto 1997; Smith 2010; for a discussion of bidimensional vs. unidimensional approaches, see Sasse 2001). Most importantly, the distinction between aspect class (such as state, activity etc.) and the marking of the perfective/imperfective opposition by grammatical aspect markers is of concern here. Aspect class and grammatical aspect marking contribute to the aspectual interpretation of a sentence in different ways. Aspectual class membership is taken to be an inherent characteristic of predicates. For this reason, this level is sometimes called lexical aspect. I will refer to the aspectual class of a predicate as its situation type. Since situation type is not a matter of single verbs, or even lexemes, but rather of predicates in the sense of verb phrases or constructions, categorization into situation type classes pertains to phrasal aspect. The marking of a perfective/imperfective opposition is referred to by various terms in the literature, such as aspect proper, grammatical aspect, viewpoint aspect etc.; I will call it viewpoint.

200

Theresa Hanske

Aspectuality is typically described in terms of boundedness: States of affairs may be presented as including their endpoints (bounded) or as an ongoing situation, without reference to any endpoints (unbounded). This applies to situation type distinctions, such as terminative predicates (accomplishments and achievements) and non-terminative predicates (states and activities), as well as to the viewpoint distinctions, namely the perfective/imperfective opposition. Though the notion of boundedness is relevant for the definition of situation type classes and of viewpoint, it manifests itself in diffent ways at the two levels. Boundedness associated with situation type is conceived of as inherent boundedness. It contrasts with boundedness of viewpoint, which is a grammatically established boundedness (Comrie 1976: 21–22; Sasse 2001: 7–10). In connection with the necessity to delimit situation type from viewpoint, a further distinction which is crucial for the analysis of the aspectuality domain exists. It pertains to the difference between predicates as the units of categorization for situation type class as opposed to the realization of predicates in sentences. A predicate exhibits an actual aspectual reading when instantiated on the sentence level. Its situation type class, however, is a generalization from the various readings that it can have when actualized in sentences. As mentioned above, the units of categorization are not lexemes or verb senses. They are in fact abstract phrases, for which the aspectually relevant arguments are fully specified, but which abstract away from inflection. The constructions push a cart, draw a circle and draw circles are different abstract phrases in this sense. The situation type of an abstract phrase is understood as its potential for being realized in certain contexts and with certain interpretations. Abstract phrases and their instantiations in sentences therefore differ in level of abstraction. The same is true for lexemes and abstract phrases; the situation types of abstract phrases must not to be equated with the lexically-determined aspectual potential of the verbal lexemes (for a detailed discussion of the conceptual distinction between abstract and actualized phrases in the literature on aspectuality, see Behrens 1998: 294).

3.2 The selectional framework For my discussion of the aspectuality of caused change of location SVCs in Vietnamese, I use the selectional framework of Breu and Sasse (Breu 1985, 1994, 1996; 2000; Sasse 1991; Sasse 2000). Situation type and viewpoint are located on two levels of representation in this framework. Situation type is associated with the phrasal level and viewpoint is attributed to the sentence level. They are assumed to interact systematically and their interaction yields predictable meanings on the sentence level.

Serial verbs and change of location constructions in Vietnamese

201

Situation type is regarded as an inherent characteristic of abstract phrases12, which are categorized according to a set of time schemata. Most importantly, the time schemata are explicitly defined as configurations of phases (situations: S) and endpoints (boundaries: B). Schematically, this can be presented as follows (Sasse 2000: 224, slightly modified). (27)

Situation type distinctions are based on the four Vendlerian situation type classes, but the inventory is augmented by splitting up states into two stative classes (i.e. total state (TSTA) vs. inceptive state (ISTA)). Another modification pertains to the terminative situation type classes. In the Breu/Sasse approach, the decisive criterion used to distinguish between gradual terminatives (GTER) and total terminatives (TTER) is whether or not the abstract phrases conceptualize a prephase before the right boundary. Gradual terminatives have a prephase while total terminatives do not (for the situation type classes within the Breu/Sasse approach and the Vendlerian notions (Vendler 1967), see Sasse 2000). The viewpoint distinction, i.e. the perfective/imperfective opposition, is also defined in terms of phases and boundaries: Viewpoint markers select for the typical phases and boundaries that the situation types provide. Selection means that viewpoint markers profile the phases and boundaries of situation types according to their specialization and mark them as being relevant at the time of reference (Breu 1996: 44, FN 11). Accordingly, it is assumed that an operatoroperand relationship holds between viewpoint and situation type. The imperfective viewpoint profiles the phases (S), whereas the perfective viewpoint selects the typical boundaries (B1 and/or B2) of the situation type time schemata. Due to the interaction of situation type and viewpoint, certain readings systematically present themselves on the sentence level. For example, the imperfective viewpoint profiles the (pre-)phase within a gradually terminative situation type and thus, the sentence gets a processual meaning. When the perfective view-

12 The classification of abstract phrases into situation type classes is an extension of the original framework by Breu and Sasse, for which verb meanings are the units of categorization. However, see Sasse (2001; especially FN 28) for qualifications of the original conception.

202

Theresa Hanske

point operates on a gradually terminative phrase, it profiles its right boundary and yields a completed meaning on the sentence level.13

3.3 Perfect viewpoint and the perfect of result reading of caused change of location SVCs with stative V2 As a third viewpoint category, the perfect viewpoint is distinguished from the perfective and imperfective. In the perfect reading, the resultant state of a situation is emphasized. This reading is attained because the perfect viewpoint profiles the result while simultaneously referring to the situation that brought about the result (Breu 1988).14 This can be illustrated by the different implications that hold for sentences with perfect viewpoint marking (as in 28a) compared to nonperfect marking (as in 28b). (28)

a.

I have lost my penknife.

b. I lost my penknife. In (28a), the perfect marking indicates that the penknife is still lost at the time of reference; in (28b), with non-perfect marking, no such implication is made (for a more detailed discussion of (28) and further examples, see Comrie 1976: 52). Breu and Sasse conceive the relation between the situation and its result in terms of grammatically-established boundedness. The perfect is a viewpoint that combines perfective and imperfective viewpoint. Its perfective component selects the right boundary of the situation and the imperfective component profiles the postphasal resultant state which persists into the present (Breu 1988; Sasse 1991).

13 The Breu/Sasse apporach accounts primarily for the interaction of situation type and viewpoint in languages with obligatory marking of the perfective/imperfective opposition. This means that systems with optional viewpoint marking – as in Vietnamese – require some qualification, since the optionality has consequences for the relationship between abstract phrases and their aspectual interpretations on the sentence level. When viewpoint is not obligatory, the aspectual meaning of sentences is not always due to an interaction of situation type and viewpoint. The operator-operand relation only holds when a viewpoint marker is present. Where this is not the case, a viewpoint-neutral instantiation of the situation type phrases should be assumed (for a discussion of a “light” version of the selectional approach along these lines, see Sasse 2001: 21). 14 Usually, several semantic perfect types are subsumed under the perfect viewpoint (e.g. perfect of result, experiential perfect, perfect of persistent situation). The perfect of result is only one of them, but may be the most prototypical (for a more comprehensive discussion of the range of meanings, see for example Breu 1988, Comrie 1976).

Serial verbs and change of location constructions in Vietnamese

203

Caused change of location constructions with a stative V2 have a perfect of result reading, as was demonstrated in section 1.2.2. This suggests that the construction itself or the stative V2 might be an exponent of the perfect category. However, assuming that the perfect of result reading of caused change of locations with stative V2 comes about due to an interaction of situation type and viewpoint proves problematic. If the stative V2 is analyzed as a viewpoint operator, it is unclear what kind of abstract phrase is assumed to be in its scope. Abstracting away from the stative V2 in order to obtain the abstract phrase is not possible, because the stative V2 (plus its locational argument) explicitly refers to the result as a subpart of the overall state of affairs. It is an integral part of the abstract phrase. Thus, the stative V2 differs considerably from real viewpoint markers, such as đang.15 The perfect of result reading is not effected by viewpoint. Caused change of location SVCs with stative V2 s nonetheless combine a bounded and a persistent unbounded situation. Below, I will argue for an inherent boundedness account of this SVC-type, i.e. its boundedness characteristics will be attributed to the level of the abstract phrase. At this point, I take up the earlier discussion of why the progressive marker đang is not compatible with caused change of location SVCs with a stative V2 . The blocking of đang is consistent with the claim that this construction type is characterized by a combination of boundedness and unboundedness. The progressive viewpoint marker đang conflicts with the inherent boundedness characteristics of caused change of location SVCs with stative V2 s. Đang qualifies as an exponent of the imperfective viewpoint and grammatically establishes unboundedness. It selects situations provided by the situation type and suppresses their inherent boundaries, if there are any (see (11)). In caused change of location SVCs with a stative V2 , the situation is understood as being necessarily bounded for the result to come about. This is incompatible with the function of đang as suppressing inherent boundaries. The viewpoint marker cannot apply to the resultant state, either. In Vietnamese, this is not due to the fact that stative predicates may never be within the scope of the progressive marker (see (10)). Rather, it is because viewpoint markers do not operate just on one V(P) in a SVC. It is a key characteristic of serial verb constructions that viewpoint markers must have scope over a serial verb construction as a whole (Durie 1997, Aikhenvald 2006). When the progressive đang cannot apply to V1 because the situation it refers to has a necessary inherent boundary, it also cannot apply to V2 , consistent with more general constraints on serial verb constructions. 15 Note that the viewpoint marker đang ‘be in the process of’ is also treated as a serial verb in the literature. Yet, it is generally agreed upon that the uses of đang in SVCs represent a TAM category.

204

Theresa Hanske

4 Aspectuality and the alternation of caused change of location SVCs 4.1 The caused change of location SVC with dynamic V2 s My categorization of the caused change of location construction draws upon the distinction between abstract phrases and actualized phrases. Abstract phrases and actualized phrases can be characterized in terms of a type-token relationship, and manifest different degrees of abstraction in analysis.16 The caused change of location constructions with dynamic V2 s are the units of categorization for situation type class, i.e. abstract phrases consist of the SVC as a whole. As already mentioned, abstract phrases are categorized for situation type class according to their potential to be actualized in certain contexts. Aspectual tests (Dowty 1979), such as the progressive test, give information about the actualization potential of abstract phrases. They test the compatibility and interpretation of actualized phrases in contexts that are assumed to be diagnostic for situation type class. Because its situation type is the potential of an abstract phrase, realizations of the abstract phrase in different test contexts must be taken into consideration. This is why aspectual tests consist of a battery of different successively applied tests (for the methodology of aspectual tests, see Behrens 1998: 285–301). Taxis relations in discourse can be used as test contexts, since they systematically trigger a processual or completed reading of actualized phrases. Caused change of location SVCs with dynamic V2 s are gradually terminative, as has become clear from their readings discussed in section 1.2.1. (29)

Caused change of location SVCs with dynamic V2 s Situation type of abstract phrases: gradually terminative Interpretations of actualized phrases: – with the progressive marker đang: processual (cf. (11)) – in sequential and incidential taxis: completed (cf. (15)–(18)) – interpretation with adverbial trong hai phút: inherently bounded (cf. (13)–(14))

They depict a state of affairs that consists of a situation and a right boundary. When occurring with the progressive marker đang, the situation within their 16 In English, abstract phrases are usually represented by the citation forms, and actualized phrases by the respective inflected forms. Since Vietnamese is an isolating language, the conceptual difference between abstract phrases and actualized phrases cannot be represented by different word forms.

Serial verbs and change of location constructions in Vietnamese

205

time schema is profiled and yields a processual reading. In sequential taxis and as the foregrounded event in incidential taxis, these constructions have a completed reading which is indicative of an inherent boundary. This contrasts, for example, with activities which cannot have a completed meaning. Activities may be stopped but not finished, since they refer to situations without a boundary in the sense of a change-of-state. When they appear in a sequence of actions, they have a delimitative reading, i.e. they are interpreted as temporally bounded. Their interpretation with the adverbial trong hai phút ‘in/for two minutes’ likewise demonstrates that caused change of location SVCs with dynamic V2 s have an inherent right boundary. A processual, i.e. atelic, interpretation of the sentences is not possible. In this case, the phrases also pattern with gradual terminatives (the difference between (13) and (14) with regard to the resultative reading is addressed further below).

4.2 The caused change of location SVC with the stative V2 As with abstract dynamic V2 caused change of location phrases, it is also assumed that the abstract stative V2 caused change of location phrases consist of the SVC as a whole. Actualized stative V2 caused change of location SVCs always have a perfect of result reading. Compared with the dynamic V2 alternant, their reading is strongly restricted (see section 1.2). In section 3.3, I argued that they feature a bounded-unbounded configuration, and that this must be inherent, as it cannot be viewpoint. I will proceed to show why it is justified to allocate the boundedunbounded configuration of stative V2 caused change of location SVCs to the phrasal level. At the beginning of this chapter, I presented two sets of verbs (put verbs and caused motion verbs). They were distinguished according to their distribution in the alternating subtypes of the caused change of location SVC. The restriction on caused motion verbs occurring in caused change of location SVCs with a stative V2 is a lexical constraint on verbs whose lexical information does not specify a result component. These kinds of restrictions in respect to possible combinations support the hypothesis that caused change of location SVCs are “lexical” rather than “grammatical”.17 For stative V2 caused change of location SVCs, I propose a time schema that involves a situation, a right boundary, and a resultant situation. Furthermore, I assume that the perfect of result reading is due to their situation type. That is,

17 For a critical discussion of treating phenomena such as serial verb constructions as pertaining either to the lexicon or to grammar, see section 5.

206

Theresa Hanske

the situation type determines that the result is relevant at the time of reference. My proposal is tantamount to introducing an additional gradually terminative subclass, which may account for the stative V2 caused change of location SVCs. (30)

Situation type of abstract stative V2 caused change of location phrases:

As mentioned in section 3.3, the perfect of result reading, i.e. the relevance of the result at the time of reference, cannot be an effect of external profiling associated with viewpoint. I hypothesize this specific reading is due to lexical enforcement which occurs when a stative V2 combines with a V1 that itself is lexically specified for a resultative component. Whether the result is enforced, implied, or not significant at all does not depend exclusively on the stativity or dynamism of V2 s. The lexical information inherent to V1 plays a role for the interpretation of the degree of resultativity. Thus, the semantic type of both verbs, V1 and V2 , influences the degree of resultativity. Examples (13) and (14) are contrastive in respect to resultativity, even though a caused change of location phrase with a dynamic V2 is present in both sentences. It is only in (13), where a put verb is in V1-position, that the temporal adverbial trong hai phút ‘in/for two hours’ is additionally understood to modify the implicit result of the caused change of location SVC (cf. (13ii) vs. (14ii)). With the caused motion V1 (in (14)), the caused change of location does not imply a result, thus ruling out a resultative reading. Put verbs and caused motion verbs also differ in respect to resultativity in other construction types. Put verbs are compatible with a resultative state reading when used in a Theme-subject location SVC, irrespective of the stative or dynamic character of V2 , as demonstrated in (31). In contrast, caused motion verbs refer to a process, and a resultative state reading is therefore excluded in (32). (31) a.

Cái cốc đặt ở / lên bàn. CL cup put be.at ascend table ‘The cup is put (i.e. remains put) on the table.’

b.

Cái tem dán ở / lên phong bì. CL stamp stick be.at ascend envelope ‘The stamp (is) stuck to the envelope.’

Serial verbs and change of location constructions in Vietnamese

(32)

207

a. Cái thùng lăn vào phòng. CL barrel roll enter room ‘The barrel rolled into the room.’ b. Hai xe đâm nhau. two vehicle crash each other Người lái xe tung ra ngoài. driver throw exit RN ‘The two vehicles crashed. The drivers were hurled out (of the car.).’

Furthermore, caused motion verbs are generally constrained with regard to their occurrence with stative V2 s in resultative SVCs. In (33), the verb đánh ‘beat, hit’ is used in its caused motion sense, but it cannot co-occur with a stative verb in V2 -postion, such as xa ‘be far away’ (see (33a)). Stative verbs may occur with caused motion verbs in resultative constructions, but only within a different construction type. A result subevent can only combine with caused motion verbs when a dynamic V2 intercedes. Example (33b) illustrates that the unacceptable sentence in (33a) can be corrected by insertion of a dynamic V2 , here bay ‘fly’. (33)

a. *Vận động viên đánh quả bóng xa. player hit CL ball be.far.away intended: ‘The player hit the ball far.’ b. Vận động viên đánh quả bóng bay xa. player hit CL ball fly be.far.away ‘The player hit the ball far.’

5 Lexicon expansion and the grammatical category viewpoint Finally, I want to elaborate on the question of how the alternation with caused change of location SVCs in Vietnamese relates to more general issues discussed in the literature on aspectuality. Apart from the impact it may have on grammar, verb serialization works as a mechanism for lexicon expansion. This is evident in the inchoative alternation (cf. (26)), where the semantic difference would commonly justify the assumption of two lexical units. It is considered common knowledge in the literature on serial verb constructions for some SVC types to exhibit a tendency to lexicalize

208

Theresa Hanske

(Aikhenvald 2006; Durie 1997). However, lexicalization is usually not claimed for caused change of location SVCs – at least for Vietnamese. Presumably, their relevance for the lexicon has been obscured by the dominant view that, since their V2 s are coverbs, they grammaticalize (rather than lexicalize). In Mandarin Chinese, a similar type of serialization exists, namely the resultative verb (RV) construction. For distributional reasons, it is analyzed as a word formation process. Some authors treat these as compounds and not as true serial verb constructions (for a discussion of the different types of RV constructions and their analyses in the literature, see Bisang 1992: 101–112; Paul 2005: 17–20.). The class of resultative verbs consists of three subclasses for which the directional verb subclass by and large conforms to the dynamic V2 s considered here (directional RVs: e.g. shàng ‘ascend’, xià ‘descend’, jìn ‘enter’). The second subclass contains mainly stative verbs, which are seen as resultative verbs proper (resultative RVs: e.g. gānjiing ‘clean’, sĭ ‘dead’, kāi ‘open’). The third subclass contains completive verbs (completive RVs: e.g. wán ‘finish’, dào ‘reach’, jiàn ‘perceive’). The contribution of resultative verbs to aspectuality is widely acknowledged (Kang 2001; Smith 2010; Li 1991). Resulative verb constructions convey different degrees of resultativity, such that their readings in sentences range from a mere completed to a strongly resultative interpretation, as (34a) and (34b) illustrate (Chen 2008: 28). (34) a.

Tā xĭ-wán le yīfu b. Tā xĭ-gānjing le yīfu he wash-complete PFV clothes he wash-be.clean PFV clothes ‘He finished washing the clothes.’

‘He washed the clothes clean.’

The variability of possible readings – such as completed, implied result, or entailed result – is primarily associated with the lexical semantics of the resultative verbs. Their lexical impact on the phrasal or word level is generally accepted in the literature on Mandarin Chinese. It is their aspectual readings in sentences that cause problems for analysis. The aspectual effect of resultative verbs is either attributed to sentence aspect or to the lexical level. Smith (2010) claims that resultative verbs contribute to lexical semantics, e.g. resultative verbs may lexically add a result. Additionally, she assumes they perform a perfective viewpoint function in sentence aspect. The analysis in terms of viewpoint attempts to capture the fact that resultative verbs determine the aspectual readings on the sentence level. The completed reading of actualized accomplishment phrases with resultative verbs cannot be overridden by imperfective viewpoint marking. A processual interpretation is only possible for the corresponding phrases without the resultative verb. These phrases are compatible with the progressive viewpoint marker zài (cf. (35a) vs.

Serial verbs and change of location constructions in Vietnamese

209

(35b) from Smith 2010: 289). Although she advocates a strict delimitation of situation type aspect and viewpoint aspect18, there is no mention in Smith (2010) why this analysis does not undercut the distinction. (35)

a. Zhāngsān xiĕ-wán-le yì-fēng xìn. Zhangsan write-finish-PFV oneCL letter ‘Zhangsan wrote a letter.’ b. Zhāngsān zài xiĕ yì-fēng xìn. Zhangsan PROG write oneCL letter. ‘Zhangsan was writing a letter.’

Li (1991) deals with resultative verb constructions exclusively on the lexical level. He proposes that the resultative verbs take part in word formation and effect a change of situation type class with regard to V1. The following sentence pair illustrates a shift from totally stative to inceptive stative, which is said to be productive. The similarity to the inchoative alternation in Vietnamese is conspicuous (see (26)). Li (1991) also points out a parallelism between resultative verbs and morphological perfective viewpoint in languages that obligatorily mark viewpoint. Compare the Mandarin Chinese example in (36) (adapted from Li 1991: 174) with example (37) from Modern Greek (adapted from Behrens 1998: 309). (36)

a. Xiăoling ài /*ài-le Zhāngsān Xiaoling love love-PFV Zhangsan ‘Xiaoling loves/loved Zhangsan.’ b. Xiăoling ài-shàng-le Zhāngsān. Xiaoling fall.in.love-PFV Zhangsan ‘Xiaoling fell in love with Zhangsan.’

(37)

a. Ton aghapúsa polí. him love (1.sg.IMPFV ) very ‘I love him very much.’

b.

Ton aghápisa polí. Him love (1.sg.PFV ) very ‘I fell in love with him very much.’

18 Smith’s use of situation type aspect and viewpoint aspect is not equivalent to my use of the terms situation type and viewpoint; instead, her terms correspond roughly to my use of phrasal aspect and sentence aspect.

210

Theresa Hanske

Li nonetheless argues that viewpoint and resultative verbs are distinct phenomena: In the Greek example, the perfective viewpoint operates on a predicate that already pertains to the inceptive states and profiles the initial boundary, yielding an ingressive meaning. The resultative verb, however, lexically sets an initial boundary with a totally stative predicate and so effects a shift in situation type. His example is well chosen, since treating the inceptive construction as word formation is relatively uncontroversial. It is not clear how he would account for examples in which the resultative verb constructions form new lexical entries of the gradually terminative class. In this case, their actualizations are necessarily interpreted as completed (see (35)). The ascription of a necessarily completed reading here, however, does not meet the usual definiton of the gradually terminative situation type. In contrast to total terminatives, gradual terminatives normally do not need to be interpreted as bounded on the sentence level. Resultative verb constructions in Mandarin Chinese cause problems of analysis because they qualify as relatively clear cases of formation of lexical units but they also seem to determine sentence aspect. These difficulties are akin to the issues discussed in the context of preverbs in Slavic languages, but also, for example, in Hungarian. Preverbs are clear cases of derivational morphology in Hungarian and therefore should at most affect the situation type class of their base. However, some preverbs effect a completed reading on the sentence level, even when this cannot be ascribed to the situation type of the abstract phrase that is headed by the derived verb. Not surprisingly, some authors analyze preverbs as perfective viewpoint (e.g. de Groot 1995). However, such an approach is incompatible with the necessity to distinguish between the two aspectual levels (Sasse 2001: 32). Hungarian preverbs have not developed into a consistent system of aspectual verb pairs comparable to the Russian viewpoint. Yet in Russian, preverbs exist alongside viewpoint marking with aspectual verb pairs. The issue of how preverbs can be distinguished from viewpoint in Russian is still a matter of debate (Bertinetto 1997; Lehmann 1988; Breu 2000). The derivational character of the Slavic viewpoint system poses problems of analysis, because a strategy usually associated with word formation is employed here to convey a grammatical category. This in turn, seems to challenge the distinction of separate levels of representation, in general, i.e. the lexicon vs. grammar as distinct levels. Verbal aspect pairs are frequently treated as grammar by lexical means, i.e. “grammaticalized lexical categories” (Dahl 1985: 89). Yet, it is questionable whether this postulate is really helpful for delimiting preverbs from viewpoint. Sasse (2001: 50) proposes to approach these issues from the perspective of lexicon-grammar interaction rather than lexicon-grammar opposition. Although the principles of this interaction in the respective languages

Serial verbs and change of location constructions in Vietnamese

211

would still have to be specified, the study of aspectuality with resultative verbs in Mandarin Chinese and serial verb constructions in Vietnamese could only benefit from such a change in perspective. My proposal to differentiate the situation type class of gradual terminatives in terms of relevance of resultativity may be seen as an attempt along these lines. Abstract phrases in this class are more varied than the Vendlerian definition (including the selectional framework used here) suggests. As for significant differences within the gradually terminative class, I refer to a well-known example, namely abstract phrases with an incremental Theme in the sense of Dowty (1991). Abstract phrases, such as read a book, build a house, paint the wall differ from “true” gradual terminatives, e.g. draw a circle, with regard to the grade of inherent boundedness.19 Whether finer distinctions with inherently-bounded situation types (gradual terminatives and total terminatives) are necessary has also been discussed in the literature (Bertinetto and Squartini 1995). The role that resultativity may play in the categorization of inherently-bounded predicates is discussed by Breu (1996). Ehrich (1992: 74–99) explicitly proposes a feature [+resultative] that subdivides the German verbal lexicon in order to account for the different readings of the German perfect (tense). Sasse (2000) expounds the relations of stativity and resultativity, and argues for a finer subclassification of inherently-bounded predicates in Cayuga (Sasse 2000 and the literature cited there).

6 Conclusion The observation that the stative V2 vs. dynamic V2 caused change of location SVCs yield different readings at the sentence level has been our point of departure. While the former have a perfect of result reading, the latter may be interpreted as processual or completed. I have argued that this aspectual alternation is triggered on the phrasal level of caused change of location SVCs, because the two types of caused change of location SVCs differ in situation type.

19 Both subtypes are inherently bounded, because they are compatible with in-x-time-type adverbials, cf. The child drew the circle in two minutes and The child read the book in an hour. However, if they are cross-checked, abstract phrases with incremental Themes do not necessarily prove to be inherently bounded. ‘For-x-time’ adverbials are acceptable with them and the actualized phrases yield “activity”-readings, i.e. they behave as if they were inherently unbounded, cf. *The child drew a circle for two minutes vs. The child read a book for an hour. For the treatment of the bounded/unbounded ambiguity with gradual terminatives in different approaches, see Behrens (1998: 291–293).

212

Theresa Hanske

I have shown that the abstract dynamic V2 caused change of location phrases belong to the gradually terminative situation type. In contrast, I have argued that stative V2 caused change of location SVCs involve an explicit result at the phrasal level, such that the abstract phrases of this construction type are inherently bounded-unbounded. This leads to a finer differentiation within the gradually terminative situation types. I have proposed an additional terminative situation type which accounts for the particular configuration of bounded and unbounded phases specific to the abstract stative V2 phrases. This type is characterized by a time schema that specifies a phase, a right boundary, and a necessary result phase. Furthermore, the result phase is marked as relevant on the phrasal level due to lexical enforcement. I regard lexical enforcement here to be a semantic effect of serialization that can only be understood if the compositionality of serial verb constructions is taken into consideration. I have shown that lexical information of both V1 and V2 is relevant in caused change of location SVCs: When V1 s with a lexical resultative component combine with the stative V2 ở ‘be at’ in a caused change of location SVC, the stative V2 explicitly “spells out” and thereby enforces the resultative component of V1 . In the last section, I discussed some “old” controversies about the aspectuality of complex predicates and how they relate to caused change of location SVCs in Vietnamese. I have shown that these types of predicates cause particular problems for an aspectual analysis, suggesting that the terminative situation types are more differentiated than the situation type taxonomies which have previously been proposed in the literature.

References Aikhenvald, Alexandra Y. 2006. Serial Verb Constructions in Typological Perspective. In: Alexandra Y. Aikhenvald and Robert M. Dixon (eds.), Serial Verb Constructions. A Crosslinguistic Typology. Oxford: Oxford University Press, 1–68. Behrens, Leila. 1998. Ambiguität und Alternation. Methodologie und Theoriebildung in der Lexikonforschung. Habilitationsschrift, Universität München. Bertinetto, Pier Marco. 1997. Il dominio tempo-aspettuale. Demarcazioni, intersezioni, contrasti. Torino: Rosenberg & Sellier. Bertinetto, Pier Marco and Mario Squartini. 1995. An Attempt at Defining the Class of ‘Gradual Completion’ Verbs. In: Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham and Mario Squartini (eds.), Temporal Reference, Aspect and Actionality. Vol. I: Semantic and Syntactic Perspectives. Torino: Rosenberg & Sellier, 11–26. Bisang, Walter. 1992. Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai und Khmer: vergleichende Grammatik im Rahmen der Verbserialisierung, der Grammatikalisierung und der Attraktorpositionen. Tübingen: Narr.

Serial verbs and change of location constructions in Vietnamese

213

Breu, Walter. 1985. Handlungsgrenzen als Grundlage der Verbalklassifkation. In: Werner Lehfeldt (ed.), Slavistische Linguistik 1984. München: Otto Sagner, 9–34. Breu, Walter. 1988. Resultativität, Perfekt und die Gliederung der Aspektdimension. In: Jochen Raecke (ed.), Slavistische Linguistik 1987. München: Otto Sagner, 42–74. Breu, Walter. 1994. Interactions between Lexical, Temporal and Aspectual Meanings. Studies in Language 18(1): 23–44. Breu, Walter. 1996. Komponentenmodell der Interaktion von Lexik und Aspekt. In: Wolfgang Girke (ed.), Slavistische Linguistik 1995. München: Otto Sagner, 37–74. Breu, Walter. 2000. Zur Position des Slavischen in einer Typologie des Verbalaspekts (Form, Funktion, Ebenenhierarchie und lexikalische Interaktion). In: Walter Breu (ed.), Probleme der Interaktion von Lexik und Aspekt (ILA). Tübingen: Niemeyer, 21–54. Chen, Jidong. 2008. The Acquisition of Verb Compounding in Mandarin Chinese. Unpublished doctoral dissertation, Free University Amsterdam. Clark, Marybeth. 1978. Coverbs and Case in Vietnamese. Canberra: Australian National University, Dept. of linguistics. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Dahl, Östen. 1985. Tense and Aspect Systems. Oxford: Basil Blackwell. De Groot, Casper. 1995. Aspect, Mood, and Tense in Functional Grammar. In: Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham and Mario Squartini (eds.), Temporal Reference, Aspect and Actionality. Vol. I: Semantic and Syntactic Perspectives. Torino: Rosenberg & Sellier, 29–44. Dowty, David R. 1979. Word Meaning and Montague Grammar: The semantics of verbs and times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel. Dowty, David R. 1991. Thematic Proto-Roles and Argument Selection. Language 67(3): 547–619. Durie, Mark. 1997. Grammatical Structures in Verb Serialization. In: Alex Alsina, Joan Bresnan and Peter Sells (eds.), Complex Predicates. Stanford: Center for the Study of Language and Information, Stanford University, 289–354. Emeneau, Murray B. 1951. Studies in Vietnamese (Annamese) Grammar. Berkeley: University of California Press. Ehrich, Veronika. 1992. Hier und Jetzt. Studien zur lokalen und temporalen Dexis im Deutschen. Tübingen: Niemeyer. Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41: 9–45. Kang, Jian. 2001. Perfective aspect particles or telic Aktionsart markers? Studies of the directional verb compounds. Journal of Chinese Linguistics 29(2): 281–339. Kuhn, Wilfried. 1990. Untersuchungen zum Problem der seriellen Verben: Vorüberlegungen zu ihrer Grammatik und exemplarische Analyse des Vietnamesischen. Tübingen: Niemeyer. Lehmann, Volkmar. 1988. Der russische Aspekt und die lexikalische Bedeutung des Verbs. Zeitschrift für slavische Philologie 48(1): 170–181. Li, Chor-Shing. 1991. Beiträge zur kontrastiven Aspektologie. Das Aspektsystem im Modernen Chinesisch. Frankfurt a.M.: Peter Lang. Nguyễn, Dình-Hòa. 1997. Vietnamese. Amsterdam/Philadelphia: John Benjamins. Paul, Waltraud. 2005. The serial verb construction in Chinese: A Gordian knot. In: Bernard Oyharçabal and Waltraud Paul (eds.), Proceedings of the workshop La notion de « construction verbale en série » est-elle opératoire? December 9, 2004, EHESS, Paris. Published in electronic form at: http://www.typologie.cnrs.fr/spip.php?article107. Sasse, Hans-Jürgen. 1991. Aspekttheorie. In: Hans-Jürgen Sasse (ed.), Aspektsysteme. Arbeitspapier Nr. 14. NF. Köln: Institut für Sprachwissenschaft, Universität zu Köln.

214

Theresa Hanske

Sasse, Hans-Jürgen. 2000. Aspektsemantik und Lexikonorganisation im Cayuga. In: Walter Breu (ed.), Probleme der Interaktion von Lexik und Aspekt (ILA). Tübingen: Niemeyer, 191–246. Sasse, Hans-Jürgen. 2001. Recent Activity in the Theory of Aspect: Accomplishments, Achievements, or just Non-progressive State? Arbeitspapier 40. NF. Köln: Institut für Sprachwissenschaft, Universität zu Köln. Slobin, Dan I. 2004. The Many Ways to Search for a Frog: Linguistic Typology and the Expression of Motion Events. In: Sven Strömquist and Ludo Verhoven (eds.), Relating Events in Narrative: Typological and Contextual Perspectives. Mahwah, NJ: Lawrence Erlbaum, 219–257. Smith, Carlota. 2010. The Parameter of Aspect. Second Edition, reprod. Dordrecht/Boston/ London: Kluwer. Thompson, Laurence C. 1987. A Vietnamese Reference Grammar. [Mon-Khmer Studies 13/14]. Honolulu: University of Hawaii Press. Vendler, Zeno. 1967. Verbs and times. In: Zeno Vendler, Linguistics in Philosophy. Ithaca, NY: Cornell University Press, 97–121. Wechsler, Stephen M. 2003. Serial Verbs and Serial Motion. In: Dorothee Beermann and Lars Hellan (eds.), Proceedings of the Workshop on Multi-Verb Constructions. Trondheim Summer School 2003. Norwegian University of Science and Technology, Trondheim, 1–27. Zwarts, Joost. 2008. Aspects of a typology of direction. In: Susan Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. Amsterdam: John Benjamins, 79–106.

IV Pronouns and minor word classes

Tran Thuan and Benjamin Bruening

8 Wh-phrases as indefinites: A Vietnamese perspective* Vietnamese, like many languages of the world, uses wh-phrases as indefinites. When these wh-indefinites appear bare, without a demonstrative, they are only licensed in certain syntactic contexts, and only take low scope. When they occur with a demonstrative, however, they can appear in any context, and may take widest scope, even outside of islands. We propose a formal account of the uses of wh-phrases in questions and as these two types of indefinites that accounts for these facts, as well as cross-linguistic patterns involving wh-phrases used as indefinites. Keywords: wh-questions, indefinites, alternative semantics, wide-scope indefinites, choice functions

1 Introduction Vietnamese allows wh-words to be used as indefinites in certain contexts. For instance, in a simple positive sentence, a wh-word must be interpreted as a question (1), but in a yes-no question the same wh-word can be interpreted as an indefinite (2):1 (1) Cô ấy gặp ai? she meet who ‘Who did she meet?’ *‘She met anyone.’ (2)

Cô ấy có gặp ai không? she Q meet who Q ‘Did she meet anyone?’

* This chapter significantly expands on material in Tran (2009). All Vietnamese data come from the first author (Tran). The work of the second author (Bruening) was partially supported by the National Science Foundation (grant no. BCS-0518308). We would like to thank Satoshi Tomioka for significant help with the semantics and logical formulae. 1 Abbreviations: CL = classifier; Comp = complementizer; Dem = demonstrative; Fut = future; Neg = negation; Part = particle; PL = plural; Q = question particle; R = relative clause marker.

218

Tran Thuan and Benjamin Bruening

We refer to wh-words used as indefinites as wh-indefinites (we justify treating wh-words as underlyingly question elements in section 3). The contexts where a wh-word can be interpreted as an indefinite in Vietnamese include yes-no questions, negative sentences, the protasis of a conditional, and some others. A complete list is given in section 2. A wh-word can also be used as an indefinite when it appears with the element đó, which in other contexts is a demonstrative (3). When this element follows a wh-word, it can be used as an indefinite in any context. For instance, unlike a bare wh-indefinite, a wh-word with đó can be used in a simple positive declarative, as in (4): (3)

Lan có mua quyển sách đó đâu. Lan Neg buy CL book Dem Neg ‘Lan did not buy that book.’

(4) Tân vừa gặp ai đó. Tan just meet who Dem ‘Tan just met someone.’ In this chapter, we show that the bare wh-indefinites and the non-bare ones (those with đó) differ in two ways. First, bare wh-indefinites require licensing, but non-bare ones do not; and second, bare wh-indefinites may only take low scope, but non-bare ones may take widest scope. We propose an analysis that captures these differences. In our analysis, wh-words universally denote sets of individual alternatives (Hamblin 1973, Kratzer and Shimoyama 2002). To be used as an indefinite, a wh-word must be turned into a set. A null element performs this conversion, and this null element must be licensed by being in the scope of a licenser. Since it must occur within the scope of a licenser, a wh-indefinite can only take low scope. In contrast, the demonstrative đó introduces a choice function, which is then bound by an existential quantifier inserted at any CP layer (Reinhart 1997). Since the existential quantifier occurs high, a wh-word with đó may take very wide scope. Section 2 describes the facts of Vietnamese whindefinites that we are concerned with. Section 4 outlines our analysis. Our analysis is also concerned with some cross-linguistic generalizations regarding wh-words used as indefinites, which we point out and discuss along the way. We also compare Vietnamese to Mandarin Chinese, which has been very well described and for which several analyses have been proposed. The licensing condition that we propose for the null Vietnamese existential quantifier is adapted from one proposed for Mandarin Chinese by Lin (1998b). Further cross-linguistic tendencies are discussed in the conclusion.

Wh-phrases as indefinites: A Vietnamese perspective

219

2 The Distribution and Scope of Wh-Indefinites in Vietnamese Bare wh-indefinites only appear in certain contexts, and seem to require some kind of licensing. In contrast, a wh-indefinite plus đó can appear in any context, with wide scope possible. We first describe bare wh-indefinites, and then turn to the non-bare ones.

2.1 Bare Wh-Indefinites The distribution of bare wh-indefinites in Vietnamese is very similar to Mandarin Chinese, as described by Lin (1998b). Like Mandarin, there are contexts where wh-indefinites prefer to appear without a nominal classifier, and contexts where they prefer to appear with a classifier. We will have no account of this preference, and the presence or absence of the classifier seems to have no effect on the semantics (with or without the classifier, the wh-word is interpreted as an indefinite). For the purposes of this chapter we will ignore the classifier, although the examples will include them where they are preferred. (We continue to refer to wh-indefinites without the particle đó as “bare wh-indefinites,” even though they are not technically bare when they occur with a classifier.) The following contexts allow a wh-word to be interpreted as an indefinite, and they typically appear without a classifier in this use: negation, yes-no questions, the protasis of a conditional, epistemic adverbs, existential ‘have’, the complement of a nonfactive verb, and the complement of a negated factive verb. We illustrate them in turn, with the licensing operator underlined. The first context is negation. Positive sentences do not license wh-indefinites (1 above), but negative ones do: (5)

Tân không gặp ai. Tan Neg meet who ‘Tan does/did not meet anyone.’

The second context is yes-no questions: (6)

không? Cô ấy có gặp ai she Q meet who Q ‘Did/Does she meet anyone?’

220

Tran Thuan and Benjamin Bruening

The third context is the protasis of a conditional: (7)

Nếu ai đến thì Anh Thơ sẽ rất vui. if who arrive then Anh Tho Fut very happy ‘If anyone arrives, Anh Tho will be very happy.’

The fourth context involves what we will classify as epistemic adverbs, which have meanings like ‘seemingly’ and ‘probably’. These license wh-indefinites: (8) a.

Hình như ai vừa gặp Tân. seemingly who just meet Tan ‘It seems someone just met Tan.’

b.

Chắc ai mới bắt nạt anh ta. probably who just bully him ‘Probably someone just bullied him.’

Fifth is what we will call existential ‘have’, which asserts existence: (9)

Có ai gặp Tân. have who meet Tan ‘Someone met/meets Tan.’

The sixth context where wh-indefinites are licensed, and prefer to lack a classifier, is in the complement to a nonfactive verb: (10)

là tôi mới mua gì cho Lan. Tân nghĩ/tin/*biết Tan think/believe/*know Comp I just buy what for Lan ‘Tan thought/believed/*knew I just bought something for Lan.’

Factive verbs like ‘know’ do not license wh-indefinites in their complements. However, negated factive verbs do: (11) Tôi không nhớ cô ấy đã gặp ai rồi. I Neg remember she ASP meet who already ‘I don’t remember that she already met someone.’ We will return to this context below, and its counterpart in Mandarin Chinese.

Wh-phrases as indefinites: A Vietnamese perspective

221

In the next set of contexts, wh-indefinites are licensed, but the wh-word prefers to appear with a classifier (see Lin 1998b, 225–227 on Mandarin2 ). The first such context is in the scope of a modal: (12) a.

cho anh ta. Tôi phải đi mua ?(cái) gì I must go buy CL what for him ‘I must go buy something for him.’

b.

ăn ?(cái) gì trước, nhưng đừng ăn nhiều quá. Anh nên you should eat CL what beforehand but do.not eat much too ‘You can eat something beforehand, but don’t eat too much.’

The second is imperative sentences: (13) Lại đây ăn ?(cái) gì đã! come here eat CL what Part ‘Come here to eat something!’ And the third is in the complement to verbs like ‘want’ and ‘plan’, which have non-realized complements: (14)

để giúp cô ấy. Tôi muốn làm ?(cái) gì I want do CL what to help her ‘I want to do something to help her.’

As said above, we have nothing to say about why some contexts prefer classifiers. We take all wh-indefinites, whether they occur with a classifier or not, to be equivalent semantically, and treat them identically in our analysis. Although the distribution of wh-indefinites is very similar in Vietnamese and Mandarin Chinese, there is one difference. This is that existential ‘have’ licenses a wh-indefinite by itself in Vietnamese (9), but it does not in Chinese. Our analysis of Vietnamese ‘have’ is given below; we also suggest what is different about Chinese.

2 An example from Mandarin is the following: (i) Nǐ kěyǐ xiān chī *(diǎn) shénme (dōngxi ), kěshì bié chī tài duō. you may beforehand eat *(CL) what thing but don’t eat too much ‘You are allowed to eat something beforehand but don’t eat too much.’ (Lin 1998a, 225, (20b))

222

Tran Thuan and Benjamin Bruening

As shown above, bare wh-indefinites are only licensed in certain contexts. We now turn to structural relations that have to hold between a bare wh-indefinite and its licenser. In Vietnamese, a wh-indefinite does not have to be in the same clause as its licenser, and in fact they can be separated by island boundaries: (15)

Hình như Nim mới ăn [cái bánh mà [Gi vừa mua cho ai]]. seemingly Nim just eat [CL cake Rel [Gi just buy for who]] ‘It seems Nim just ate the cake that Gi just bought for someone.’

However, it is necessary for the licenser to take scope over the wh-indefinite. In the sentence above, the licensing epistemic adverb is in the matrix clause, and takes scope over the embedded clause which contains the wh- indefinite. In the following, in contrast, negation is in an embedded clause (a sentential subject), while the wh-word is in the matrix clause, and the sentence is ungrammatical on the indefinite reading: (16) *[Anh ta không đến] làm ai rất buồn. [he Neg arrive] make who very sad ‘That he did not arrive makes anyone very sad.’ Importantly, whether a classifier is present or not, bare wh-indefinites always take scope below their licenser. In (5), repeated below, the only interpretation has negation taking scope over the existential quantifier (Neg > p); it is impossible for the existential quantifier to take scope over negation: (17)

Tân không gặp ai. Tan Neg meet who ‘Tan does/did not meet anyone.’ *‘There is someone such that Tan does/did not meet that person.’

Wh-indefinites are not limited to narrowest possible scope, however. If there are two licensers, intermediate scope is possible (see Lin 2002 on Mandarin): (18)

Nếu anh không muốn mời ai thì báo cho tôi biết. if you Neg want invite who then report for I know ‘If you do not want to invite anyone, let me know.’ or ‘If there is someone you do not want to invite, let me know.’

Wh-phrases as indefinites: A Vietnamese perspective

223

(19) Hình như cô ấy không thích ai. seemingly she Neg like who ‘It seems she does/did not like anyone.’ or ‘It seems there is someone she does/did not like.’ Hence, the generalization is that bare wh-indefinites must be within the scope of a licenser. If there is more than one potential licenser, they just need to take scope beneath one of them. We hypothesize that intermediate scope arises through covert movement of the wh-indefinite (plus a null existential quantifier that combines with it in the analysis that we outline below). We saw above that the licenser and the whindefinite can be separated by an island boundary (15); however, it turns out that intermediate scope cannot be obtained when crossing the lower licenser would have to cross an island boundary: (20)

Hình như Nim không ăn [cái bánh mà Gi mua cho ai]. seemingly Nim Neg eat [CL cake Rel Gi buy for who] ‘It seems Nim did not eat the cake that Gi bought for anyone.’ *‘It seems that there is someone such that Nim did not eat the cake that Gi bought for that person.’

(21) Nếu Nim không ăn [cái bánh mà Gi mua cho ai] thì Gi rất vui. if Nim Neg eat [CL cake Rel Gi buy for who] then Gi very happy ‘If Nim does not eat the cake that Gi bought for anyone, then Gi will be very happy.’ *‘If there is someone such that Nim does not eat the cake that Gi bought for that person, then Gi will be very happy.’ In these examples, there are again two potential licensers, but now the whindefinite is inside an island to movement (a complex noun phrase), while the lower licenser is outside the island. The wh-indefinite would have to cross the island boundary in order to take scope over the lower licenser. In such configurations, intermediate scope is barred, which indicates that syntactic movement is crucially involved. Islands only block movement, they do not block other sorts of dependencies such as operator-variable binding (Chomsky 1977). There is also a different form for sentential negation, which does not permit a wh-indefinite to take scope over it:

224 (22)

Tran Thuan and Benjamin Bruening

Hình như Tân chẳng gặp ai. seemingly Tan Neg meet who ‘It seems Tan did not meet anyone.’ *‘It seems that there is someone that Tan did not meet.’

(23)

Nếu anh chẳng muốn mời ai thì cho tôi biết. if you Neg want invite who then let I know ‘If you do not want to invite anyone, let me know.’ *‘If there is someone that you do not want to invite, let me know.’

This form of negation also blocks covert wh-movement (which we have hypothesized takes place in the absence of a question particle; see Bruening and Tran 2006): (24) Tân chẳng mời ai. Tan Neg invite who ‘Tan didn’t invite anyone.’ *‘Who didn’t Tan invite?’ (25)

Tân không mời ai./? Tan Neg invite who ‘Tan didn’t invite anyone.’ or ‘Who didn’t Tan invite?’

We take this to indicate, again, that wh-indefinites must move in order to take higher scope. They do so covertly, without any visible effect. In summary, bare wh-indefinites have to take scope lower than their licenser; when there are two potential licensers, intermediate scope is obtained by covert movement of the wh-indefinite. This movement is blocked by syntactic islands, and by certain elements like negative chẳng.

2.2 Non-Bare Wh-Indefinites In contrast to bare wh-indefinites, non-bare wh-indefinites (those with the particle đó) can appear in any type of clause. For instance, they can appear in simple positive declaratives, unlike bare wh-indefinites (as in example 4, above, and all the examples here). They are also unlike bare wh-indefinites in that they can take scope over negation and conditional and yes-no question operators:

Wh-phrases as indefinites: A Vietnamese perspective

(26)

225

Tân không gặp ai đó. Tan Neg meet who Dem ‘There is someone that Tan does/did not meet.’

(27)

Nếu Tân không mời ai đó thì Thơ sẽ rất buồn. if Tan Neg invite who Dem then Tho Fut very sad ‘Someone is such that if Tan does not invite them, then Tho will be very sad.’

(28)

Cô ấy có gặp ai đó không? she Q meet who Dem Q ‘Did she meet someone (a particular person)?’

Scope-taking ignores movement islands, again unlike bare wh-indefinites: (29)

đó]. Tân không ăn [cái bánh mà Thơ mua cho ai Tan Neg eat [CL cake Rel Tho buy for who Dem] ‘Someone is such that Tan did not eat the cake that Tho bought for them.’ (widest scope)

We will account for non-bare wh-indefinites as involving choice functions (Reinhart 1997, Kratzer 1998), which take scope by binding, not by movement. Before we turn to our analysis of these facts, we first discuss some crosslinguistic considerations that favor one approach to wh-indefinites over other possible approaches.

3 Cross-Linguistic Considerations: Haspelmath’s Generalization We have seen above that Vietnamese uses wh-words both as interrogatives in whquestions and as indefinites. This is a very common situation cross-linguistically (e.g., Haspelmath 1997). We see three possible ways to analyze this alternation: one can analyze wh-words and indefinites as being identical; one can derive wh-words as questions from a basic indefinite use; or, one can analyze whindefinites as being derived from the question use. In many analyses of wh-questions, wh-words are treated as identical to indefinites. For instance, a wh-question like Who left? is analyzed as identical

226

Tran Thuan and Benjamin Bruening

to the declarative Someone left, except that the whole clause denotes a set of propositions rather than a single proposition (e.g., Karttunen 1977). In both cases, however, the subject is simply an existential quantifier, restricted to range over humans. In many analyses, then, wh-words are indefinites; for instance, Cheng (1991, 1994) analyses Mandarin Chinese wh-words as restricted free variables, exactly like indefinites. Other analyses, like that of Cole and Hermon (1998), take the second route and treat wh-questions as derived from indefinites, by the addition of some kind of question operator. However, we think both of these approaches are on the wrong track, for two reasons. First, wh-words used as indefinites are special, and do not have the distribution of lexical indefinites in any language. For instance, only wh-indefinites can appear in pairs in conditionals (“bare conditionals” in Cheng and Huang 1996). They also often need extra morphology to be used as indefinites, or require special licensing conditions. We take this to indicate that wh-indefinites are special, and require a special treatment. Second, a generalization noticed by Haspelmath (1997) points to the conclusion that the indefinite use of a wh-word is derived from the question use. That is, the question use is basic, from which the indefinite use is derived. In a survey of several hundred languages, Haspelmath found that many languages have a wh-word that occurs bare in its question use (e.g., where), while its indefinite use involves additional morphology (e.g., somewhere, anywhere). There are also many languages where the two are identical (like Vietnamese bare whindefinites). In contrast, there is no language where the indefinite use is basic, but the question use involves additional morphology: no language has an indefinite like place that becomes a wh-question word by adding a morpheme, like *wh-place. This generalization points to the wh-question use of wh-words being basic, and the indefinite use being derived, rather than the other way around. Therefore, the analysis that we build will start with the wh-question use as basic, with the indefinite use arising from the addition of something. This will either be an existential quantifier, which can be null (Vietnamese) or overt (English somewhere, anywhere); or it will be a morpheme that turns the whword into a choice function (Vietnamese đó).

4 A Neo-Hamblin Account Following Kratzer and Shimoyama (2002), based on an original idea of Hamblin (1973), we take wh-words to denote sets of individuals. They are not properties, though (type ); rather, they are individual alternatives (type e). In type, they

Wh-phrases as indefinites: A Vietnamese perspective

227

are individuals rather than functions. The alternatives that they denote expand via pointwise function application: a function that takes an individual argument combines in turn with each member of the alternative set. Spelling this out formally, we propose the following denotations for Vietnamese ‘who’ and ‘what’ (ignoring world variables): (30) a.

[[ai]] = {x: person(x)}

b. [[gì]] = {x: thing(x)} So, ai is the set of people. Suppose that there are three people in the context, a, b, and c; then ai denotes the set {a,b,c}. We now show how this will work in the various contexts where wh-words occur.

4.1 Wh-Questions Wh-questions are the basic context for wh-words. Consider the following example: (31) Tân gặp ai thế? Tan meet who Q ‘Who did Tan meet?’ In this example, the wh-word is the object. It will combine with the verb, which we assume is a two-place function (λxλy.y meets x), via pointwise function application. If {a, b, c} is the set of people, combining ‘who’ with ‘meet’ gives the set of properties {λy.y meets a, λy.y meets b, λy.y meets c}, as in (32b): (32)

a.

[[gặp]] = λxλy.y meets x

b. [[gặp ai]] = {λy.y meets x: person(x)} c.

[[(31)]] = {p: p= Tan meets x, x a person}

The verb will then combine with the external argument, Tân, to give a set of propositions (32c). A set of propositions is exactly the denotation of a question in a Hamblin semantics (there is probably a speech act operator that operates on this set of propositions, but we ignore this for present purposes). So the question meaning arises from the pointwise function application that occurs because of the wh-word. The question particle does not contribute to the question semantics at all; we assume that it serves to mark where pointwise function application stops (and it may have other functions, for instance marking the question as realis; see Bruening and Tran 2006).

228

Tran Thuan and Benjamin Bruening

In a Hamblin semantics, wh-movement may take place or not; it makes no difference to the semantics. If wh-movement does take place, it is not driven by the semantics, but rather by the syntax. Note further that pointwise function application can occur at every node of the tree, continuing up the tree and across clause boundaries; so wh-questions can be long-distance, and they can also violate islands (if no movement takes place for syntactic reasons; again, see Bruening and Tran 2006). So, treating wh-words as individual alternatives accounts neatly for whquestions. It also accounts for Haspelmath’s generalization: individual alternatives lead automatically to a question semantics (a set of propositions); something has to be done to get some other interpretation. We will turn to what that might be shortly, but as background we first have to consider embedded questions.

4.2 Embedded Questions Before spelling out how wh-words might be turned into indefinites in a Hamblin semantics, we have to first consider how the pointwise function application that is triggered by a wh-word might be stopped. In an embedded question like ‘I wonder who Tan met’, for instance, it would not do to continue pointwise function application beyond the embedding verb, otherwise the entire sentence would incorrectly be interpreted as a question. What we want is for the verb ‘wonder’ to take a set of propositions as its complement, and stop the pointwise function application. One way to do this is to recognize elements that are sets of alternatives as distinct types. So, for instance, an ordinary individual x is type e, but a set of individual alternatives {x} is a different type, which we will notate as eH . Same for all other types: λx.P(x) is type , while {λx.P(x)} is type H , a proposition p is type t, but {p} is type tH , and so on. Now, we restrict pointwise function application to cases where one of the elements involved is a Hamblin type (σH ): (33)

(Ordinary) Function Application: If α is a branching node with daughters β and γ, and [[β]]w,g M Dσ and [[γ]]w,g M D , then [[α]]w,g = [[γ]]([[β]]).

(34) Hamblin Function Application (HFA; adapted from Kratzer and Shimoyama 2002): a. If α is a branching node with daughters β and γ, and [[β]]w,g M DσH and [[γ]]w,g M D , then [[α]]w,g = {a a Dτ : p b p c [b a [[β]]w,g & c a [[γ]]w,g & a = c(b)]}

Wh-phrases as indefinites: A Vietnamese perspective

b.

229

If α is a branching node with daughters β and γ, and [[β]]w,g M Dσ and [[γ]]w,g M DH , then [[α]]w,g = {a a Dτ : p b p c [b a [[β]]w,g & c a [[γ]]w,g & a = c(b)]}

So, pointwise function application (HFA) is only triggered when one of two elements that must combine is a Hamblin type; otherwise, ordinary Function Application takes place. In our example above, the Vietnamese wh-in-situ ‘Tan met who?’, HFA is triggered when ‘meet’ combines with ‘who.’ The result of the combination is also a Hamblin type (because the output of the rule is a set of alternatives). The derivation takes place exactly as described above. However, in the embedded question case (‘I wonder who Tan met’), we want HFA to stop with ‘wonder’. This will happen because, as a distinct type, a Hamblin type can be the argument of a function. So, a question-embedding verb will take a Hamblin set of propositions as its first argument, and output a non-Hamblin type. Simplifying greatly, it will be type . Note that this is distinct from H , which is what is called for by the rule of HFA. What we have is the following tree: (35)

We assume that ordinary Function Application always takes place if it can. In this case, it can, because one of the sisters is exactly the type called for as the first argument of the other. HFA is only called upon when ordinary Function Application will not work. So HFA will not take place in this instance, even though one of the elements involved is a Hamblin type. The output will therefore be an ordinary type, , which can then combine with the external argument of ‘wonder’ to yield a proposition.3 With this in place, we can now turn to wh-indefinites.

4.3 Bare Wh-Indefinites We hypothesize that indefinites are created from wh-words by merging an existential quantifier with the wh-word. However, it is not a straightforward matter 3 Note that this account predicts intervention effects across the board. For instance, a whphrase inside an embedded wh-question should be unable to take higher scope. This is not true; see Tran (2009). It is probably necessary to augment this approach with indexing, as in Beck (2006).

230

Tran Thuan and Benjamin Bruening

to merge an existential quantifier with a Hamblin question word. Existential quantifiers are commonly conceived of as taking two sets as arguments, but a question word in a Hamblin semantics is an individual, not a set. We therefore propose that something special has to be done, and the fact that something special has to be done explains why wh-indefinites are special in many languages. For instance, the element that performs the special operation requires special licensing in Vietnamese. We posit the existence of a null element that mediates between the question word and the existential quantifier, which we notate as H. This element is like a question-embedding verb as described above in taking a Hamblin type as its argument and returning a non-Hamblin type. In this case, it takes a wh-phrase that denotes individual alternatives and returns an ordinary type : (36) [[H]] = λz M DeH .λx.x a z. (37)

a.

[[H(ai)]]= λx.x a {z: person(z)}

b. [[H(gì)]]= λx.x a {z: thing(z)} The output of this combination is then the appropriate type for combination with the existential quantifier, which we propose is null in Vietnamese. This null existential quantifier (“Øp”) merges with the [H wh-word] constituent to form a quantificational noun phrase: (38) a.

b.

[[Øp]] = λQλP.px.Q(x)=1 & P(x) = 1

c.

[[NP]] = λP.px.x a {z: person(z)} & P(x) = 1

In the following example, repeated from (5) above, the quantificational noun phrase will move and abstract over the proposition. Movement will take it to a position above the (base position of the) subject but below negation, so that negation will take scope over the entire proposition (we assume that movement of the subject to a pre-negation position is related to topicality and does not affect the denotational semantics): (39)

Tân không gặp ai. Tan Neg meet who ‘Tan does/did not meet anyone.’

Wh-phrases as indefinites: A Vietnamese perspective

231

(40) [[(39)]] = ¬px.x a {z: person(z)} & Tan meets x As stated above, we hypothesize that H, as a special syntactic/semantic element, requires licensing in Vietnamese. The question of what the licensing environments listed above have in common was addressed by Lin (1998b) for Mandarin Chinese, who proposed the following condition: (41)

The Non-Entailment of Existence Condition (Lin 1998b, 230, (34)) The use of [a wh-indefinite] is felicitous iff the proposition in which the [wh-indefinite] appears does not entail the existence of a referent satisfying the description of the [wh-indefinite].

Other possible semantic notions do not match the licensing environments listed above. For instance, yes-no questions and epistemic operators are not downward-entailing, while epistemic operators and nonfactive verbs are veridical (see Giannakidou 2001). It does seem that non-entailment of existence is generally the right notion, as we see from the following list of licensing contexts in Vietnamese: 1. Negation: A sentence like ‘Tan didn’t meet anyone’ does not entail the existence of someone Tan met; 2. Yes-no questions: A question like ‘Did Tan meet anyone?’ does not entail the existence of someone Tan met; 3. ‘If ’ (protasis of conditional): A clause like ‘If Tan meets anyone, . . .’ does not entail the existence of someone Tan will meet or has met; 4. Epistemic operators (‘seemingly’, ‘probably’): A sentence like ‘It seems that Tan met someone’ does not entail the existence of someone Tan met (it is only likely or probable); 5. Nonfactive verbs: A sentence like ‘Lan thinks Tan met someone’ does not entail the existence of someone Tan met; 6. Modals: A sentence like ‘You should eat something’ does not entail the existence of something the addressee will eat in the real world; 7. Imperatives: A sentence like ‘Eat something!’ does not entail the existence of something the addressee will eat in the real world; 8. Verbs like ‘want’ and ‘plan’: A sentence like ‘I want to eat something’ does not entail the existence of something I will eat in the real world. The contexts that do not license wh-indefinites all entail existence: 1. Positive declaratives: A sentence like ‘Tan met someone’ does entail the existence of someone Tan met;

232 2.

3.

Tran Thuan and Benjamin Bruening

Complements of factive verbs: A sentence like ‘I remember that someone came to look for you’ does entail the existence of someone that came to look for you (factive verbs presuppose the truth of their complements); Wh-questions.

The last item on the list is wh-questions, which we have not yet discussed. In a wh-question, all but the wh-phrase is presupposed. The fact that everything else is presupposed entails the existential closure of the variable contributed by the wh-indefinite. Therefore, wh-questions should not license wh-indefinites in Vietnamese. It appears that this is correct: (42) Cô ấy muốn biết [ai vừa mua gì]. she want know who just buy what a.

‘She wants to know who just bought what.’

b. *‘She wants to know who just bought anything.’ c. *‘She wants to know what anyone just bought.’ However, there are a couple of problems for the Non-Entailment of Existence Condition, which will lead us to revise it. First, complements of negated factive verbs do entail existence: A sentence like ‘I don’t remember that she met someone already’ entails the existence of someone that she met, since negated factive verbs still presuppose the truth of their complements. Mandarin and Vietnamese speakers judge that in the following examples (Vietnamese repeated from above), it is presupposed that someone did come to look for you and that she did in fact already meet someone: (43) Wǒ bú jìdé (yǒu) shéi lái zhǎo-guo nǐ. I not remember (have) who come look-for you ‘I do not remember that anybody came to look for you.’ (Mandarin; Lin 1998a, 236, (55)) (44)

Tôi không nhớ cô ấy đã gặp ai rồi. I Neg remember she ASP meet who already ‘I don’t remember that she already met someone.’

A second problem for Lin’s licensing condition is the consequent clause of a conditional sentence. This environment does not entail existence, yet whindefinites are disallowed (they are disallowed in consequents of conditionals in Mandarin Chinese, too):

Wh-phrases as indefinites: A Vietnamese perspective

233

(45) *Nếu Anh Thơ đến [thì ai sẽ rất vui]. if Anh Tho arrive [then who Fut very happy] ‘If Anh Tho arrives, anyone will be very happy.’ If Lin’s Non-Entailment of Existence Condition were correct, wh-indefinites should be licensed anywhere in a conditional. The third problem is existential ‘have’, which entails existence because it asserts it. Nevertheless, it licenses a wh-indefinite in Vietnamese (9, repeated from above): (46)

Có ai gặp Tân. have who meet Tan ‘Someone met/meets Tan.’

Because of these problems, we modify Lin’s licensing condition to be syntactic, but with a semantic basis. We state the following, syntactic, licensing condition on our null element H: (47)

Licensing Condition on H: H is licensed if and only if it is in the scope of an operator with an [NE] feature.

In this statement of the licensing condition, particular operators license H, ones with an [NE] feature. The following, semantic, condition endows certain operators with this feature: (48)

[NE] Operators: Let p be a proposition of the form px.P(x) & Q(x). Then a propositional operator OP has an [NE] feature if and only if OP(p) does not entail px.P(x) & Q(x).

So, operators have an [NE] feature if, when they operate on a proposition with an existential quantifier, they are non-veridical. Let us take as such a proposition our example ‘Tan met someone/anyone.’ Negation will have an [NE] feature, because negation operating on px.person(x) & Tan met x does not entail px.person(x) & Tan met x. The same holds for the yes-no question operator, epistemic operators (‘seemingly’, ‘probably’), nonfactive verbs, modals, imperatives, and verbs like ‘want’ and ‘plan’. In contrast, the assertion operator of a positive declarative is not [NE]; neither is a factive verb or a wh-question operator.

234

Tran Thuan and Benjamin Bruening

As for the three problematic contexts, negated factive verbs are the simplest to explain. Factive verbs do not have an [NE] feature, and so do not license whindefinites (in our terms, H); but negation does. H is therefore in the scope of an operator with an [NE] feature when it occurs in the complement of a negated factive verb, and is licensed. Since our licensing condition is purely syntactic, [NE] operators will always license wh-indefinites in their scope, regardless of the semantics of the actual sentence. This is an important difference between our licensing condition and Lin’s, and it straightforwardly explains why negated factive verbs license a wh-indefinite while at they same time they presuppose the existence of a referent for that wh-indefinite. Turning now to existential ‘have’, we hypothesize that có is the existential quantifier in an existential sentence, and there is no null quantifier. Notice that this makes it a non-propositional operator, since it does not take a proposition as an argument (it takes two properties). So the semantic condition in (48) does not apply. In the case of non-propositional operators that are not covered by (48), we hypothesize that some can be lexically specified as licensing H, that is, as having the feature [NE]. As we will see below, this is necessary for the demonstrative đó, too. So, we lexically endow có, ‘have’, and đó, the demonstrative that is analyzed below, with the [NE] feature. Existential có is therefore lexically specified as licensing H, by virtue of having the feature [NE] inherently, independently of the licensing condition above. It therefore licenses H by itself, and can always occur with a wh-indefinite. Since this is a lexical specification, a similar element in another language might not have the [NE] feature. This will give us the difference between Vietnamese and Mandarin Chinese, where existential ‘have’ does not license a whindefinite by itself. Mandarin ‘have’ (you) does not have the [NE] feature, and so a wh-indefinite that occurs with it will have to be within the scope of another [NE] operator. This nicely accounts for the difference between Vietnamese and Mandarin as a difference in lexical items.4 The third problematic context was conditionals. A wh-indefinite is licensed in the protasis of a conditional, as expected, but not in the consequent, which is unexpected. We think this difference will once again fall out from the licensing condition being syntactic rather than semantic. We contend that the right analysis of conditionals must involve an operator that takes scope only over the protasis and not over the consequent clause. Suppose that this operator is ‘if’ or its counterpart in other languages, which may be null. ‘If’ takes scope only over the 4 In another language, the null existential quantifier that we have posited in (38) might be lexically specified as licensing H by itself, and such a language would then have no limitation on where a wh-indefinite can occur. Such a language is Passamaquoddy, as described in Bruening (2007).

Wh-phrases as indefinites: A Vietnamese perspective

235

clause it occurs in, the protasis, and not over the consequent clause. Since this operator does not entail existence when it operates on a proposition of the form px.P(x) & Q(x), it has the [NE] feature. As such, it licenses H in its scope. Since only the protasis is in its scope, H is only licensed in the protasis of a conditional. While we do not have an analysis of conditionals to offer that would meet this characterization, we think that it is justified by cross-linguistic facts pointed out to us by Satoshi Tomioka. Cross-linguistically, consequent clauses of conditionals seem to act like simple declaratives. As matrix clauses, they have the form and morphology appropriate to a matrix clause in the language. Only the ‘if’ clause has a special form and acts like an embedded clause. In Passamaquoddy, for instance, ‘if’ clauses take a special morphological form, the Unchanged Conjunct (“3Conj” is the Conjunct form of 3rd person agreement), while the consequent clause has the form appropriate to a simple matrix clause:5 (49)

Nit olu sameht-aq, kat=te=hc that.Inan Top touch.Inan-3Conj Neg=Emph=Fut woli-y-uku-w-on. 3.good-make-Inv-Neg-InanSubj ‘If he touches it, it will not have a good effect on him.’ (Mitchell 1921/1976, line 68)

In English, the protasis of a conditional may have subject-auxiliary inversion, but the consequent clause may not, acting in this respect like a declarative matrix clause: (50)

a.

Had he done that, we would not be in this mess.

b. *Had he done that, would we not be in this mess. Given this robust cross-linguistic generalization, consequent clauses are essentially matrix declarative clauses and are not in the scope of an [NE] operator. The [NE] operator in the protasis takes scope only over that clause. Hence, whindefinites are only licensed in the protasis of a conditional, and not in the consequent clause (unless the consequent clause has another [NE] operator, like negation or a modal).

5 Conj = Conjunct inflection; Emph = Emphatic particle; Fut = Future; Inan = Inanimate; Inv = Inverse (object higher in animacy than subject); Subj = Subject.

236

Tran Thuan and Benjamin Bruening

Turning to the scope of the quantificational NP built out of a wh-phrase, when a bare wh-indefinite takes scope above one licenser but below another (intermediate scope), we posit movement of the quantificational noun phrase, as described above: (51)

Hình như ◊ cô ấy không thích ai. seemingly she Neg like who ‘It seems there is someone she does/did not like.’ (one interpretation of (19))

This is allowed only when the movement keeps the wh-indefinite within the scope of a licensing operator; otherwise the licensing condition is violated, since H must occur in the scope of an [NE] operator to be licensed. There are some data that suggest that the licensing condition as stated above is not good enough, and that there is a surface c-command condition in addition. A wh-indefinite is licensed as a subject only when it occurs after negation, and not before it: (52)

a. *Ai không gặp Tân. who Neg meet Tan ‘Someone doesn’t/didn’t meet Tan.’ gặp Tân. b. Không ai Neg who meet Tan ‘No one met/meets Tan.’

Below we state a version of the licensing condition that incorporates c-command: (53)

Licensing Condition on H (c-command version): H is licensed if and only if it is c-commanded by an operator with an [NE] feature.

It is not yet clear to us that this reformulation is necessary, however. It is also possible that (52a) is ungrammatical because the pre-negation position is a topic position, and wh-indefinites (indefinites generally) do not make good topics. At this point we do not attempt to decide between the two formulations.

4.4 Non-Bare Indefinites The following example, repeated from (4) above, illustrates a wh-word with đó, which can occur in any context, including a simple positive declarative:

Wh-phrases as indefinites: A Vietnamese perspective

(54)

237

Tân vừa gặp ai đó. Tan just meet who Dem ‘Tan just met someone.’

We hypothesize that the demonstrative đó is like có, above, in licensing H by itself. It can do this because it is lexically endowed with an [NE] feature. So, đó merges with the constituent [H wh-phrase], and licenses H. To explain the scopal properties of đó detailed above, we take đó to introduce a choice function, as follows: (55)

a.

b. [[đó]] = λP.f(P) c. [H ai đó] = f(λx.x a {z: person(z)}) The choice function f is bound by a null existential quantifier inserted high in the clause, in the CP layer: (56)

đó]. Tân không ăn [cái bánh mà Thơ mua cho ai Tan Neg eat [CL cake Rel Tho buy for who Dem] ‘Someone is such that Tan did not eat the cake that Tho bought for them.’ (widest scope)

(57) [[(56)]] = pf.CH(f) & ¬Tan eat ιy.cake(y) & Tho bought y for f(λx.x a {z:person(z)}) In general, non-bare wh-indefinites take very wide scope. We capture this by inserting the existential quantifier high, generally above other scopal operators. Interestingly, however, a non-bare wh-indefinite like ai đó obligatorily takes scope above a local negation, but it may take scope below a higher negation: (58)

Tân không gặp ai đó. Tan Neg meet who Dem ‘There is someone that Tan does/did not meet.’ *‘Tan did not meet anyone.’

(59)

đó]. Thơ không biết [Tân gặp ai Tho Neg know [Tan meet who Dem] ‘There is someone that Tho does not know that Tan met.’ or ‘Tho does not know that there is someone that Tan met.’

238

Tran Thuan and Benjamin Bruening

We follow Reinhart (1997) in allowing the existential quantifier to be inserted at any CP level. This means that the scope of a wh-indefinite will always be above local negation, because CP is higher than the position where negation occurs. However, a wh-indefinite can take scope lower than negation when there is a lower CP. The existential quantifier can be inserted in the embedded CP, giving it scope below negation in a higher clause. (It can also be inserted higher, giving it widest scope; the two options lead to ambiguity, as shown in 59.) In the following example, ai đó can apparently take narrow, intermediate, or wide scope: (60)

Tất cả sinh-viên phải đọc [tất cả các quyển sách [mà ai đó all student must read all PL CL book Rel who Dem đã giới-thiệu]]. Asp recommend ‘Every student must read every book that someone recommended.’

The wide scope and intermediate scope readings are expected, but the narrowest scope reading is not. But note that this example involves other quantificational noun phrases, in particular universal quantifiers. We follow Kratzer (1998) in allowing the choice function to be parameterized to include a variable that can be bound by a universal quantifier elsewhere in the sentence. This permits the following representations for the wide, intermediate, and narrow scope readings, respectively: (61) a.

pf such that every student1 must read every book 2 that f(λx.x a {z:person(z)}) recommended

b.

pf such that every student1 must read every book 2 that f 1 (λx.x a {z:person(z)}) recommended

c.

pf such that every student1 must read every book2 that f 2 (λx.x a {z:person(z)}) recommended

The existential quantifier actually takes widest scope, but the choice function variable f includes a variable that can be bound by another quantifier, allowing it to vary according to that quantifier. Hence, our analysis accounts for the various possible readings of non-bare wh-indefinites, and how they differ from bare whindefinites. One thing that our analysis does not account for is the fact that a non-bare wh-indefinite prefers to appear with existential ‘have’ when it is in subject position, just like indefinites generally:

Wh-phrases as indefinites: A Vietnamese perspective

(62)

239

Có ai đó gặp Tân. have who Dem meet Tan ‘Someone met/meets Tan.’

Above we hypothesized that ‘have’ is the existential quantifier when it occurs with a bare wh-indefinite. In the interests of a uniform analysis, we hypothesize that it is also an existential quantifier with non-bare wh-indefinites. In this case, it is the spellout of the existential quantifier over choice functions that is inserted in the CP layer. We suggest that it occurs in the C position, adjacent to the subject. It is only overt when the wh-indefinite is adjacent to it, as a subject. If this hypothesis is correct, it predicts that, unlike non-bare wh-indefinites generally, a non-bare wh-indefinite that occurs in an embedded clause adjacent to ‘have’ will only take narrow scope. We derived wide scope above by allowing the null existential quantifier to be merged high, at the highest CP. But if ‘have’ is the existential quantifier, we can see that it is low in (63), in the lowest CP: (63) Nếu có ai đó đến thì Anh Thơ sẽ rất vui. if have who Dem arrive then Anh Tho Fut very happy ‘If anyone arrives, Anh Tho will be very happy.’ *‘There is someone such that if that person arrives, Anh Tho will be very happy.’ Since the existential quantifier is below ‘if’, it should only take scope below ‘if’. This is correct. Example (63) only permits scope within the conditional clause, unlike (64), which lacks ‘have’: (64)

Nếu ai đó đến thì Anh Thơ sẽ rất vui. if who Dem arrive then Anh Tho Fut very happy ‘If anyone arrives, Anh Tho will be very happy.’ or ‘There is someone such that if that person arrives, Anh Tho will be very happy.’

We take this to be strong evidence in support of our analysis.

5 Conclusion To summarize our analysis, we have suggested that wh-words are universally sets of individual alternatives. Indefinites are created from wh-words by the

240

Tran Thuan and Benjamin Bruening

addition of two things: a null element H that turns a set of individual alternatives into a property; and either an existential quantifier or a choice function. The null element H in Vietnamese is only licensed by operators with an [NE] feature. These are operators that do not entail existence. Our analysis is consistent with cross-linguistic patterns indicating that indefinite uses of wh-words are derived from the more basic question use (Haspelmath’s generalization). In addition, however, there are some other crosslinguistic tendencies (possibly universals) that we have no account of yet. Vietnamese conforms to these tendencies, and is a good illustration of them. First, null existential quantifiers tend to require licensing. That is, if a language has two series of indefinites based on wh-words, one bare, one with additional morphology, the bare one is the one that requires licensing. We saw this above with Vietnamese: bare wh-indefinites require licensing, but non-bare ones do not. Cross-linguistically, there are non-bare wh-indefinites that require licensing (like English anywhere), but they are not typically opposed to a bare series of wh-indefinites in the same language (English does not use where as an indefinite). The second cross-linguistic tendency is that overt morphology yielding regular contrasts between bare wh-expressions and morphologically complex wh-expressions tends to introduce choice functions. That is, wh-words with additional morphology can be wide-scope indefinites (Bruening 2007). Conversely, null morphology tends to be an existential quantifier, not a choice function. In other words, bare wh-indefinites do not take wide scope; they are limited to lowest (or sometimes intermediate) scope (Bruening 2007). Our analysis does not account for these tendencies, if indeed they hold up as valid cross-linguistic generalizations. We leave verification and explanation of them to future research.

References Beck, Sigrid. 2006. “Intervention Effects Follow from Focus Interpretation.” Natural Language Semantics 14: 1–56. Bruening, Benjamin (2007), “Wh-in-Situ Does Not Correlate with WhIndefinites or Question Particles.” Linguistic Inquiry 38: 139–166. Bruening, Benjamin and Thuan Tran. 2006. “Wh-Questions in Vietnamese.” Journal of East Asian Linguistics 15: 319–341. Cheng, Lisa L.-S. 1991. On the Typology of Wh-Questions. Ph.D. thesis, Massachusetts Institute of Technology. Distributed by MIT Working Papers in Linguistics, Cambridge, Mass. Cheng, Lisa L.-S. 1994. “Wh-Words as Polarity Items.” In: Ren-Kui Li, ed., Chinese Languages and Linguistics II, pp. 614–640, Taipei: Academica Sinica. Cheng, Lisa L.-S. and C.-T. James Huang. 1996. “Two Types of Donkey Sentences.” Natural Language Semantics 4: 121–163.

Wh-phrases as indefinites: A Vietnamese perspective

241

Chomsky, Noam. 1977. “On WH-Movement.” In: Peter Culicover, Thomas Wasow, and Adrian Akmajian, eds., Formal Syntax, pp. 71–132, New York: Academic Press. Cole, Peter and Gabriella Hermon. 1998. “The Typology of Wh-Movement: Wh-Questions in Malay.” Syntax 1: 221–258. Giannakidou, Anastasia. 2001. “The Meaning of Free Choice.” Linguistics and Philosophy 24: 659–735. Hamblin, C. L. 1973. “Questions in Montague English.” Foundations of Language 10: 41–53. Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Clarendon. Karttunen, Lauri. 1977. “Syntax and Semantics of Questions.” Linguistics and Philosophy 1: 3–44. Kratzer, Angelika. 1998. “Scope or Pseudoscope: Are There Wide-Scope Indefinites?” In: Susan Rothstein, ed., Events and Grammar, pp. 163–196, Dordrecht: Kluwer. Kratzer, Angelika and Junko Shimoyama. 2002. “Indeterminate Pronouns: The View from Japanese.” In: Yukio Otsu, ed., Proceedings of the Third Tokyo Conference on Psycholinguistics, Tokyo: Hituzi Syobo. Lin, Jo-Wang. 1998a. “Distributivity in Chinese and Its Implications.” Natural Language Semantics 6: 201–243. Lin, Jo-Wang. 1998b. “On Existential Polarity Wh-Phrases in Chinese.” Journal of East Asian Linguistics 7: 219–255. Lin, Jo-Wang. 2002. “On Choice Functions and Scope of Existential Polarity Wh-Phrases in Mandarin Chinese.” Paper presented at GLOW Asia 2002, Taiwan. Mitchell, Lewis. 1921/1976. Koluskap Nekotok Skitkomiq (‘When Koluskap Left the Earth’). Indian Township, Maine: Wabnaki Bilingual Education Program. Edited and updated version of text in J. D. Prince (1921), “Passamaquoddy Texts,” Volume X of the Publications of the American Ethnological Society. Reinhart, Tanya. 1997. “Quantifier Scope: How Labor is Divided between QR and Choice Functions.” Linguistics and Philosophy 20: 335–397. Tran, Thuan. 2009. Wh-Quantification in Vietnamese. Ph.D. thesis, University of Delaware. Thuan Tran: [email protected] Benjamin Bruening: [email protected]

Marie-Claude Paris and Lê Thị Xuyến

9 On conjunction and comitativity in Vietnamese* In this paper, we describe some aspects of the syntax/semantics of và ‘and’, với ‘with’/‘and’, cùng ‘together’, (cùng) với ‘together with’ as well as their interaction with nhau ‘each other’. While và ‘and’ can only be labelled as a (true) conjunction — whether it is distributive or collective, với ‘with’/‘and’ presents two distributions: it can either be a (comitative) conjunction or a (comitative) adjunct. The coocurrence restrictions between với ‘with’ and cùng ‘together’ are studied with respect to three classes of predicates as well as those between và ‘and’ and với nhau. In conclusion, a parallelism is sketched between Vietnamese, on the one hand, and French et/avec and Chinese gēn, on the other. An asymmetry between the pre-verbal and the post-verbal positions is found to exist both in Chinese and in Vietnamese, but it works in a symmetric fashion. Keywords: conjunction, preposition, comitative, và ‘and’, với ‘with’/‘and’, cùng ‘together’, nhau, gēn ‘with’/‘and’, Chinese

Introduction We have chosen to study the following lexical items và ‘and’, với ‘with’/‘and’, cùng ‘together’, (cùng) với ‘together with’ for two reasons. First, because the labels these items are provided are seldom (or even never) justified syntactically in Vietnamese grammar books; either translations are given in French or in English – as for example in Trương (1970: 240) and Nguyễn (1997: 162) – or Chinese historical sources are provided. Và originates from hé, hùo (和) meaning ‘to mix’, với from hùi (會) ‘to gather’ and cùng from gòng (共) ‘to be in common, together’, cf. Alves (2005) and Trương (1970: 242–244). If such pieces of information are quite interesting, they are not of much use to the nonnative speaker, when s/he tries to construct well-formed sentences containing these elements. Moreover và ‘and’, với ‘with’/‘and’, cùng ‘together’, (cùng) với ‘together with’ are often found in construction with nhau ‘each other’, for which, * We would like to thank three anonymous reviewersfor their detailed remarks and suggestions. We address our thanks to Nguyễn Văn Huệ and Nguyễn Hoàng Trung (National University, Ho Chi Minh City) for sharing their intuitions with us. The usual disclaimers apply.

244

Marie-Claude Paris and Lê Thị Xuyến

again, translations are provided (such as (reciprocal) se in French or each other in English), but for which no syntactic/semantic characteristics are given. Second, because a good deal of research on coordination, comitativity and reciprocity has been carried out during the last four decades, and particularly between 2000 and 2010, but the contrast between và and với Vietnamese has not been dealt with as such in these publications.1 We hope to that this paper will contribute to a better understanding of the syntax and semantics of both và ‘and’ and với ‘with’/‘and’ as well that of their co-occurrence with nhau ‘each other’. This paper is organized as follows. First, we describe some2 syntactic characteristics of và and of với. Then, we study the co-occurrences of cùng and với and we investigate some properties of postverbal với nhau. Finally, we consider that nhau is a marker of co-presence rather than of (weak or strong) reciprocity. In conclusion, we draw a parallelism between the (synchronic) behavior of và, với, cùng, (cùng) với and French et/avec, on the one hand, and Mandarin Chinese gēn, on the other hand.

1 The Distribution of và First, we briefly study the distribution of và and, then, as certain it is a conjunction. When necessary, we compare và and với. Và forms part of the subject in the pair (1)–(2), and of the object in (3)–(4). (1) Kim và Tam làm việc. Kim and Tam do work ‘Kim and Tam work.’

(2)

Kim và Hoa cãi nhau. Kim and Hoa quarrel ‘Kim and Hoa quarrel.’

1 In our view, some of the most important pieces of research in functional/typological linguistics in the last decade are: Arkhipov (2009), Choi-Jonin (2002), Evans (2008), Frajzyngier & Curl, Tracy S. (eds.) (2000), Haspelmath (2004a, b), Lehmann and Shin (2005), Stassen (2000), Stolz et al. (2006); and in formal linguistics: Progovac (2003) and Zhang (2006, 2007). In the special issue of Language and Linguistics (2011, 12.1), edited by Stacy Fang-Ching Teng and devoted to Austronesian languages, both formal and functional approaches are used. We wrote this paper long before this issue of Language and Linguistics came out. To our knowledge, in all the publications mentioned above, the contrast between và ‘and’ and với ‘and’/‘with’ is not studied. In the remainder of this paper, we will not gloss the grammatical morphemes we are studying any longer. 2 We do not study all the distributional characteristics of và and với. Suffice it to say that và has a broader distribution than với. It is co-occurs with DPs, PPs, VPs, CPs, but (conjunctive) với typically co-occurs with DPs.

On conjunction and comitativity in Vietnamese

(3)

Kim ăn cơm và cá. Kim eat rice and fish

(4) Ăn táo và lê mỗi ngày. eat apple and pear every day

‘Kim eats rice and fish.’3

‘Eat apples and pears every day.’

245

In (5)–(8), với substitutes with và as used in (1)–(4). Note that while in the pair (5)–(6) the substitution of và and với does not yield much meaning difference, this is not the case in the pair (7)–(8). The readings of và in (3)–(4) and of với in (7)–(8) are different. In the former pair và is interpreted as a distributive conjunction, in the latter với is interpreted as a comitative/collective conjunction. (3) indicates to two separate eating events, but (7) only one. (5)

Kim với Tam làm việc. Kim and/with Tam do work ‘Kim and Tam work.’ ‘Kim works with Tam.’

(6)

Kim với Hoa cãi nhau. Kim and/with Hoa quarrel ‘Kim and Hoa quarrel.’ ‘Kim quarrels with Hoa.’

(7)

Kim ăn cơm với cá. Kim eat rice and/with fish ‘Kim eats rice (together) with fish.’ ‘Kim eats rice and fish.’

(8) Ăn táo với lê mỗi ngày. Eat apple and/with pear every day ‘Eat apples (together) with/and pears every day.’

3 (3)–(4) would be best translated by the distributive conjunction et . . . et . . . in French, as in (i) and (ii), respectively. (i) Kim mange et du riz et du poisson ‘Kim eats both rice and fish.’ (ii) manger tous les jours et des pommes et des poires ‘eat both apples and pears every day’ The equivalent to French et . . . et . . . , which would be and . . . and . . . in English, does not exist. It does not exist in Vietnamese either: *và . . . và. . . .

246

Marie-Claude Paris and Lê Thị Xuyến

The standard tests4 used to ascertain that a lexeme functions as a coordinator work well in Vietnamese. In a [NP1 và NP2 ] complex constituent, the subconstituent [và NP2 ]5 cannot 1. occupy different positions. Compare the difference in grammaticality between (9) and (10), 2. be preceded by the negative marker không, (11)–(12), 3. be preceded by an adverb such as vẫn ‘always’, (13)–(14), 4. be cleft, (13), (15). (9)

Kim và Tam làm việc. Kim and Tam do work ‘Kim and Tam work.’

(10) *Kim làm việc và Tam. Kim do work and Tam (11) Kim và Tam không đi ăn. Kim and Tam Neg. go eat ‘Kim and Tam do not go to eat.’ (12) *Kim không và Tam đi ăn. Kim Neg. and Tam go eat (13) Kim và Tam vẫn đi ăn ở quán kia. Kim and Tam always go eat Loc restaurant deictic ‘Kim and Tam always go to eat in this restaurant.’ (14) *Kim vẫn và Tam đi ăn ở quán kia. Kim always and Tam go eat Loc restaurant deictic (15) *Kim không phải là và Tam đi ăn ở quán kia. *Kim Neg. true be and Tam go eat Loc restaurant deictic 4 See Ross (1968). As noted by a reviewer, these tests belong to common knowledge in syntax, but as they are seldom used in Vietnamese linguistics books, we deemed it necessary to use them at length. Note that, contrary to standard assumptions, “extraposition of coordinands to the end of the clause” is attested, in German for example, Haspelmath (2004: 7). 5 In a coordinate construction [A & B], the two elements A and B stand in an asymmetrical (syntactic) relationship, cf. Progovac (2003) and Zhang (2006). [A & B] should be bracketed [A [& B]].

247

On conjunction and comitativity in Vietnamese

The results of the tests mentioned above can be summed up as in Table 1 below:

và + NP

Movement

Negation

Adverbial modification

Clefting

*

*

*

*

Table 1: Diagnostic profile of và + NP

In section 1 we have as certained that và is a coordinator and mentioned that, in some contexts, và and với can be substituted for each other. In that case, a meaning difference can be noticed: và indicates distributive coordination,6 but với comitative/collective coordination. In the following section we will study the distribution of với. We will compare với and và when necessary.

2 The distribution of với In this section, we will oppose two different uses of với.

2.1 Postverbal với NP In the three pairs (16)–(21) below, we will test whether the substitution between và and với – already noted above in (5)–(8) – always leads to grammatical results. Recall that we attributed the difference between NP1 và NP2 and NP1 với NP2 constituents to a different semantic property. When opposed, và indicates a distributive conjunction and với a comitative/collective conjunction.7 The alternation between với and và in preverbal position in (16), (18) and (20) yields grammatical results. But the positioning of với NP2 in postverbal position does not work alike. (17) and (19) are acceptable, but (21) is not.8 Thus, while predicates such as làm việc ‘to do work’ or cãi nhau ‘to quarrel’ allow both for the substitution of và and với and for two positionings of the với NP2 constituent, a

6 In opposition to với, và can be a distributive conjunction. Và is underspecified: it indicates either distributivity or collectivity, depending on the context. 7 In numerous languages distributive and collective conjunctions are marked differently. See, for instance, Hetzron (1973) for Hungarian, McNally (1993) for Russian and Skrabalova (2003) for Czech and also Arkhipov (2009: 228–229, 234–235). 8 Sentences with postverbal với, (17) and (19), are more natural than those with preverbal với, (16) and (18). But we need to study với in subject position, in order to compare it with và in (1) = (9). For other uses of postverbal với, see Nguyễn (1997: 191–192).

248

Marie-Claude Paris and Lê Thị Xuyến

predicate such as là cô giáo ‘to be a teacher’ does not.9 Compare the grammatical pairs (16)–(17) and (18)–(19) to (20)–(21). (21) is ungrammatical, contrary to (17) and (19). (16) Tôi với (/và) Tam làm việc. I with/and Tam do work

(17)

‘I work with Tam’. ‘I and Tam work.’ (18)

Kim với (/và) Hoa cãi nhau. Kim with/and Hoa quarrel ‘Kim and Hoa are quarrelling.’

(20)

Tôi làm việc với Tam. I do work with Tam ‘I work with Tam.’

(19)

Kim cãi nhau với Hoa. Kim quarrel with Hoa ‘Kim is quarreling with Hoa.’

Kim với (/và) Hoa là cô giáo. (21) *Kim là cô giáo với Hoa. Kim with/and Hoa be a teacher Kim be a teacher with Hoa ‘Kim and Hoa are teachers.’

Before we try to explain the ungrammaticality of (21), let us investigate some syntactic properties of với in greater detail. First, we will test the deletability of the với NP constituent and, second, we will modify với by means of the adverb vẫn ‘always’.

2.2 Deletion of the với NP constituent An asymmetry was noted above between predicates like làm việc ‘to work’ and cãi nhau ‘to quarrel’ on the one hand, and là cô giáo ‘to be a teacher’ on the other, with respect to the postverbal position of [với NP]. A second asymmetry can be established now between làm việc ‘to work’ and là cô giáo ‘to be a teacher’, on the one hand, and cãi nhau ‘to quarrel’, on the other. From (17) and (21), we can derive (22)–(23), but from (19) we cannot derive (24). 9 In other constructions than those under study, in comparisons of equality for example, [với NP] can occupy different positions. Compare (i) and (ii). (i) A ở một nhà [với B]. A live one house with B ‘A lives in the same house as B.’ (ii) A ở [với B] một nhà. A live with B one house ‘A lives in the same house as B.’

On conjunction and comitativity in Vietnamese

(22)

Tôi làm việc. I do work ‘I work.’

(23) Kim là cô giáo. Kim be a teacher

249

(24) *Kim cãi nhau. Kim quarrel

‘Kim is a teacher.’

Hence, from the grammaticality of (22)–(23) and the ungrammaticality of (24), we can draw the following conclusions: 1. the three predicates chosen all behave differently, 2. [với NP] is an optional constituent in non-symmetrical predications only, i.e. (22)–(23). In symmetrical predications, such as cãi nhau ‘to quarrel’, the subject NP must be plural. Plurality can be indicated either by a plural NP or by a coordinate NP, as in (18). As the NP Kim in (24) is singular, the unacceptability of (24) is accounted for. 3. (17) and (19) show that a sequence NP1 với NP2 in (16), (18) should be bracketed as follows: [NP1] [với NP2 ]. NP1 and với NP2 are two separate constituents. 4. If với can be labelled as a comitative/collective conjunction in (7), it can be labelled as comitative preposition in (17). In (17) với NP is an adjunct, in (18) it is a comitative conjunct. 5. A comitative adjunct is found in a predication which involves an event performed by an Agent and another participant in a ‘joint participation’ (Teng 1970; Lehmann and Shi 2005), as in (17). With stative predications, no ‘joint participation’ is possible, hence (21) is ungrammatical. Table 2 sums up the optionality of với NP and its postverbal distribution: Types of predicates

postverbal [với NP]

[với NP] Optional argument

làm việc ‘to work’

Z

Z

cãi nhau ‘to quarrel’

Z

*

là cô giáo ‘to be a teacher’

*

Z

Table 2: [với NP] and types of predicates

2.3 The co-occurrences between với NP and the adverb vẫn ‘always’ When modified by the adverb vẫn ‘always’, preverbal với in (25) behaves like the conjunction và in (14). We can thus conclude that preverbal với is a comitative conjunction. (26) is the well-formed counterpart of (25).

250

Marie-Claude Paris and Lê Thị Xuyến

(25) *Kim vẫn với Hoa cãi nhau. Kim always and/with Hoa quarrel (26)

Kim với Hoa vẫn cãi nhau. Kim and/with Hoa always quarrel ‘Kim and Hoa are always quarrelling.’ ‘Kim is always quarrelling with Hoa.’

The well-formedness of (27) is interesting. Postverbal [với NP2 ] behaves differently from [và NP2 ], as the contrast in grammaticality between (27) and (28) shows. Postverbal với NP is acceptable, postverbal [và NP2 ] is not. From (27), we can label với in a [với NP2 ] constituent in postverbal position as a comitative preposition. (27)

Kim vẫn cãi nhau với Hoa. Kim always quarrel with Hoa ‘Kim and Hoa are always quarrelling.’

(28) *Kim vẫn cãi nhau và Hoa. Kim always quarrel and Hoa To sum up, in this section we have shown that [và NP2 ] and [với NP2 ] have different properties. Và is a coordinator, với most probably either a comitative conjunction or a comitative preposition, depending on the (semantic) type of predicates it co-occurs with. If such a hypothesis is true, we can proceed one step further and now explain the ungrammaticality of (21). The predicate là cô giáo ‘to be a teacher’ is stative. Statives do not allow comitative PP, only activity predicates do. Hence the ungrammaticality of (21) is due to a semantic incongruence. Là cô giáo ‘to be a teacher’ allows a preverbal conjunctive [với NP2 ] phrase, but not a postverbal comitative prepositional phrase [với NP2 ]. So far we have we shown that: (i) và is only a coordinator, (ii) conjunctive với and prepositional với behave differently, (iii) there is an asymmetry between the preverbal and the postverbal positions: comitative PPs are allowed only postverbally. In the next section we will study the distribution of cùng as a modifier of với. We will take up the three classes of predicates mentioned above: làm việc ‘to work’, cãi nhau ‘to quarrel’ and là cô giáo ‘to be a teacher’.10 10 The claims we are making here do not hold for just the three predicates we use as examples, but for the whole semantic classes these specific predicates belong to. We thank an anonymous reviewer for pointing this out to us.

On conjunction and comitativity in Vietnamese

251

3 The distribution of cùng ‘together’ before [với NP2 ] Grammar books mention without giving any constraints that với can be modified by cùng, yielding a cùng với NP constituent.11 We will show that such a statement is far too general. The co-occurrence between cùng and với is constrained by the status of với. Our conclusion is that comitative/prepositional với can be modified by cùng, but that comitative/conjunctive với cannot. Below we study the distribution of both preverbal and postverbal với phrases modified by cùng in sentences containing the predicates làm việc ‘to work’, cãi nhau ‘to quarrel’ and là cô giáo ‘to be a teacher’, respectively. Note that the twoway distinction between predicates like làm việc, on the hand, and predicateslike cãi nhau and là cô giáo, on the other hand, established in (22)–(24) above finds a confirmation. When với indicates comitativity (in activity predications), it can be modified by cùng ‘together’, both in pre-and post-verbal positions, cf. (29)–(30). A strict parallelism can be drawn between (16)–(17) above and (29)–(30). (29)

Kim cùng với Tam làm việc. Kim together with Tam do work ‘Kim works with Tam.’

(30)

Kim làm việc cùng với Tam. Kim do work together with Tam ‘Kim works with Tam.’

Contrary to activity predications, both symmetrical and stative/distributive predicates disallow cùng với NP, whether in pre- or post-verbal positions, cf. (31)–(32) and (33)–(34). (31) *Kim cùng với Hoa cãi nhau. Kim together with Hoa quarrel ‘Kim is quarreling with Hoa.’ (32)

*Kim cãi nhau cùng với Hoa. Kim quarrel together with Hoa

(33)

*Kim cùng với Hoa là cô giáo. Kim together with Hoa be a teacher

11 Trương (1970: 242), Nguyễn (1997: 163).

252

Marie-Claude Paris and Lê Thị Xuyến

(34) *Kim là cô giáo cùng với Hoa. Kim be a teacher together with Hoa With symmetrical predicates, there is a discrepancy between the pre-and post-verbal distribution of với NP and cùng với NP. Whereas với NP is always acceptable, cùng với NP never is. Compare the well-formed pair (18)–(19) to the ill-formed (31)–(32). With distributive/stative predicates, there is a discrepancy in the preverbal position between với NP and cùng với NP. The former is accepted, the latter is not. Compare (20), which is grammatical, to (33), which is not. As is expected, when postverbal với NP is unacceptable, cùng với NP is unacceptable, too. Both (21) and (34) are ungrammatical. The distribution of pre-and post-verbal [cùng với NP] is summed up in Table 3 below. Types of predicates

Preverbal [cùng với NP]

Postverbal [cùng với NP]

làm việc ‘to work’

Z

Z

cãi nhau ‘to quarrel’

*

*

là cô giáo ‘to be a teacher’

*

*

Table 3: [cùng với NP] and types of predicates

4 The distribution of cùng ‘together’ + NP In the previous section we studied the co-occurrence between cùng and với. Now we describe examples where cùng ‘together’ follows an NP. In (35), one could say that cùng functions as a (comitative) preposition. As expected from its meaning, it is only comitative/collective conjunctions that cùng ‘together’ can reinforce, hence its ungrammatical co-occurrence with (distributive) và, (36). (35)

Kim cùng Tam làm việc. Kim together Tam do work ‘Kim works together with Tam.’

(36) *Kim cùng và Tam làm việc. Kim together and Tam do work

On conjunction and comitativity in Vietnamese

253

Cùng cannot be substituted for preverbal với neither in distributive nor in symmetrical12 predications, cf. (36)–(37). Hence cùng is not a conjunction. (37)– (38) stand in opposition to (18) and (20). (37) *Kim cùng Hoa là cô giáo. Kim together Hoa be a teacher (38) *Kim cùng Hoa cãi nhau. Kim together Hoa quarrel Table 4 sums up the distribution of preverbal [cùng NP]. Types of predicates

preverbal [cùng với NP]

làm việc ‘to work’

Z

cãi nhau ‘to quarrel’

*

là cô giáo ‘to be a teacher’

*

Table 4: Preverbal [cùng với NP] and types of predicates

It is often claimed that in the cùng với NP sequence, cùng is optional: it simply reinforces the meaning of với.13 Such a statement is only partially true. It is not the case that all với NP constituents can be transformed into ‘emphatic’ cùng với NP. Only activity verbs which allow an optional comitative phrase allow the two variants với NP and cùng với NP. Neither in symmetrical nor in stative/ distributive predicates do với NP and cùng với NP alternate. Cf. (31), (33) and (36)–(37). In order to keep the distribution of cùng regular, rather than saying that it is both a (comitative) preposition – cf. (35) – and a (comitative) adverbial modifier – cf. (29) – one could say that it always is an adverbial modifier whose comitative partner với is deleted in activity predications, because it is semantically redundant. Thus, cùng với Tam would be the underlying form of cùng Tam in (35).

5 The distribution of post verbal với nhau Most grammar books equate postverbal nhau with a reciprocity marker,14 translating it into English ‘each other’ or French (reciprocal) ‘se’. If true, one would 12 This is also noted by Lehmann and Shin (2005: 40). “No language in our sample uses a comitative adverb (meaning ‘together’) to express the relation of reciprocal partner.” 13 Nguyễn (1988: 185–189). 14 “The mutual relationship is expressed by the reciprocal substitute nhau ‘(with) each other’, ‘(with) one another’ (Nguyễn 1997: 137)”. We disagree with a reviewer who suggested to treat nhau as a marker of weak reciprocity, because nhau is found in context where no reciprocity – whether weak or strong – is at stake.

254

Marie-Claude Paris and Lê Thị Xuyến

expect nhau to occur only in symmetrical predications, i.e. with verbs such as là bạn ‘to be friends with’, là đồng nghiệp ‘to be colleagues’ etc. This is indeed the case — see (39)–(40), but activity predicates such as làm việc ‘to work’ or ăn ‘to eat’ also allow với nhau, cf. (41)–(42). As expected distributive predications, such là cô giáo ‘to be a teacher’, mang thai/có thai/có chửa ‘to be pregnant’ etc. do not allow với nhau, cf. (43)–(44). (39)

Các anh là bạn với nhau. Pl. Sg2 be friend with each other ‘You are (mutual) friends.’

(40)

A và B cười với nhau. A and B smile with each other ‘A and B smile to each other.’

(41)

Chúng tôi cùng làm việc với nhau. Pl. I also do work with each other ‘We work together.’

(42) Kim và Tam cùng ăn với nhau. Kim and Tam also eat with each other ‘Kim and Tam eat together.’ (43) *A và B là cô giáo với nhau A and B be a teacher with each other (44) *A và B là mang thai/có thai/có chửa với nhau A and B be pregnant/have pregnant/have pregnant with each other Table 5 sums up the distribution of với nhau in construction with three different types of predicates. Types of predicates

postverbal [với nhau]

làm việc ‘to work’

Z

là bạn ‘to be friends’

Z

là cô giáo ‘to be a teacher’

* *

mang thai/có thai/có chửa ‘to be pregnant’ Table 5: Postverbal [với nhau] and types of predicates

On conjunction and comitativity in Vietnamese

255

6 The distribution of postverbal cùng với nhau From what we now know of the distribution and meaning of the comitative adverbial cùng, one expects that the distribution of postverbal với nhau and cùng với nhau be similar. This is exactly what is the case. Only (comitative) activity predicates, i.e. comitative adjuncts, allow postverbal cùng với nhau. Both symmetrical and distributives predicates do not. Compare the difference in acceptability between (45) and (46)–(47). (45)

?Chúng tôi làm việc cùng với nhau.15 Pl. I do work with each other ‘We work together.’

(46)

??/*A và B là đồng nghiệp cùng với nhau. A and B to be colleagues with each other

(47)

*A và B là cô giáo cùng với nhau. A and B to be a teacher with each other

Table 6 sums up the distribution of both postverbal với nhau and cùng với nhau. Types of predicates

Postverbal [với nhau]

Postverbal [cùng với nhau]

làm việc ‘to work’

Z

Z

cãi nhau ‘to quarrel’

Z

*

là cô giáo ‘to be a teacher’

*

*

Table 6: Postverbal [với nhau]/[cùng với nhau] and types of predicates

7 More on the marker nhau Rather than saying that nhau translates as ‘each other’/‘one another’ or that it marks reciprocity, we would like to claim that it simply indicates co-presence.16

15 (i) below, where cùng is preverbal, is more natural than (45). (i) Chúng tôi cùng làm việc với nhau. Pl. I also do work with each other ‘We work together.’ 16 See Seiler (1974). The reader is referred to Lehmann and Shin (2005: 18) for a definition of the notion of concomitance.

256

Marie-Claude Paris and Lê Thị Xuyến

Nhau is (optionally) found both in symmetrical/collective predications and in non-symmetrical/distributive predications, as in (48), (50). If (48) does indicate reciprocity, i.e. ‘the participants bear identical pairs of roles in identical relations (of reverse directionality)’17, in (49)–(50) the two participants Kim and Hoa bear identical relations to the predicate, but these relations are not converse at all. (48)

A và B là đồng nghiệp (với nhau). A and B be colleagues (with each other) ‘A and B are colleagues.’

(49)

Kim và Hoa là cô giáo cùng làng. Kim and Hoa be a teacher same village ‘Kim and Hoa are teachers in the same village.’

(50)

Kim và Hoa là cô giáo cùng làng (với nhau). Kim and Hoa be a teacher same village (with each other) ‘Kim and Hoa are teachers in the same village.’

Table 7 summarizes the facts studied so far in postverbal position. Columns one and two show the symmetrical behaviors of với NP and với nhau. Columns three, four and five illustrate the regular distribution of cùng. Types of predicates

postverbal postverbal postverbal postverbal preverbal [với NP] [với nhau] [cùng với NP] [cùng với nhau] [cùng NP]

làm việc ‘to work’

Z

cãi nhau ‘to quarrel’

Z

là cô giáo ‘to be a teacher’ *

Z

Z

Z

Z

Z

*

*

*

*

*

*

*

Table 7: The diagnostic profile of the different predicate classes

In sum, so far we have studied and opposed và NP and với NP constituents. If và is undoubtedly a conjunction, với can either introduce a comitative conjunct or a comitative adjunct. Comitative adjuncts are, by definition, optional/ external arguments. They In (i) below, which contains a reciprocal predicate, nhau ‘each other’ is not compulsory. Hence it is difficult to maintain that reciprocity should be marked by means of nhau. (i) Kim cùng cãi (nhau) với Hoa. Kim together quarrel (each other) with Hoa ‘Together Kim and Hoa are quarreling with someone else.’ 17 Lichtenberk (2000: 56).

On conjunction and comitativity in Vietnamese

257

(i) are preferably found in postverbal position, (ii) can be modified by cùng, both pre- and post-verbally, (iii) can also take với nhau postverbally. A với comitative conjunct can appear in preverbal position, but cannot be modified by cùng in this position. It is an internal argument, typically found in symmetrical/collective predications. The behavior of và and với NP constituents is not an idiosyncrasy of Vietnamese. Both French and Mandarin Chinese exhibit similar phenomena. Before we sketch a comparison between these languages in section 9 below, we will propose a syntactic structure for the different uses of và and với studied above.

8 The syntactic structures of và and với The two conjunctions và and với share a similar syntactic structure, but they differ in that they accept different types of constituents. As mentioned in note 2, và ‘and’ has a wide distribution: it admits DPs, PPs, VPs and CPs; on the contrary, conjunctive với ‘with’ only admits [+human] DPs. To account for the behavior of và, we adopt Progovac’s (2003) and Zhang’s (2006) general structure for coordination, cf. (51). Và is the head of the construction, A the specifier of & and B its complement. A and B stand for possibly different lexical categories. (51)

[A [& B]].

For với, we adopt Zhang’s (2007) two-fold analysis of with and posit two different syntactic structures (52) and (53), respectively. With symmetrical/collective với, DP1 is the specifier of với and DP2 its complement. Both DP1 and DP2 are [+human]. The whole DP is (nominal and) plural. D cannot be modified by the adverbg cùng, cf. (31) above. With non-symmetric/comitative với, the constituent [với DP2 ] is a PP, which is adjoined to DP1. P can, as expected, be modified by cùng, cf. (29). (52)

(53)

258

Marie-Claude Paris and Lê Thị Xuyến

cf. (31) verb làm việc ‘to work’ *Kim cùng với Hoa. . . Kim together with Hoa

cf. (29) verb cãi nhau ‘to quarrel’ Kim cùng với Tam. . . Kim together with Tam

9 French et/avec and Mandarin Chinese gēn In French two morphemes et ‘and’ and avec ‘with’ share similarities with Vietnamese conjunctive và and conjunctive/prepositional với, respectively. In Mandarin the morpheme gēn ‘and’/‘with’ also presents syntactic commonalities with và and với. Gēn functions as a distributive or a comitative conjunction and as a preposition, depending on the predicates it co-occurs with. As noticed by Seiler (1974), French avec, English with and German mit are the only prepositions which have negative counterparts.18 Sans, without and ohne are the negative counterparts of avec, with and mit. Note that while the (sometimes synonymous) counterparts of avec, with and mit — that is et, and and und — are undoubtedly conjunctions, the (syntactic) status of avec, with and mit is much more problematic. A clear distinction can be made, though, between two uses of avec, with and mit19: a comitative and non comitative usage. We will show this in what follows. In the synonymous pair (54) and (55)20, conjunctive et and prepositional avec share semantic properties. The negative counterpart of (55) is (56). Moreover, the avec NP of (52) can be deleted, cf. (54).

18 See also Stolz, Stroh, and Urdze (2006: 167–170). 19 The instrumental or adverbial uses of avec, with and mit are studied in (Seiler 1974), among others, but not here. Note, in passing, that contrary to English, French and German, the comitative and the instrumental markings are not identical in Vietnamese (với vs. bằng). For more examples of bằng, cf. Nguyễn (1997: 162, 192). Lehman and Shin (2005: 81) give an interesting example (E 207b) where preverbal với and bằng alternate. We have slightly changed the word by word gloss as well as the translation of this example. (i) Với/bằng sự quen biết tôi được làm việc tại and/with Cl acquaintance I get do work Loc cơ quan nhà nước. [E 207b] office house state ‘I got a job in the public service thanks to an acquaintance.’ 20 Such a synonymy is, of course, only partial. See Choi-Jonin (2002) for more details.

On conjunction and comitativity in Vietnamese

(54)

Jean et Marie travaillent. ‘John and Mary work.’

(55)

Jean travaille avec Marie. ‘John works with Mary.’

(56)

Jean travaille sans Marie. ‘John works without Mary.’

(57)

Jean travaille Ø Ø. ‘John works Ø Ø.’

259

The quadruplet (58)–(61) below does not behave like (54)–(57) above. (60)– (61) are built on the same model as (56)–(57), but they are not acceptable. The reason is that se quereller ‘to quarrel’ is a symmetrical predicate, whereas travailler ‘to work’ is not. Contrary to travailler ‘to work’, se quereller ‘to quarrel’ requires (at least) two (internal) participants. This is shown by the unacceptability of (60)–(61). From (60)–(61), it can be concluded that the avec Marie constituent in (59) is not a comitative/external argument. Avec Marie in (59) is a compulsory/ internal argument. (58)

Jean et Marie se querellent. ‘John and Mary quarrel.’

(59)

Jean se querelle avec Marie. ‘John quarrels with Mary.’

(60)

*Jean se querelle sans Marie. ‘John quarrels without Mary.’

(61) *Jean se querelle Ø Ø. *‘John quarrels Ø Ø.’ Let us now study the behavior of et and avec in a stative/distributive predication, like être professeur ‘to be a teacher’. The affirmative avec ‘with’ and the negative counterparts sans of et ‘and’ in (60) are not acceptable: (60)–(61) are ill-formed. Hence one can conclude that être professeur ‘to be a teacher’ does not allow a comitative/external argument. Moreover, as (62) is grammatical — unlike (58), we can ascertain that être professeur ‘to be a teacher’ is not a symmetrical predication: it is distributive.

260

Marie-Claude Paris and Lê Thị Xuyến

(59)

Jean et Marie sont professeurs. ‘John and Mary are teachers.’

(60)

*Jean est professeur avec Marie *‘John is a teacher with Mary.’

(61) *Jean est professeur sans Marie *‘John is a teacher without Mary.’ (62)

Jean est professeur Ø Ø. ‘John is a teacher Ø Ø.’

In sum, in the examples chosen, French avec presents a dual syntactic behavior. On the one hand, it is a (comitative) preposition, in which case it introduces an external/optional argument, i.e. an adjunct. On the other hand, avec introduces an internal/non optional argument: it marks a comitative conjunction.21 The French facts presented above are summarized in Table 8 below. They are paired with their Vietnamese and English counterparts. type of predicates

examples

conjunctive/prepositional behaviour

– symmetrical predicate ± activity

Viet.: là cô giáo Engl.: to be a teacher Fr.: être professeur

và and/*with/*without et/*avec/*sans

+ symmetrical predicate ± activity

Viet.: cãi nhau Engl.: to quarrel Fr.: se quereller

với/*cùng với with/*without avec/*sans

– symmetrical predicate + activity

Viet.: làm việc Engl: to work Fr.: travailler

cùng với with/without avec/sans

Table 8: Vietnamese, English and French compared

Contrary to what is the case in both French and Vietnamese, Mandarin Chinese has only one morpheme gēn, which corresponds either to conjunctives et/và and et/avec/với, on the one hand, or to prepositional avec/với, on the other hand. Like Vietnamese, Mandarin evidences an asymmetry between the pre- and post-

21 See Schapira (2002).

On conjunction and comitativity in Vietnamese

261

verbal positions.22 In postverbal position gen can only function as a conjunction, whereas in preverbal position it can be analyzed either as a (±distributive) conjunction, or as a comitative conjunction, or as a preposition, depending on the predicates gēn co-occurs with. The standard tests of negation, adverbial modification, clefting, relative clause formation etc. allow one to easily tease apart conjunctive gen and prepositional gen. What is interesting is that with symmetrical predicates, gēn behaves syntactically in a dual way. It obeys both the tests of conjunctions and of prepositions, under negation and question formation, for example. We will not present the facts here, but refer the reader to Paris (2008, 2010).

10 Conclusion In this paper we have tried to go beyond the simple translations or the descriptions of the two morphemes và and với found in Vietnamese grammar books. Basing ourselves both on syntactic tests and on results obtained in general linguistics, we have tried to show how the notion of comitativity/collectivity can help understand the behavior of với. Both conjunctive với and prepositional với can carry the feature of comitativity, but the syntactic and semantic characteristics of conjunctive với and prepositional với are different. Conjunctive với does not co-occur with cùng and is found in symmetrical/collective predications. Prepositional với co-occurs with cùng and is found in non-collective predications. Finally, we have tried to show that nhau does not always indicate reciprocity (or pluractionality). Nhau is also found in comparisons of equality, cf. (64). In such cases it simply indicates that the same property is attributed equally (but not conversely) to two participants, as evidenced by the co-occurrence of nhau and đều ‘equally’. (64)

Anh ấy và anh đều chạy nhanh như nhau. Sg3 and Sg2 equally run fast each other ‘He runs as fast as you.’

22 But the two asymmetries are not parallel. Vietnamese allows cùng với comitative PPs both pre-and post-verbally, as in (29)–(30) above. Mandarin allows only conjunctive gēn NPs ‘and’ both in pre- and post-verbal positions. A comitative PP cannot appear postverbally in Mandarin. For a study of gēn in Mandarin and a comparison between gēn and French et/avec, see Paris (2008) and (2010), respectively.

262

Marie-Claude Paris and Lê Thị Xuyến

References Alves, Mark. 2005. Sino-Vietnamese grammatical vocabulary and triggers for grammaticalization. The 6th Pan-Asiatic International Symposium on Linguistics: 315–332. Hanoi: Social Sciences Publishing House. Arkhipov, Alexandre. 2009. Comitative as a cross-linguistically valid category. In: Patience Epps and Alexandre Arkhipov (eds.), New Challenges in Typology. Transcending the Borders and Refining the Distinction, 223–246. Berlin: Mouton de Gruyter. Borillo, Andrée. 1971. Remarques sur les verbes symétriques du français. Langue française 11: 17–31. Choi-Jonin, Injoo. 2002. Comitatif et jonctif en français et en coréen. Cahiers de grammaire 27: 11–28. Evans, Nick. 2008. Reciprocal constructions: toward a structural typology”. In: König, E. and Gast, V. (eds.) Reciprocity and Reflexivity: Cross-linguistic explorations. Berlin: Mouton de Gruyter. Frajzyngier, Zygmunt and Curl, Tracy S. (eds.). 2000. Reciprocals: Form and Function. Amsterdam: Benjamins. Haspelmath, Martin. 2004a. Coordinating constructions: an overview. In: Martin Haspelmath (ed.), Coordinating constructions, 3–39. (Typological studies in language 58.) Amsterdam: Benjamins. Haspelmath, Martin. 2004b. Coordinating constructions. (Typological Studies in Language 58.) Amsterdam: Benjamins. Hetzron, Robert. 1973. Conjoining and comitativization in Hungarian. A study of rule ordering. Foundations of language 10: 493–507. Lichtenberk, Frantisek. 1999. Reciprocals without reflexives. In: Zygmunt Frajzyngier and Traci S. Curl (eds.), Reciprocals: Forms and Function, 31–62. Amsterdam: Benjamins. Lakoff, George and Stanley Peters. 1967. Phrasal conjunction and symmetric predicates. In: David A. Reibel and Sanford A. Schane, (eds.), Modern Studies in English, 113–142. New Jersey: Prentice-Hall. Lehmann, Christian and Yong-Min Shin. 2005. The functional domain of concomitance: A typological study of instrumental and comitative relations. In: Christian Lehmann (ed.), Typological Studies in Participation, 9–104. (Studia Typologica 7.) Berlin: Akademie Verlag. Mchombo, Sam A. 1999. Quantification and verb Morphology: The case of reciprocals in African languages. Linguistic Analysis 29: 182–221. Mcnally, Louise. 1993. Comitative coordination: a case study in group formation. Natural Language and Linguistic Theory 11: 347–379. Nguyễn, Anh Quế. 1988. Les mots vides du vietnamien moderne. [Hư từ trong tiếng Việt hiện đại]. Hà Nôi: Ed. Sciences Sociales. Nguyễn, Đình-Hoà. 1997. Vietnamese. Amsterdam: Benjamins. Paris, Marie-Claude. 2008. On parts of speech in Chinese: gēn. The Linguistic Review 25(3–4): 347–366. Paris, Marie-Claude. 2010. Mandarin gen and French et/avec. Another look at distributivity and collectivity. In: Dingfang Shu and Ken Turner (eds.), Contrasting Meaning in Languages of the East and West, 517–530. (Contemporary Studies in Descriptive Linguistics 14.) Berne: Peter Lang.

On conjunction and comitativity in Vietnamese

263

Progovac, Ljiljana. 2003. Structure for coordination. In: Lisa Cheng and Rint Sybesma (eds.), The second GLOT International state-of-the article book. The latest in linguistics, 241–287. Berlin: Mouton de Gruyter. Seiler, Hansjakob. 1974. The principle of concomitance: instrumental, comitative and collective. Foundations of Language 12: 215–247. Skrabalova, Hana. 2003. Comitative constructions in Czech. In: Peter Kosta and Jens Frasek (eds.), Current Approaches to Formal Investigations into Slavic languages, 685–696. (Linguistik International. Band 9.) Frankfurt: P. Lang. Ross, John. 1968. Constraints on variables in syntax. Ph.D. thesis, Department of Linguistics, MIT. Schapira, Charlotte. 2002. Préposition et conjonction? Le cas de avec. Travaux de linguistique 44: 89–100. Stassen, Léon. 2000. And-languages and with-languages. Linguistic Typology 4(1): 1–54. Stolz, Thomas, Cornelia Stroh and Aina Urdze. 2006. On Comitatives and related Categories. A Typological Study with Special Focus on the Languages of Europe. Berlin: Mouton de Gruyter. Teng, Shou-hsin. 1970. Comitative versus phrasal conjunction. Papers in linguistics 2(2): 315– 358. Teng, Stacy Fang-Ching (guest ed.). 2011. Coordination and comitativity in Austronesian languages. Language and Linguistics 12(1): (i)–(iv). Trương, Văn Chình. 1970. Structure de la langue vietnamienne. Paris: Paul Geuthner. Zhang, Niina. 2006. On the configuration issue of coordination. Language and Linguistics 7(1): 175–223. Zhang, Niina. 2007. The syntax of the English comitative constructions. Folia Linguistica 41: 135–169.

Daniel Hole

10 Focus particles and related entities in Vietnamese* The chapter analyzes the system of focus-sensitive particles in Vietnamese. EVEN/ ALSO/ONLY foci are discussed across syntactic categories, and Vietnamese is found to organize its system of focus-sensitive particles along three dimensions of classification: (i) EVEN vs. ALSO vs. ONLY; (ii) particles c-commanding foci vs. particles c-commanding backgrounds; (iii) adverbial focus-sensitive particles vs. particles c-commanding argument foci only. Towards the end of the chapter, free-choice constructions and additional sentence-final particles conveying ONLY and ALSO semantics are briefly discussed. The peculiar Vietnamese system reflects core properties of the analogous empirical domain in Chinese, a known source of borrowings into Vietnamese over the millennia. Keywords: focus particles, background particles, free-choice, Chinese, association-with-focus, structured propositions

1 Introduction: Focus-sensitive particles This chapter discusses strategies of expressing EVEN foci, ALSO foci and ONLY foci in Vietnamese (frequently referred to as AEO foci in the following). The article combines descriptive and analytical parts to get a grip on the empirical domain, which has, to the best of my knowledge, never been investigated in any detail before. The data presented in this article, if not indicated otherwise, comes from elicitation work with native speakers.1

* This article was partly written in the context of project A5 of SFB 632 “Information structure – the linguistic means for structuring utterances, sentences and texts” funded by the Deutsche Forschungsgemeinschaft. It is a slightly revised version of Hole (2008). I would like to thank Mark Alves, Andreas Dufter, Volker Gast, Shinichiro Ishihara, Hạ Kiêu Phương, Elisabeth Löbel, Svetlana Petrova, Nguyễn Thu Trang, Laurent Sagart, Tran Thuan and, particularly, Malte Zimmermann and Stavros Skopeteas for comments and discussion. 1 I worked with two consultants: (i) HẠ Kiêu Phương, female, 28 years old, from Hanoi/ Vietnam, a student in Germany since she was 18; (ii) NGUYỄN Thu Trang, female, 24 years old, from Hanoi/Vietnam, a student in Germany since she was 20; Trang moved from Vietnam to the Czech Republic with her parents when she was ten years old.

266

Daniel Hole

The following semantic background assumptions concerning AEO foci are made.2 ALSO foci presuppose the truth of an alternative proposition that differs in the position of the focus. For an English sentence like Peter ate also the beans this means that this sentence is felicitously uttered only if a proposition of type ‘Peter ate x’, with x ≠ the beans, was part of the common ground before it was uttered (‘Peter ate the onions’, for instance). ONLY foci entail the falsity of all (contextually relevant) alternative proposition that differ in the position of the focus. For an English sentence like Peter ate only the beef this means that this sentence is true if and only if Peter ate nothing from the set of contextually salient alternatives to the beef. A different way of stating the same entailment would be to say that all the things that Peter ate (from the set of contextually salient alternatives) were identical to the beef. EVEN foci typically presuppose the truth of alternative propositions that have alternative values in the position of the focus. If one says “Even the firstyear students solved this problem”, then this typically means that some more senior students likewise solved the problem. There is a complication here in that it needn’t necessarily be the case that other students did solve the problem if the sentence is to be uttered felicitously. This may, e.g., be the case in a context where lazy third-year students are contrasted with hard-working first-year students. It is sufficient if one just expects the more experienced students to be able to solve the problem to make the use of even felicitous in our example. This means that the existential quantification hypothesized to underlie the semantics of even (‘the same holds true of some alternative’) only holds with respect to possible states of affairs, but not necessarily with respect to a given state of affairs. For this reason the generalization concerning alternatives with EVEN foci was hedged when we first introduced it above (“EVEN foci typically presuppose the truth. . .”). A second component of meaning tied to EVEN foci has to do with scalarity. EVEN foci have to mark the endpoint on a scale to be felicitous. It is typically assumed that the ordering underlying EVEN scales is expectedness or likelihood. If even the first-year students solved the problem, then these students were, among the relevant members of the comparison class, least likely or least expected to solve the problem.

2 I assume familiarity with basic notions of information structure. Cf. König (1991), Krifka (2007), or, for the more formally inclined, Rooth (1996) for overviews of the empirical domain from a theoretical perspective.

Focus particles and related entities in Vietnamese

267

The literature on AEO foci is voluminous, but for the purpose of the survey in the present chapter, the informal characterizations of meaning just presented will be sufficient.3 To the best of my knowledge, no studies with a comparable empirical scope have been written to date. For this reason, the present article strives to carve out the major descriptive generalizations organizing the field of AEO foci in Vietnamese. Special problems tied to individual focus types or particles are noted throughout the chapter, but are, for the most part, left for future treatment. As will become clear shortly, Vietnamese has a very rich system of AEOparticles. Most notably, a set of argument focus markers is opposed to a set of non-argument, or adverbial, focus markers. A second distinction can be drawn between particles interacting with foci on the one hand, and particles interacting with backgrounds on the other. A third distinction that will only concern us towards the end of the article has to do with sentence-final particles. In contradistinction to the particles that are discussed in the bulk of the chapter, viz. particles preceding foci or backgrounds, the particles discussed later come last in a sentence. The chapter introduces the association-with-focus pattern of expressing AEO foci in section 2. Section 3 familiarizes the reader with the partition pattern of focus-background marking of Vietnamese. Ideally, the focus and the background are syntactically opposed to each other in this pattern, and both the focus and the background are morphologically marked as such. Section 3 likewise contrast focus-background partition structures with clefts. Section 4 reviews the expression of AEO foci across syntactic categories in Vietnamese; foci on direct objects, indirect objects, subjects, adjuncts and verbs are treated separately, and foci on subjects with intransitive verbs receive a discussion of their own. There is a Vietnamese free-choice construction involving indefinite pronominals in which background markers are used and which makes regular use of the partition pattern; this construction is discussed in section 5. Section 6 reviews the generalizations arrived at from a more general perspective. Section 7, finally, summarizes the main findings and puts the Vietnamese system in context before the background of the surprisingly similar system of focus-background marking in Mandarin Chinese. Language contact is identified as the

3 Classical references for ONLY include Horn (1969), Jacobs (1983) and von Fintel (1994). For a survey of the research on ONLY, cf. Horn (1996). See Krifka (1998) for an important take on ALSO. Influential contributions to the semantics of EVEN include Karttunen and Peters (1979), Kay (1990) and Krifka (1995). König (1991) gives a valuable overview of the entire empirical domain.

268

Daniel Hole

likely source of the similarity between Mandarin and Vietnamese, but the exact conditions of the language contact operative here must be left open.

2 The association-with-focus pattern (AwF) Vietnamese has adverbial focus-sensitive particles to express AEO readings. These particles often occur in a sentence-medial position behind the subject and before the predicate as in (1). ‘Predicate’ is here taken to refer to a verbal projection comprising at least the VP and (non-epistemic) modal verbs, if there are any. Sentences where the particles are used in other positions, especially in sentence-initial position, will be discussed in subsequent sections. I call the resulting pattern of expressing AEO foci ‘association-with-focus’, or ‘AwF’, for short.4 (1) and (2) provide one example each for chỉ ‘only’ and thậm chí ‘even’. (There is a syntactic complication with the adverbial ALSO particle, which we will turn to after the discussion of (1) and (2).)5,6 (1) Hôm qua Nam chỉ [ăn thịt bò] thôi. yesterday Nam only eat meat beef PRT ‘Nam only [ate beef ] yesterday.’ (2)

Hôm qua Nam thâm ̣ chí [ăn thịt bò]. yesterday Nam even eat meat beef ‘Nam even [ate beef ] yesterday.’

Much like their English translations, (1) and (2) are compatible with foci comprising any subconstituent, or the whole, of the bracketed constituents. (1), for instance, has at least the three potential interpretations (i) ‘The only thing that Nam did yesterday was to eat beef’ (VP focus), (ii) ‘The only thing that Nam ate yesterday was beef’ (object focus), and (iii) ‘The only thing that Nam did with 4 The term ‘association-with-focus’ goes back to Rooth (1985). We will return to the theoretical significance of this terminological choice in the concluding section 6. 5 The following abbreviations are used in examples: ANT – anterior tense; ASP – aspect marker; CL – classifier/determiner; CONT. CONJ – contrastive conjunction; COP – copula; FC – free-choice particle; PL – plural; POST – posterior tense; PRT – particle; PRT FOC – particle preceding foci; PRT BG – particle preceding backgrounds; Q – sentence-final question particle. 6 We will discuss thôi in section 7. Thôi is a sentence-final ONLY marker which frequently co-occurs with other ONLY words. Since it is the ONLY word of Vietnamese that I know least about it is not discussed before the concluding section of the chapter.

Focus particles and related entities in Vietnamese

269

the beef yesterday was to eat it’ (verb focus). The same holds, ceteris paribus, for (2). Prosody partially disambiguates these different readings. Specifically, a focus accent on the verb will, under most circumstances, enforce a narrow verb focus, whereas a focus accent on the object is compatible with a wider array of readings.7,8 The difficulty arising with adverbial cả ‘also’ alluded to above is that this particle follows the verb instead of preceding it, as was the case with chỉ ‘only’ and thậm chí ‘even’. This is shown in (3). (3)

[Bác nông dân nuôi lợn.] (Bác ấy) trồng cả cà chua. the farmer raise pig he grow also tomatoes ‘The farmer raises pigs. He also grows tomatoes.’

The context provided for (3) makes it clear that the entire VP trồng cà chua ‘grow tomatoes’, as opposed to nuôi lợn ‘raise pigs’, is in focus. The focus particle separates the two parts of the focus. This is incompatible with the idea that adverbial focus-sensitive particles should c-command their foci (König 1991; Büring and Hartmann 2001). It is possible, however, to state a generalization with reference to the left edge of the VP if one says that cả ‘also’ as an adverbial particle must follow the first word of the VP, i.e. the main verb. As Thompson (1987: 271) puts it for the class of function words under which he subsumes cả: “Postpositive particles are movable particles occurring as complement after their immediate constituent partners.” Even though this wording doesn’t take into account the fact that the object together with the verb constitutes the relevant interacting category in this construction, the quote makes it clear that cả belongs to a distributional class whose members follow items with which they interact. In movement terms one could say that cả is in a syntactic position comparable to that of chỉ ‘only’ and thậm chí ‘even’ as in (1) and (2), except that for some idiosyncratic reason tied to cả the verb must move to a position immediately preceding the particle.9 There may well be a phonological motivation 7 Cf. Schwarzschild (1999) or Büring (2006) for the conditions under which focus accents on verbs are compatible with wide foci. 8 Cf. Đỗ Thế Dũng et al. (1998) or Jannedy (2007) for studies on intonation in Vietnamese. According to Jannedy (2007), who bases her conclusions on experimental work, focus accents in Vietnamese can probably be described in terms familiar from intonation languages like English (among them segment duration, f0 excursions and amplitude). 9 Note that the V2-requirement of German, which is underlyingly OV, leads to similar patterns in main clauses. This is shown in (ia) with the derived main clause position of the inflected verb as opposed to the more basic linearization in subordinate clauses as in (ib). (Largest possible foci are marked by bracketing.)

270

Daniel Hole

for such a movement if cả is an enclitic.10 At the moment I lack evidence to settle the issue, but this would certainly be a research question worth pursuing. If the analysis is correct that cả may follow (parts of ) its associating focus, we predict that, in the extreme case, cả should be possible with a narrow focus on the preceding verb. This pattern is indeed attested, as is witnessed by (4). (4) Bác nông dân không chỉ ăn cà chua mà trồngF cả cà chua. the farmer not only eat tomato but grow also tomato ‘The farmer doesn’t just eat tomatoes, he also growsF tomatoes.’ The assumption of preposed verbs with cả receives further support from a similar pattern arising with a certain use of the modal element được ‘can’. In this pattern, too, the canonical order between main verb and functional element is reversed (Duffield 2001; Cheng and Sybesma 2004 discuss parallel facts for Cantonese dak). The SVO character of Vietnamese would generally seem to predict the order MODAL – MAIN VERB as attested in (5). But with the modal verb được as in (6) the reverse order MAIN VERB – MODAL occurs. (5)

Nam có thể ăn thịt bò. Nam can eat meat beef ‘Nam can eat beef.’

(6)

Nam ăn được thịt bò. Nam eat can meat beef ‘Nam can eat beef.’ (he’s not allergic to it, or otherwise adversely affected by it)

This constitutes a parallel with the adverbial cả case in (3) where the main verb precedes the adverbial focus-sensitive particle. I conclude that there is some

(i) a. Der Bauer [züchtet auch Tomaten]. the farmer grows also tomatoes ‘The farmer also [grows tomatoes].’ b. . . . dass der Bauer auch [Tomaten züchtet]. that the farmer also tomatoes grows ‘. . . that the farmer also [grows tomatoes].’ 10 Thanks to Stavros Skopeteas for pointing this possibility out to me.

Focus particles and related entities in Vietnamese

271

support for the idea that the unexpected order of main verb and particle in (3) is derived and ultimately irrelevant to interpretation.11 The sentences in (5′) and (6′), which combine the structures of (5) and (6) with an adverbial focus-sensitive particle, provide evidence for another pertinent generalization: The predicative constituent to the right of an adverbial focus-sensitive particle need not be a bare VP, but may include modal morphemes as well. (5′)

Nam chỉ có thể ăn thịt bò. Nam only can eat meat beef ‘Nam can only eat beef.’

(6′) Nam chỉ ăn được thịt bò. Nam only eat can meat beef ‘Nam can only eat beef.’ (he’s allergic to other things, or otherwise adversely affected by other things) Besides modal elements, which always follow adverbial focus sensitive particles, the temporal particles đã ‘ANTERIOR TENSE ’ and sẽ ‘POSTERIOR TENSE ’ occur adjacent to adverbial focus-sensitive particles. Thậm chí ‘even’ precedes the temporal particles, whereas chỉ ‘only’ follows them. This is shown in (7). (7)

a. Nam (thậm chí) đã/sẽ (*thậm chí) ăn pho mát. Nam even ANT / POST even eat cheese ‘Nam even ate cheese.’/‘Nam will even eat cheese.’ b. Nam (*chỉ ) đã/sẽ (chỉ ) ăn pho mát. Nam only ANT / POST only eat cheese ‘Nam only ate cheese.’/‘Nam will only eat cheese.’

The position of thậm chí to the left of chỉ’s position fits in well with an observation that can be made in languages like English: if EVEN and ONLY occur in a single clause and their foci are nested, EVEN must take scope over ONLY (cf. Paul even bought only flowers vs. *Paul only bought even flowers). Moreover, it is known that EVEN foci generally take wide scope (Krifka 1995). If we generalize over the different cases surveyed in (6) through (7), we arrive at the schematic structure in (8). In terms of the sequence of TAM markers, 11 In generative terms this amounts to saying that the verb reconstructs at LF and adverbial cả ‘also’ c-commands all parts of its focus at this level of representation.

272

Daniel Hole

it matches with the analogous tree-geometric architecture of functional verbal categories known, e.g., from Beck and von Stechow (2006). (8) EVEN + tense + ONLY + modal (+ASP) + VP Depending on one’s theoretical choices, one may thus want to say that adverbial focus-sensitive particles are not, or need not be, sisters of VPs. Instead they may be said to adjoin to ModPs or TPs, i.e. to modality-marked or tense-marked constituents larger than VP. Alternatively, one could speak of the left edge of the extended VP domain as the structural position of thậm chí, cả and chỉ. Summarizing the discussion in this section, and evading the theoretical issue just mentioned, we can state the generalizations in (9). (9)

Adverbial focus-sensitive particles in Vietnamese (i) Adverbial focus-sensitive particles in Vietnamese associate with a constituent in the extended VP-projection of a sentence; (ii) the adverbial focus-sensitive particle for EVEN foci is thậm chí; (iii) the adverbial focus-sensitive particle for ALSO foci is cả; (iv) the adverbial focus-sensitive particle for ONLY foci is chỉ.

3 The partition pattern 3.1 Prototypical instantiations of the partition pattern The prototypical partition pattern used to express AEO foci syntactically opposes an argument focus part and a background part. Either part may contain a particle yielding AEO focus readings. The structure in (10) depicts this state of affairs. (10)

(prototypical case) [[PRT FOC Focus argument] [PRT BG Background]]

THE PARTITION PATTERN

The particles preceding the focus in the partition pattern (PRT FOC in (10)) are different from the adverbial focus-sensitive particles discussed in section 2, and the background particles (PRT BG in (10)) constitute yet another distinct paradigm. In the clearest cases, as exemplified in (11), the focus precedes the background, and each part begins with the respective particle. Here and in the following, postverbal material which is to be construed as given is parenthesized. While its non-realization is the norm in actual discourse, its rendering in the examples is hoped to facilitate the accommodation of appropriate contexts.

Focus particles and related entities in Vietnamese

(11) a.

[[Đến PRT FOC even

273

NamF] [cũng [ăn (thịt bò)]BG]]. Nam PRT BG even/also eat meat beef

‘Even NamF ate beef.’ [ăn (thịt bò)]BG]]. NamF] [cũng Nam PRT BG even/also eat meat beef ‘NamF, too, ate beef.’

b.

[[Cả

PRT FOC also

c.

[[Mỗi PRT FOC only

NamF] [mới [ăn (thịt bò)]BG]]. Nam PRT BG only eat meat beef

‘Only NamF eats beef.’ In (11a), the EVEN focus is preceded by đến, and the EVEN background by cũng. In (11b), the ALSO focus is preceded by cả, while the background begins with the same particle cũng that was used in (11a). Note that cả in (11b) is analyzed as an instance of PRT FOC (i.e. as a particle which precedes arguments in focus), and not as an adverbial focus particle. The latter categorization was assumed for the homophonous form in section 2. I assume that the non-canonical adverbial syntax discussed there allows us to make this distinction. As a focus particle in the partition pattern, cả behaves just as the other particles of its paradigm. As an adverbial focus-sensitive expression, cả features the special verb-preposing behavior discussed above. (11c) makes use of the ONLY-particle mỗi preceding the subject focus, while the background begins with mới (the orthographic similarity between the two particles is misleading; we are dealing with two distinct words). The background particle mới is distinct from the background particle in the EVEN/ALSO cases in (11a/b). It was stated above that the cases in (11) constitute prototypical cases with clear partitionings into focus and background. We will now turn to patterns where the partition turns out less neatly.

3.2 Subjects/Topics preceding background markers One factor obscuring the picture is that, with non-subject foci, the background particle must follow the subject if there is one, even if the subject forms part of the background. This is illustrated in (12). (12) Đến PRT FOC even

[ pho mát]F [Nam cũng thích]BG . cheese Nam PRT BG even/also like

‘Nam likes even cheeseF.’

274

Daniel Hole

I take this less clear-cut surface pattern of focus-background partition to reflect another information-structural partition, viz. that into topic and comment. While the fact that Nam likes, or doesn’t like, certain things is under discussion and is, therefore, background, the discourse address under which this information is stored is Nam. In other words, Nam is the topic of (12) (this amounts to Reinhart’s 1982 notion of ‘aboutness’ topics). There is a further complication here in that the rule requiring Nam to precede the background marker cũng is sensitive to subjects, and not to topics. It is, however, well known that the subject function is frequently the grammaticalized counterpart of the discourse function of topics. I therefore conclude that sentences like (12) don’t just instantiate the focus background partition at the surface, but also the partition into subject/ topic and predicate/comment.

3.3 Mixed structures and optional use of markers Two more factors tend to render partition structures less transparent. Often either PRT FOC or PRT BG may be dropped, or adverbial particles may be used together with PRT FOC or PRT BG . (13)–(15) present relevant examples. (13) a.

[[(Đến) PRT FOCeven

NamF] [*(cũng) [ăn (thịt bò)]BG]]. (cf. (11a)) Nam PRT BGeven/also eat meat beef

‘Even NamF ate beef.’/‘NamF, too, ate beef.’ b.

[ăn (thịt bò)]BG]]. [[(Thậm chí) (đến) NamF] [*(cũng) even PRT FOCeven Nam PRT BGeven/also eat meat beef ‘EvenF Nam ate beef.’/‘NamF, too, ate beef.’

(14)

[[(Cả) PRT FOCalso

[ăn (thịt bò)]BG]]. (cf. (11b)) NamF] [*(cũng) Nam PRT BGeven/also eat meat beef

‘NamF, too, ate beef.’ (15)

[ăn (thịt bò)]BG]]. (cf. (11c)) a. [Chỉ [(mỗi) NamF] [(mới) only PRT FOConly Nam PRT BGonly eat meat beef ‘Only NamF eats beef.’ [ăn (thịt bò)]BG]]. b. [(Chỉ ) [(mỗi) NamF] [mới only PRT FOConly Nam PRT BGonly eat meat beef ‘Only NamF eats beef.’ [ăn (thịt bò)]BG]]. c. [(Chỉ ) [mỗi NamF] [(mới) only PRT FOConly Nam PRT BGonly eat meat beef ‘Only NamF eats beef.’

Focus particles and related entities in Vietnamese

275

The options in (13) illustrate the fact that either PRT FOC đến or the adverbial marker thậm chí, or both, may be dropped without necessarily changing the interpretation. My consultants share the intuition, however, that the variants with thậm chí are less colloquial than those without. In contradistinction to the uses of adverbial thậm chí seen so far in (2) and (7), thậm chí precedes the subject in (13b). In analogy to the EVEN cases in (13), PRT FOCalso cả in (14) may be dropped without influencing the interpretation. Note, though, that with cả dropped (14) is string-identical to (13a) with đến dropped. Nevertheless a distinction can probably be drawn between (13) and (14) with the relevant particles left out. This is because the EVEN reading of (13) is felt to go along with a stronger focus accent on Nam and a more emphatic sentence intonation irrespective of whether đến is present or not. Put differently, it is not just the particles đến and cả that, if present, allow one to distinguish between (13) and (14), but also the more emphatic prosody of (13) if compared with (14). In contradistinction to the focus particles cả and đến, and the background particle mới, the background particle cũng may not be left out if a focus interpretation of the ALSO or EVEN kind is aimed at. The ONLY-cases in (15) are different from the standard ALSO-case in (14) for at least three reasons. First, while all variants in (15) are grammatical, those that employ adverbial chỉ, with or without other overt markers, seem to be most natural and colloquial. In the case of the EVEN foci in (13), by contrast, the versions with adverbial thậm chí were identified as less colloquial above. Second, with ONLY foci in the partition pattern it is possible to leave out any one of the particles of the maximal structure. In the cases of ALSO foci and EVEN foci as in (13) and (14), PRT BGeven/also cũng is used no matter whether cả, or đến, precede its position or not. At present, I cannot account for these differences between ONLY-marking and ALSO/EVEN-marking, but from a general perspective the different patterns are in line with observations made for other languages and in the theoretical literature. Too, also, even and only in English each have their peculiarities in English, and the same may be said about translational equivalents in other languages. From a theoretical perspective, such differences are to be expected for the contrast between additive focus semantics as with ALSO and EVEN as opposed to restrictive focus semantics as with ONLY. It was pointed out in section 1 that ONLY sentences entail the exclusion of alternatives, while ALSO and EVEN presuppose the inclusion of alternatives. Moreover, the necessarily emphatic nature of utterances with EVEN foci (Krifka 1995) sets these foci apart from ONLY foci and ALSO foci. What must remain a task for the future is to

276

Daniel Hole

match the observed distributional peculiarities of each Vietnamese particle with the general properties of each focus type.

3.4 Partition structures with in-situ foci A further confounding factor in the domain of the partition pattern is that the foci marked by PRT FOC need not be syntactically opposed to the background, but may also be embedded within the background. This pattern occurs with VPinternal material as illustrated in (16). (16) a.

[Lam cũng cho Nam cả tiềnF]BG . Lam PRT BGeven/also give Nam PRT FOCalso money ‘Lam gave Nam also moneyF.’

b.

[Nam chỉ đọc mỗi sáchF thôi]BG. Nam only read PRT FOConly book PRT ‘Nam read only [books/a book]F.’

In (16a) the object tiền ‘money’ is preceded by PRT FOCalso cả, but the whole expression is embedded within the background predicate which is marked as such by PRT BGeven/also cũng. We will see more examples of such structures in sections 4.1.1 and 4.1.2. A further peculiar fact about (16) concerns the ONLYparticle mỗi in (16b). We classify it as belonging to the partition pattern, but it is not embedded in a predicate background-marked by mới. Instead, the adverbial focus-sensitive ONLY-particle chỉ is used. The generalization seems to be that background-marking mới may precede only background material. Abstracting away from the complications just stated, we find the preliminary topological system of focus-background partition summarized in (17). (17)

AEO FOCI (to be revised) a. The general pattern [PRT FOC FOCUS] [PRT BG BACKGROUND]

TOPOLOGY OF THE PARTITION PATTERN FOR

b. Instantiations EVEN: đến ALSO: cả ONLY: mỗi

cũng FOCUS cũng mới

BACKGROUND

Focus particles and related entities in Vietnamese

277

We will refine our generalizations for the partition pattern in section 4.3 below. At that point it will be shown that the partition pattern interacts with the adverbial particles in a yet more general way than was discussed in connection with examples (13) through (15). Having introduced the two basic patterns of focus construal (AwF vs. partition), we will now turn to a discussion of individual syntactic functions that may instantiate AEO foci.

4 AEO foci with different syntactic functions 4.1 Object foci 4.1.1 Direct objects There are two ways to arrive at AEO foci on direct objects. One way is to make use of the AwF pattern, the other one is to apply the partition pattern. We have seen examples of the AwF pattern in (1)–(3) in section 2 already. These examples are repeated in (18) for convenience (with a trivial adaptation in the case of (18c)). In contradistinction to the discussion in section 2, the representations in (18) have been specified so as to restrict the readings to object foci. (18)

DIRECT OBJECT + AwF STRATEGY

a. Hôm qua Nam thâm ̣ chí ăn [thịt bò]F. yesterday Nam even eat meat beef ‘Nam even ate beefF yesterday.’ b. Bác nông dân trồng cả [cà chua]F. the farmer grow also tomatoes ‘The farmer also grows tomatoesF.’ c.

Hôm qua Nam chỉ ăn [thịt bò]F thôi. yesterday Nam only eat meat beef PRT ‘Nam only ate beefF yesterday.’

A second set of sentences exemplifying the same AwF pattern is found in (19). (19) DIRECT OBJECT + AwF STRATEGY a. Nam thâm ̣ chí đã đọc [quyển sách]. Nam even ANT read the book ‘Nam even read [the book]F.’

278

Daniel Hole

b.

Nam ăn cả [thịt gà]F. Nam eat also meat chicken ‘Nam also eats chickenF.’

c.

Nó chỉ ghét tôiF thôi. he only hates me PRT ‘He only hates meF.’

(20) is a first set of examples of the partition pattern for AEO foci on direct objects. In these examples the objects in focus have been preposed. (20)

DIRECT OBJECT + PARTITION STRATEGY + PREPOSED FOCUS

a.

Đến PRT FOCeven

[ pho mát]F Nam cũng thích. cheese Nam PRT BGeven/also like

‘Nam likes even cheeseF.’ b.

Cả PRT FOCalso

[quyển sách] Nam cũng đọc. the book Nam PRT BGalso read

‘Nam read even the bookF.’ c.

ăn thôi. (Chỉ ) mỗi [thịt bò]F Nam mới only PRT FOConly meat beef Nam PRT BGonly eat PRT ‘Only beefF does Nam eat.’

My consultants report a strengthening effect for (20b) such that an EVEN reading is arrived at if the ALSO focus is preposed. This effect was absent with the subject focus in (14), presumably because that example involved no preposing. Cf. the discussion of (15) above for the fact that the ONLY focus in the partition pattern as in (20c) is, in contradistinction to EVEN foci and ALSO foci, additionally preceded by the adverbial particle chỉ. As stated in 3.4 above, the foci in the partition pattern need not precede their backgrounds in each and every case if the focus is constituted by material that originally belongs in the VP. Since direct objects originate inside VP, (20b/c) have the in-situ variants in (20′b/c). EVEN foci on direct objects indicated by đến, by contrast, regularly trigger the clear partition pattern of (20a). The in-situ variant of (20a) in (20′a) is ungrammatical. (20′)

DIRECT OBJECT + PARTITION STRATEGY + IN SITU FOCUS

a. *Nam (cũng) thích đến [ pho mát]F. Nam PRT BGeven/also like PRT FOCeven cheese int.: ‘Nam likes even cheeseF.’

Focus particles and related entities in Vietnamese

279

b. Nam cũng đọc cả [quyển sách]F. Nam PRT BGeven/also read PRT FOCalso the book ‘Nam read also [the book]F.’ c.

Nam chỉ ăn mỗi [thịt bò]F thôi. Nam only eat PRT FOConly meat beef PRT. ‘Nam ate only beefF.’

A second asymmetry concerns the use of background marking cũng alongside cả in (20′b), whereas no background marking particle is used in the ONLY case in (20′c) (recall that chỉ is the adverbial ONLY particle; the background marker would be mới). Concerning the non-use of mới in such configurations it was stated in connection with ex. (16) above that mới may probably c-command backgrounded material only. This would predict that it cannot be used in in-situ partition structures like (20′c).

4.1.2 Indirect objects The picture that emerges for indirect objects with AEO focus interpretations is parallel to the one found with direct objects. As in the case of direct objects above, I will present paradigms for the AwF pattern and for the partition pattern. In the case of the AwF pattern, the foci are again restricted to the indirect object constituent despite the fact that identical strings are also compatible with verb foci, or VP foci. The verb figuring in examples (21)–(23) is cho ‘give’. Just as in the English construction give s.o. s.th, the goal argument precedes the theme argument. (21) INDIRECT OBJECT + AwF STRATEGY a. Nam thâm ̣ chí/chỉ cho [học sinh]F tiền. Nam even/only give student money ‘Nam only/even gives studentsF money.’ b. Nam cho cả [học sinh]F tiền. Nam give also student money ‘Nam also gives studentsF money.’ (22) provides the paradigm for preposed indirect objects in the partition pattern, and (23) assembles the in-situ variants. Preposing of the ALSO focus in (22b) triggers the same strengthening effect observed with the direct object in (20b) above.

280 (22)

Daniel Hole

INDIRECT OBJECT + PARTITION STRATEGY + PREPOSED FOCUS

a. Đến PRT FOCeven

[học sinh]F Nam cũng cho (tiền). student Nam PRT BGeven/also give money

‘Even to [the student(s)]F, Nam gives money.’ b. Cả PRT FOCalso

cho (tiền). [học sinh]F Nam cũng student Nam PRT BGeven/also give money

‘Even to [the student(s)]F, Nam gives money.’ cho (tiền). c. Chỉ mỗi [học sinh]F Nam mới only PRT FOConly student Nam PRT BGonly give money ‘Only to [the student(s)]F does Nam give money.’ (23)

INDIRECT OBJECT + PARTITION STRATEGY + IN SITU FOCUS

a. *Nam cũng cho đến [học sinh]F (tiền). Nam PRT BGeven/also give PRT FOCeven student money int.: ‘Nam gives even [students]F money.’ b. Nam cũng cho cả [học sinh]F (tiền). Nam PRT BGeven/also give PRT FOCalso student money ‘Nam gives also [students]F money.’ c. Nam chỉ cho mỗi [học sinh]F (tiền) thôi. Nam only give PRT FOConly student money PRT ‘Nam gives only [students]F money.’ As is the case in English and many other languages, Vietnamese has a second argument frame for ditransitive predications. Instead of strings of type V IO DO, we also find strings of type V DO P IO as in English give the present to Bertha. I call this the prepositional IO pattern. The Vietnamese prepositional IO pattern is đưa DO cho IO. The preposition used (cho) is identical in form to the verb cho of the V IO DO pattern.12 In (24a), an example with focus on a prepositional object is given for the AwF pattern.

12 This kind of polysemy between verbs of giving and directional prepositions occurs in many languages that employ verb serialization (with this term taken in a broad sense here), and it is the norm in the language area where Vietnamese is spoken (cf. Bisang 1992).

Focus particles and related entities in Vietnamese

281

(24) PREPOSITIONAL OBJECT + AwF STRATEGY a. Nam thâm cho [học sinh]F. ̣ chí/chỉ đưa tiền Nam even/only give money to student ‘Nam even/only gives money to studentsF.’ b. *Nam đưa cả tiền cho [học sinh]F. Nam give also money to student ‘Nam also gives money to studentsF.’ It is not clear to me why the structure with postverbal adverbial cả cannot be used if narrow focus on the indirect/prepositional object is intended. While I conjecture that this has something to do with the non-canonical syntax of adverbial cả, I’m unable to state the exact reason for the unavailability of (24b) with the intended reading. The partition pattern with preposed foci in the prepositional IO pattern produces degraded structures with preposition-marked IOs, or at least these structures have more specific requirements than the preposing partition patterns with the V IO DO pattern. (25) bears witness of this. (25)

PREPOSITIONAL OBJECT + PARTITION STRATEGY + PREPOSED FOCUS

a.

*?Đến

PRT FOCeven

học sinh giàuF Nam cũng đưa tiền cho. student rich Nam PRT BGeven/also give money to

‘Nam gave money even to the rich F students.’ b.

mỗi học sinh nghèoF Nam mới đưa tiền cho thôi. only PRT FOConly student poor Nam PRT BGonly give money to PRT

?Chỉ

‘Only to the poor F students did Nam give money.’ What appears to contribute to the difficulties in the preposing structure in the prepositional IO pattern of (25a) is the fact that the preposition is stranded. Moreover, and possibly unrelatedly, one of my consultants provided the additional adjective nghèo ‘poor’, which will typically yield a narrow focus on this adjective within the larger pied-piped DP học sinh nghèo ‘the poor student(s)’. At the moment, I lack further information concerning the exact reasons for the degraded status of (25a), and why (25b) is rated a lot better by my consultants. Due to the strengthening generally observed with preposed ALSO foci, (26) with cả instead of đến, if it is good, means the same as (25a) (cf. the discussion of (20b) and (22b) above).

282 (26)

Daniel Hole

??Cả

PRT FOCalso

học sinh giàuF Nam cũng đưa tiền cho. student rich Nam PRT BGeven/also give money to

‘Nam gave money even to the richF students.’

4.2 Subject foci with transitive verbs (Intended) AEO subject foci with transitive verbs in the AwF pattern are presented in (27). (27)

SUBJECT + TRANSITIVE VERB + AwF STRATEGY

a. Thậm chí NamF ăn cả (thịt bò). even Nam eat also meat beef ‘Even NamF eats beef.’ b. *NamF ăn cả (thịt bò). Nam eat also meat beef int.: ‘NamF eats beef, too.’ [b′. Cả PRT FOCalso

ăn (thịt bò). NamF *(cũng) Nam PRT BGeven/also eat meat beef

‘NamF, too, eats beef.’] c.

Chỉ NamF ăn (thịt bò). only Nam eat meat beef ‘Only NamF eats beef.’

(27b) shows that ALSO foci on subjects cannot be signaled by the adverbial ALSO particle cả with its peculiar verb-preposing property (cf. (3)/(4) in section 2). Given the use of cả in the initial position of the bracketed (27b′), one may be tempted to analyze this example as a case where the adverbial particle cả – which is homophonous with the ad-argument particle – embeds the complete sentence just like chỉ in (27c). The impossibility to drop background marking cũng, though, indicates that (27b′) is to be analyzed as an instance of the partition pattern with cả instantiating the ad-argument particle. In contradistinction to the ALSO case, the ONLY focus on the subject with the adverbial particle chỉ in (27c) yields a grammatical structure. The partition patterns for subject foci look as in (28) and (29).

Focus particles and related entities in Vietnamese

(28)

283

SUBJECT + PARTITION STRATEGY + PREPOSED FOCUS

a. Đến

NamF cũng ăn (thịt bò). Nam PRT BGeven/also eat meat beef

PRT FOCeven

‘Even NamF eats beef.’ ăn (thịt bò). NamF cũng Nam PRT BGeven/also eat meat beef

b. Cả PRT FOCalso

‘NamF, too, eats beef.’ ăn (thịt bò). c. (Chỉ ) mỗi NamF mới only PRT FOConly Nam PRT BGonly eat meat beef ‘Only NamF eats beef.’ (29)

SUBJECT + PARTITION STRATEGY + IN SITU FOCUS

a.

Đến PRT FOCeven

NamF *(cũng) ăn (thịt bò). Nam PRT BGeven/also eat meat beef

‘Even NamF eats beef.’ b. Cả PRT FOCalso

ăn (thịt bò). NamF *(cũng) Nam PRT BGeven/also eat meat beef

‘NamF, too, eats beef.’ ăn (thịt bò). c. Chỉ mỗi NamF (mới) only PRT FOConly Nam PRT BGonly eat meat beef ‘Only NamF eats beef.’ As before, the partition patterns for ONLY foci in (28) and (29) are peculiar in that adverbial chỉ is preferably used in sentence-initial position alongside the ad-argument focus marker. It is doubtful whether an in-situ partition pattern for subjects with EVEN foci and ALSO foci really exists, because the grammatical variants in (29a/b) are string-identical to (28a/b). A similar question may be raised in connection with (29c) with the ONLY focus in the in-situ partition pattern, except that, here, the background marker may be dropped.

4.3 Adjunct foci For adjunct foci in the partition pattern, I will provide data of two different structural types: adjunct foci in simple sentences, and foci in adjunct clauses within complex sentences. Before turning to those structures, the AwF pattern for adjunct foci in simple clauses is covered. I have no data illustrating the AwF pattern for complex clauses with foci in adjunct clauses whose focus markers take matrix scope.

284

Daniel Hole

4.3.1 Adjunct foci in simplex sentences (30) is a paradigm of AEO foci on adjuncts in the AwF pattern. (30)

ADJUNCT +

a.

AwF STRATEGY Năm ngoái Nam làm việc thậm chí vào [chủ nhật]F. last year Nam do work even on Sunday ‘Last year Nam worked even on SundaysF.’

b.

Năm ngoái Nam làm việc cả vào [chủ nhật]F. last year Nam do work also on Sunday ‘Last year Nam worked also on SundaysF.’

c.

Năm ngoái Nam làm việc chỉ vào [thứ ba]F. last year Nam do work only on Tuesday ‘Last year Nam worked only on TuesdaysF.

In (30) the adverbial focus-sensitive expressions occur syntactically close to the adjuncts with which they interact. In addition, thậm chí in the preverbal position does seem to allow for EVEN readings on adjuncts.13 The partition pattern for adjunct foci in simplex clauses yields the paradigm in (31). (31) ADJUNCT + PARTITION STRATEGY + SIMPLEX SENTENCE a. Năm ngoái thậm chí vào [chủ nhật]F Nam cũng làm việc. last year even on Sunday Nam PRT BGeven/also do work ‘Last year Nam worked even on SundaysF.’ b.

làm việc. Năm ngoái cả vào [chủ nhật]F Nam cũng last year also on Sunday Nam PRT BGeven/also do work ‘Last year Nam also worked on SundaysF.’

c.

làm việc. Năm ngoái chỉ vào [thứ ba]F Nam mới last year only on Tuesday Nam PRT BGonly do work ‘Last year Nam worked only on TuesdaysF.’

13 I.e., sentences like (i) with the interpretation given in the translation are grammatical. (i) Năm ngoái Nam thậm chí làm việc vāo [chủ nhật ]F. last year Nam even do work on Sunday ‘Last year Nam even worked on [Sundays]F .’

Focus particles and related entities in Vietnamese

285

The examples in (31) all involve preposing. In-situ partition structures are not provided, but they are possible with cả ‘also’. It is worth pointing out that the automatic strengthening effect that we observed with preposed ALSO foci that are arguments is probably absent with non-arguments (i.e., (31b) is not necessarily interpreted as ‘Last year Nam worked even on Sundays’).14 If compared with the other partition structures discussed so far, an important difference emerges. The particles marking the foci in previous examples have all been from class PRT FOC , i.e. from the class of focus markers for argument expressions. The general pattern of these pairings of focus and background particles is repeated in (32)(= (17)). (32)

AEO FOCI (to be revised) The general pattern [PRT FOC FOCUS] [PRT BG BACKGROUND]

TOPOLOGY OF THE PARTITION PATTERN FOR

a.

b. Instantiations EVEN: đến ALSO: cả ONLY: mỗi

cũng FOCUS cũng

BACKGROUND

mới

What we find in (31), though, is that the adverbial focus-sensitive particles that have figured in the AwF patterns of previous sections now combine with the background markers that were so far only matched with the ad-argument focus particles of class PRT FOC . Our topology of the partition pattern for AEO foci should thus be modified as in (33) to allow for either possibility depending on whether arguments or non-arguments are in focus in the partition pattern. 14 Stavros Skopeteas (p.c.) has suggested to carve out the difference between arguments and adjuncts with a minimal pair corresponding to The cat jumped only onto the table vs. The cat slept only on the table. I tested these sentences, but the result was inconclusive. Both sentences may have mỗi in them, the particle hypothesized here to mark argument foci only; cf. (i) and (ii). (i) Con mèo chỉ nhảy mỗi lên bàn. the cat only jump PRT FOConly onto table ‘The cat jumped only onto the table.’ (ii) Con mèo chỉ ngủ mỗi trên bàn. the cat only sleep PRT FOConly onto table ‘The cat slept only on the table.’ The parallel construal of the PPs in (i) and (ii) with mỗi need not be counterevidence to the claim defended in the main text, viz. that arguments have focus markers of their own, among them mỗi. ‘Sleeping’-verbs frequently classify as verbs of posture with PP complements that are subcategorized for (like ‘live’, ‘stand’, ‘lie’, ‘sit’; Chinese is a case in point). Therefore one would have to construe a minimal pair with a different set of verbs. I haven’t done this.

286 (33)

Daniel Hole

AEO FOCI (revised) The general pattern [PRT FOC / PRTADV FOCUS] [PRT BG BACKGROUND]

TOPOLOGY OF THE PARTITION PATTERN FOR

a.

b. Instantiations ALSO:

cả/cả

EVEN:

đến/thậm chí

ONLY:

mỗi/chỉ

cũng FOCUS cũng

BACKGROUND

mới

4.3.2 Adjunct foci in complex sentences Complex sentences with foci in adjunct clauses are found in (34). (34) ADJUNCT + PARTITION STRATEGY + COMPLEX SENTENCE a. {Ngay cả/Thậm chí (cả)} khi thời tiết đẹpF Nam cũng even also/even also when weather good Nam PRT BGeven/also đi ôtô. drive car ‘Even when/if the weather is goodF Nam still drives with his car.’ b.

đi ôtô. Cả khi thời tiết đẹpF Nam cũng also when weather good Nam PRT BGeven/also drive car ‘Nam also drives with his car when/if the weather is goodF.’

c.

đi ôtô. Chỉ khi thời tiết xấuF Nam mới only when weather bad Nam PRT BGonly drive car ‘Only when/if the weather is badF does Nam drive with his car.’

With the exception of ngay in (34a), the complex sentence patterns employ exactly those markers that we have seen in the simple sentences already. We may say, by way of summary, that Vietnamese adjunct foci in simplex sentences may be encoded in the AwF pattern, or in the partition pattern. In complex sentences with foci in adverbial or adjunct clauses, only examples in the partition pattern were presented. Background particles with adjunct foci are not matched with focus particles from class PRT FOC as in the case of argument foci, but with particles from the adverbial paradigm. I.e., the split in the system that separates partition structures from non-partition structures cannot be aligned with the use of adverbial particles as opposed to particles from class PRT FOC if adjunct foci are taken into consideration. We will return to the issue in section 6, where the resulting system will also be represented schematically.

Focus particles and related entities in Vietnamese

287

4.4 Verb foci With verbs in AEO focus, we find the sole availability of the AwF pattern. The partition pattern seems to be excluded. Accordingly, the examples in (35) through (38) all involve adverbial association-with-focus by means of thậm chí ‘even’, cả ‘also’ (with its characteristic preposing of the verb) and chỉ ‘only’. (35)

VERB + AwF STRATEGY

Hôm qua Nam thậm chí ănF pho mát(, chứ không chỉ đứng nhìn). yesterday Nam even eat cheese CONTR . CONJ not only stand see ‘Yesterday Nam even ateF the cheese(, he didn’t just look at it).’ (36)

VERB + AwF STRATEGY

Nam thậm chí không thèm nhìnF pho mát. Nam even not want see cheese ‘Nam didn’t even want to lookF at the cheese.’ (37)

VERB + AwF STRATEGY

Bác nông dân không chỉ ăn cà chua mà trồngF cả cà chua. the farmer not only eat tomato but grow also tomato ‘The farmer doesn’t just eat tomatoes, he also growsF tomatoes.’ (38) VERB + AwF STRATEGY Q: Có phải hôm qua Nam nấu và ăn thịt bò không? is.it.true yesterday Nam cook and eat meat beef Q ‘Did Nam cook and eat the beef yesterday?’ A:

Không, nó chỉ nấuF (thịt bò) thôi. no he only cook meat beef PRT ‘No, he only cookedF the beef/it.’

(35) is a sentence in which ăn ‘eat’ is an EVEN focus; eating is construed as the contextually identified superlative relation in terms of unexpectedness that may hold between Nam and cheese; by contrast, just looking (at cheese) is the contextually given more likely relationship. (36) shows that negation intervening – and possibly scoping between – the focus operator and the focus does not alter the picture.15 From the perspective of 15 Cf. Gast and van der Auwera (2010) for discussion of analytic options in the typology of scalar additive operators with respect to the interaction with negation and other entailmentreversing operators.

288

Daniel Hole

English, this is not much of a surprise (cf. the English translation of (36)). But languages like German or Dutch have special EVEN markers that must be used in such configurations (nicht einmal, auch nicht ‘not even’ in German, zelfs niet, niet eens ‘not even’ in Dutch; cf. König 1991). (37) with its special verb-preposing syntax is identical to (4). The discourse in (38), finally, enforces a narrow ONLY focus on the verb nấu ‘cook’.

4.5 Sentences with intransitive verbs In this subsection, we will take a look at AEO foci with intransitive verbs. We will discuss how narrow AEO argument focus and broad AEO sentence focus are expressed in these structures. Since the expression of argument focus is as with transitives, we will concentrate on the differences between sentences with unaccusatives and unergatives in their potential to express narrow focus or broad (sentence) focus. There is no difference between sentences with unaccusative and unergative verbs in terms of the availability of different readings in the AwF pattern (allnew/thetic vs. subject in focus vs. verb in focus). Thetic readings and subject foci are available while verb foci are excluded. (39) illustrates this for ONLY foci. (39)

a. INTRANSITIVE VERB + UNACCUSATIVE + AwF STRATEGY Chỉ cây đổ. only tree topple.over (i) ‘The only thing that was the case was that [the tree toppled over.]F.’ (the chair wasn’t blown away) (ii) ‘Only [the tree]F toppled over.’ (the lamp post didn’t) *(iii) ‘The tree only [toppled over]F.’ (it didn’t burst in addition) b. INTRANSITIVE VERB + UNERGATIVE + AwF STRATEGY Chỉ thầy giáo nhảy. only teacher dance (i) ‘It was only the case that [the teacher danced]F.’ (nothing else happened) (ii) ‘Only [the teacher]F danced.’ (the students didn’t) *(iii) ‘The teacher only [danced]F.’ (he didn’t smile happily at the same time)

If a narrow focus on the verb is intended, the particles must immediately precede the verbs as in (40).

Focus particles and related entities in Vietnamese

(40)

289

a. INTRANSITIVE VERB + UNACCUSATIVE + AwF STRATEGY Cây chỉ đổ thôi. tree only topple.over PRT ‘The tree only [toppled over]F.’ (it didn’t burst in addition) b. INTRANSITIVE VERB + UNERGATIVE + AwF STRATEGY Thầy giáo chỉ nhảyF thôi. teacher only dance PRT ‘The teacher only dancedF.’ (he didn’t smile happily at the same time)

There are at least two non-ambiguous ways to narrow the focus down to the subject. These two ways are (i) partition structures with background markers and (ii) là-clefts (not covered here; but cf. Hole 2008: 17–20; 36–37). Special intonation patterns may be a further possibility. Strategy (i) alone, partition structures with background markers in nonmodalized contexts, is generally available with EVEN foci and with ALSO foci (cf. (41a/a′/b/b′). In accordance with our generalizations about the partition pattern for EVEN foci and ALSO foci we always find background-marking cũng in (41a/a′/b/b′). For ONLY foci, we get a split. Unaccusative đổ ‘topple over’ yields ungrammatical results in a partition structure with background marking mới (cf. (41c)), whereas unergative nhảy ‘dance’ yields a grammatical sentence (cf. (41c′)). (41)

INTRANSITIVE VERB + PARTITION STRATEGY

a. Đến PRT FOCeven

[cái cây]F cũng đổ. the tree PRT BGeven/also topple.over

‘Even [the tree]F toppled over.’ a′. Đến PRT FOCeven

nhảy. [thầy giáo]F cũng teacher PRT BGeven/also dance

‘Even [the teacher]F danced.’ b. Cả PRT FOCalso

đổ. [cái cây]F cũng the tree PRT BGeven/also topple.over

‘[The tree]F, too, toppled over.’ b′. Cả PRT FOCalso

nhảy. [thầy giáo]F cũng teacher PRT BGeven/also dance

‘[The teacher]F danced, too.’

290

Daniel Hole

c.

Chỉ mỗi [cái cây]F (*mới) đổ. only PRT FOConly the tree PRT BGonly topple.over int.: ‘Only [the tree]F toppled over.’

nhảy. c′. Chỉ mỗi [thầy giáo]F (mới) only PRT FOConly teacher PRT BGonly dance int.: ‘Only [the teacher]F danced.’ While one expects the difference between unaccusatives and unergatives to surface somewhere, the interpretation of the contrast between (41c) and (41c′) is by no means trivial. First, it is unclear why the contrast arises with ONLY foci only. Second, one would like to know whether the differing availability of background-marking mới reflects different structural positions of the foci. One could imagine that, due to their agentive semantics, subjects of unergatives like nhảy ‘dance’ must surface higher, i.e. in a position more to the left than subjects of unaccusatives like đổ ‘topple over’. It could then be the case that just the position more to the left actually precedes the structural position of mới, and that mới with unaccusatives is ungrammatical for that reason. In the absence of further evidence this is just a speculation, though. This concludes our survey of AEO foci in sentences with intransitive verbs.

5 Partition structures and free-choice The background marker for AE foci, cũng, occurs in at least one more construction expressing universal quantification with specific restrictions in terms of information structure, viz. in free-choice constructions. The present section discusses this construction, but I am not aiming at an exhaustive coverage of the empirical domain. (42a) is an example of a free-choice construction obligatorily employing cũng. (42) a.

{Đứa nào/ Ai} Nam *(cũng) thích (cả). person which/ who Nam PRT BGalso/even like FC ‘Nam likes everyoneF.’/‘Nam likes whoever there is.’

b. *Nam cũng thích {đứa nào/ ai} (cả). Nam PRT BGalso/even like person which/ who FC int.: ‘Nam likes everyoneF.’/‘Nam likes whoever there is.’

Focus particles and related entities in Vietnamese

291

In (42a) the object constituent contains an indefinite pronominal (glossed by a wh-word; cf. Tran and Bruening in this volume for more discussion of indefinite pronominals/wh-words), and it must be preposed (cf. the ungrammaticality of (42b) with the in-situ object). (There are two ways to encode the human indefinite, either analytically with the phrase đứa nào ‘which person’, or with a singleword indefinite for humans ai ‘who’.) Cũng must not be dropped. I analyze this construction as a free-choice construction, where universal quantification is over arbitrary valuations of the person variable. This means that a sentence like (42a) asserts that for the (arbitrarily) chosen value from the domain of persons we get the truth-value 1 for the sentence, and choosing any other value would likewise yield 1.16 (43) demonstrates how things change under negation. (43) a′. Nam (*cũng) chảng thích {đứa nào/ ai} (cả). Nam PRT BGalso/even not.EMPH like person which/ who FC ‘Nam likes [nobody (whatsoever)]F.’ (good with cũng as ‘It is also the case that Nam likes nobody (whatsoever).’) a′. Chảng {đứa nào/ ai} là Nam thích (cả). not.EMPH which person/ who COP Nam like FC ‘Nam likes [nobody (whatsoever)]F.’ b. *{Đứa nào/ Ai} Nam chảng thích (cả). person which/ who Nam not.EMPH like FC int.: ‘Nam likes nobodyF.’

16 More precisely, this analysis amounts to saying that the focus in free-choice constructions in Vietnamese is on the relevant operator, i.e. that device that picks out a particular referent from the relevant domain, and that alternative operators would pick out other referents with the same truth-functional outcome. This construal of free-choice semantics allows us to identify the operator in free-choice constructions with the choice function, i.e. the ε-operator (von Heusinger 1997; cf. also Giannakidou 2001 on the analysis of free-choice constructions). This is an indirect way of arriving at universal quantification over the entire domain. Cf. Hole (2004: sect. 4.3.4, 2006: 344–5) for the parallel case in Mandarin. A more widely adopted analysis of free-choice semantics was developed by Kratzer and Shimoyama (2002). Kratzer and Shimoyama analyze free-choice pronouns as denoting sets of type-identical elements. The crosslinguistic tendency to have a single pronominal form for pronouns with a negative polarity semantics and with a free-choice semantics speaks in favor of the analysis sketched above in terms of quantification over choice functions. Since NPIs are not typically interpreted as sets (Krifka 1995), one may wish to maintain a parallel semantics for free-choice pronouns, too.

292

Daniel Hole

(43a) is the negation of (42a). The sentence features the emphatic negative particle chảng; non-emphatic không ‘not’ may not be used. The indefinite pronominals must stay in situ if nothing else changes alongside (cf. the ungrammaticality of (43b), where the indefinite pronominals have moved). With this syntax, the use of background marking cũng is deviant (unless an ALSO-reading with wide scope is aimed at where cũng is not part of the construction under discussion, i.e. a reading like ‘Nam, too, likes nobody (whatsoever)’.). (43a′) is a variant of (43a) where the pronominal has been preposed and which is grammatical. The reason for the grammaticality is that the negation precedes the pronominals as in (43a) because the negative particle has likewise been preposed. With preposing of the pronominals, the copula là must be used before the predicate. This copula is the same element that occurs in clefts (cf. Hole 2008: 17–20; 36–37). The pattern instantiated by (43a′) is special in that it has an indefinite pronominal in what appears to be a clefted position. An English translation as It is NO body who Nam likes is deviant because the clefted constituent may not be a quantifier.17 A more adequate structure to mimic the preposing syntax in English would seem to be one involving do-support (Nobody does Nam like). This, in turn, would cast doubt on a free-choice analysis for the Vietnamese structure under scrutiny, viz. structures with preposed negation and indefinite pronominal plus là as a functional equivalent of the in-situ structures as in (43a). This is so because the universal quantification relevant to the interpretation of a sentence like Nobody does Nam like derives from the quantifier alone. In the analysis of the free-choice construction that we have sketched above and in footnote 16, the effect of universal quantification arises in the focus-semantic domain: the arbitrarily chosen valuation of the assertion yields a true sentence, and so would any alternative valuations. For the time being, I will continue to treat the preposing structure in (43a′) as a free-choice construction, but the issue needs to be revisited. The sentences in (42) and (43) have an optional free-choice particle cả at the end of the sentence. Note that we have discussed cả as an adverbial focus sensitive particle ‘also’ and as a homophonous ad-focus particle in previous sections. I assume that the free-choice marker cả is at least related to these uses by polysemy. Free-choice cả does not seem to form a constituent with the pronominals since it occurs in sentence-final position in (43b), a sentence in which the pronominals have been preposed; cả would be predicted to move along if it formed a constituent with the indefinite arguments. 17 The cleft structure with a focus accent on NO body should not be confounded with an acceptable English sentence like It is nobody who Nam LIKES [; it is someone who he deSPISES ]. In the latter construction the relative clause restricts the person variable and forms a constituent with the pronominal. Any verb may embed the pronominal in such a construction (cf. I met nobody who Nam likes). Cleft constructions as discussed in the text are restricted to cooccur with copulae.

Focus particles and related entities in Vietnamese

293

Table 1 summarizes the properties of Vietnamese free-choice-constructions with positive and negative polarity that we have discussed. Positive polarity

Negative polarity

position of indefinite pronominal

Preposed

preposed (with negation)/in situ

use of background marker cũng

Yes

No

use of COP là

No

yes (with preposing of indefinite pronominal)

use of free-choice marker cả

Possible

Possible

form of negation

d.n.a.

emphatic negation chả

Table 1: Properties of free-choice constructions with positive and negative polarity

Examples with indefinite/free-choice pronominals other than đứa nào ‘which person’ and ai ‘who’ are found in (44) through (46). The a-examples feature positive polarity, the b-examples negative polarity. The b′-examples involve preposing of the negation particle and the pronominal. (44)

PLACE

a. {Chỗ nào/ Đâu} Nam *(cũng) lau chùi (cả). place which/ where Nam PRT BGeven/also clean FC ‘Nam cleans up everywhereF.’ b. Nam chảng lau chùi {chỗ nào/ đâu} (cả). Nam not.EMPH clean place which/ where FC ‘Nam cleans up [nowhere (whatsoever)]F.’ b′. Chảng {chỗ nào/ đâu} là Nam lau chùi (cả). FC not.EMPH place which/ where COP Nam clean ‘Nam cleans up [nowhere (whatsoever)]F.’ (45)

TIME

a.

Lúc nào Nam *(cũng) lau chùi. time which Nam PRT BG even/also clean ‘Nam cleans up [at any time]F.’

b.

Nam chảng lúc nào lau chùi (cả). FC Nam not.EMPH time which clean ‘Nam [never (ever)]F cleans up.’

b′. Chảng lúc nào là Nam lau chùi (cả). not.EMPH time which COP Nam clean FC ‘Nam [never (ever)]F cleans up.’

294 (46)

Daniel Hole

THING

a.

Cái gì Nam *(cũng) lau chùi. thing what Nam PRT BG even/also clean ‘Nam cleans up everythingF.’

b.

Nam chảng lau chùi cái gì (cả). Nam not.EMPH clean thing what FC ‘Nam cleans up [nothing (whatsoever)]F.’

b′.

Chảng cái gì là Nam lau chùi (cả). not.EMPH thing what COP Nam clean FC ‘Nam cleans up [nothing (whatsoever)]F.’

There are probably further specialized indefinite pronominal expressions that figure in free-choice constructions like the ones in (42) through (46). They are used to express free-choice meanings of other semantic types, e.g. manner or cardinality. Since I lack sufficient evidence to exclude that some, or all, of these additional expressions instantiate constructions that are not free-choice constructions I must leave the exact delimitation of free-choice constructions in Vietnamese for future research.

6 The patterns of focus marking in Vietnamese: three orthogonal dimensions of classification Table 2 presents a first classification of Vietnamese focus-sensitive expressions as it has emerged from the discussion above.18

EVEN

Adverbial particles

Argument focus markers

Background markers

thậm chí

đến

cũng

ALSO

cả

cả

cũng

ONLY

chỉ

mỗi

mới

Table 2: Vietnamese focus-sensitive expressions with AEO foci (to be revised)

18 Table 2 deliberately refrains from making use of a representation format with more underspecification. To be sure, one could also have a single instance of cả and cũng, respectively, and use it to fill two adjacent positions. Since I’m not sufficiently confident about the nature of the observed identities on the signifier side (and whether both identities should be treated on

Focus particles and related entities in Vietnamese

295

In a sense, Table 2 constitutes an idealization. To sharpen the picture in the previous sections, I have not usually represented those variants of EVEN foci which have đến or thậm chí immediately followed by cả without changing the interpretation; cf. (34a), repeated here as (47). (47)

đi ôtô. {Ngay cả/Thậm chí (cả)} khi thời tiết đẹpF Nam cũng even also/even also when weather good Nam PRT BGeven/also drive car ‘Even when/if the weather is good F Nam still drives with his car.’

In fact, this pattern occurs frequently in spontaneous utterances provided by my consultants. From the perspective of what we have assumed about the semantic relationship between EVEN foci and ALSO foci in section 1, this co-occurrence is not much of a surprise. Still, since I am not sure about how to analyze cả in individual instances of those combinations (argument vs. non-argument focus?), I have decided in favor of an exposition which maximizes the signaling contrast between EVEN foci and ALSO foci. Recall from section 4.3 that it is not right to treat the background markers as necessarily co-occuring with the argument focus markers, even though most examples that we have discussed would support this pairing. What we have seen in connection with adverbially focus-marked adjuncts, which may also trigger background marking, is that it is more adequate to oppose the background markers to the set of focus-sensitive expressions as a whole. Put differently, we have three dimensions of classification, and not just two. These dimensions of classification are listed in (48). (48)

a. EVEN vs. ALSO vs. ONLY b. particles preceding argument foci only vs. particles also preceding non-argument foci c. particles preceding foci vs. particles preceding backgrounds19

a par), I have decided in favor of maximum specification in Table 2. The alternative not favored here is given in (i). (i)

EVEN

Adverbial particles

Argument focus markers

thậm chí

đến

ALSO ONLY

Background markers cũng

cả chỉ

mỗi

mới

19 Recall from the discussion of the partition structures that the generalization in terms of c-command or precedence is an idealization in the case of the background markers (at least if

296

Daniel Hole

Particles c-commanding foci

Particles c-commanding backgrounds (at some level of representation)

EVEN

argument non-argument

đến thậm chí

cũng

ALSO

argument non-argument

cả cả (plus preposing of the verb)

cũng

ONLY

argument non-argument

mỗi chỉ

mới

Table 3: Vietnamese focus-sensitive expressions with AEO foci (final)

This concludes the discussion of the core system of focus-sensitive and backgroundsensitive expressions in Vietnamese as it has been laid out in the present chapter.

7 Conclusions and outlook This chapter has surveyed the distribution of elements signaling EVEN foci, ALSO foci and ONLY foci in Vietnamese. We have found variation along three major dimensions. The first dimension concerns the difference between adargument markers and adverbial markers: there is one set of particles combining with arguments in focus, or with arguments containing a focus, and another set combining with non-arguments in focus, or with non-arguments containing a focus. Another dimension of variation separates particles preceding foci from particles preceding backgrounds. The third dimension of variation is a classification of foci into EVEN foci, ALSO foci and ONLY foci. The general architecture of this system was discussed in the preceding section 6. Table 4 summarizes the special properties of each kind of focus type that we have identified in this chapter. This chapter has only paid cursory attention to the register sensitivity of individual particles. It seems to be the case that thậm chí has a more formal flavor to it than chỉ or cả in the same paradigm. The same holds true of thậm chí in comparison with đến and cũng in the orthogonal EVEN paradigm.

one looks at the surface patterns only). While the particles in the left column reliably c-command their foci at the surface (with the sole exception of adverbial cả; cf. section 2), the particles in the right column may c-command both (the largest portion of) the background and the focus. The clear partition is only visible at the surface if the focus has been preposed, or if constituents with a canonically preverbal position are in focus.

Focus particles and related entities in Vietnamese

EVEN

ALSO

ONLY

preposing/topicalization of foci unrestricted (except for verb foci)

no preposing/topicalization of foci (preposing triggers EVEN readings)

preposing/topicalization frequently possible

BG-marking cũng obligatory in partition structures with foci preceding their backgrounds

BG-marking cũng obligatory in partition structures with foci preceding their backgrounds

adverbial particle thậm chí either precedes the predicate or the whole sentence

adverbial particle cả follows the verb syncretism/homonymy of adverbial focus-sensitive particle cả and argument focus particle cả use of sentence-final nữa alongside other ALSOparticles attested (see below)

297

BG-marking mới frequently optional in partition structures with foci preceding their backgrounds adverbial chỉ either precedes predicate or the whole sentence frequent use of adverbial chỉ alongside FOC and BG markers. frequent use of sentencefinal thôi alongside other ONLY-particles (see below)

Table 4: Special properties of EVEN foci, ALSO foci and ONLY foci in Vietnamese

Another interesting issue left undiscussed in the main parts of the chapter concerns the fact that there is at least one more position in which particles signaling AEO foci may occur, namely the sentence-final position. With ONLY foci, in particular, we find the frequent use of a particle, thôi ‘only’, in sentence-final position.20 With ALSO foci we sometimes find nữa in that position. Cf. (49) for one example each; the sentence-final particles have been highlighted. (49)

a.

Chỉ mỗi [thịt bò]F Nam mới ăn thôi. only PRT FOConly meat beef Nam PRT BGonly eat only ‘Only beefF does Nam eat.’

b.

nữa. Nam ăn thịt bò và cũng ăn cả [thịt gà]F Nam eat meat beef and PRT BGeven/also eat PRT FOCalso meat chicken also ‘Nam eats beef, and he eats also chickenF.’

Thôi occurs frequently in my data and its use is often considered, if not obligatory, then at least strongly preferred. One of my consultants reports the intuition that the use of thôi interacts with the use of chỉ in the following way. Both 20 Thôi is used in (1), (16b), (18c), (19c), (20c), (20′c), (23c), (25b), (38) and (40).

298

Daniel Hole

particles may be used simultaneously, as is the case in (56a), but if both are dropped, at least one of them is felt to be missing. Sentence-final thôi occurs in many examples in this chapter, but for reasons of exposition I just glossed it as PRT when it occurred.21 While I’m unable to state anything precise about restrictions or triggers of thôi (or nữa) at the present moment, it is immediately evident that the existence of these additional particles enhances the analytical challenge posed by the “particle proliferation” that we find in the domain of AEO foci in Vietnamese. In (49a), for instance, four words are used that we could, with some justification, translate as ‘only’. In the present chapter, and except for a comparative remark below, I will have nothing else to say about the intriguing property of particle proliferation of Vietnamese. From the perspective of Standard Average European languages, the various strategies for expressing AEO foci in Vietnamese appear exotic and highly peculiar. In the areal context, however, there is at least one more language with a similarly complex pattern of AEO focus marking. This language is Mandarin Chinese, and chances are high that more instances of such systems can be found in Chinese dialects. (50) provides a set of examples to illustrate the AwF pattern. The partition pattern is exemplified in (51). (50)

M ANDARIN C HINESE + AwF

STRATEGY

a. Lăo Wáng shènzhì bù hē cháF. old Wang even not drink tea ‘Old Wang doesn’t even drink teaF.’ b. Lăo Wáng yĕ hē cháF. old Wang also drink tea ‘Old Wang also drinks teaF.’ c. Lăo Wáng zhĭ hē cháF. old Wang only drink tea ‘Old Wang only drinks teaF.’

21 In the English translation of 1b), the translation of the focus constituent is a contrastive topic, and either is in focus (cf. Krifka 1998). This is an indirect result of the obligatory postposing of either in English. In the Mandarin sentence, the information-structural partitioning may indeed be as indicated. Cf. also the German translation, which has been added for (51b) and which mimics the Chinese information structure more straightforwardly.

Focus particles and related entities in Vietnamese

(51)

M ANDARIN C HINESE + PARTITION a. Lián PRT FOCeven

299

STRATEGY

[zhèi zhŏng shū]F Lăo Wáng *(dōu) măi-guo. this kind book old Wang PRT BGeven buy-ASP

‘Old Wang has bought even [this kind of book]F before.’ wŏ *( yĕ) bú qù. b. Jiùsuàn DéniánF lái, if.PRT FOCalso Denian come I PRT BGalso not go ≈ ‘DenianCT coming won’t make me go, [either]F.’22 cf. German Auch wenn DenianF kommt, gehe ich nicht hin. c. Zhĭyŏu PRT FOConly

măi-guo. [zhèi zhŏng shū]F Lăo Wáng *(cái) this kind book old Wang PRT BGonly buy-ASP

‘Only [this kind of book] has Old Wang bought before.’ Without going into any detail here, it is evident that Mandarin instantiates a system that is very similar to that of Vietnamese. Table 5 duplicates Table 3 for Mandarin. Particles c-commanding foci

Particles c-commanding backgrounds

EVEN

lián (partition) shènzhì (AwF )

dōu

ALSO

[ jiùsuàn (partition)]22 yĕ (AwF )

yĕ

ONLY

zhĭyou (partition) zhĭ (AwF )

cái

Table 5: Particles c-commanding foci vs. particles c-commanding backgrounds in Mandarin

One difference between the Mandarin and the Vietnamese systems should be pointed out, though. In Mandarin, the adverbial particles are restricted to an adverbial position at the left edge of VPs/tense phrases/modal phrases. Subjects invariably precede them. The Vietnamese adverbial particles thậm chí, cả and chỉ, by contrast, may also head complete sentences, simplex and complex. This could either be interpreted as evidence to the effect that Vietnamese adverbial particles are more flexible in terms of possible adjunction sites; or it could be taken to mean that the adverbial particles occur in identical positions in Mandarin 22 Jiùsuàn ‘If . . . too’ has been bracketed because it is a focus marker and simultaneously a complementizer. I have no clear evidence of any ALSO particle in Mandarin which obligatorily precedes/c-commands ALSO foci in the Mandarin partition pattern of simplex sentences; but cf. Hole (2006: 353, fn. 14) for a possible instance in the Mandarin counterpart of the rather . . . than-construction.

300

Daniel Hole

and Vietnamese, but that Mandarin can move material across this position to the left more easily. I’ll have to leave this matter for future research. For a second similarity between Vietnamese and Chinese turn to (52) and (53). These Chinese sentences feature the (highlighted) sentence-final ‘only’words éryĭ and bàle. I.e. Mandarin, just like Vietnamese, has a sentence-final position that may host ONLY-particles. (52)

Qĭtú zhìzào bú yòng néngyuán-de yŏngdòngjī try construct not need source.of.energy-MOD perpetuum.mobile zhĭ shì yī zhŏng huànxiăng éryĭ. (adapted from Hou (ed.) 1998: 190) only COP 1 CL :kind illusion only ‘To try and construct a machine capable of perpetual motion which is not in need of a source of energy is just a chimera and no more.’

(53)

Wŏ zhĭ shì shuō shuō bàle, nĭ zĕnme jiù dāngle zhēn ne! I only COP say say only you how at.once take.as true PRT ‘I just said it [without really meaning it], how could you take it for granted right away?’ (adapted from Hou (ed.) 1998: 13)

To the best of my knowledge, the exact distribution of these particles hasn’t been investigated yet. What may be said with some certainty is that bàle is more colloquial than éryĭ, and that éryĭ with its classical origin literally means ‘then stop’. Moreover, there is an intuition of speaker orientation and downtoning present in the Chinese sentence-final ONLY-words that parallels certain uses of just in English (cf. [Don’t scold him.] He’s just a boy/[She didn’t mean to interfere.] She just wanted to offer her help). I hypothesize that the same shade of meaning is also present with thôi in Vietnamese. The parallels to Vietnamese in terms of syntax and “particle proliferation” are again striking. To be sure, Chinese and Vietnamese are not genetically related. Chinese is Sino-Tibetan, while Vietnamese is an Austro-Asiatic Language. It is well-known, however, that Chinese has exerted strong influence on Vietnamese over the last two millennia. For this reason, one could easily imagine that there has been structural borrowing from Chinese to Vietnamese in addition to the well-attested numerous lexical borrowings (cf. Lương 1994 with his list of 2316(!) borrowed monosyllabic morphemes/characters).23 In fact, according to Lương (1994: 176, 23 Note that contemporary research in contact linguistics no longer assumes structural borrowings to have their source in substrate languages only. If the contact situation is close enough, structural borrowings with their source in superstrate languages (Chinese in our case) do occur (Thomason 2001).

Focus particles and related entities in Vietnamese

301

192) and Alves (2006), from among the function words discussed in this chapter, at least the following are of Chinese origin: thậm chí ‘even’ (cf. Mod. Chinese shènzhì ‘even’ as in (50a), chỉ ‘only’ (cf. Mod. Chinese zhĭ ‘only’ as in (50c/52)) and mỗi ‘PRT FOC only’ (cf. Mod. Chinese mĕi ‘every’24). It is quite likely that the number of loans in our domain is even bigger than that, but at present I lack reliable information about the diachrony of other particles. I hope that this chapter, despite the many questions that had to be left unresolved, will serve as a point of departure for further studies dealing with the empirical intricacies and theoretical implications of AEO foci in Vietnamese and in general. There is some hope that the rich Vietnamese system can shed new light on the modeling of the focus background partition. The co-existence, and reliable distinguishability, of different paradigms of expressions signaling AEO foci may, for instance, be used to argue for a less-than-minimal theory of focus syntax. Given that an association-with-focus strategy competes with a partition strategy in Vietnamese, the theoretical divide between adverbial approaches (Jacobs 1983; Büring and Hartmann 2001) and partition approaches (von Stechow 1982) to the syntax and semantics of focus particles appears in a new light. This is so because Vietnamese would seem to lend support to both theories. The detailed argumentation for such a theory is beyond the scope of this chapter and must be left for a future occasion.

References Alves, Mark. 2006. Grammatical Sino-Vietnamese vocabulary. Ms. Montgomery College, Rockville, Maryland. (Last updated April 4, 2006; URL: http://www.geocities.com/ malves98/Alves_Sinovietnamese_grammatcalvo cabulary_CHART.pdf ) Beck, Sigrid and Arnim von Stechow. 2006. Handout zu LFs für Ereignissemantik. Ms. Universität Tübingen. Bisang, Walter. 1992. Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai und Khmer. Vergleichende Grammatik im Rahmen der Verbserialisierung, der Grammatikalisierung und der Attraktorposition. Tübingen: Narr. Büring, Daniel. 2006. Focus Projection and Default Prominence. In: Valéria Molnár and Susanne Winkler (eds.), The Architecture of Focus, 321–46. Berlin/New York: Mouton De Gruyter. Büring, Daniel and Katharina Hartmann. 2001. The syntax and semantics of focus-sensitive particles in German. Natural Language and Linguistic Theory 19: 229–81. Cheng, Lisa L.-S. and Rint Sybesma. 2004. Postverbal ‘can’ in Cantonese (and Hakka) and Agree’. Lingua 114: 419–45. 24 It is certain beyond doubt that mỗi is a Chinese loan as a quantifier with the meaning ‘each’ (Lương 1994; Alves 2006). The semantic connection with the focus particle use of mỗi ‘only’ is not obvious.

302

Daniel Hole

Đố Thế Dũng, Trần Thiên Hương and Georges Boulakia. 1998. Intonation in Vietnamese. In: Daniel Hirst and Albert Di Cristo (eds.), Intonation Systems. A Survey of Twenty Languages, 395–416. Cambridge: CUP. Duffield, Nigel. 2001. On certain head-final effects in Vietnamese. In: Karine Megerdoomian and Leora Anne Bar-el (eds.), WCCFL 20 Proceedings, 101–14. Somerville, MA: Cascadilla Press. von Fintel, Kay. 1994. Restrictions on quantifier domains. Ph.D. dissertation, University of Massachusetts at Amherst. Gast, Volker and Johan van der Auwera. 2010. Scalar additive operators in the languages of Europe. Language 87(1): 1–53. Giannakidou, Anastasia. 2001. The meaning of free choice. Linguistics and Philosophy 24: 659–735. Hedberg, Nancy and Lorna Fadden. 2007. The Information Structure of It-clefts, Wh-clefts and Reverse Wh-clefts in English. In: Nancy Hedberg and Ron Zacharski (eds.), The GrammarPragmatics Interface: Essays in Honor of Jeanette K. Gundel, 49–76. Amsterdam/ Philadelphia: Benjamins. von Heusinger, Klaus. 1997. Salienz und Referenz. Der Epsilonoperator in der Semantik der Nominalphrase und anaphorischer Pronomen. Berlin: Akademie. Hole, Daniel. 2004. Focus and Background Marking in Mandarin Chinese. System and theory behind cái, jiù, dōu and yĕ. London/New York: Routledge Curzon. Hole, Daniel. 2006. Mapping VPs to restrictors: Anti-Diesing effects in Mandarin Chinese. In: Klaus von Heusinger and Ken Turner (eds.), Where Semantics Meets Pragmatics, 337–80. Amsterdam: Elsevier. Hole, Daniel. 2008. EVEN, ALSO and ONLY in Vietnamese. In: Shinichiro Ishihara, Svetlana Petrova and Anne Schwarz (eds.), Interdisciplinary Studies in Information Structure 11, 1–54. Universitätsverlag Potsdam. Horn, Laurence R. 1969. A presuppositional analysis of only and even. Proceedings of the Fifth Annual Meeting of the Chicago Linguistics Society, 97–108. Horn, Laurence R. 1996. Exclusive company: only and the dynamics of vertical inference. Journal of Semantics 13: 11–40. Hou, Xuechao (ed.). 1998. Xiandai Hanyu xuci cidian. (Dictionary of Function Words in Contemporary Chinese). Beijing: Beijing Daxue Chubanshe. Jacobs, Joachim. 1983. Fokus und Skalen. Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer. Jannedy, Stefanie. 2007. Aspects of prosody in Vietnamese. Workshop on “Intonational Phonology: Understudied or Fieldwork Languages”, August 5th, 2007, University of California at Los Angeles (ICPhS 2007 Satellite Meeting). Kay, Paul. 1990. Even. Linguistics and Philosophy 13: 59–111. Karttunen, Lauri and Stanley Peters. 1979. Conventional Implicatures. In: C.-K. Oh and D. A. Dineen (eds.), Syntax and Semantics 11: Presupposition, 1–56. New York: Academic Press. König, Ekkehard. 1991. The Meaning of Focus Particles. A comparative perspective. London/ New York: Routledge. Kratzer, Angelika and Junko Shimoyama. 2002. Indeterminate Pronouns: The view from Japanese. In: Yukio Otsu (ed.), The Proceedings of the Third Tokyo Conference on Psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Krifka, Manfred. 1995. The semantics and pragmatics of polarity items. Linguistic Analysis 25: 209–57.

Focus particles and related entities in Vietnamese

303

Krifka, Manfred. 1998. Additive particles under stress. In: Devon Strolovitch and Aaron Lawson (eds.), Proceedings of Semantics and Linguistic Theory (SALT ) 8, 111–28. CLC, Ithaca. Krifka, Manfred. 2007. Basic notions of information structure. In: Caroline Féry, Gisbert Fanselow and Manfred Krifka (eds.), Working Papers of the SFB 632. Interdisciplinary Studies on Information Structure, 13–56. Potsdam: Universitätsverlag Potsdam. Lương, Văn Kế. 1994. Der chinesische Einfluß auf die vietnamesische Sprache. Münster/ Hamburg: Lit. Reinhart, Tanya. 1982. Pragmatics and linguistics: an analysis of sentence topics. Philosophica 27: 53–94. Rooth, Mats. 1985. ‘Association with focus’. Ph.D. dissertation, University of Massachusetts at Amherst. Rooth, Mats. 1996. Focus. In: Shalom Lappin (ed.), The Handbook of Contemporary Semantic Theory, 271–97. Oxford: Blackwell. Schwarzschild, Roger. 1999. Givenness, AvoidF and other constraints on the placement of accent. Natural Language Semantics 7: 141–77. von Stechow, Arnim. 1982. Structured propositions. Technical report 59. SFB 99, Universität Konstanz. Thomason, Sarah. 2001. Language Contact: An Introduction. Edinburgh: Edinburgh University Press. Thompson, Laurence C. 1987. A Vietnamese Reference Grammar. Honolulu: University of Hawai‘i Press.

Index accommodation 10–12, 14, 272 acquisition, first language 87–98, 101, 115– 116 activity 199, 250–251, 253–255, 260 agent 157–158, 159–160, 169, 249 – agent demotion 168, 173, 186 – agentive 290 alternation 139, 149, 179, 185, 187, 198, 204, 207, 209, 225 anaphor 93, 114, 157, 161 argument – argument focus 267, 272–273, 283, 285– 286, 288, 294–297 – external argument 256, 259–260 – optional argument 249, 260 article, lexical 57, 71, 73, 77, 84 – definite article, cf. determiner, classifier 72–73 – indefinite article 71 aspect 131, 147–148, 185, 187–188, 190, 192, 197–200, 204, 207–209, 211– 212 – aspectual alternation 185, 210, 211 association-with-focus 265, 267–268, 287, 301 Austro-Asiatic 57, 76, 300 Austro-Tai 57, 76 Autosegmental Phonology 35–36, 47 background marker 194, 267, 274, 279, 285, 289–290, 292–295, 297 backgrounded event 194–195 bare noun 59, 72, 97 bị ‘passive marker’ 156, 160–162, 164–166, 168, 171–182 case 87, 168, 169–170, 174, 179 change of location 185–189, 191–193, 195– 199, 203–208, 211–212 child language 87–98 child data – cross-sectional 87, 95, 97–98, 109–110, 113–114 – longitudinal 87, 89, 91, 95, 96–98, 101, 110, 113

Chinese 65, 77, 132, 134, 138, 140, 151–152, 155–156, 160–164, 166–171, 173–182, 208–211, 218–221, 226, 231–232, 234, 243–244, 257–258, 260, 265, 267, 298– 301 choice function 217, 225–226, 237–240 classifier 57–59, 64, 87–123, 219, 221–222 – anaphoric use 93, 95 – classifier omission 110–111, 113 – classifier phrase 62–63, 87–94, 103, 109 – deixis and classifier 91 – demonstrative and classifier 88, 91–92, 97 – double classifier 91 – extra cái 65–66, 93 – focus marker cái, cf. demonstrative reinforcer 65, 67, 69 – general classifier 90, 98, 113–114 – measure words and classifiers 61, 63, 84 – non-numeral constructions 87, 91, 95, 109, 114–115 – numeral classifier 88, 93, 95, 97–98, 101– 102, 109, 111–112, 114, 115–116. – specific classifier 98–99, 113–114 – wh-words and classifiers 88, 92, 101 Clausal Typing Hypothesis 151–152 có ‘exist’ 130, 138–139, 142–143, 146–148, 234, 237 comitative – adjunct 243, 249, 255–256 – comitative PP 250, 252–253, 260 complementizer 130, 132–135, 152, 179–180, 299 – CP structure 127, 130, 137, 152, 257 complex predicates 212 conditionals 226, 232, 234–235 conjunction 194, 243, 253, 256–258, 261 – distributive conjunction 245, 247, 258 – comitative/collective conjunction 247, 249–250, 252, 256, 258, 261 control 155 – control structures 173 convergence 9–11, 14 Conversation Analysis 35, 37–39, 48

306

Index

– telephone conversation 35, 46, 48–49 count noun, cf. mass noun 57, 59 coverbs 186, 197–198, 208 đã ‘anterior’ 130–133, 138–139, 143, 145– 148, 271 đang ‘progressive’ 190–195, 199, 203–204 demonstrative 57–58, 67–70, 87–88, 90– 93, 97–98, 101, 114–116, 134, 217–218, 234, 237 – demonstrative reinforcer, cf. focus marker cái 57–58, 69–70, 84 determiner, cf. article 57–58, 72–73, 101, 109 dialect 9, 13, 15–16, 18, 21–23, 25–29, 32, 298 divergence 9–10, 13–14 được ‘abilitative modal’/‘adversity passive marker’ 134, 164–166, 172, 174, 176, 182, 270 English 11–14, 35–36, 38, 59, 137, 140–141, 151, 161, 168, 171, 174, 177–180, 187– 188, 197, 204, 226, 235, 240, 243–245, 253, 258, 260, 266, 269, 271, 275, 279– 280, 288, 292, 298, 300 existential quantifier 218, 222–223, 226, 230, 233–234, 237–240 experimenter 9, 11–12, 18, 21, 23–28, 32–34, – experimenter effect 9, 18, 25, 27 focus – adjunct focus 67, 283–284, 286, 295 – focus marker 84, 267, 283, 285, 294–295 – focus marker cái, cf. demonstrative reinforcer 65, 67, 69 – focus particle 265, 267–270, 273, 275, 285, 292, 301 – focus-sensitive particle 270–272, 276, 285, 292, 296 – verb focus 269, 279, 287, 297 free-choice 265, 267, 290–294 French 70, 130, 179, 243–244, 253, 257– 258, 260 German 35–36, 38, 87, 168, 173–174, 211, 258, 288

glottalization 16, 18–19, 26–28 head-initial 127–128, 130–134, 136–138, 143, 149, 151–152 head-final 88, 127–128, 131–132, 151 haplology 65 indefinites 217–218, 226, 228–229, 236, 238, 240 – wide-scope indefinites 218–219, 225, 237–240 interaction – talk-in-interaction 35–36, 38 – tone-intonation interaction 35, 46–49, 52 intonation 35–36, 45–51, 275 island 158, 161, 217, 222–225, 228 kin term 128, 163, 171, 178 Korean 89, 156, 170, 173 mass noun, cf. count noun 57, 59, 66, 72 measure word 57, 60–61, 84 – measure phrase 59, 63 modifier 65, 68–69, 79, 134, 250, 253 – head-modifier 133 movement 57, 68 – A-movement 159–160 – A’-movement 158, 179, 181 – N-movement 69, 79 – operator movement 178–179 – phrasal movement 57, 78, 84 – pitch movement 36, 54 – roll-up movement 84, 131 – wh-movement 151, 224, 228 negation 127, 129, 131, 143–144, 146, 148, 150, 219, 222, 223–224, 230–231, 233– 238, 247, 261, 287, 291 – sentential negation 129, 143–144, 148, 223 Northern Vietnamese (NVN), cf. dialect null operator 155, 179, 233, 240 omission errors 87, 91, 101, 109–111, 113–116 outer object 169–171 passive – adversity passive 170–171, 182 – indirect passive 156, 162–163, 169, 171

Index

– intransitive passive 156–166 patient 157, 161, 169, 178 – patient promotion 168, 173 perception 9–12, 13–14, 18, 22, 25, 27–28, – tonal perception 9, 15, 19 placeholder 90, 113, 115 position – postverbal position 247, 250–251, 256– 257, 261 – preverbal position 247, 251–253, 284 predicate-raising 127, 131–132, 149, 150 predication 249, 261 – non-symmetrical predication 249, 256 – stative predication 249, 251, 253, 259 – symmetrical predication 249, 256–257, 261 progressive 190–193, 195, 199, 203–204, 208 prosody 35–37, 39, 42, 45, 269, 275 question – embedded question 135, 228–229 – tag question 132, 135, 140–141, 144–145, 149, 152 – wh-question 152, 217, 227–228 – yes-no question 37, 40–41, 128, 130, 132, 135–136, 141, 143–149, 151–152, 217– 219, 231, 233 reciprocity 244, 253, 255–256, 261 – reciprocity marker 244, 253, 255 repeater 90 repair – next turn repair initiator 39–40, 42 – other-initiated self-repair 39 – self-repair 38 resultativity 185, 193, 205–211 – perfect of result reading 185, 187, 192, 195, 202–203, 205–206, 211 reverse directionality 256 sẽ ‘posterior’ 130, 133, 145, 147–148, 194, 271 selection 176, 201 – selectional restriction 61 – selectional framework 211

307

serial verbs 185, 197 – serial verb construction (SVC) 185–187, 197, 203, 207–208, 211–212 Sino-Tibetan 57, 76, 300 situation type 185, 200–201, 203–206, 209–212 Southern Vietnamese (SVN), cf. dialect specifier 57, 59, 68–69, 144, 147, 149, 152, 257 sprachbund 76 strengthening effect 278–279, 285 structured proposition 265 tense-aspect interaction 127, 147, 187, 211 terminative 201, 211–212 – gradually terminative 202, 204, 206, 210–212 Thai 77, 80, 83, 88–93, 115, 173–174, 176– 178, 180 time schema 201, 205, 212 tone – intonational tone 35, 48 – lexical tone 14–29, 35–37, 40, 42–43, 45, 46–51, 87 topic 128, 131, 142, 145, 152, 236, 273–274, 297 – topicality 230 – topicalization 127, 297 unaccusative 288–290 unergative 173–174, 288–290 universal quantifier 149, 238, 292 viewpoint 168, 185, 199–203, 205–210 wh-word 87–88, 92–95, 101, 114, 149, 151, 217–219, 226–228, 230, 240, 291 – wh-in situ 149–150 – wh-indefinite 218–219, 221–222, 225, 236, 239–240 – wh-question word 36–37, 40, 43–44, 151 word order 57–58, 68–69, 70, 76–79, 81– 83, 88, 90, 115–116, 132–133, 186 – word order typology 77, 84