Constructions across Grammars 9027242704, 9789027242709

Citation preview

.) s d t e r ( pe tman l i H s in t Ö r a a l M -O n a J and

c ro s

s

B E N J A M I N S C U R R E N T TO P I C S

a ions truc t Cons mars Gram



Constructions across Grammars

Benjamins Current Topics issn 1874-0081 Special issues of established journals tend to circulate within the orbit of the subscribers of those journals. For the Benjamins Current Topics series a number of special issues of various journals have been selected containing salient topics of research with the aim of finding new audiences for topically interesting material, bringing such material to a wider readership in book format. For an overview of all books published in this series, please see http://benjamins.com/catalog/bct

Volume 82 Constructions across Grammars Edited by Martin Hilpert and Jan-Ola Östman These materials were previously published in Constructions and Frames 6:2 (2014).

Constructions across Grammars Edited by

Martin Hilpert Université de Neuchâtel

Jan-Ola Östman University of Helsinki

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

doi 10.1075/bct.82 Cataloging-in-Publication Data available from Library of Congress. isbn 978 90 272 4270 9 (Hb) isbn 978 90 272 6708 5 (e-book)

© 2016 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Company · https://benjamins.com

Table of contents

Introduction Reflections on Constructions across Grammars Martin Hilpert and Jan-Ola Östman On the borrowability of subject pronoun constructions in Turkish– Dutch contact A. Seza Doğruöz On the universality of frames: Evidence from English-to-Japanese translation Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore Phonological elements and Diasystematic Construction Grammar Steffen Höder

1

7

34 67

Clause combining across grammars: A contrastive analysis of L1 and L2 construal of discourse organization Bracha Nir

97

Constructional tolerance: Cross-linguistic differences in the acceptability of non-conventional uses of constructions Florent Perek and Martin Hilpert

131

Constructions do not cross Languages: On cross-linguistic generalizations of constructions Philipp Wasserscheidt

169

Index

203

Introduction

Reflections on Constructions across Grammars Martin Hilpert and Jan-Ola Östman

1. Towards a multilingual Construction Grammar This book addresses constructions across grammars, that is, speakers’ knowledge of constructions in multilingual contexts. As a matter of its intellectual heritage, Construction Grammar has traditionally tried to model the knowledge of monolingual speakers. The general index to the Oxford Handbook of Construction Grammar (Hoffmann & Trousdale 2013) does not list terms such as bilingualism, multilingualism, translation, or language contact. It is probably fair to say that to this day, most constructional research continues to focus on single languages, and here it is English that receives the most attention. However, these tendencies are changing. Moving away from a predominant focus on English and an idealized view of native-speaker competence in a homogeneous linguistic community (Chomsky 1965: 3), the contributions in this book investigate whether Construction Grammar can be used to model the linguistic knowledge of speakers in decidedly heterogeneous communities of speakers, including bilingual and multilingual speakers, learners and native speakers, beginners and experts. The idea that Construction Grammar may lend itself particularly well to such an undertaking receives support from an existing body of research that has taken constructional thinking into cross-linguistic domains. The contributions in Fried and Östman (2004) have demonstrated that Construction Grammar can be usefully applied across different languages; Croft (2001, 2005) derives strong typological arguments for Construction Grammar from cross-linguistic comparisons; and the papers in Boas (2010) provide evidence that the notion of constructions is of great value in contrastive linguistics. The contributions in this book can thus build on a foundation of ideas that has been firmly established. Yet, much remains to be done. So far, few studies have approached phenomena such as multilingualism or language contact from a constructional perspective,

doi 10.1075/bct.82.01int 2016 © John Benjamins Publishing Company

2

Martin Hilpert and Jan-Ola Östman

and the interesting theoretical proposals that have been made remain to be worked out in detail. For example, drawing on ideas of Weinreich (1954), Höder (2012) pioneers the idea of diasystematic links that capture generalizations that speakers make across corresponding constructions in their repertoire of languages. Diasystematic constructions of the kind that Höder envisages would include not only similar-sounding lexical elements across two or more languages (i.e. German Haus and English house), but also syntactic patterns such as yes-no questions or cleft sentences. Höder’s suggestion is intriguing, and it holds a number of implications for current conceptions of constructional knowledge. Taking multilingualism seriously may force us to re-think some of the basic assumptions of Construction Grammar. It also may lead us to adapt our views of constructional knowledge, away from the idea of a relatively static system of constructions, towards the concept of a flexible network of emerging and emergent practices (cf. Auer & Pfänder 2011). In order to promote a discussion of a multilingual Construction Grammar, we, the editors of this book, organized a conference at the Freiburg Institute for Advanced Studies in 2012. The idea of the FRIAS workshop was to explore the role of constructions in scenarios where speakers are competent in multiple grammars, either speaking several varieties of the same language or being able to switch between altogether different languages. These scenarios, as suggested above, include situations of language contact, bilingualism, language learning and attrition, and second language acquisition, amongst others. The crucial question is how speakers in these situations make use of constructional generalizations. Are constructions from one language or variety projected to another one? If speakers form diasystematic meta-generalizations that establish correspondences between constructions across languages, how should those be represented in a network of constructions? How are issues of variability handled? Do different social scenarios lead to different strategies of grammatical representation? What are the theoretical implications of such representations? These are questions that ought to be taken very seriously in the further development of Construction Grammar as a theory of linguistic knowledge. The questions also show that researchers in Construction Grammar ought to cooperate with scholars from other subfields within linguistics, in particular experts working on language contact (Thomason 2001), contrastive and comparative linguistics (König & Gast 2007), variationist sociolinguistics (Tagliamonte 2012), second language acquisition (Ortega 2009), bilingualism (Romaine 1995), as well as code-switching and code-mixing (Auer 1998). Furthermore, research on the sociolinguistics of globalization (Blommaert 2010) has made the point that speakers are not only bi- or multilingual, but actually super-diverse (Blommaert & Rampton 2011) in their access to different languages and modes of communication. A necessary step for Construction Grammar to be a realistic model of linguistic knowledge is the adoption of some of these ideas into a truly dynamic model of linguistic



Reflections on Constructions across Grammars

competence. Such a model would have to include several aspects of linguistic knowledge that current models of constructional knowledge do not acknowledge enough. First, it would have to have a place for knowledge of intra-linguistic variation, that is, an account of speakers’ ability to use different varieties, and their ability to accommodate to different interlocutors. Second, constructional knowledge also includes speakers’ crosslinguistic competence. Bilingual or multilingual speakers have the ability to form meta-generalizations that connect corresponding constructions from different languages in their linguistic repertoires. Third, in keeping with the idea of superdiversity, the knowledge of language that multilingual speakers have is not tidily divided into separate language systems, but rather represents an arrangement of ‘bits and pieces’, which would mean a ragbag of resources from several languages that are put to use in the situations that require them. What is traditionally seen as a switch between different codes may thus, upon closer inspection, turn out to be a transition between very intimately related resources, even if those resources include linguistic material from different languages. Thus, there clearly is a need for Construction Grammar to address the questions that are raised by the issues of multilingualism and language contact. Can a constructional view be usefully applied to research questions in these areas? How will a reflection on these issues feed back into the theoretical development of Construction Grammar? The contributions in this book initiate a discussion of those matters. The next section briefly presents each of them. 2. The contributions in this book This book brings together six contributions that shed light on constructions across grammars in different ways, drawing on a typologically diverse set of languages and contact situations. Seza Doğruöz opens the discussion with an analysis of the borrowability of subject pronoun constructions in Turkish-Dutch language contact. Focusing on Turkish as spoken in the Netherlands, her study investigates uses of first person pronouns. A quantitative analysis reveals that these do not differ in frequency across Standard Turkish and Netherlandic Turkish. However, a qualitative analysis shows that the constructional contexts of those pronouns differ across the two varieties. Unusual Netherlandic Turkish pronoun uses include idiomatic phrases with meanings such as ‘I don’t know’ or ‘as far as I know’, which correspond to Dutch idioms with overt pronouns. Another unusual habitat for a Turkish subject pronoun is the context of a subordinate clause. This, too, is explained through Dutch interference. Finally, the use of pronouns in left dislocation and in polar questions also accounts for some unusual patterns in Netherlandic Turkish. The

3

4

Martin Hilpert and Jan-Ola Östman

main conclusion of the study is that Netherlandic Turkish does not undergo dramatic changes at the most schematic levels of grammatical organization, but instead on several lower levels, including the level of fully specified constructions. Yoko Hasegawa, Russell Lee-Goldman, and Charles J. Fillmore analyze the universality of frames based on evidence from English-to-Japanese translation. Their study investigates nominal and verbal modes of expression in Japanese. In comparison to English, abstract nominal style, referred to by the authors as objectcentered encoding, is dispreferred in Japanese. This dispreference manifests itself for instance in the fact that English transitive constructions are often translated into Japanese as sequences of intransitive clauses. The reverse happens much less often. The main goal of the article is to contrast object- and event-centered encoding by offering frame-based analyses of English sentences and their Japanese translational equivalents. In the center of attention is the question whether the frames that are evoked by the English originals are also evoked by the Japanese translations. As the discussion shows, sometimes there are substantial parallelisms, sometimes the picture is more complex. The analysis serves to illustrate how FrameNet can be used to investigate cross-linguistic contrasts and to assess the quality of translations. In his contribution to this book, Steffen Höder proposes a new type of relation between constructions, namely a link between corresponding constructions in different languages or language varieties that form part of a speaker’s linguistic repertoire. The view of constructional knowledge that encapsulates this idea is called Diasystematic Construction Grammar. The motivation behind this proposal is that (i) language contact has been largely ignored so far in Construction Grammar and (ii) knowing and using two languages evidently has massive consequences for the organization of linguistic knowledge in speakers’ minds. In this sense, Diasystematic Construction Grammar is really just Construction Grammar applied to issues of multilingualism, language contact, and the co-occurrence of several varieties. The study focuses on phonological elements, which have been neglected in Construction Grammar up to now. Höder’s main claim is that phonological elements need to be integrated into a satisfactory model of constructional knowledge. Höder goes on to argue that diasystematically distinctive phonemes carry pragmatic meaning and are thus constructions. Bracha Nir studies clause combining across grammars, investigating how advanced learners of English, with Hebrew as their native L1, use clause-combining structures. Their performance is contrasted with the behavior of English and Hebrew natives. To this end, written narratives are elicited and analyzed with regard to clausal structure. A number of differences between native and learner texts emerge. Native English narratives are longer, make greater use of different- and same-subject coordination, and show more frequent omission of same-subjects in complement clauses. However, differences between Native Hebrew texts and



Reflections on Constructions across Grammars

Learner English texts show that the learner profile is not due to L1 influence. On the whole, both Native and Learner English show more frequent use of strongly integrated biclausal constructions, whereas Hebrew shows a greater tendency towards paratactic devices. The study thus illustrates the structural complexities that arise in language contact situations. Also Florent Perek and Martin Hilpert analyze the behavior of second language learners. They advance a claim they call the ‘constructional tolerance hypothesis’. The main question that drives the analysis is whether typological differences in learners’ native languages influence the tolerance of those learners towards nonconventional uses of constructions in the L2. Whereas languages such as English grant speakers considerable freedom to combine constructions and verbs in nonconventional ways, so that you can ‘sneeze the napkin off the table’, other languages, such as French, impose stronger restrictions on the way in which constructions and lexical items can be combined. In order to test the idea of constructional tolerance, two groups of L2 learners of German perform a grammaticality judgment task. These groups differ in their L1s, which are English and French respectively. According to the constructional tolerance hypothesis, speakers of a constructionally tolerant language (such as English) should judge non-conventional examples in an L2 with more lenience than speakers of a valency-driven language (such as French). The results suggest that grammaticality judgments are influenced by the availability of a productive L1 construction that shows functional overlap. In his study entitled ‘Constructions do not cross languages’, Philipp Wasserscheidt offers a different perspective on some of the issues that are discussed by the other contributions to this book. Drawing on data from typologically distinct languages, specifically code-switching between Slavic, Finno-Ugric, and Turkic languages as well as English, the paper critically assesses the importance of language form for diasystematic, cross-linguistic representations of constructional knowledge. In developing this argument, the study outlines a new approach to syntactic transfer phenomena, arguing that syntactic transfer by itself is not to be taken as prima facie evidence for diasystematic generalizations that involve linguistic form. 3. The way ahead We hope that the contributions to this book will spark further interest in a multilingual understanding of Construction Grammar. It goes without saying that we are in the early stages of developing such an understanding, which means that, for the moment, our discussion will leave us with more questions than answers. That, however, is not in itself a bad thing. Thankfully, more research addressing constructions across grammars is already in the making: For instance, at the

5

6

Martin Hilpert and Jan-Ola Östman

2014 International Construction Grammar Conference in Osnabrück, Hans Boas and Steffen Höder co-organized a theme session on Construction Grammar and Language Contact. As Construction Grammar continues to make inroads into fields such as Second Language Acquisition, Sociolinguistics, Translation Studies, and Linguistic Area Studies, chances are good that we will be seeing more events of that kind in the future.

References Auer, P. (1998). Code-switching in conversation. London: Routledge Auer, P., & Pfänder, S. (Eds.) (2011). Constructions: Emerging and emergent. Berlin: de Gruyter. ​ doi: 10.1515/9783110229080

Blommaert, J. (2010). The sociolinguistics of globalization. Cambridge: Cambridge University Press. ​doi: 10.1017/CBO9780511845307 Blommaert, J., & Rampton, B. (2011). Language and superdiversity. Diversities, 13(2), 1–21. Boas, H. C. (Ed.). (2010). Contrastive studies in Construction Grammar. Amsterdam: John Benjamins. ​doi: 10.1075/cal.10 Chomsky, N. A. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press. Croft, W. A. (2001). Radical Construction Grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press. ​doi: 10.1093/acprof:oso/9780198299554.001.0001 Croft, W. A. (2005). Logical and typological arguments for Radical Construction Grammar. In JO. Östman & M. Fried (Eds.), Construction grammars: Cognitive grounding and theoretical extensions (pp. 273–314). Amsterdam: John Benjamins. ​doi: 10.1075/cal.3.11cro Fried, M., & Östman, J-O. (Eds.). (2004). Construction Grammar in a cross-language perspective. Amsterdam: John Benjamins. ​doi: 10.1075/cal.2 Höder, S. (2012). Multilingual constructions: A diasystematic approach to common structures. In K. Braunmüller & Ch. Gabriel (Eds.), Multilingual individuals and multilingual societies (pp. 241–257). Amsterdam: John Benjamins. ​doi: 10.1075/hsm.13.17hod Hoffmann, T., & Trousdale, G. (Eds.). (2013). The Oxford handbook of Construction Grammar. Oxford: Oxford University Press. ​doi: 10.1093/oxfordhb/9780195396683.001.0001 König, E., & Gast, V. (2007). Understanding English-German contrasts. Berlin: Erich Schmidt Verlag. Ortega, L. (2009). Understanding second language acquisition. London: Routledge. Romaine, S. (1995). Bilingualism. Oxford: Blackwell. Tagliamonte, S. (2012). Variationist sociolinguistics: Change, observation, interpretation. Malden/ Oxford: Wiley-Blackwell. Thomason, S. G. (2001). Language contact: An introduction. Edinburgh: Edinburgh University Press. Weinreich, U. (1954). Is a structural dialectology possible? Word, 10, 388–400.

On the borrowability of subject pronoun constructions in Turkish–Dutch contact A. Seza Doğruöz

Netherlands Institute for Advanced Studies

Turkish spoken in the Netherlands (NL-Turkish) sounds different in comparison to Turkish spoken in Turkey (TR-Turkish) due to Dutch influence. In addition to borrowed Dutch words/phrases, Dutch influence on NL-Turkish is also observed through literally translated constructions. This study investigates the Dutch influence on NL-Turkish constructions with subject pronouns. Analyses of NLTurkish and TR-Turkish spoken corpora do not reveal any significant differences in terms of subject pronoun frequency. However, qualitative analysis of the data reveals some unconventional cases of subject pronoun use in NL-Turkish. In these cases, subject pronouns do not lead to unconventionality on their own but as parts of larger constructions that are copied from Dutch as chunks. Following the principles of usage-based approaches, these unconventional constructions are further analyzed in terms of their level of schematicity and flexibility. Keywords: Turkish, Dutch, subject pronouns, language change, usage-based approaches, constructions

1. Introduction Language change has been widely studied in contact settings (Weinreich 1953; Thomason & Kaufman 1988; Boeschoten 1994; Croft 2000, 2006; Milroy 2003; Leino & Östman 2005). Turkish has been in contact with Dutch in the Netherlands for the last fifty years. There are differences between Turkish spoken in the Netherlands (NL-Turkish) and Turkish spoken in Turkey (TR-Turkish) even when NL-Turkish speakers do not use Dutch words/phrases in their Turkish (Backus 2005; Doğruöz 2007; Doğruöz & Backus 2007, 2009). In this study, Dutch is the ‘model language’ for the contact-induced changes in Turkish, the ‘replica language’ (Heine & Kuteva 2005). In contact settings, several types of linguistic borrowing (e.g. lexicon, structures, grammar, phonology

doi 10.1075/bct.82.02dog 2016 © John Benjamins Publishing Company

8

A. Seza Doğruöz

etc.) take place. According to Thomason & Kaufman (1988), borrowing starts with lexical items and gradually spreads to more structural aspects (e.g. syntax). Dutch words and phrases are commonly used in NL-Turkish (Backus 2004) but Dutch influence is also observed when NL-Turkish speakers do not use overt Dutch words/ phrases. Although word order turns out to be quite stable (Doğruöz & Backus 2007), NL-Turkish hosts different types of Dutch-like constructions with lexical, morphological, and structural variation (Doğruöz & Backus 2009). Subject pronoun use is also expected to change in the case of contact between a pro-drop language (e.g. Turkish) and a non pro-drop language (e.g. Dutch). More specifically, subject pronouns in the pro-drop language are expected to increase their frequency and also change their discourse-pragmatic properties due to contact with a non pro-drop language (e.g. Montrul 2004; Silva-Corvalán 1994; Sorace & Filiaci 2006; Tsimpli et al. 2004; Polinsky 1995; Pease-Alvarez, Hakuta & Bayley, 1996). This cross-linguistic influence is assumed to be especially prevalent if the pro-drop language is sociolinguistically the weaker one. Traditionally, Turkish is described as a pro-drop language (Enç 1986; Erguvanlı-Taylan 1979; Özsoy 1987; Kerslake 1987; Turan 1996; Öztürk 2002; Gürel 2006) which makes it possible to have clauses without overt subject pronouns. In the case of Turkish-Dutch contact, Dutch is expected to influence NLTurkish subject pronoun use but systematic studies are still lacking. Following usage-based approaches to language analysis, this study explores subject pronouns as part of whole constructions rather than as independent lexical items in both NL-Turkish and TR-Turkish corpora. When we consider a subject pronoun as part of a construction, the Dutch influence becomes visible mostly in lexically fixed and semi-fixed constructions. Although both corpora are limited in size for diachronic generalizations, this study captures the Dutch influence on NL-Turkish subject pronouns in a subtle way which would not be possible through a syntactic analysis only. By analyzing variation in NL-Turkish through combining theories from sociolinguistics and Cognitive Linguistics (Langacker 1987; Croft 2001; Tomasello 2003), this study contributes to the developing field of Cognitive Sociolinguistics (Geeraerts, Kristiansen & Peirsman 2010; Kristiansen & Dirven 2008). Instead of focusing only on the cognitive aspects of language use and analyzing language out of context, Cognitive Sociolinguistics combines social aspects of variation (e.g. contact vs. non-contact settings) with empirical linguistic analysis through naturalistic data sets. The structure of the paper is as follows: Section 2 describes the sociolinguistic background of the Turkish community in the Netherlands. Section 3 provides an overview of subject pronoun change across different contact settings. Section 4 focuses on the problems of syntactic and discourse-pragmatic generalizations.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

Sections 5 and 6 introduce usage-based approaches to subject pronoun analyses. Section 7 exemplifies unconventional constructions with subject pronouns in NLTurkish. Section 8 discusses the implications of constructional analysis for subject pronouns in comparison to existing literature and Section 9 explores possibilities for future studies across different language contact situations. 2. Turks/Turkish in the Netherlands Starting in the 1960s, Western European countries invited workers from various countries, including Turkey, to come and work in Europe. Central Anatolia has been the main area of emigration due to a high rate of unemployment in the 1960s. Today, everyone in this area has a relative or acquaintance who has immigrated to Europe. The initial intention of the first immigrants was to stay for a few years and earn enough money to improve the lives of their families in Turkey. After family reunifications became common in the 1970s, full-fledged immigrant communities started to form. Currently, Turks are one of the largest minority groups in the Netherlands (CBS 2011). Although the first generation did not really learn Dutch beyond the basic level, second and third generation Turks who grew up in the Netherlands are fluent speakers of Dutch since they have gone through schooling. However, this does not necessarily lead to a language shift. The perceived value of Turkish in the Netherlands is still very high (Yağmur & Van de Vijver 2012). Changes in the replica language are highly related to social factors such as the type and size of the immigrant community, relations within and outside the community, additional languages spoken by the people in the community, attitudes of the speakers as well as the interactions between speakers of different languages (Aikhenvald & Dixon 2001; Thomason 2001). Some of the factors that play a role in the maintenance of NL-Turkish could be summarized as follows (Backus 2004: 695): – – – – –

few exogamous marriages; possibility of returning back to Turkey; frequent summer-long holidays in Turkey; easy access to Turkish media through internet, TV and radio programs; many opportunities for intra-group contact through Turkish organizations and social networks; – some marginalization and physical segregation in the urban areas where most Turks live (cf. Akıncı & Backus 2005).

9

10

A. Seza Doğruöz

3. Subject pronoun use in language contact and change In generative approaches to linguistics, syntax is immune to changes, no matter how intense the contact is (Montrul 2004; Toribio 2004; Sanchez 2004). Syntactic change refers to the alterations in the syntax-semantics or syntax-pragmatics interfaces. Although there are contradictory results about the increase of overt subject pronoun frequency in contact situations, zero subject use is usually maintained (except Polinsky 1995; Rothman 2009). Spanish in the US is by far the best studied case of subject pronoun use in contact situations. Silva-Corvalán (1994), Flores-Ferrán (2004), Pease-Alvarez, Hakuta & Bayley (1996) do not find an increase in the use of overt subject pronouns in the Spanish of Spanish-English bilinguals in the US. However, most report some loosening of the pragmatic constraints on subject pronouns (SilvaCorvalán 1994; Lapidus & Otheguy 2005). Montrul (2004) reports a difference among the Spanish heritage users in the US based on their proficiency level. In comparison to the Spanish monolinguals, heritage speakers with an intermediate level of Spanish proficiency increase their use of overt subject pronouns. Similar to these findings, Schmitt (2000) and Bolonyai (2000) also report an increase in the use of subject pronouns in Russian and Hungarian speakers who are residents in the US. Tsimpli et al. (2004) compare the use of overt subjects by Greek-English bilinguals with Greek monolinguals and Italian-English bilinguals with Italian monolinguals. The results of the analyses reveal that the Greek-English bilinguals prefer overt preverbal subjects more than the monolinguals do. For the Italian group, the differences between monolinguals and bilinguals concern mostly the pragmatic aspects of the overt subjects rather than the syntactic ones. In the same vein, Sorace & Filiaci (2006) find English speakers learning Italian as a second language to be more tolerant of overt subject pronouns in complex Italian clauses. There are some studies exploring the use of subject pronouns in child language acquisition as well. Paradis & Navarro (2003) indicate a slight overuse of subject pronouns for Spanish-English bilingual children in comparison to Spanish monolingual children. On the other hand, Özcan et al. (2000) find no differences between the bilingual Danish-Turkish children and Turkish monolingual children in terms of the frequency of subject pronouns or their pragmatic functions. For the Dutch-Turkish contact setting, both Schaufeli (1991) and Aarsen (1994) report a slight increase of subject pronoun use in the narratives of bilingual children in comparison to the ones by monolingual children. However, these differences are not statistically significant. Similarly, there are some reports of a tendency to use redundant subject pronouns for Turkish spoken in Germany (Pfaff 1992; Rehbein 2001) but there are no systematic accounts.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

As far as Turkish as a second language is concerned, learners of Turkish are reported to use null and overt subject pronouns in a way that is similar to the native Turkish speakers and grasp the subtle pragmatic differences, except for the third person subject pronoun (Gürel 2006). In sum, all these studies, be they grounded in contact linguistics or second language acquisition, confirm the fact that the syntactic pro-drop feature is not easily lost in contact situations. Bilingual speakers are able to form sentences without subject pronouns. However, when they use subject pronouns, the pragmatic meaning attached to a particular form does not always conform to the conventions of the monolingual variety. Although the previous studies have provided us with insights about subject pronoun use by bilinguals in controlled experimental settings (e.g. judgment tests), we have little direct evidence (except narratives to a certain extent) about how bilinguals use subject pronouns in daily spontaneous conversations. In order to fill this gap, this study investigates the subject pronoun use in NL-Turkish through natural spoken conversations. Subject pronouns in the above-mentioned studies are assumed to be syntactically free and optional but there are discourse-pragmatic contexts which require or prohibit their use. This duality is the result of analyzing language in separate modules (e.g. lexicon, syntax, pragmatics etc.) which are assumed to operate (somewhat) independently from each other (Hulk & Müller 2000). Analyzing language in modules is problematic since it leads to overgeneralizations at the syntactic and discourse-pragmatic levels. In order to analyze subject pronoun use at a finer level, this study follows usage-based approaches (Tomasello 2003; Goldberg 2006) to language analyses which do not compartmentalize language into different modules. Instead, the strict division between lexicon and syntax is abolished and language is assumed to be made up of constructions which have unique forms and meanings. Section 4 describes the problems with the modular view for subject pronoun analysis and Section 5 discusses how usage-based approaches can offer solutions for these problems. 4. How to analyze subject pronouns? In generative linguistics, utterances with subject pronouns are represented syntactically (e.g. [Spro O V] and their discourse-pragmatic features are described in terms of a few general functional categories (e.g. contrast and topic shift). Sections 4.1 and 4.2 describe the problems with these generalizations.

11

12

A. Seza Doğruöz

4.1 Problems with syntactic generalizations Syntactically, the grammatical roles of the lexical items in an utterance are represented by a schematic template which does not have any lexical and morphological specifications (e.g. [S (O) V]). In terms of subject pronoun analyses, this representation is problematic since quite a few utterances may have similar schematic templates with the same subject pronoun, although they instantiate different constructions. This problem is illustrated with examples (1), (2), and (3), which all have the schematic [Spro (O) V] template, but in which the subject pronoun belongs to a different construction in each case. The utterance in example (1) has an out-of-the-blue meaning since it reveals an unexpected piece of information (e.g. someone suddenly enters a room and announces that s/he has fallen down). It is the whole utterance which rejects pragmatic presuppositions and is therefore associated with the presentational focus rather than the subject pronoun on its own (Göksel & Özsoy 2003, also cf. sentence focus by Lambrecht 1994). (1) Ben düş-me-di-m. I fall-neg-past1-1sg ‘I fell down.’

In (2), the [Spro (O) V] template yields a contrastive meaning since two subject referents (ben ‘I’ and kızkardeş-im ‘sister-poss.1sg’) are being contrasted with each other in terms of falling down or not. Due to this contrastive meaning, the subject pronoun is not only part of the [Spro V] template, but also part of the [Spro V1-neg S2 V1] template. Therefore, the contrastive meaning should not be attributed to the subject pronoun on its own but to the template representing the whole construction instead. (2) Ben düş-me-di-m kızkardeş-im düş-tü. I fall-neg-past-1sg sister-poss.1sg. fall-past ‘I did not fall down, my sister did.’

The subject pronoun in example (3) is part of a fixed expression illustrating how submissive ladies consult their husbands for everything. This expression is often cited by ladies as a humorous excuse to decline doing something. Note that this expression is made up of two clauses with two subject referents (ben ‘I’ and beyim ‘husband-poss.1sg’) and two verbs which share the same root (i.e. bil ‘know’) but one of them has a negation marker (i.e. –me ‘neg’). 1.  abl: Ablative, acc: Accusative, com: Commitative, dat: Dative, dm: Discourse Marker, gen: Genitive, inf: Infinitive, loc: Locative, nom: Nominalization, neg: Negation, opt: Optative, past: Past tense, pl: Plural, pres: Present Tense, pro: Pronoun, prog: Progressive, poss: Possessive, rel: Relative, qp: Question Particle, sg: Singular.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

(3) Ben bil-me-m bey-im bil-ir. I know-neg-1sg husband-poss.1sg. know-pres.3sg ‘I don’t know (anything), my husband knows (better).’

Syntactically, both utterances in examples (2) and (3) share the [S1 V1-neg S2 V1] template. However, the subject pronoun is part of a formulaic expression with fixed lexical elements in example (3) and it is part of a contrastive construction without any fixed lexical elements in example (2). When we focus only on the schematic representation of the constructions with subject pronouns (e.g. [Spro O V]), we could easily overlook the subtle cues of form and meaning in other constructions. 4.2 Problems with discourse-functional categories Traditionally, topic shift and contrast are assumed to be the discourse-pragmatic functions of subject pronouns in Turkish (Enç 1986). However, these associations are not necessarily exhaustive of subject pronoun use in contextualized settings (e.g. daily conversations) and they tend to overlook how subject pronouns convey a certain meaning in collaboration with their lexical and structural neighbors in the same utterance. In sum, focusing only on the form or the discourse-pragmatic functions of subject pronouns leads to overgeneralizations (cf. the criticism of the ‘lumper’ approach by Croft 2001) and overlooks subtle cues about how and when subject pronouns are used in daily conversations. These overgeneralizations are also problematic for cross-linguistic analyses (see Evans & Levinson 2009 for further discussion about problems with language universals) since constructions with subject pronouns differ across languages that are traditionally categorized as pro-drop. In order to overcome these problems, usage-based approaches provide us with a framework that considers subject pronouns as parts of constructions. 5. Analyzing subject pronouns through usage-based approaches Utterances produced by actual speakers are the starting point for usage-based approaches. When we speak, we do not build utterances from scratch by picking up morphemes and words randomly. Instead, we either utter completely prefabricated units with fixed lexical items and meanings (e.g. formulaic expressions) or recycle and combine the different constructions which may have some fixed and some non-fixed lexical items associated with a certain meaning (Erman & Warren, 2000). Language is assumed to be made up of constructions which combine both structural and functional aspects simultaneously (Langacker 1987; Croft

13

14

A. Seza Doğruöz

2001; Tomasello 2003; Fried & Östman 2004; Bybee 2006; Goldberg 2006; Bybee & Beckner 2010). Conventionality and fixedness are the common features of constructions in usage-based approaches (Croft 2001; Langacker 1987; Tomasello 2003; Goldberg 1995, 2006). Conventionality refers to a tacit agreement among the members of a speech community about the use of individual constructions (e.g. fixed vs. free items, structural features, intonation, meaning etc.). Within every construction, there is at least one fixed lexical/structural (e.g. a morpheme, word, word order) aspect and/or a conventionalized meaning. Native speakers recognize the constructions based on their fixed and/or conventionalized aspects which contribute to the unit status of the constructions in return. Each utterance in a language is made up of several constructions, which could be represented at different levels of schematicity (Croft & Cruse 2004; Fried & Östman 2004; Kay & Michaelis 2012). The utterances in examples (2) and (3) exemplify how different constructions with subject pronouns could be placed on the Specificity Continuum (see Figure 1), based on their fixed and non-fixed features (Doğruöz & Backus 2009). The right hand side of the continuum represents syntax and hosts maximally schematic representations such as [Spro V] [S O V]. The left hand side is reserved for the conventionalized maximally specific instantiations (i.e. lexicon) of the schematic representations. Some of these specific instantiations appear so frequently with each other that they are recognized as fixed expressions by native speakers. For example, the subject pronoun ben ‘I’ occurs with various verbs in Turkish. When it is part of a fixed expression, it has to appear with a specific verb (e.g. bilmek ‘know’) such as example (3). In between are the partially schematic units which have characteristics of both lexicon and syntax. Although the subject pronoun ben ‘I’ unifies with the verb in example (2), there is no restriction that this verb has to be a particular one. Therefore, it is represented as ‘V’ to indicate its partially schematic (not fixed) union with the subject pronoun.

Partially Schematic

Max.specific [Ben I bey-im husband-1SG.

bil-me-m know-NEG-1SG. bil-ir] know-PRES.

V1-NEG] [ben V1-NEG” “I V1] [kızkardeş-im “sister-POSS.1SG V1”

Max.schematic

[S.PRO V] [S O V]

Figure 1.  Representation of constructions with subject pronouns on the Specificity Continuum (Doğruöz & Backus 2009)



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

When we speak, we utter neither isolated individual words (e.g. kızkardeş-im ‘sister-poss.1sg’) nor maximally schematic templates (e.g. [Spro V]). Instead, we build our utterances using maximally specific (e.g. fixed expressions or specific instantiations of more schematic constructions) and partially schematic constructions. So far, subject pronouns in Turkish have only been theoretically analyzed based on their maximally schematic representations (i.e. on the right side of the Specificity Continuum). This study argues that subject pronouns are parts of larger constructions rather than individual lexical items on their own. In situations of language contact, subject pronouns are copied from the model language into the replica language through the constructions they are part of. Section 6 provides a quantitative overview of subject pronoun use in NL-Turkish and TR-Turkish spoken corpora and Section 7 describes the usage-based analysis of unconventional subject pronoun constructions in NL-Turkish. 6. How to detect Dutch influence on NL-Turkish 6.1 Description of data In order to detect on-going changes in a contact variety, it is desirable to make comparisons with the non-contact variety (Bullock & Toribio 2004). This study compares NL-Turkish and TR-Turkish in terms of subject pronoun use. The subset of the NL-Turkish spoken corpus (18.461 words) that is used for this study was collected in Tilburg (NL) among the second generation TurkishDutch bilingual speakers (six speakers in all) who were born in the Netherlands or immigrated to the Netherlands before they turned 6. At the time of the recordings, they were between the ages of 18 and 30. The subset of the TR-Turkish spoken corpus (20.772 words) was collected in a small Central Anatolian town in Turkey. All five participants were monolingual Turkish speakers and they were between the ages of 18 and 30 at the time of the recordings. Collecting and transcribing two types of spoken corpora in two countries required intense labor and time. However, objective comparisons of the two corpora also prevented the author (TR-Turkish speaker from Istanbul) from making premature judgments about NL-Turkish subject pronoun use. Although the author was present at some of the conversations, her conversational contributions are excluded from the data analysis.

15

16

A. Seza Doğruöz

6.2 Frequency analysis This section compares the frequency of subject pronoun use in NL-Turkish and TR-Turkish spoken corpora. Although this analysis may not indicate how subject pronouns behave within the constructions they are part of, it provides an opportunity to make cross-linguistic comparisons with respect to subject pronoun use in Turkish and other pro-drop languages. First, the utterances of the NL-Turkish and TR-Turkish speakers (excluding the interviewer) were coded according to their clause types. As illustrated in Table 1, simplex clauses with one finite verb were the most frequent clause types for both corpora. Table 1.  Frequency of clause types in NL-Turkish and TR-Turkish corpora Types of clauses

NL-Turkish

Simplex clauses

1805

(52%)

1849

(49%)

Copula clauses

1104

(32%)

1136

(30%)

Complex clauses   473

(14%)

  652

(17%)

Reported speech    76 clauses

(  2%)

  153

(  4%)

Total

3458

TR-Turkish

3790

Subject pronoun use in simplex clauses has been explained in Section 4. In complex clauses, subject pronoun use is rather restricted (Turan 1996) and the subject of the main clause cannot be overtly expressed when it is co-referential with the subject of the subordinate clause (see example 4). (4) Anne-m gel-ince hemen yemek yaptı. mother-poss.1sg come-ger immediately food make-past-1sg ‘My mother prepared some food as soon as she came.’ * Anne-m gel-ince hemen o yemek yap-tı. mother-poss.1sg come-ger immediately she food make-past-1sg ‘My mother prepared some food as soon as she came.’

Similar restrictions also hold for reported speech clauses which make use of null subject expressions when the subject referents of the subordinate clause and the reporting verb are the same (see example 5). (5) Ben [baba-m-a ev-e gid-iyor-um] de-di-m. I father-poss.1sg-dat house-dat go-prog-1sg say-past-1sg ‘I told my father that I am going home.’



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

*/? Ben baba-m-a ben ev-e gid-iyor-um de-di-m. I father-poss.1sg-dat I house-dat go-prog-1sg say-past-1sg ‘I told my father that I am going home.’

Subject pronoun use in copula clauses depends on the type of copula clause (adjectival/nominal and existential copula clauses). In adjectival/nominal copula clauses, the copula is expressed through the person inflection on the adjective/noun. The subject expression could be overt or non-overt (see examples 6a and 6b). (6) a. b.

Ben çok yorgun-um. I very tired-1sg ‘I am very tired.’ Çok yorgun-um. very tired-1sg ‘I am very tired.’

In existential copula clauses, genitive pronouns are used instead of subject pronouns, creating possessive third person subject NPs (see example 7). (7) [Ben-im yeni bir araba-m] var. I-poss.1sg new one car-poss.1sg exist ‘I have a new car.’ (literally ‘my new car exists’)

Table 2 illustrates the use of subject pronouns per clause type in NL-Turkish and TR-Turkish spoken corpora. This study focuses only on subject pronoun use in simplex clauses (with one finite verb) for the following reasons: – More than half of the subject pronouns in both data sets occur in simplex clauses (224 out of 418 in NL-Turkish and 235 out of 456 in TR-Turkish). – Within the simplex clauses, imperatives, discourse-marker like elements and fixed expressions are excluded from the analyses for the following reasons: Table 2.  Subject pronoun frequency per clause type NL-Turkish and TR-Turkish corpora NL-Turkish

TR-Turkish

Clauses with subject pronouns

Clause Frequency

Clauses with subject pronouns

Clause frequency

Simplex clauses

224

1805

235

1849

Copula clauses

103

1104

103

1136

Complex clauses   77

  473

  88

  652

Reported speech   14 clauses

   76

  30

  153

Total

3458

456

3790

418

17

18

A. Seza Doğruöz



– Imperatives do not make use of subject pronouns. – The use or non-use of the subject pronoun is already part of the fixed form in fixed expressions or discourse-marker-like elements (see appendix A and B).

Simplex clauses in both corpora are further analyzed in terms of the types of subject references (Null, NP, Subject Pronoun). The analyses reveal that null subjects are used much more frequently than the pronominal or nominal subjects in both NL-Turkish and TR-Turkish corpora (see Table 3). Table 3.  Proportions of subject expressions in NL-Turkish and TR-Turkish Data

NP

NULL

PRO

Total number of simplex clauses

NL-Turkish

221 (15%)

1025 (70%)

215 (15%)

1461 (100%)

TR-Turkish

252 (16%)

1130 (70%)

229 (14%)

1611 (100%)

There are statistically no significant differences in terms of the ratios of null, pronominal and nominal subject expressions across the two corpora. The probability of error of the chi-square test statistic was greater than the alpha level of significance (0.05), χ2 (2) = 0.27, p = .88. The ratios of subject pronouns in both Turkish corpora are also similar to those in Spanish (20%, Davidson 1996) and in Italian (17%, Oliveira 2000). In sum, these analyses of simplex clauses do not reveal any differences between NL-Turkish and TR-Turkish in terms of the frequency of subject pronoun use. However, further qualitative analyses of the NL-Turkish corpus reveal subject pronoun constructions that sound unconventional for TR-Turkish speakers. Section 7 discusses the method of analysis (7.1) and examples of unconventional NL-Turkish constructions based on their level of schematicity (Sections 7.2 and 7.3). 7. Unconventional NL-Turkish constructions with subject pronouns 7.1 Constructional analysis Some constructions in NL-Turkish data sound unconventional to TR-Turkish ears since they are often literal translations from Dutch. These translations are often clearly recognizable as such, although there is rarely a perfect one-to-one correspondence between a Dutch original and the Turkish rendition. In this study, unconventional NL-Turkish constructions with subject pronouns are detected in the following way:



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

– All the potential cases of unconventional constructions with subject pronouns in the NL-Turkish corpus are identified by the author (a TR-Turkish speaker). – A panel of five TR-Turkish speakers is consulted to confirm or disconfirm the unconventionality in each case. For the confirmed cases, the judges also provide the conventional equivalents in TR-Turkish. When there is a disagreement among the judges, these constructions are categorised as ‘uncertain’. – In order to investigate the possible Dutch influence, the Dutch equivalents of the unconventional NL-Turkish constructions were established with a panel of five Dutch speakers. Based on the level of schematicity, qualitative analyses of NL-Turkish spoken data reveal two types of unconventional constructions with subject pronouns: fixed expressions and partially schematic constructions. 7.2 Dutch influence on maximally specific constructions The unconventionality in these constructions could be attributed to the translation of lexically fixed Dutch constructions which include a subject pronoun. Since all the lexical items in these constructions are fixed, they are placed on the left hand side of the Specificity Continuum (Figure 1) which hosts ‘maximally specific constructions’. 7.2.1 [I don’t know] construction In the example (8), the NL-Turkish speaker explains how his mother and father differ in their Dutch proficiency. The expression ben ne bil-eyim ‘I what knowopt-1sg’ is identified as unconventional by TR-Turkish speakers. This is probably due to the strong contrast perceived by the TR-Turkish speakers although it is not intended by the NL-Turkish speaker. (8) NL-TR: Annem çok kötü # babam biraz daha iyi ama onunki de süper değil # yani kendini anlatacak kadar biliyo ama çok zor durumları tabi anlatamaz kendi yani. ‘My mother is really bad # my father is a bit better but his (Dutch) is also not that super # I mean he could explain himself but he cannot talk about difficult situations I mean.’ Teknik şey-ler mesela [ben ne bil- eyim] technical thing-pl for.example. I what know-opt. # bil-iyor ama o kadar fazla bil-mi-yor. know-prog.3sg but that quantity much know-neg-prog.3sg ‘Technical things, for example, [I don’t know] # he knows a bit but not that much.’

19

20 A. Seza Doğruöz

Both TR-Turkish and Dutch make use of the fixed ‘I don’t know’ expressions which function as fillers to help the speaker gain time while thinking about what to say in a natural conversation. Dutch makes use of the following expression in the same context: (9) NL: Ik weet het niet.2 I know.pres it not ‘I don’t know.’

TR-Turkish has three possible translations for the same expression: (10) a. b. c.

Ne bil-e-yim. what know-opt-1sg ‘I don’t know.’ Ne bileyim ben. what know-opt-1sg I ‘I don’t know.’ Ben ne bil-e-yim. I what know-opt-1sg ‘I don’t know.’

When they are used as fillers, forms (9) and (10a) are the best matches in Dutch and TR-Turkish, respectively. Note that the TR-Turkish form (10a) does not include a subject pronoun. The meaning difference between (10a) and (10b) is not very clear for the TR-Turkish judges3. However, there is a marked difference of frequency according to the Google search engine. (10a) has many more hits (i.e. is used more frequently) than (10b) (405.000 vs. 42.500). Example (10c), on the other hand, conveys a completely different meaning. It is used as a rhetorical question and could be translated into English as in example (11). It is also the least frequent form (9.830) according to the Google search engine. (11) A: Where are my socks? B: How can I know? (with a bit of annoyance)

Since the TR-Turkish speakers prefer using Ne bil-e-yim ‘what know-opt-1sg’ instead of Ben ne bil-e-yim ‘I what know-opt-1sg’ in example (8), it is possible to

2.  In spoken Dutch, the subject is often reduced to /k/ and the object to /t/, which reflects the unit status of the expression. 3.  However, when they were detecting unconventionality in NL-Turkish, all of them pointed out the presence of the subject pronoun rather than its position. Therefore, we take the null subject form (10a) as the conventional form in this context.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

claim a Dutch influence on the NL-Turkish version. This construction was used three times by one NL-Turkish speaker. 7.2.2 [As far as I know] construction In example (12), the NL-Turkish speaker explains why a local product (leblebi ‘roasted chickpeas’) from Çorum (a city in Central Anatolia) is popular all over Turkey. The unconventionality stems from the use of the subject pronoun which conveys an unintended contrastive meaning for TR-Turkish speakers in the benim bildiğim kadarıyla ‘as far as I know’ construction. (12) NL-TR: Sonuçta Türkiye’de öyle bir kültür var. Her vilayetin kendine göre has bir ürünü olur falan # yani Çoruma da leblebiyi yakıştırmışlar. ‘As a matter of fact there is such a culture in Turkey. Every city has a product of its own. They have assigned to Çorum the leblebi.’ Ama [ben-im bil-diğ-im kadar-ıy-la] but I-poss.1sg know-nom-1sg. amount-poss.3sg-com. Çorum-dan Çorum-abl daha iyi daha fazla üret-il-en yer-ler var. more good more plenty produce-pass-nom place-pl exist.pres ‘But as far as I know there are other places that produce leblebi more than Çorum.’

The fixed expression ‘as far as I know’ has two forms in Turkish, one with and one without a subject pronoun. A Google search for the two expressions reveals that (12b) is much more frequent than (12a) (684.000 vs. 77.700 hits). (12) a. b.

ben-im bil-diğ-im kadar-ı-yla I-poss.1sg know-nom-poss.1sg amount-poss.3sg-com ‘as far as I know’ bil-diğ-im kadar-ı-yla know-nom-poss.1sg amount-poss.3sg-com ‘as far as (I) know’

TR-Turkish judges assign unconventionality to the use of subject pronoun in example (12). However, it is not easy to identify a semantic difference between (12a) and (12b). According to one judge, ben-im bil-diğ-im kadar-ı-yla ‘I-poss.1sg know-nom-poss.1sg amount-poss.3sg-com’ sounds as if the speaker has more doubts about the truth value of what s/he will say next, indicating that there might be some counterevidence.

21

22

A. Seza Doğruöz

The use of the subject pronoun in the Dutch version (13) of the ‘as far as I know’ construction suggests a Dutch influence on the NL-Turkish version. This construction is used twice by the same NL-Turkish speaker. (13) voor zo ver ik weet for so far I know.pres ‘as far as I know’

Both I don’t know and as far as I know constructions are used by NL-Turkish and TR-Turkish speakers to take a distance from the topic of discussion in an indirect way. Therefore, they may be considered to function as evidentials (Johanson 2000). In addition to the morphological evidential markers (e.g. -mIş) on Turkish verbs, further research is needed to describe and classify the indirect constructions with lexically fixed and non-fixed items in NL-Turkish and TR-Turkish. 7.3 Dutch influence on partially schematic constructions In addition to the maximally specific constructions, subject pronouns could also be parts of partially schematic constructions. The examples below illustrate the Dutch influence on partially schematic constructions with subject pronouns that sound unconventional to TR-Turkish speakers. 7.3.1 Subordinate constructions with subject pronouns In example (14), the NL-Turkish speaker complains about the programs on Turkish TV channels. He detests watching them because of the violence they show. (14) NL-TR: [Ben bak-tığ-ım zaman] moral-im bozul-uyor. I look-nom-1sg time spirit-poss.1sg damage-prog-3sg ‘When I look at (those TV shows), I get depressed.’ TR-TR: [Bak-tığ-ım zaman] moral-im bozul-uyor. look-nom-1sg time spirit-poss.1sg damage-prog-3sg ‘When I look at (those TV shows), I get depressed.’ NL: [Als ik ernaar kijk] word ik depressief when I there look.pres become.pres I depressive ‘If I look at it, I become depressed.’

The use of the subject pronoun in ben bak-tığ-ım zaman ‘I look-nom-1sg’ construction sounds unconventional to TR-Turkish speakers since it conveys a contrastive meaning (e.g. ‘I get depressed but others do not’) although the NL-Turkish speaker does not intend it. In the same context, TR-Turkish speakers would not use a subject pronoun since there is no need to change the topic or signal contrastive information.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

The use of subject pronoun in NL-Turkish could be explained through copying the [Als S V] ‘when S V’ partially schematic construction from Dutch. The subordination marker als ‘when’ is the fixed lexical item whereas the subject and verb slots are open. This construction was used once in the NL-Turkish data. The second example in this category also involves the copying of a partially schematic Dutch construction. The subordinate clause [dat S V] hosts a subordination marker (i.e. dat) as a lexically fixed item but the subject and the verb slots are free. In example (15), the NL-Turkish speaker explains the difficulties of living in Amsterdam as a big city. (15) NL-Turkish: Ora-nın bir şey-i var [o-nu ben sev-me-m]. there-gen one thing-poss.3sg exist that-acc I like-neg-1sg ‘There is something there that I do not like.’ TR-Turkish: Ora-nın sev-me-diğ-im bir şey-i var. there-gen like-neg-rel-1sg one thing-poss.3sg exist ‘There is something about that place that I don’t like.’ NL: Er is daar iets wat ik niet leuk vind. there is there something that I not nice find.pres ‘There is something there that I don’t like.’

It seems that the NL-Turkish speaker literally translated the Dutch equivalent [Er is ADV NP wat S ADJ V] ‘There is ADV NP what S ADJ V’ into Turkish with two finite clauses, one of which makes use of a subject pronoun. This construction sounds unconventional to TR-Turkish ears since the use of an overt subject pronoun (ben ‘I’) conveys an unintended contrastive meaning. To convey a noncontrastive meaning, TR-Turkish speakers would use a finite clause and a copula clause instead of a relative clause which would combine both clauses with a nonfinite verb (sev-me-diğ-im ‘like-neg-rel-1sg’). This unconventional construction was used once in the data set. 7.3.2 Left Dislocation: addition of a new construction NL-Turkish also makes use of the Left Dislocation construction, which requires the placement of a lexical item in the left margin of the clause, in a syntactically independent way. The detached part (usually an NP) introduces the topic of the following clause and its referent (the lexically fixed item dat/die ‘that/those’) surfaces as the subject of that clause (Lambrecht 1994; Manetta 2007). In example (16), a NL-Turkish speaker (who is a local politician) explains the difficulty of speaking with the Turkish press. According to him, the journalists would like him to speak only in Turkish.

23

24

A. Seza Doğruöz

(16) NL-Turkish: Gazeteci-ler onlar sadece Türkçe konuş-mak isti-yor-lar. journalist-pl they only Türkçe speak-nom want-prog-3pl ‘The journalists, they only want to speak in Turkish.’ TR-Turkish (a): Gazeteci-ler sadece Türkçe konuş-mak isti-yor-lar. journalist-pl only Turkish speak-nom want-prog-3pl ‘Journalists only want to speak Turkish.’ TR-Turkish (b): Gazeteci-ler-i bil-iyor-sun sadece Türkçe konuş-mak journalist-pl-acc know-prog-2sg only Turkish speak-nom isti-yor-lar. want- prog-3pl ‘You know the journalists, they only want to speak Turkish.’ NL: Journalist-en, die willen alleen maar in het Turks praten. journalist-pl they want-3pl only but in the Turkish talk.inf. ‘The journalists, they only want to speak in Turkish.’ Instead of the Left Dislocation construction, TR-Turkish speakers would use the following constructions in the same context: – A regular finite clause with the topic surfacing as a nominal subject (TRTurkish (a)) – Two finite clauses (TR-Turkish (b)) Although this partially schematic [NP dat/die V] ‘SNP dat/die V’ construction is quite common in Dutch, it does not exist in TR-Turkish. This construction was used four times by the same NL-Turkish speaker in the data. 7.3.3 [Do you mean?] construction: an uncertain case of unconventionality In example (17), the interviewer and the NL-Turkish speaker discuss the origins of Turkish families in the Netherlands. The NL-Turkish speaker asks a clarification question [X diyosun sen?] ‘Do you mean X?’ in order to clarify whether the interviewer would like to know where his grandmother comes from in Turkey.



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

(17) Interviewer: Sizin ya da ailenizin Türkiye’den geldiği yer? ‘Where do you or your family come from in Turkey?’ NL-Turkish: Nevşehir # Yani babaannem-in falan mi Nevşehir (city) # That.is.to.say grandmother-gen etc. qp di-yo4-sun sen? say-prog-2sg you ‘Nevşehir. You mean like the one of my grandmother’s or what?’ TR-Turkish: Nevşehir # Yani babaannem-in falan mı Nevşehir # That.is.to.say grandmother-gen etc. qp di-yo-sun? say-prog-2sg ‘Nevşehir. (You) mean like the one of my grandmother’s or what?’ NL: Bedoel je die van mijn oma of zo? Mean you one of my grandmother or what ‘You mean like the one of my grandmother’s or what?’

When asked for their judgments, TR-Turkish speakers identified the partially schematic NL-Turkish construction (17) as unconventional since it implies some criticism (and/or disapproval) although it is not intended. Example (18) illustrates the TR-Turkish version of this construction when the criticism is actually intended. (18) Bu-na çorba mı di-yo-sun sen? this-dat soup qp say-prog-2sg you ‘You call this soup?’

This unconventional construction was only observed five times in NL-Turkish corpus and all of them were used by the same speaker. Since the partially schematic [Bedoel je Sub. Clause] ‘Mean you Sub. Clause’ is literally translated from Dutch into NL-Turkish, it is rather tempting to assume a Dutch influence in the example (17). However, not all TR-Turkish judges agreed on the unconventionality of this NL-Turkish construction. For example, one judge explained his doubts as “it is not completely wrong to use a subject pronoun, but there is still something that does not sound right to me”. Perhaps, it is just the personal style of the speaker to address the other speaker with a subject pronoun and create intimacy. It may be relevant to know that this informant is a politician and often uses informal terms of address to establish intimacy with the potential voters in the Turkish community.

4.  Morphemically, -yor is the progressive marker, but the /r/ sound at the end of this morpheme is usually devoiced in spoken Turkish.

25

26 A. Seza Doğruöz

7.3.4 Yes/No question constructions When answering simple Yes/No questions (a typical topic-maintenance situation), subject pronouns are not normally used in TR-Turkish unless there is something contrastive in the context. In example (19), this convention is violated. In this part of the conversation, the NL-Turkish speaker describes what his uncles do for a living. (19) Interviewer: Çalış-ıyor mu onlar? work-prog qp they ‘Do they work?’ NL-Turkish: Onlar çalış-ıyor. they work-prog.3pl ‘They work.’ TR-Turkish: Evet/çalış-ıyor. yes/work-prog.3sg NL: Ja/Ze werk-en. yes/they work.pres-3pl

In this example, a TR-Turkish speaker would only reply with the verb (i.e. çalışmak ‘work’) without an overt subject expression. In Dutch, a subject pronoun is also used in such contexts. This construction occurred once in the NL-Turkish data. Although there are no lexically fixed items, the template of this construction is somewhat fixed and the verb root (i.e. çalış “work”) has to be the same in both turns of the question-answer pair in TR-Turkish as illustrated in example (20). Therefore, it is considered as a partially schematic construction. (20) Question: Answer:

[V1 S QP]? [V1]

In sum, 17 cases of unconventional constructions with subject pronouns were found in the NL-Turkish corpus. Figure (2) illustrates the placement of these constructions on the Specificity Continuum (Doğruöz & Backus 2009) based on their level of schematicity.

Max.specific Fixed Expressions (5)

Partially Schematic

Max.schematic

Subordinate Constructions (2) Yes/No Questions (1) Left Dislocation Construction (4) [Do you mean X?] Construction (5)

Figure 2.  Unconventional subject pronoun constructions on the Specificity Continuum



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

So far, subject pronoun use seems to be highly maintained in NL-Turkish just like the word order (Doğruöz & Backus 2007). Dutch influence is mostly visible through fixed and partially schematic subject pronoun constructions that are literally translated from Dutch into NL-Turkish. 8. Discussion The innovative way of analyzing subject pronouns as part of whole constructions gave us the opportunity to reveal some degree of Dutch influence on NL-Turkish which would not have been possible through analyzing subject pronouns as individual lexical items and based only on their syntactic role. Although this study sets an example for analyzing contact effects through constructions, it is too early to claim a systematic change for NL-Turkish constructions with subject pronouns. First of all, in most cases, the specific type of unconventionality is unique to only one of the NL-Turkish speakers in the corpus. Secondly, the type and token frequency is very low. As Rostila (2006) points out, syntactic changes require a wide range of constructions to adopt the new form and increase their frequency of use. In case of Turkish-Dutch contact, there should be a whole range of unconventional constructions that adopt the Dutch-like subject pronoun use and they should increase their frequency in comparison to the conventional (i.e. TRTurkish) constructions. Unconventional constructions with subject pronouns in the data meet none of these conditions in NL-Turkish so far. In all conversations, the interviewer and the informant did not know each other very well. This may have had an effect on the types of subject pronoun constructions that were used. It is possible that different types of subject pronoun constructions would have been observed in less formal settings and in different contexts. In the case of unconventional constructions, direct evidence for Dutch influence is often hard to come by. Resemblance to the Dutch equivalent is only circumstantial evidence. Analysis of a larger data set with more diverse groups of informants (e.g. different generations, backgrounds) could lead to firmer empirical conclusions. There is also a need for experimental evidence to confirm/disconfirm the conventionalization of unconventional subject pronoun constructions within the NLTurkish speech community. Information on this aspect would shed some light on how far the putative changes are spread within the NL-Turkish community.

27

28

A. Seza Doğruöz

9. Conclusion and further research This study investigates the on-going changes in NL-Turkish subject pronoun use through quantitative and qualitative methods. The comparative analyses of NLTurkish and TR-Turkish spoken corpora do not reveal statistically significant differences in terms of subject pronoun frequency. However, the qualitative analyses of the NL-Turkish corpus reveal some contexts where subject pronoun sounds unconventional to TR-Turkish speakers. In these cases, the subject pronouns do not lead to unconventionality on their own as individual lexical items but as parts of larger constructions which are copied as chunks from Dutch. Figure 2 illustrates these constructions based on their fixed and non-fixed elements on the Specificity Continuum. If it is possible to reveal a Dutch influence on NL-Turkish constructions, is it possible to classify TR-Turkish constructions with subject pronouns based on their levels of schematicity? If subject pronouns are indeed parts of fixed and partially schematic constructions in TR-Turkish, the use or non-use of a subject pronoun may not be optional (see also Öztürk 2002). In that case, it is possible to hypothesize that exactly similar utterances with and without subject pronouns belong to different constructions with unique forms and meanings. In line with this argument, Ono & Thompson (2003) reveal that subject pronouns are parts of maximally specific and partially schematic constructions in Japanese as well. To conclude, we need more cross-linguistic comparisons (e.g. Spanish-English contact) to outline the inventories of subject pronoun constructions in different languages and to explain which ones are attractive in contact settings. It would particularly be interesting to explore possible changes in contact situations concerning two or more pro-drop languages (e.g. contact among Turkic languages).

References Aarsen, J. (1994). Relating two events in two languages: Acquisition of cohesive devices by TurkishDutch bilingual children at school age, Ph.D. thesis. Tilburg: Tilburg University Press. Aikhenvald, A. Y., & Dixon, R. M.W. (2001). Areal diffusion and genetic inheritance: Problems in comparative linguistics. Oxford: Oxford University Press. Akıncı, M.A., & Backus, A. (2005). The structure and the role of code-switching in TurkishFrench conversations. In J.N. Jorgensen, S. Talayman & C. Dabelsteen (Eds.), Languaging and language practices (pp. 156–166). Copenhagen: University of Copenhagen. Backus, A. (2004). Convergence as a mechanism of language change. Bilingualism: Language and Cognition, 7, 179–181. ​doi: 10.1017/S1366728904001567 Backus, A. (2005). Codeswitching and language change: One thing leads to another? International Journal of Bilingualism, 9, 307–340. ​doi: 10.1177/13670069050090030101



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

Bolonyai, A. 2000. Elective affinities: Language contact in the abstract lexicon and its structural consequences. International Journal of Bilingualism, 4, 81–106.   d ​ oi: 10.1177/13670069000040010601 Boeschoten, H. E. (1994). Second language influence on first language acquisition: Turkish children in Germany. In G. Extra & L. Verhoeven (Eds.), The cross-linguistic study of bilingual development (pp. 253–263). Amsterdam: North Holland. Bullock, B., & Toribio, A.J. (2004). Introduction: Convergence as an emergent property in bilingual speech. Bilingualism: Language and Cognition, 8, 303–320. Bybee, J. (2006). From usage to grammar: The mind’s response to repetition. Language, 82, 711–733. ​doi: 10.1353/lan.2006.0186 Bybee, J. L., & Beckner, C. (2010). Usage-based theory. In B. Heine & H. Narrog (Eds.), The Oxford handbook of linguistic analysis (pp. 827–856). Oxford: Oxford University Press. Campbell, L. (1993). On proposed universals of grammatical borrowing. In H. Aartsen & R. J. Jeffers (Eds.), Historical linguistics 1989: Papers from the 9th International Conference on Historical Linguistics (pp. 91–109). Amsterdam: John Benjamins.   ​doi: 10.1075/cilt.106.08cam CBS. (2011). Central Bureau voor Statistiek. http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=37296ned&D1=a&D2=0,10,20,30,40,50,%28l-1%29l&HD=120315-1354&HDR=G1&STB=T Croft, W. (2000). Explaining language change: An evolutionary approach. Harlow (Essex): Longman. Croft, W. (2001). Radical Construction Grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press. ​doi: 10.1093/acprof:oso/9780198299554.001.0001 Croft, W. (2006). The relevance of an evolutionary model to historical linguistics. In O. Nedergaard Thomsen (Ed.), Competing models of linguistic change: Evolution and beyond (pp. 91–133). Amsterdam: John Benjamins. ​doi: 10.1075/cilt.279.08cro Croft, W., & Cruse, A. (2004). Cognitive linguistics. Cambridge: Cambridge University Press. ​ doi: 10.1017/CBO9780511803864

Davidson, B. (1996). ‘Pragmatic weight’ and Spanish subject pronouns: The pragmatic and discourse uses of ‘tu’ and ‘yo’ in spoken Madrid Spanish. Journal of Pragmatics, 26, 543–565. ​ doi: 10.1016/0378-2166(95)00063-1

Doğruöz, A. S. (2007). Synchronic variation and diachronic change in Dutch Turkish: A corpus based analysis, Ph.D. Thesis. Tilburg: Tilburg University Press. Doğruöz, A. S., & Backus, A. (2007). Postverbal elements in immigrant Turkish: Evidence of Change? International Journal of Bilingualism, 11(2), 185–220.   d ​ oi: 10.1177/13670069070110020301 Doğruöz, A. S., & Backus, A. (2009). Innovative constructions in Dutch Turkish: An assessment of on-going contact induced change. Bilingualism: Language and Cognition, 12(1), 41–63. ​ doi: 10.1017/S1366728908003441

Enç, M. (1986). Topic switching and pronominal subjects in Turkish. In D. I. Slobin & K. Zimmer (Eds.), Studies in Turkish linguistics (pp. 195–209). Amsterdam: John Benjamins. ​ doi: 10.1075/tsl.8.11enc

Erguvanlı-Taylan, E. (1979). The function of word order in Turkish grammar, Ph.D. Thesis, UCLA. Erman, B., & Warren, B. (2000). The idiom principle and the open choice principle. Text, 20(1), 29–62.

29

30

A. Seza Doğruöz Evans, N., & Levinson, S. (2009). The myth of language universals: Language diversity and its importance for cognitive science. Behavioral and Brain Sciences, 32, 429–492.   d ​ oi: 10.1017/S0140525X0999094X Flores-Ferrán, N. (2004). Spanish subject personal pronoun use in New York City Puerto Ricans: Can we rest the case of English contact? Language Variation and Change, 16, 49–73.   d ​ oi: 10.1017/S0954394504161048 Fried, M., & Östman, J-O. (2004). Construction Grammar: A thumbnail sketch. In M. Fried & J.-O. Östman (Eds.), Construction Grammar in a cross-language perspective (pp. 11–87). Amsterdam: John Benjamins. ​doi: 10.1075/cal.2.02fri Geeraerts, D., Kristiansen, G., & Peirsman, Y. (2010). Introduction. In D. Geeraerts, G. Kristiansen & Y. Peirsman (Eds.), Advances in cognitive sociolinguistics (pp. 1–19). Berlin & New York: Mouton De Gruyter. ​doi: 10.1515/9783110226461.1 Goldberg, A. (1995). Constructions. A construction grammar approach to argument structure. Chicago: The Chicago University Press. Goldberg, A. (2006). Constructions at work: The nature of generalizations in language. Oxford: Oxford University Press. Göksel, A., & Özsoy, A.S. (2003). dA: A focus/topic associated clitic in Turkish. Lingua, 113, 1143–1167. ​doi: 10.1016/S0024-3841(03)00016-0 Gürel, A. (2006). L2 acquisition of pragmatic and syntactic constraints in the use of overt and null subject pronouns. In R. Slabakova, S. A. Montrul & P. Prevost (Eds.), Inquiries in linguistic development: In honor of Lydia White (pp. 259–282). Amsterdam: John Benjamins. ​ doi: 10.1075/z.133.15gur

Heine, B., & Kuteva, T. (2005). Language contact and grammatical change. Cambridge: Cambridge University Press. ​doi: 10.1017/CBO9780511614132 Hulk, A., & Müller, N. (2000). Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition, 3, 227–244.   d ​ oi: 10.1017/S1366728900000353 Johanson, L. (2000). Turkic indirectives. In L. Johanson & B. Utas (Eds.), Evidentials: Turkic, Iranian and neighbouring languages (pp. 61–89). Berlin & New York: Mouton de Gruyter Kay, P., & Michaelis, L.A. (2012). Constructional meaning and compositionality. In C. Maienborn, K. von Heusinger & P. Portner (Eds.), Semantics: An international handbook of natural language meaning, Vol. 3. (pp. 2271–2296). Berlin: de Gruyter. Kerslake, C. (1987). Noun phrase deletion and pronominalization in Turkish. In H. E. Boeschoten & L. T. Verhoeven (Eds.), Studies on modern Turkish: Proceedings of Third Conference on Turkish Linguistics (pp. 91–103). Tilburg: Tilburg University Press. Kristiansen, G., & Dirven, R. (2008). Introduction. Cognitive sociolinguistics: Rationale, methods and scope. In G. Kristiansen & R. Dirven (Eds.), Cognitive sociolinguistics: Language variation, cultural models, social systems (pp. 1–21). Berlin & New York: Mouton de Gruyter. ​ doi: 10.1515/9783110199154.0.1

Lambrecht, K. (1994). Information structure and sentence form: Topic, focus and mental representations of discourse referents. Cambridge (UK): Cambridge University Press.   d ​ oi: 10.1017/CBO9780511620607 Langacker, R. W. (1987). Foundations of cognitive grammar. Volume I: Theoretical prerequisites. Stanford: Stanford University Press. Lapidus, N., & Otheguy, R. (2005). Contact induced change? Overt nonspecific Ellos in Spanish in New York. In L. Sayahi & M. Westmoreland (Eds.), Selected Proceedings of the Second



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

Workshop on Spanish Sociolinguistics (pp. 67–75). Somerville (MA): Cascadilla Proceedings Project. Leino, J., & Östman, J-O. (2005). Constructions and variability. In M. Fried & H. C. Boas (Eds.), Grammatical constructions: Back to the roots (pp. 191–212). Amsterdam: John Benjamins. ​ doi: 10.1075/cal.4.12lei

Manetta, E. (2007). Unexpected left dislocation: An English corpus study. Journal of Pragmatics, 39, 1029–1035. ​doi: 10.1016/j.pragma.2007.01.003 Milroy, J. (2003). On the role of speaker in language change. In R. Hickey (Ed.), Motives for language change (pp. 143–157). Cambridge: Cambridge University Press. Montrul, S. (2004). Subject and object expression in Spanish heritage speakers: A case of morphosyntactic convergence. Bilingualism: Language and Cognition, 7(2), 125–142.   ​doi: 10.1017/S1366728904001464 Oliveira, M. (2000). The pronominal subject in Italian and Brazilian Portugese. In M. A. Kato & E. V. Negrão (Eds.), Brazilian Portugese and the null subject parameter (pp. 37–53). Frankfurt am Main: Vervuert Verlag, Ono, T., & Thompson, S. A. (2003). Japanese (w)atashi/ore/boku ‘I’: They’re not just pronouns. Cognitive Linguistics, 14, 321–347. ​doi: 10.1515/cogl.2003.013 Özcan, F. H, Keçik, I., Topbaş, S., & Konrot, A. (2000). A comparative study in pronominal use in the discourse of monolingual Turkish speaking and bilingual Turkish-Danish speaking children. In A. Holmen & J. N. Jorgensen (Eds.), Det er conversation 801 değil mi? Perspectives on the bilingualism of Turkish speaking children and adolescents in North Western Europe (pp. 101–121). Copenhagen: Denmarks Larerhojskoles Reproduktionafdeling. Özsoy, A. S. (1987). The null subject parameter and Turkish. In Hendrik E. Boeschoten & Ludo T. Verhoeven (Eds.), Studies on modern Turkish: Proceedings of The Third Conference on Turkish linguistics (pp. 82–90). Tilburg: Tilburg University Press. Öztürk, B. (2002). Turkish as a non-pro-drop language. In E. Erguvanli-Taylan (Ed.), The verb in Turkish (pp. 239–259). Amsterdam: John Benjamins. ​doi: 10.1075/la.44.10ozt Paradis, J., & Navarro, S. (2003). Subject realization and crosslinguistic interference in the bilingual acquisition of Spanish and English: What is the role of input? Journal of Child Language, 30, 371–393. ​doi: 10.1017/S0305000903005609 Pfaff, C. W. (1992). Turkish in contact with German: Language maintenance and loss among immigrant children in West Berlin. International Journal of Sociology of Language, 90, 97–129. Pease-Alvarez, L., Hakuta, K., & Bayley, R. (1996). Spanish proficiency and language use in a California Mexico community. Southwest Journal of Linguistics, 15(1/2), 137–151. Polinsky, M. (1995). Cross-linguistic parallels in language loss. Southwest Journal of Linguistics, 14(1/2), 87–123. Rehbein, J. (2001). Turkish in European societies. Lingua E Stile, XXXVI, 317–334. Rothman, J. (2009). Pragmatic deficits with syntactic consequences ? L2 pronominal subjects and syntax-pragmatics interface. Journal of Pragmatics, 41, 951–973.   d ​ oi: 10.1016/j.pragma.2008.07.007 Rostila, J. (2006). Storage as a way to grammaticalization. Constructions, 1, 1–59. Sanchez, L. (2004). Functional convergence in the tense, evidentiality and aspectual systems of Quechua Spanish bilinguals. Bilingualism: Language and Cognition, 7, 147–162.   d ​ oi: 10.1017/S136672890400149X Schaufeli, A. (1991). Turkish in an immigrant setting: A comparative study of the first language of monolingual and bilingual Turkish children, Ph.D. Thesis, University of Amsterdam.

31

32

A. Seza Doğruöz Schmitt, E. (2000). Overt and covert codeswitching in immigrant children from Russia. International Journal of Bilingualism, 4, 9–28. ​doi: 10.1177/13670069000040010201 Silva-Corvalán, C. (1994). Language contact and change: Spanish in Los Angeles. Oxford: Clarendon Press. Sorace, A., & Filiaci, F. (2006). Anaphora resolution in near-native speakers of Italian. Second Language Research, 22(3), 339–368. ​doi: 10.1191/0267658306sr271oa Thomason, S. G., & Kaufman, T. (1988). Language contact, creolization and genetic linguistics. Berkeley/Los Angeles: University of California Press. Thomason, S. G. (2001). Language contact: An introduction. Washington, D.C.: Georgetown University Press. Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. Cambridge & London: Harvard University Press. Toribio, A. J. (2004). Convergence of as an optimizing strategy in bilingual speech: Evidence from code-switching. Bilingualism: Language and Cognition, 7, 165–173.   ​doi: 10.1017/S1366728904001476 Tsimpli, I. M., Sorace, A., Heycock, C., & Filiaci, F. (2004). First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism, 8(3), 257–277. ​doi: 10.1177/13670069040080030601 Turan, Ü. D. (1996). Null versus overt subjects in Turkish discourse: A centering analysis, Ph.D. Thesis, University of Pennsylvania. Weinreich, U. (1953). Languages in contact: Findings and problems. The Hague: Mouton. Yağmur, K., & van de Vijver, F. (2012). Acculturation and language orientations of Turkish immigrants in Australia, France, Germany, and the Netherlands. Journal of Cross-Cultural Psychology, 43(7), 1–21. ​doi: 10.1177/0022022111420145

Appendix A.  Discourse-marker-like elements Discourse-marker-like elements

Example from corpus

bil-mi-yor-um know-neg-prog-1sg

Bilmiyorum# sen nasıl buluyorsun? ‘I don’t know# how do you find it?’

bil-iyo mu-sun know-prog qp-2sg

Sanki ulusal bi şey biliyo musun. ‘It is like something national you know.’

bak-a-lım look-opt-1pl

Bakalım artık öğretmenlere sığınıyoruz. ‘We will see we ask for the gratitude of teachers.’

bak-iy-im look-opt-1sg

Ehm bakiyim beş yaşında galiba. ‘Ehm let me see, she is five years old I guess.’

yani nasıl anlat-ay-ım I.mean how tell-opt-1sg

Yani nasıl anlatayım çok karışık iş. ‘I mean how can I tell, it is very complicated.’

anla-dı-n mı understand-past-2sg qp

Mesela biz burdayız anladın mı ama onlar orda. ‘For example we are here you see but they are there’



On the borrowability of subject pronoun constructions in Turkish–Dutch contact

Appendix B.  Fixed expressions Allah koru-sun God protect-imp.3sg.

‘may God forbid’

Allah kısmet eder-se God possible make-cond.3sg.

‘may God permit’

Allah-a şükür-ler ol-sun God-dat thank-pl cop-imp.3sg.

‘we thank Allah’

Ben teşekkür eder-im I thank make-1sg.

‘I thank you’ (as a reply to ‘thank you’)

Ne yalan söyli-y-im why lie say-opt-1sg.

‘to tell the truth’

Dur bak-al-ım stop look-opt-1sg.

‘let’s see’

Ben-den geç-ti artık I-abl pass-past anyway.

‘I am too old for these things’

Zaman göster-ecek time show-fut.3sg.

‘time will show’

Ne yap-al-ım what do-opt-1pl.

‘we can’t do anything else’

Hayırlısı ol-sun good become.imp.3sg

‘hopefully good will happen’

Gerisini sen hesap et rest you calculate do.

‘imagine the rest’

Ben bil-di-m bil-eli I know-past-1sg know-ger

‘since the times I know/am aware’

O da ayrı mesele that also separate issue.

‘that is another story’

Rica ed-er-im please do-pres-1sg.

‘you are welcome.’

Hiç alakam yok any attention-poss.1sg exist-not

‘I don’t have anything to do with that’

33

On the universality of frames Evidence from English-to-Japanese translation Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore University of California, Berkeley and International Computer Science Institute

This paper investigates the cross-linguistic applicability of the concept of frame as developed in the Berkeley FrameNet project. We examine whether the frames created for the annotation of English texts can also function as a tool for the assessment of the accuracy of English-to-Japanese translations. If the semantic structure of a source text is analyzed in terms of the frames evoked by its constituent words and the ways in which the elements of those frames are realized, then those frames, their constituent elements, and their interconnections must somehow be present in the translation. The paper concentrates on passages involving causation, as causal relationships are considered by many to exhibit the most salient differences in rhetorical preference between the two languages. Keywords: causation, Frame Semantics, FrameNet, noun-centered vs. verbcentered typology, parallel-text corpora, rhetorical structure, topic-worthiness, transitivity, translation assessment

1. Introduction This paper investigates the cross-linguistic applicability of the concept of frame as developed for English in the FrameNet project (Fillmore & Baker 2010; Fillmore et al. 2003). In particular, we will examine whether or not the frames created for the annotation of English texts can also function as an assessment tool for the accuracy of English-to-Japanese translation. The hypothesis to be tested is that if the semantic structure of an original English text is carefully analyzed in terms of the frames evoked by its constituent words and the ways in which the elements of

doi 10.1075/bct.82.03has 2016 © John Benjamins Publishing Company

36

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

those frames are realized, such frames and frame elements, as well as their interconnections, must somehow be shared by the translation.1 FrameNet is an online lexical resource based on the principles of Frame Semantics (Fillmore 1982, 1985, 1994; Fillmore & Atkins 1992). A frame is a schematic understanding of types of events, situations, individuals, and things, including the participants, props, parts, and their relations to each other and to the larger situation. Words are understood with a particular frame as background, or, in FrameNet terms, words evoke frames. The FrameNet database groups words with the same background knowledge into frames, and defines in prose these frames and the parts of the frame (the props, participants, etc.; they are called the frame elements in the FrameNet project). For instance, the words vend, sell, vendor, seller, auction, and retail are grouped in the Commerce_sell frame (frame names are mono-spaced), as they all have as background a commercial transaction and take the point of view of the seller of goods. Note that a single frame may contain both nouns and verbs (or any part of speech). The frames are arranged in a hierarchy, connected by several types of relationships. The Commerce_sell frame is a more specific type of the Giving frame,2 so it inherits features from the Giving frame. At the same time, it is a perspective on a more general Commercial_scenario frame. Another perspective on that general frame is Commerce_buy, which includes the words buy and purchase. Alongside inheritance and perspective, we expect that the relations causative_of and inchoative_of will play a major role in considering English-to-Japanese translations. To give two examples, Endangering (imperil, endanger) is the causative counterpart of Being_at_risk (safe, at risk), and Death (die, pass away) is the inchoative counterpart of the stative Dead_or_alive (dead, deceased, alive, living). We expect frames will aid greatly in establishing relations between word senses in different languages, and in understanding one way in which translations may differ from originals by evoking different, but related, frames.

1.  Empirically, the acceptable ranges of translation accuracy vary significantly by the type of text and the initiator’s purpose for translating the text (see Hasegawa 2011 for further discussion on this topic). As explained below, we adhere in this paper to scientific texts for which factual accuracy is normally expected. 2.  There are many interesting issues to investigate in the cross-linguistic applicability of frames. For example, the verb give, the archetypal member of the English Giving frame, does not entail a transfer of ownership. That is, Give it to me can mean simply ‘hand it over to me’, where the speaker does not intend to keep the item indefinitely. By contrast, the Japanese verbs ageru and kureru, which are most commonly used as translational equivalents of give, necessarily entail a transfer of ownership. The FrameNet database can serve as a precious resource for this kind of subtle yet significant cross-linguistic difference.



On the universality of frames

It is widely agreed that translation must be regarded as an art, not a science (Newmark 1981: 137), and that translations, especially of passages anchored in the author’s culture, are bound to lose some meanings due to different expectations and experience on the part of the target-text readers. In order to minimize the effect of cultural differences for our investigation, we have examined a parallel-text corpus consisting of selected passages from the Scientific American magazine and their Japanese translations appearing in the Nikkei Saiensu magazine. This decision is based on the assumption that scientific writing is a genre in which considerations of factual specificity and conceptual clarity are mandatory, while those of aesthetic elegance and cultural nuances are normally less relevant. Therefore, we expect scientific translations to be a base-line testing ground for the utility of FrameNet tools: if the frames needed for scientific texts turn out not to be crosslinguistically applicable, there will be little reason to expect FrameNet frames to serve as a tertium comparationis (i.e. a common platform of comparison) for texts in such areas as esthetics, social structure, religion, or art. Influenced by common formulations of salient differences between English and Japanese, which will be explained in detail in Section 2, we concentrate on passages involving causation. In short, many Japanese linguists see expressions of causal relationships as a context for one of the most salient differences in rhetorical preference, or perhaps even differences in general cognitive tendencies, between the speakers of English and Japanese. The organization of this paper is as follows: Section 2 illustrates the rhetorical differences in causal expressions (including transitivity), attempts to reformulate as empirical hypotheses the cognitively deep typological characteristics claimed to differentiate the two languages, and explores ways of testing these hypotheses with data taken from our parallel texts. Section 3 analyzes source-and-translation pairs exhibiting the kinds of differences discussed in Section 2, using descriptive notions developed in FrameNet. Section 4 is devoted to a discussion of our findings and their implications for broader issues in translation. 2. Typological differences in framing causal events In 1930 Japan, a silent movie entitled Nani ga kanojo o soo saseta ka made a sensation and achieved box-office success. (1) Nani ga kanojo o soo saseta ka. what nom she acc so made.to.do q ‘What made her do it?’

37

38

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

This success was reportedly due in significant part to its linguistically eccentric title; it used familiar vocabulary and familiar grammatical structure, but it juxtaposed an abstract subject (nani ‘what’) to a VP headed by a verb with causative morphology, and that was just not possible in normal Japanese. Even today, after decades of noticeable rhetorical-style changes influenced by English, sentences that pile up abstract nouns, such as (2), continue to sound strange or foreign to Japanese ears. (2) Kono jijitsu no ninshiki ga mondai no kaiketsu ni kooken-suru. this fact gen awareness nom problem gen solution dat contribution-do ‘The recognition of this fact will contribute to the resolution of the problem.’

(Even the denominal verb kooken-suru ‘contribute’ permits an analytic gloss as ‘makes a contribution’, adding another abstract noun.) A more idiomatic formulation of the content intended in (2) would be along the lines of (3). (3) Kore ga wakar.eba, mondai wa zutto kaiketsu-shi.yasuku naru. this nom if.understand problem top much solution-do.easily become ‘If we understand this, the problem will become more manageable.’

Many Japanese grammarians have offered descriptions of this type of rhetorical difference between Japanese and English, but their characterizations are sometimes so impressionistic that researchers whose native language is not Japanese find them inscrutable. Among such claims, Ikegami (1988: 9) appears to assert that the different ways of encoding are derived from a deep-seated difference in cognition: in Japanese text, “[a]n individuum is not seen in isolation; it is not clearly separated from what it stands contiguous with. It is merely a part of a larger whole, with which it may become merged to the extent of losing its identity.” Adapting Ikegami’s idea, Maynard (1997: 172) characterizes the same phenomenon as scene-orientation (as in Japanese) vis-à-vis agent-orientation (as in English). She contends, “[f]or Japanese people, the scene of an event as a whole assumes the primary focus of attention (in comparison to English, where the agent is the primary focus).” We find it necessary to examine the phenomena that led to such speculation in less impressionistic ways, in the hope of providing testable hypotheses about differences in rhetorical preferences between the two languages. To this end, we have selected a parallel-text corpus consisting of the first several paragraphs of a number of Scientific American articles published between October 2005 and October 2006 and their Japanese translations in the Nikkei Saiensu magazine. This corpus contains 266 English sentences, mostly multi-clausal, accompanied by translations into Japanese created by professional translators and judged to be reflective of idiomatic Japanese.



On the universality of frames

Let us begin our examination with Seidensticker & Anzai’s (1983) claim that English uses transitive constructions with significantly greater frequency than Japanese. We found that 382 English and Japanese clause-pairs agreed in transitivity (i.e. transitive or intransitive in both languages), whereas 119 clause-pairs did not.3 That is, of all the English clauses that were translated fairly straightforwardly into Japanese, approximately 75% maintained their transitivity, but 25% switched it.4 Of the 119 unmatched clause-pairs, 99 English transitive clauses were translated into Japanese intransitive clauses, e.g. (4):5 (4)

Initially the brain can function normally as it loses dopaminergic neurons. Doopamin-sadoosei nyuuron ga shooshitsu-shite mo toosho wa dopaminergic neuron nom disappear even initially top seijoo.ni kinoo-suru. normally function Backtranslation: ‘Even if dopaminergic neurons disappear, [the brain] functions normally at first.’

By contrast, only 20 English intransitive clauses were translated into Japanese transitive clauses, e.g. (5): (5)

Gleevec [a drug] has been a huge clinical success. Guribekku wa rinshoo no ba de ooki.na seikoo o osameta. Gleevec top clinical gen place in huge success acc accomplished Backtranslation: ‘Gleevec accomplished a huge success in the area of clinical trials.’

The ratio of “English transitive into Japanese intransitive” vs. “English intransitive into Japanese transitive” is approximately 5:1. Therefore, as Seidensticker & Anzai claim, we conclude that transitive clauses are indeed significantly less preferred in Japanese than in English. The next hypothesis to consider is Ikegami’s (1981) typology of DO-language (suru no gengo) vs. BECOME-language (naru no gengo). Citing Bloomfield (1933), 3.  Categorization between transitive and intransitive was made based on semantic, rather than morphosyntactic, criteria. Point to, for example, is morphosyntactically intransitive, but semantically transitive. We used paraphrasability as a diagnostic test. That is, if the predicate can be paraphrased into transitive, e.g. point to into suggest, we categorized it as transitive. To simplify our analysis, we considered passive clauses to be intransitive, although we are aware that they could be categorized as transitive because, in most cases, two entities are involved. 4.  Some clauses are either not translated at all or translated into significantly different constructions. We did not count those cases. 5.  Most example sentences are somewhat simplified for expository purposes.

39

40 Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

Ikegami argues that the most favored sentence structure in English is actor-action, and, thus, it is a DO-language, where events are described as actions involving actors. Japanese, by contrast, is said by Ikegami to be a BECOME-language, preferring to describe events as a chain of state-changes. When checking this hypothesis with our parallel texts, we could not simply count transitive and intransitive clauses and assign them respectively to the DOtype and the BECOME-type descriptions. Many transitive verbs in English indicate states (e.g. have, entail, know, represent, suggest) or non-agentive events (e.g. experience, fail, lose, reach, complete, undergo), and many intransitive verbs indicate acts and processes (e.g. function, pervade, pass through, run, work). Therefore, we first identified predicates that denote a change of state and then determined whether the depicted situation is given a DO type or BECOME type expression.6 The sentences in (6) exemplify the former, and those in (7), the latter. (6)

I had built a model of a room that was part of my lab.  [DO] Watashi wa kenkyuu.shitsu no naka ni heya no shukushoo-mokei I top lab gen inside loc room gen miniature-model o tsukutta. acc made Backtranslation: ‘I made in my lab a miniature model of the room.’

(7)

The world’s population is stabilizing.  Sekai jinkoo wa anteeka-shi.tsutsu-aru. world population top is-stabilizing Backtranslation: ‘The world’s population is stabilizing.’

[BECOME]

The results of sorting clauses under these criteria, summarized in Table 1, do not support Ikegami’s hypothesis that English favors the DO-type and Japanese the BECOME-type of description (χ2 = 0.03, df = 1, p > 0.1). Table 1 DO

English

Japanese

110

104

BECOME

255

248

Total

364

352

The reason why there are fewer clauses in Japanese than in English in Table 1 is that some English transitive clauses are translated into stative clauses in Japanese, as shown in (8): 6.  When the clause in question is negated, we considered the affirmative counterpart.



(8)

On the universality of frames

These therapies alleviate symptoms, not causes. Chiryoohoo wa izure.mo taishoo-ryoohoo de, konponteki.na treatment top each-one symptomatic-treatment cop fundamental chiryoohoo de wa nai. treatment cop top not Backtranslation: ‘The therapies are all symptomatic treatments, not fundamental ones.’

In (8), the English original uses the transitive VP alleviate symptoms, but the Japanese translation is stative, taishoo-ryoohoo de ‘are symptomatic treatments’. We discarded Japanese clauses that do not denote any change of state in our analysis. As shown in Table 1, English does not necessarily use DO-type descriptions significantly more frequently than Japanese does. Regarding this phenomenon, Kondo’s commentary (1986: 2, cited by Uchimura 1991: 408) is particularly relevant: “One salient feature of English syntax, although often neglected by native speakers of English (and of other European languages) is a frequent and almost unlimited use of inanimate entities (things, time, space, collectives, abstract concepts etc.) as subjects for verbs that indicate intentional acts. To treat these uses as peripheral to mainstream English constructions fails to reflect an important characteristic of English, especially as viewed from the vantage point of a Japanese speaker.” (Translation by Uchimura)

If we take Kondo’s phrase “verbs that indicate intentional acts” as referring to verbs that can express an agentive act when occurring with a human subject, as in (9), then, Ikegami’s claim is at least interpretable.

(9) This lightning-fast channel jumping should permit [DO] cognitive radio systems to transmit voice and data streams at reasonable speeds. Kono denkoo-sekka.no chiiki-henkoo ni.yotte, onsei ya deeta o this lightning-fast channel-jumping by voice and data acc shikarubeki sokudo de yaritori dekiru yoo.ni-naru. [BECOME] reasonable speed at exchange be.able become Backtranslation: ‘By this lightning-fast channel jumping, we will become able to exchange voice and data at reasonable speeds.’

Of course, the meaning of permit in (9) is quite different from a situation of one human being giving another human being permission to do something. By recognizing permit as meaning ‘enable’ we can see that ‘X enables Y’ and ‘by X, Y becomes possible’ are two shapings of the same propositional form. The new count reflecting this re-categorization of the DO-type and the BECOME-type of encoding (i.e. if the verb can be used to depict an intentional act, the clause is categorized as the DO-type) is shown in Table 2.

41

42

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

Table 2 English Japanese DO (re-categorized)

260

187

BECOME (re-categorized)

104

165

Total

364

352

In Table 2, the ratio of DO to BECOME in English is 2.5:1, whereas in Japanese is 1.1:1. That is, if we re-categorize the predicates in the way explained above, we find that, while Japanese uses DO type and BECOME type equally frequently, English uses more than twice as many DO type descriptions than BECOME type ones. Therefore, Table 2 would support Ikegami’s hypothesis (χ2 = 25.56, df = 1, p  Animate Nonhuman > Inanimate

Therefore, when translated into Japanese, sentences like the following, in which a non-human entity is selected as the subject (occupying a more salient syntactic position) and a human is downgraded as the object as exemplified in (21), are likely to be reconstructed in such a way that the human occupies the subject position: (21) But nothing prepared me for the curious challenges involved in figuring out what Washington actually looked like. Shikashi, Washinton ga jissai.ni dono-yoo.na sugata o but nom actually what kind-of appearance acc shite-ita ka o suitei-suru to-iu kyoomi-bukai choosen o was q acc imagine quot interesting challenge acc hajimeru ni-atari, watashi wa nan.no yobi.chishiki-mo nakatta. start at I top no preparation there.was.not Backtranslation: ‘But when I started the interesting challenge of imagining what Washington actually looked like, I had no preparatory knowledge.’

Similarly, in (16), the subject of attack is Parkinson’s disease and the direct object is humans; therefore, we need to paraphrase it along the lines of those younger than 40 can acquire Parkinson’s disease, which evokes the Getting_disease frame:15 Getting_disease definition: A Victim starts off without the Disease, and then comes to suffer from it. If the Disease is infectious, then the Source from which the Disease is transmitted may be mentioned.

The Getting_disease frame includes as its frame evokers (byooki ni) kakaru ‘contract (a disease)’, (byooki ni) naru ‘become (sick)’, kansen-suru ‘get infected’, hasshoo-suru ‘acquire (symptoms)’, etc. (22) Shindan hoohoo ga shinpo-shita koto ni.yotte [40-sai.miman]Victim diagnosis method nom advanced nmlz by 40.below (Givón 1979), the agency hierarchy of Silverstein (1976) and Comrie (1981), and the topic acceptability scale of Lambrecht (1986). 15.  The frames of Attack and Getting_disease are related, but not closely. Attack (in its literal sense) is related via inheritance to Transitive_action, and finally to Objective_influence, which is a perspective on a general frame of Influence. The other perspective on Influence is Be_influenced. This is the frame from which Getting inherits; Getting_disease is a subtype of Getting. The metaphorical attack in the example sentence describes a situation more general than the literal Attack frame does. A full metaphorical analysis of this case will place the English sentence closer to the general Transitive_action frame and thereby closer to the evoked Japanese frame of Getting_disease.



On the universality of frames

demo paakinson-byoo o]Disease hasshoo-suru koto ga wakatte.kita even Parkinson’s disease acc acquire nmlz nom aware.became

As demonstrated, all major frames and frame elements of the original sentence (12) are encoded in its translation; therefore, according to the FrameNet’s frame test, this translation is judged as highly accurate. 3.2 Example 2 The translation of the second example is less straightforward than the first one: (23) Developments over the past decade have given new credibility to the idea that Earth’s biosphere could have arisen from an extraterrestrial seed. Kono 10 nen de kenkyuu ga susunde-kuru to, chikyuu no seibutsu this year over research nom has.progressed as earth gen life wa chikyuu.gai no seimei.tai kara hassei-shita to-iu aidea top earth.outside gen life.form from emerged quot idea mo hi.genjitsuteki.na o.hanashi to wa ienakunatte-kita. also unrealistic story quot top cannot.say-became Backtranslation: ‘As research has progressed over these 10 years, the idea that life on Earth sprang from extraterrestrial organisms can no longer be said to be a fanciful tale.’

The matrix predicate of (23) is give (new) credibility, which evokes the Evidence frame: Evidence definition: The Support, a phenomenon or fact, lends support to a claim or proposed course of action, the Proposition, where the Domain_of_ relevance may also be expressed. Some of the words in this frame (e.g. argue) are communication words used in a non-communicative, epistemic sense.16

(24) [Developments over the past decade]Support have given new credibility to [the idea that Earth’s biosphere could have arisen from an extraterrestrial seed]Proposition

This frame assignment is notable in that neither of the words give nor credibility evokes the Evidence frame. Rather, credibility evokes the Trust frame, which describes situations in which some source of information is believable (cf. that idea has no credibility). 16.  As discussed in Section 2, some Japanese scholars assume that words that are used as communication verbs always express communication even when they occur with an inanimate subject. It is essential to recognize that most such verbs are polysemous and can also be used as epistemic markers without any assumption of a communication agent.

51

52

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore Trust definition: A Cognizer thinks that the Information given by a particular

Source is correct. The specific Content or Topic of the Information may also be described.

(25) [Developments over the past decade]Source have given new credibility [to the idea that Earth’s biosphere could have arisen from an extraterrestrial seed]Information

Give is here acting as what FrameNet identifies as a causative support verb. As a support verb, it allows a noun’s semantic arguments to be expressed as its own arguments. It is a causative support because it additionally evokes the Causation frame (A Cause causes an Effect).17 In this case, the fact that some information is correct (i.e. the idea that … in (23)) is the Effect, and the Cause is the subject of give, namely developments over the past decade. (26) [Developments over the past decade]Cause have given new credibility to [the idea that Earth’s biosphere could have arisen from an extraterrestrial seed]Effect

The combination of Causation and Trust is semantically equivalent to Evidence, as described above. We thus analyze the multiword expression give credibility as evoking the Evidence frame.18 This complex construction can be realized in Japanese as follows (details will be discussed shortly): (27) [X] wa [Y] ni aratana shinpyoosei o ataeta. top to new credibility acc gave X = developments over the past decade Y = the idea that Earth’s biosphere could have arisen from an extraterrestrial seed ‘X have given Y new credibility.’

Development as an abstract noun normally evokes the Progress frame; however, developments (plural) here refers to research results, evoking the Achieving_first frame. 17.  The phrase have credibility evokes only the Trust frame. Have is a (non-causative) support verb, allowing the Information frame element to be expressed as its subject: [That idea] Information has CREDIBILITY. In general we find have, get, and give combine with nouns to describe related events: have an idea, get an idea, give someone an idea. Some nouns allow only one or two of these verbs: have one’s revenge, get one’s revenge, ??give one’s revenge. 18.  FrameNet currently does not have the capability to render this sort of complex analysis; either the sentence is analyzed separately in the Trust and Causation frames, or the multiword expression give credibility is placed directly in the Evidence frame.



On the universality of frames Achieving_first definition: A Cognizer introduces a New_idea into society.

Both the Cognizer and New_idea frame elements are null-instantiated.19 The evokers of this frame include: coin.v, coinage.n, discover.v, discoverer.n, discovery.n, invent.v, invention.n, inventor.n, originate.v, originator.n, pioneer.n, pioneer.v, pioneering.a. Decade evokes the Calendric_unit frame: Calendric_unit definition: Words in this frame name the different parts of the

calendric cycle, both man-made and natural. Frame elements include Whole for the whole of which the target is a part, Relative_time for locating the time with respect to an identifiable reference point, and Name for the name of the day (month, etc.) of a specially named unit. Words in this frame figure into a variety of temporal schemas, realized as constructions.

(28) Developments over the [past]Relative_time decade have given new credibility to the idea that Earth’s biosphere could have arisen from an extraterrestrial seed

Japanese does not have a lexical equivalent of decade; therefore, the term must be interpreted as 10 years and then translated. The past decade can be translated straightforwardly as kako ‘past’ 10-nen ‘10 years’; however, the translator of this magazine article preferred kono ‘this/these’ 10-nen. Japanese words that evoke the Achieving_first frame include hakken(suru) ‘discovery, discover’ and hatsumei(suru) ‘invention, invent’. Thus, “developments over the past decade” can be translated as kono 10 nen no hakken: (29) Kono 10 nen no hakken wa [Y] ni aratana shinpyoosei o ataeta this year gen discovery top to new credibility acc gave Y = the idea that Earth’s biosphere could have arisen from an extraterrestrial seed ‘Developments over these 10 years have given Y new credibility.’

However, as discussed in Section 2, causative sentences with an abstract subject are highly marked in Japanese. Therefore, the original text is first paraphrased along the lines of (30):

19.  A frame element which is conceptually salient may go unexpressed in a sentence (Fillmore et al. 2003). This is called null instantiation. It may be licensed by a construction (e.g. imperatives in English allow omission of the subject) or by a lexical item, as in the case of development. If the missing frame element is necessarily recoverable from linguistic or extralinguistic context, this is definite null instantiation. In contrast, if the filler of the frame element is not recoverable, or need not be located in the context, this is called indefinite null instantiation.

53

54

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

(30) As research has progressed over these 10 years, the idea that Earth’s biosphere could have arisen from an extraterrestrial seed has gained new credibility.

When translated into Japanese, this paraphrase is still somewhat unnatural because the subject of gain is abstract. We, therefore, paraphrase (30) further: (31) As research has progressed over these 10 years, the idea that Earth’s biosphere could have arisen from an extraterrestrial seed can no longer be said to be a fanciful tale.

This construction can be transferred into Japanese by means of the connective to ‘as’: (32) Kono 10 nen de kenkyuu ga susunde-kuru to, [Z] this year in research nom has.progressed as Z = the idea that Earth’s biosphere could have arisen from an extraterrestrial seed can no longer be said to be a fanciful tale ‘As research has progressed over these 10 years, Z.’

Idea evokes the Opinion frame: Opinion definition: A Cognizer holds a particular Opinion, which may be por-

trayed as being about a particular Topic.

(33) Developments over the past decade have given new credibility to the idea [that Earth’s biosphere could have arisen from an extraterrestrial seed]Opinion

The Cognizer frame element here is an instance of indefinite null instantiation. This frame is realized in the Japanese translation as: (34)

[chikyuu no seibutsu wa chikyuu.gai no seimei.tai kara hassei-shita earth gen life top earth.outside gen life.form from emerged to.iu]Opinion aidea quot idea ‘the idea that life on Earth sprang from extraterrestrial organisms’

Within the Opinion frame-element, the matrix predicate is could, which evokes the Possibility frame, as seen in the previous example: (35) Developments over the past decade have given new credibility to the idea that [Earth’s biosphere]Possible_event could [have arisen from an extraterrestrial seed]Possible_event



On the universality of frames

This frame is not realized in the Japanese translation, which can be backtranslated as: ‘life on Earth sprang from extraterrestrial organisms’. However, Possibility is in effect incorporated into the notion of idea. Arise evokes the Coming_to_be frame: Coming_to_be definition: An Entity comes into existence at a particular Place

and Time which may take a certain Duration_of_endstate, have a Cause, or be formed from Components.

(36) Developments over the past decade have given new credibility to the idea that [Earth’s biosphere]Entity could have arisen [from an extraterrestrial seed]Components

This frame has been transferred straightforwardly into Japanese: (37)

Kono 10 nen de kenkyuu ga susunde-kuru to, [chikyuu no seibutsu this year over research nom has.progressed as earth gen life wa]Entity [chikyuu.gai no seimei.tai kara]Components HASSEI.SHITA to.iu top earth.outside gen life.form from emerged quot aidea mo hi.genjitsuteki.na o.hanashi to wa ienaku.natte-kita idea also unrealistic story quot top cannot.say-became

As was the case with Example 1, we do not delve into minor frames that are evoked by earth, biosphere, extraterrestrial, and seed. 3.3 Example 3 This final example represents the case that involves deviation in frame correspondences. (38) Wi-Fi provides fast communications links that allow e-mail messages to appear almost instantly and Web pages to paint computer screens quickly — all with the mobility and freedom that has made cell phones nearly ubiquitous. Wi-Fi no koosoku tsuushin o riyoo-sure.ba, denshi-meeru o gen fast communication acc if.use e-mail acc sokuza.ni aite ni todokeru koto ga dekiru shi webu-peeji wa quickly addressee to deliver nmlz nom can and web-page top shunji.ni gamen ni hyooji-sareru. Wi-Fi nara, itsu.demo doko.demo tsukaeru instantly screen on is.displayed if anytime anywhere usable keitai-denwa ni hitteki-suru idoosei to jiyuu.do o arayuru mobairu- cell.phone to rival mobility and freedom acc all mobile kiki ni ataerareru-no-da. device to can.give

55

56

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

Backtranslation: ‘If (we) use the fast communication facility of Wi-Fi, (we) can deliver emails to the addressee immediately and Web pages are displayed on screens instantly. Wi-Fi can provide any mobile device with the mobility and freedom that rivals cell phones, which can be used anywhere anytime.’

The matrix verb provide evokes the Supply frame: Supply definition: A Supplier gives a Theme to a Recipient to fulfill a need or

purpose (Purpose_of_recipient) of the Recipient.

(39) [Wi-Fi]Supplier PROVIDES [fast communications links that allow e-mail messages to appear almost instantly and Web pages to paint computer screens quickly — all with the mobility and freedom that has made cell phones nearly ubiquitous]Theme

This frame is realized in the Japanese translation as: (40) … [Wi-Fi]Supplier nara, [itsu.demo doko.demo tsukaeru keitai-denwa ni if anytime anywhere usable cell.phone to hitteki-suru idoosei to jiyuu.do o]Theme [arayuru mobairu-kiki ni]Recipient rival mobility and freedom acc all mobile-device to ATAERARERU-NO-DA can.give Backtranslation: ‘… Wi-Fi can provide any mobile device with the mobility and freedom that rivals cell phones, which can be used anywhere anytime’

Within the Theme frame element, allow evokes the Make_possible_to_do frame: Make_possible_to_do definition: An Allower exists to provide the environ-

ment for which an Allowed_action may occur.

(41) Wi-Fi provides [fast communications links]Allower [that]Allower allow [e-mail messages to appear almost instantly]Allowed_action and [Web pages to paint computer screens quickly]Allowed_action — all with the mobility and freedom that has made cell phones nearly ubiquitous 20 Make_possible_to_do is the causative of the Possibility frame. That is, the words that evoke Make_possible_to_do indicate that a situation of Possibility has been brought about. Again, in order to avoid an abstract subject in a causative construction when translating into Japanese, this part of the sentence is paraphrased

20.  According to the FrameNet annotation protocol, this sentence would be bracketed as: allow [e-mail messages]Allowed_action [to appear almost instantly]Allowed_action and [Web pages] Allowed_action [to paint computer screens quickly]Allowed_action. This is because “e-mail messages” and “Web pages” are both the objects of allow as well as the subject of the corresponding predicate.



On the universality of frames

as “If we use the fast communication facility of Wi-Fi, e-mail messages appear almost instantly and Web pages paint computer screens quickly.” Appear evokes the Coming_to_be frame: Coming_to_be definition: An Entity comes into existence at a particular Place

and Time which may take a certain Duration_of_endstate, have a Cause, or be formed from Components.

(42) Wi-Fi provides fast communications links that allow [e-mail messages]Entity to appear almost instantly and Web pages to paint computer screens quickly — all with the mobility and freedom that has made cell phones nearly ubiquitous

This frame has not been transferred into the translation as such; the information has been framed based on a different perspective. Our real-world knowledge enables us to interpret e-mail messages to appear as to receive e-mail messages, but the translation takes the opposite perspective, i.e. to deliver e-mail messages, which is strictly speaking inaccurate, although these two events are factually equivalent. Delivery definition: A Deliverer hands off a Theme to a Recipient or (more indirectly) a Goal location, which is accessible to the Recipient.

(43) … [denshi-meeru o     ]Theme sokuza.ni [aite ni]Recipient e-mail acc quickly addressee to TODOKERU koto … deliver nmlz ‘… that (it) delivers emails to the addressee immediately …’

Paint evokes the Inchoative_filling frame: Inchoative_filling definition: A thing or substance, the Theme, comes to fill a

container or cover an area. The area or container can appear as the direct object with all these verbs, and is designated Goal because it is the goal of motion of the Theme. Corresponding to its nuclear argument status, it is also affected in some crucial way, unlike goals in other frames.

(44) Wi-Fi provides fast communications links that allow e-mail messages to appear almost instantly and [Web pages]Theme to paint [computer screens] Goal quickly — all with the mobility and freedom that has made cell phones nearly ubiquitous

This frame is transferred directly into Japanese: (45) Wi-Fi no koosoku tsuushin o riyoo-sure.ba, denshi-meeru o gen fast communication acc if.use e-mail acc sokuza.ni aite ni todokeru koto ga dekiru shi [webu-peeji wa]Theme quickly addressee to deliver nmlz nom can and web-page top

57

58

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

shunji.ni [gamen ni]Goal HYOOJI.SARERU instantly screen on is.displayed

In the adverbial modification at the end of the sentence, make (… made cell phones nearly ubiquitous) evokes the Cause_change frame: Cause_change definition: An Agent or Cause causes an Entity to change, either in its category membership or in terms of the value of an Attribute. In the former case, an Initial_category and a Final_category may be expressed, in the latter case an Initial_value and a Final_value can be specified.

(46) Wi-Fi provides fast communications links that allow e-mail messages to appear almost instantly and Web pages to paint computer screens quickly — all with [the mobility and freedom]Cause [that]Cause has made [cell phones] Entity [nearly ubiquitous]Final_value

This frame has not been transferred into Japanese. Rather, the translation backtranslates as ‘Wi-Fi can provide any mobile device with the mobility and freedom that rivals cell phones, which can be used anywhere anytime’. It misses the information that it was the mobility and freedom of cell phones that made them virtually ubiquitous. 4. Concluding remarks and future research directions We outlined in this paper several rhetorical differences between English and Japanese as characterized by Japanese researchers, and explored their validity using a bilingual corpus consisting of English magazine articles and their Japanese translations. Our corpus supports some of their claims, while failing to support others. We then selected from the corpus three translationally related pairs of sentences and demonstrated how the conceptual frames developed by FrameNet can be used to analyze both the English originals and their Japanese translations. We identified the major frames encoded in the source text and investigated whether they reappear in the Japanese translation. By comparing the frames evoked by major constituents of each pair of texts, we were able to assess translation accuracy more objectively than would have been possible with hitherto proposed translation evaluation methods, some of which are discussed below. This is mainly because many kinds of morphosyntactic differences between the two languages can be abstracted away from the basic frame structures. That is, frame semantic information can be expressed by using different parts of speech or — as we have seen with the causative relation — can be incorporated into the meaning of a verb in one context and expressed as a type of subordination in another, both



On the universality of frames

within the same language and across languages. We have shown that FrameNet frames are quite versatile even when applied cross-linguistically to languages that prefer different event-encoding strategies. In Example 1, Better diagnosis has made experts aware that Parkinson’s disease can attack those younger than 40, the translation contains all major frames either straightforwardly, maintaining the original grammatical structure, or, when such a method results in conflict with a stylistic norm of Japanese, a paraphrase of the source text has been translated. Therefore, in terms of the FrameNet frame test, Example 1 is judged as highly accurate. In Example 2, Developments over the past decade have given new credibility to the idea that Earth’s biosphere could have arisen from an extraterrestrial seed, we have found that the Possibility frame is absent from the translation: the idea that Earth’s biosphere could have arisen from an extraterrestrial seed is translated into Japanese that backtranslates as “the idea that Earth’s biosphere has arisen from an extraterrestrial seed.” However, the concept of possibility is part of the very notion of “idea” (vis-à-vis “fact”) in this context; thus no omission is recognized in this translation. In Example 3, Wi-Fi provides fast communications links that allow e-mail messages to appear almost instantly and Web pages to paint computer screens quickly — all with the mobility and freedom that has made cell phones nearly ubiquitous, our test has revealed that e-mail messages to appear, which is understood as a partial perspectival variant of “to receive e-mail messages” had been translated as another perspectival variant of that, i.e. “to deliver e-mail messages.” Furthermore, the information that the mobility and freedom are the causes that made cell phones nearly ubiquitous is not included in the translation. Rather, the translation backtranslates as “Wi-Fi can provide any mobile device with the mobility and freedom that rivals cell phones, which can be used anywhere anytime.” These examples have demonstrated how FrameNet frames can be utilized in assessing the accuracy of translation. Of course, accuracy is not the sole criterion for translation quality assessment, and, as discussed earlier, the importance of accuracy per se can differ significantly according to the text types. In conative texts like advertisements, for example, translation accuracy might simply be irrelevant. That is, the overall quality assessment should depend on the skopos in the sense of Vermeer (1978) and Reiß &Vermeer (1984). In the Skopos Theory, translation is viewed as a chain of human actions, not as a process of transcoding. A text is viewed as an offer of information made by a producer to a recipient. Translation is then characterized as offering to the targetlanguage audience information that is similar to the information originally offered to the source language audience. Typically, a translation project begins with an initiator who commissions a translation to accomplish a particular purpose or

59

60 Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

function when the translation is read by the target audience. Such a purpose or function is called the skopos of the translation project. In the Skopos Theory, the determiner of appropriate method and strategy is the skopos specified by the initiator, not the source text itself or the function assigned to it by the original author, nor its effect on the source-text audience (as claimed by Nida 1964).21 Although accuracy is merely one of the criteria in translation quality assessment, it is a significant one. And, arguably, the most significant criterion in assessing content-oriented texts, e.g. scientific translation. Several diagnostic tests for translation accuracy have been proposed, but, to our knowledge, they all seem to sanction the assumption that the ultimate measurements must rest on experts’ subjective judgments. Carroll (1966), for instance, evaluated the accuracy (as part of adequacy) of translation in terms of the informativeness of the original relative to the translation. That is, if the translation conveys the same amount of information, reading the original afterwards should not be informative at all. In one variation of his tests, English and Russian bilinguals first read an English translation of a Russian scientific text and then read the Russian original. He divided the translations to be measured into small parts so that a substantial number of independent judgments could be obtained (p. 56). His subjects were asked to rate the informativeness of the original of each unit on a 10-point scale (p. 58), the description of which is provided below: (47) 0. The original contains, if anything, less information than the translation. The translator has added certain meanings, apparently to make the passage more understandable. 1. Not informative at all; no new meaning is added nor is the reader’s confidence in his/her understanding increased or enhanced. 2. No new meaning is added by the original, either at the word level or the grammatical level, but the reader is somewhat more confident that s/he apprehends the intended meaning. 3. By correcting one or two possibly critical meanings, chiefly on the word level, it gives a slightly different “twist” to the meaning conveyed by the translation. It adds no new information about sentence structure. 4. In contrast to 3, adds a certain amount of information about the sentence structure and syntactic relationships. It may also correct minor misapprehensions about the general meaning of the sentence or the meaning of individual words. 5. Between 4 and 6. 21.  For example, Jonathan Swift’s Gulliver’s Travels was originally meant as a satire of contemporary social ills, but today, it is translated and read as a fantasy adventure tale. Therefore, the translation should adopt the style appropriate for fantasy tales (Reiß 1971/2000: 162).



On the universality of frames

6. The original is clearly informative. Adds considerable information about the sentence structure and individual words, putting the reader “on the right track” as to the meaning intended. 7. Between 6 and 8. 8. The original is very informative. Contributes a great deal to the clarification of the intended meaning. By correcting sentence structure, words, and phrases, it makes a great change in the reader’s impression of the intended meaning, although not so much as to change or reverse the meaning completely. 9. After reading the translation, the original is extremely informative. Makes “all the difference in the world” in comprehending the meaning intended. (A rating of 9 should always be assigned when the original completely changes or reverses the meaning conveyed by the translation.)

This is a daunting task, including highly subjective judgments (e.g. point 3: the original gives a slightly different “twist” to the meaning conveyed by the translation) as well as judgments that require adequate knowledge of linguistic analysis (e.g. point 6: the original adds no new information about sentence structure). This test may appear at first glance decently objective, but when the description of each point on the scale is examined seriously, it is deemed close to a black box. It is not inferable on what bases the experiment subjects could make such difficult decisions. It also does not seem to be very useful to a linguist-translator, since it fails to identify specific lexical or phrasal contributions to each judgment. Another assessment guideline worth mentioning is one used in the certification program of the American Translators Association. It consists of three-hour proctored examinations in a specific source-target language pair. Each examination is evaluated by two graders, who are certified translators and mark “errors” on a scale of 1, 2, 4, 8, or 16 points according to their intuition. The maximum points for deduction are, for example, 1 point if errors are not apparent to a casual sourcelanguage reader; 4 points if errors do not result in a loss of meaning; 8 points for errors whose consequences are not catastrophic; 16 points if they are. Final scores of 18 or higher are marked as ‘fail’. The checking criteria include addition, omission, word choice, too freely translated, too literal, ambiguity. However, no objective measurement guideline is available for each criterion; characterized by House (1997) as anecdotal assessment, what is depended upon is solely the experience of qualified translators. We believe that tools developed by FrameNet can be used as a tool that is at least one degree closer to objective accuracy assessment of translation, by providing the frames according to which addition and omission are identified. However,

61

62

Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

a number of striking problems emerged from this study that complicate the applicability of FrameNet concepts and practices in cross-language comparison. First, frame semantic information should be describable in a way that recognizes the difference between semantic information that is directly encoded in a lexical or grammatical form, on the one hand, and information that can be compositionally derived from the elements of a phrase, on the other. Since FrameNet itself is a lexical resource, it does not provide a complete account of frame semantics. Consider, for example, the case of comparison of degree. Comparative adjectives can be used to compare objects that are being evaluated on the same scale (this is better than that, I am older than you), but they can also be used for comparing present states with past states of the same object (this is better now, we are older now), and that interpretation requires a notion of state change; hence, in the case of better and older, the notions of improvement and aging, respectively. The quasi-paraphrase relation between better diagnosis methods (i.e. “better now than before”) and diagnostic procedures have improved (example 12) cannot be directly displayed by lexical annotations. A similar issue arises with give credibility in (23–26), where the epistemic Evidence frame was seen as the compositional result of causing ([Developments over the past decade] caused [the idea that Earth’s biosphere could have arisen from an extraterrestrial seed]) and justified belief (a Cognizer thinks that the Information given by a particular Source is correct). An analogous but simpler problem arises within the FrameNet lexicon of commercial transactions. Here the combination of Getting with an expression of money-exchange evokes the same situation as Commerce_buy: I got the book for $19.99 conveys the same situation as I bought the book for $19.99, where bought directly evokes the buying concept. A polysemy solution could treat this as a lexical problem by simply including the verb get within the Commerce_buy frame, in addition to its appearance in the Getting frame. The second problem in the applicability of FrameNet concepts in cross-language comparison is that the relation may be expressed metaphorically in one context and with frame-appropriate language in another. This is the case in the description of a disease and a person who comes to suffer the disease. Where English spoke of a disease attacking the victim, as in (16), Japanese spoke of the victim acquiring (or catching) the disease, as in (22), using a verb appropriate to just that concept, hasshoo-suru ‘acquire/have symptoms of ’. The Japanese choice is more consistent with the type-ranking that places humans over non-humans within the same clause. That a lexical solution is also possible here is suggested by the fact that some dictionary entries for attack include the case where a disease-agent is the subject.



On the universality of frames

Third, a situation can be expressed by asserting P in one case and negating the denial of P in another case. For example, we saw in (23–26) that research made a particular belief reasonable in English, but made it impossible for people to say that it could be doubted in the Japanese translation in (31–32). Most linguistic resources designed for participation in language understanding applications lack an appropriate means for interpreting negation, and the existing FrameNet database is no exception. An important result of this study is an awareness of both the utility and limitations of applying a lexical resource to analyze and compare translations. FrameNet takes a frame semantic approach to meaning description, and so it is revealing of certain types of differences between idiomatic English and Japanese, as in preferences regarding the expression of event causation. Notions such as frame-evoking expression, frame element, and frame-to-frame relations are necessary for understanding the correspondences (or lack thereof) between a translation and the original. At the same time, a lexical resource is limited in its inability to recognize the relation between a lexically-encoded meaning and that same meaning created by compositional processes, not to mention the possibility of non-lexical material (grammatical patterns, or constructions) contributing to the understanding of a sentence. Paraphrase relations such as those mentioned above, which range from relatively simple (Causation + Trust = Evidence) to quite complex (“give credibility” translated into “can no longer be said to be a fanciful tale”). Although FrameNet in the current state does not provide the means to explicitly represent these inter- and intra-language relations, by highlighting areas of great divergence across texts, it does provide a firm base upon which to conduct a deeper analysis. Moreover, turning the analytical framework embodied by FrameNet towards cross-linguistic texts highlights intriguing avenues in cross-linguistic constructional analysis. Frame semantics provides a useful dimension within which to understand how to compare similar-seeming constructions in different languages, such as comparisons or causatives. It may also highlight similarities between seemingly dissimilar constructions, or constructions that exist in a paraphrasal relationship (e.g. giving + exchange on the one hand, and commerce on the other). A full-fledged frame semantic account of sentence — and text — meaning, with FrameNet as a core component, will ideally provide a detailed enough specification or description of the meaning of a sentence such that even more detailed and precise comparative analysis can be carried out. What we have shown here is the crucial role that lexical-semantic analysis plays in this larger endeavor.

63

64 Yoko Hasegawa, Russell Lee-Goldman and Charles J. Fillmore

Acknowledgements An earlier version of this paper was presented at the 11th International Pragmatics Conference held in Melbourne, Australia in 2009. The paper was revised to its present form in 2009, but the manuscript was never published until now. In February 2014, one of the authors, Charles J. Fillmore passed away. We hope that this paper, the final collaboration among all the co-authors, retains the Fillmorean wit and wisdom that everyone has come to expect. We are grateful to the following individuals for valuable discussions on the themes addressed in this paper: Kimi Akita, Hans Boas, Michael Ellsworth, Albert Kong, and Ashlyn Moehle.

References Bloomfield, L. (1933). Language. New York: Henry Holt. Carroll, J. (1966). An experiment in evaluating the quality of translations. Mechanical Translation and Computational Linguistics, 9, 55–66. Comrie, B. (1981). Language universals and linguistic typology. Chicago: University of Chicago Press. Fillmore, C. J. (1982). Frame semantics. In Linguistic Society of Korea (Ed.), Linguistics in the morning calm (pp. 111–137). Seoul: Hanshin. Fillmore, C. J. (1985). Frames and the semantics of understanding. Quaderni di Semantica, 6, 222–254. Fillmore, C. J. (1994). The hard road from verbs to nouns. In M. Chen & O. Tzeng (Eds.), In honor of William S-Y. Wang: Interdisciplinary studies on language and language change (pp. 105–129). Taipei: Pyramid Press. Fillmore, C. J., & Atkins, B.T.S. (1992). Towards a frame-based organization of the lexicon: The semantics of RISK and its neighbors. In A. Lehrer & E. Kittay (Eds.), Frames, fields, and contrasts: New essays in semantic and lexical organization (pp. 75–102). Hillsdale: Lawrence Erlbaum Associates. Fillmore, C.J., & Baker, C.F. (2010). A frames approach to semantic analysis. In B. Heine & H. Narrog (Eds.), The Oxford handbook of linguistic analysis (pp. 313–340). Oxford: Oxford University Press. Fillmore, C. J., Johnson, C. R., & Petruck, M. R. L. (2003). Background to FrameNet. International Journal of Lexicography, 16(3), 235–250. ​doi: 10.1093/ijl/16.3.235 Givón, T. (1976). Topic, pronoun and grammatical agreement. In C. Li (Ed.), Subject and topic (pp. 149–188). New York: Academic Press. Givón, T. (1979). On understanding grammar. New York: Academic Press. Hasegawa, Y. (1996). A study of Japanese clause linkage: The connective TE in Japanese. Stanford, Calif.: CSLI Publications; Tokyo: Kurosio Publishers. Hasegawa, Y. (2011). The Routledge course in Japanese translation. London: Routledge. Hawkinson, A., & Hyman, L. (1974). Hierarchies of natural topic in Shona. Studies in African linguistics, 5, 147–170. House, J. (1997). Translation quality assessment: A model revisited. Tübingen: Gunter Narr.



On the universality of frames

Ikegami, Y. (1981). “Suru” to “naru” no gengogaku: gengo to bunka no taiporojii e no shiron [The linguistics of “do” and “become”: An essay on the typology of language and culture]. Tokyo: Taishukan Shoten. Ikegami, Y. (1982). Kotoba no shigaku [The language of poetics]. Tokyo: Iwanami Shoten. Ikegami, Y. (1988). What we see when we see flying cranes: Motion or transition. The Japan Foundation Newsletter, 15, 1–9. Koller, W. (1972). Grundprobleme der Übersetzungstheorie [Basic problems of translation theory]. Bern: Francke. Kondo, M. (1986). Eigo ni okeru museibutsu shugo no yoohoo to shakai kagaku no hoohooron [Inanimate subjects in English and the methodology of Social Science]. Daito Bunka Daigaku Kiyo [Bulletin of Daito Bunka University], 24, 1–21. Lambrecht, K. (1986). Topic, focus, and the grammar of spoken French. Ph.D. dissertation, University of California, Berkeley. Langacker, R. (1987). Nouns and verbs. Language, 63, 53–94. ​doi: 10.2307/415384 Maynard, S. (1997). Japanese communication: Language and thought in context. Honolulu: University of Hawai’i Press. Newmark, P. (1981). Approaches to translation. New York: Prentice Hall. Nida, E. (1964). Toward a science of translating. Leiden: E. J. Brill. Reiß, K. (1971). Möglichkeiten und Grenzen der Übersetzungskritik [Translation criticism — The potentials and limitations]. München: Hueber. Reiß, K. (1971/2000). Type, kind and individuality of text: Decision making in translation. In Lawrence Venuti (Ed.), The translation studies reader (translated by Susan Kitron) (pp. 160– 171). London: Routledge. Reiß, K., & Vermeer, H. (1984). Grundlegung einer allgemeinen Translationstheorie [Towards a general theory of translational action]. Tübingen: M. Niemeyer. ​doi: 10.1515/9783111351919 Seidensticker, E., & Anzai, T. (1983). Nihonbun no hon’yaku [Translating Japanese]. Tokyo: Taishukan. Silverstein, M. (1976). Hierarchy of features and ergativity. In Robert Dixon (Ed.), Grammatical categories in Australian languages (pp. 112–171). Canberra: Australian Institute of Aboriginal Studies. Tokieda, M. (1950). Nihon bunpoo: Koogo hen [Japanese grammar]. Tokyo: Iwanami Shoten. Toyama, S. (1987). Nihongo no ronri [The logic of Japanese]. Tokyo: Chuo Koronsha. Uchimura, H. (1991). Problems caused by word order when interpreting/translating from English into Japanese: The effect of the use of inanimate subjects in English. Meta, 36, 404– 413. ​doi: 10.7202/004516ar Vermeer, H. (1978). Ein Rahmen für eine allgemeine Translationstheorie [A framework for a general theory of translation]. Lebende Sprachen [Living Languages], 23, 99–102.   d ​ oi: 10.1515/les.1978.23.3.99

65

Phonological elements and Diasystematic Construction Grammar Steffen Höder

Christian-Albrechts-Universität zu Kiel

Usage-based CxG approaches share the central assumption that any grammar has to be acquired and organised through input-based abstraction and categorisation. Diasystematic Construction Grammar (DCxG) is based on the idea that these processes are not sensitive to language boundaries. Multilingual input thus results in multilingual grammars which are conceived of as constructicons containing language-specific as well as language-unspecific constructions. Within such systems, phonological structures play an important part in the identification of schematic constructions. However, the status of phonology in DCxG, as in CxG in general, yet remains unclear. This paper presents some arguments for including phonological elements systematically in the construction-based analysis of (multilingual) constructional systems. Keywords: Diasystematic Construction Grammar; language contact; multilingualism; contact-induced change; phonology

1. Constructions all the way down? Construction grammar (CxG), as represented by the different constructionist theories that have developed since the 1980s, still generates a fair amount of debate and controversy among linguists. This is mainly due to its theoretical perspective on the essence of linguistic structures, i.e. the constructionist ideas on what kind of elements linguistic systems consist of, how linguistic knowledge is acquired and organised, and whether or not grammatical organisation needs to be modular. The most far-reaching claim of CxG is the “idea that the network of constructions captures our grammatical knowledge of language in toto, i.e. it’s constructions all the way down”, as phrased by Goldberg (2006: 18; emphasis original). This is in line with standard definitions of the notion of ‘constructions’ as pairings of form and meaning, whether these definitions are strictly non-compositional as the one

doi 10.1075/bct.82.04hod 2016 © John Benjamins Publishing Company

68 Steffen Höder

in (1) or also assume compositional constructions resulting from cognitive entrenchment as the one in (2):

(1) Goldberg’s (1995: 4) definition [A] distinct construction is defined to exist if one or more of its properties are not strictly predictable from knowledge of other constructions existing in the grammar: C is a construction iffdef C is a form-meaning pair such that some aspect of Fi or some aspect of Si is not strictly predictable from C’s component parts or from other previously established constructions.



(2) Goldberg’s (2006: 5) definition Any linguistic pattern is recognized as a construction as long as some aspect of its form or function is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency […].

Generally, and particularly in the frameworks of Cognitive Construction Grammar (CCxG; Goldberg 1995, 2006) and Radical Construction Grammar (RCxG; Croft 2001), constructions are considered to be the uniform representations of all types of linguistic structures at least within the so-called syntax-lexicon continuum, allowing for different degrees of schematicity/specificity. This is illustrated by lexical (the English bread Construction, maximally specific: [bread]), morphological (the English Regular Plural, partially schematic: [noun-s]), and syntactic constructions (the English Ditransitive Construction, maximally schematic: [sbj v obj1 obj2]).1 The syntax-lexicon continuum — to be more precise: the monolingual, synchronic syntax-lexicon continuum — has indeed been the main focus of study within the CxG framework. At the same time, however, there has also been a centrifugal (in a manner of speaking) expansion of constructionist thinking into other fields, such as diachronic change and grammaticalisation (Traugott 2003; Diewald 2006, 2007; Noël 2007; Bergs & Diewald 2008; Hilpert 2008, 2011, 2013), interaction and discourse (Östman 2005; Günthner & Imo 2006; Wide 2009), linguistic variability (Leino & Östman 2008), language acquisition (Tomasello 2005), and language contact (Pietsch 2010; Höder 2012), in addition to the (centripetal) development and refinement of CxG approaches to the lexicon-syntax continuum itself (such as the work on constructional morphology by Booij 2010 or Diewald 2009).

1.  Here I am following Croft’s (2001) convention of using capitalised construction labels along with glottonyms to indicate that the constructions are not universal, but (at least) languagespecific (see, however, the discussion on language-specificity in Section 3.2).



Phonological elements and DCxG 69

The one thing that CxG apparently has steered clear of until now is phonology (cf. Boas 2013: 239). To be sure, CxG in principle assumes a phonological form for all specific, i.e. filled, constructions or constructional components such as lexemes or morphemes, and particularly intonational patterns are often mentioned explicitly as parts of the phonological form of, say, utterance-level idiomatic constructions such as the English Mad Magazine Construction (Lambrecht 1990: 217; cf. Fried & Östman 2004: 20). Still, the phonological form of constructions is often treated in an unsatisfactory way if addressed at all (see Section 2). Ideas developed in CxG have also been applied to phonological structures. One example is Vihman & Croft’s (2007) Radical Templatic Phonology, which parallels Croft’s RCxG in the bottom-up, non-reductionist way it derives syllable and segment categories from phonotactic schemas as basic units (cf. also Croft 2001: 62). Phonological elements themselves are, however, normally not thought of as constructions or included in constructional analysis. There are at least two reasons for the (as yet) low interest in phonology. One is historical: an original aim of CxG was to demonstrate that there are no clear-cut boundaries between a language’s lexicon, morphology, and syntax, and that there is more to gain than to lose if the traditional distinction is abandoned (hence the focus of early CxG studies on idiomatic, i.e. partially filled and partially schematic, structures such as [what’s X doing Y?] or [the Xer … the Yer]); phonology simply has not played a significant role in this scheme. Another reason is that phonological elements are generally taken to be purely formal features that are not associated with a particular lexical or grammatical meaning. Thus they do not qualify as constructions in the sense of definitions (1) and (2). Moreover, CCxG stipulates that constructions are motivated, whereas phonological elements are taken to lack motivation (with the rare exception of onomatopoeic sounds). When it comes to language contact, however, phonological elements do play a more significant role. Diasystematic Construction Grammar (DCxG; cf. Höder 2012; 2014a), approaches language contact situations from a CxG perspective. According to DCxG, multilingual2 speakers and communities organise their grammars in terms of multilingual systems, based on inheritance links between language-specific and common constructions, which in turn are based on formal and functional similarities. Formal similarities between lexically filled constructions depend on interlingual phonological correspondences, and language-specific phonological structures thus have a function of their own as markers of language specificity. In DCxG, this type of function is regularly analysed as pragmatic meaning. As a consequence, phonological elements as parts of the multilingual 2.  In this paper, ‘multilingualism’/‘multilectalism’ are not distinguished from ‘bilingualism’/ ‘bilectalism’, as the differences are ones of quantity, not of quality.

70 Steffen Höder

constructicon have to be included into the description of multilingual systems in some way. The following sections argue that if it is really constructions all the way down, i.e. if constructions are indeed the basic units of all grammatical knowledge, then the categorical exclusion of phonological elements from the grammar and from CxG is untenable. More specifically, it will be argued (a) that the phonological description of constructional form should be taken more seriously, (b) that phonological elements play a major role within multilingual grammars, and (c) that certain types of phonological elements that normally are viewed as pure form indeed qualify as constructions of a specific type when forming part of multilingual grammars. Section 2 discusses the place of phonology in CxG in general, while the following sections focus on DCxG: Section 3 gives a brief outline of the diasystematic approach, and Section 4 zooms in on the function of phonological elements in multilingual constructional networks, and Section 5 concludes the paper. 2. The place of phonology in construction grammar The following sections discuss the potential relevance of phonology to CxG from two different angles. Section 2.1 argues that the current practice of equating the domain of (construction) grammar with (by and large) the traditional fields of morphology and syntax while excluding phonology altogether does not follow from central assumptions in CxG, but is rather inherited from earlier theories. Section 2.2 discusses how phonological aspects can be integrated into the description of constructional form. 2.1 Meaningful vs. distinctive elements The categorical exclusion of phonological structures from the domain of (construction) grammar rests on the assumption that phonology is concerned with distinctive elements within the language system, whereas grammar is concerned with the combination of meaningful elements into complex utterances. Indeed, what Martinet (1949) famously dubs the ‘double articulation’ of language is one of the most salient semiotic features of human language as compared to other systems of symbolic communication: Utterances are composed of meaningful units, which in turn are composed of units which by themselves are meaningless, but rather distinguish different meanings. For instance, the utterance in (3) consists of several words with different meanings, such as cat ‘animal of the species Felis catus’, which can be subdivided into three distinctive units (/k/, /æ/, and /t/).



Phonological elements and DCxG

(3) The cat caught a fish. /ðə kæt kɔːt ə fɪʃ/3

The distinction between the two types of elements is reflected in the classic structuralist definitions of the ‘phoneme’ and the ‘morpheme’ as the smallest contrastive and the smallest meaningful unit in a language system, respectively. However, the observation that double articulation is a characteristic feature of human language does not imply that all types of utterances or structures actually are doubly articulated. Furthermore, despite the classic definitions of the phoneme and the morpheme as central notions in phonology and morphology, the traditional boundary between the two fields (or, in CxG terms, between phonology and the syntax-lexicon continuum) does not necessarily coincide with the distinction between meaningful and merely contrastive units. Two cases may illustrate this point. Firstly, there are submorphemic elements (i.e., elements that are not traditionally analysed as morphemes in their own right) which nevertheless can be associated with discernible semantic properties, such as the word-initial sound sequence /ɡl/ in English words such as glance, glare, glaze, gleam etc., all of which encode meanings that are related to (specific types of) visual perception, or word-initial /sn/ in, say, snack, snort, sniff, sneeze, i.e. elements that relate to (actions involving) the nose and mouth. Bergen’s (2004) experimental psycholinguistic study of such ‘phonaesthemes’ shows that these play a significant role in the acquisition and organisation of lexical knowledge into formally and functionally coherent categories, and that the frequency of phonaesthemes affects their cognitive entrenchment. In other words, in spite of their traditional analysis as being non-morphemic elements, phonaesthemes seem to qualify as constructions which are instantiated in the individual lexemes, thus forming a network as in Figure 1. ω

[ˈɡl-

‘vision, light’

glance

glare

glaze

gleam

glimmer

glimpse

glow

…

Figure 1.  English phonaestheme gl-

Secondly, there are linguistic elements which do not exhibit double, but indeed single articulation. This is the case when conventional meaningful elements cannot be (fully) subdivided into meaningless distinctive units, as they consist of (or 3.  The exact phonemic analysis and transcription is a matter of debate and controversy for any language. This is, however, irrelevant in the light of the more general arguments discussed in this article.

71

72

Steffen Höder

contain) sounds that do not occur in phonemically regular words in that language and hence do not form part of its phoneme inventory. For instance, the dental click [ǀ] is used in English to express a negative affective meaning (sometimes written as tut or tsk), whereas it does not occur in other English constructions, unlike in languages like Zulu, where it is a perfectly regular consonantal phoneme, as in icici /iːǀíːǀi/ ‘earring’. In Hebrew, in contrast, [ǀ] has an affirmative logical meaning (Gil 2011). This functional difference indicates that the meaning of [ǀ] has to be learned by speakers of either language during language acquisition, and that the association of [ǀ] with either affective or logical meanings is arbitrary and, hence, conventional. While such non-phonemic word-like utterances such as English [|] and Hebrew [|] are exceptional in terms of the traditional distinction between phonology and morphology and indeed are most likely to be treated as non-linguistic or para-linguistic elements in traditional analyses, they qualify as ordinary constructions from a CxG perspective, even though it is pointless to distinguish between a phonological and a morphological level on the formal side of such constructions. The existence of (perhaps not-so-atypical) borderline cases such as phonaesthemes and non-phonemic constructions suggests that the boundary between phonology and the syntax-lexicon continuum is not as clear-cut and impermeable as is traditionally assumed. At any rate, the categorical exclusion of phonology from the domain of CxG seems unmotivated; phonological elements do not disqualify a priori as constructional units. Rather, the status of phonological elements depends on their constructional (i.e. phonological, grammatical, pragmatic, lexical) context, and hence any such context deserves individual scrutiny. As a consequence, it might be worthwhile to reconsider the place of phonology in CxG and also to investigate the possibility of phonological elements within the constructicon. 2.2 Phonological form: simple and more complex cases It is generally accepted that constructions, defined as form-meaning pairs, include information on their phonological form, at least as far as spoken language is concerned (as opposed to other modalities such as written language or sign language). However, descriptions and analyses in CxG seldom actually specify the phonological form of the constructions they deal with, or if they do, they rarely do so in a manner that meets the demands of, say, traditional phono-lexicography. Instead, the phonological forms of constructions are often given — or rather hinted at — via their orthographic representation. In practice, this means that phonological details are usually omitted from the description of phonological forms, and that the complexities arising from a more detailed description are not taken into account at all.



Phonological elements and DCxG

While we meet this problem at all levels of constructional (and phonological) complexity, it becomes more essential when we move from phonologically filled to more schematic constructions. In lexical constructions, the differences between phonological forms and their orthographic representations tend to be few. For example, basic knowledge of phoneme-grapheme correspondence rules in Standard German is sufficient to predict the phonological form from the orthographic form in (4). (4) Standard German word [Wort /vɔrt/ ‘word’]

However, predictability is far lower with respect to irregular phoneme-grapheme correspondences and in particular suprasegmental elements, which usually lack orthographic representation. For instance, the lexical constructions record¹ and record² in (5) differ in that the former has initial stress and the latter final stress. Stress location cannot be transcribed orthographically, but is clearly part of the phonological form of these constructions.4 (5) Standard British English a. record¹ [record /ˈrekɔːd/ ‘record (noun)’] b. record² [record /rɪˈkɔːd/ ‘record (verb)’]

Similarly, the orthographic representations of the Danish lexemes mother and murder in (6) erroneously suggest a segmental difference (word-final and both represent /r/), while failing to indicate the fact that the two phonological forms, while segmentally identical, are nevertheless distinguished by the presence vs. absence of stød (cf. Basbøll 2003), i.e. a characteristic suprasegmental feature of Danish, usually realised as creaky voice on the preceding segment (hence mord /moːʔr/ [mo̰ːɐ̯]). (6) Standard Danish a. mother [mor /moːr/ ‘mother’] b. murder [mord /moːʔr/ ‘murder’]

4.  Moreover, stress location in English has a more general grammatical function as a suprasegmental word-class marker (cf. conflict, protest, insult etc., which all follow the pattern discussed here).

73

74

Steffen Höder

Another example comes from Standard Swedish, a pitch-accent language, where the distinctive contrast between two different tonal contours (known as accent 1 and 2), cannot be predicted from the orthographic form, as in the Swedish bolt and rule lexemes in (7).5 (7) Standard Swedish a. bolt [regel /²reːɡəl/ ‘bolt, latch’] b. rule [regel /¹reːɡəl/ ‘rule’]

In the case of lexical constructions, orthographically invisible phonological information can quite easily be added to the description of constructional form by providing a phonemic transcription. Things are more complex, though, in the case of more schematic constructions such as the Standard Swedish Infinitive. According to standard grammars, the infinitive (in most inflectional classes) is formed by adding a suffix -a to the verbal stem. This can easily be formalised as in (8): (8) Standard Swedish Infinitive a. [verb-a] b. kast-a throw-inf

However, if we take the tonal accents into account, it becomes obvious that the formation of the infinitive includes not only a suffix, but also a suprafix, since infinitives of morphologically simple verbs (in contrast to other inflectional forms within the verbal paradigm) are assigned accent 2 (Riad 2009: 127). A fuller description of the constructional form of the infinitive would have to include this, as well as information on the fact that the whole construction (consisting of lexical stem, suprafix, and suffix) constitutes a single phonological word (abbreviated as ω in phonological notation), which is the relevant phonological domain for the tonal accents. This is shown in (9):6 (9) Standard Swedish Infinitive a. [verb -a /(²verb-a)ω/]

5.  The exact nature of the Swedish tonal accents and accent assignment is a matter of considerable debate, and quite different analyses have been proposed. At any rate, the phonological form of Swedish constructions will have to include information on the tonal accents somehow, which is the point here. 6.  Plus signs (+) are used to indicate non-segmentability, as opposed to hyphens (-) between adjacent segmentable units.



Phonological elements and DCxG

b. kast-a /(²kasːta)/ throw+inf

The necessity of specifying phonological forms becomes even clearer in the case of non-standard varieties lacking an orthography of their own or employing one based on a related standard. For example, in a more formal register of Standard Danish, the infinitive of most verbs is formed by adding the suffix /ə/ to the stem. This is reflected in the orthography, as illustrated in (10). In a more colloquial style (as in many dialects), though, corresponding infinitives are often formed by lengthening an element in the rhyme of the stem-final syllable (either the nucleus or the coda consonant, depending on the phonological context; cf. Schachtenhaufen 2010). (In phonological terms, this suprafix is usually interpreted as compensatory lengthening resulting from /ə/ deletion.) (10) Standard Danish infinitive a. skrige [ˈsɡ̊ʁiːə, sɡ̊ʁiːː] ‘scream-inf’ b. flyve [ˈflyːu̯ə, ˈflyːu] ‘fly-inf’ c. komme [ˈkʰʌ̞̈mə, kʰʌ̞̈mː] ‘come-inf’

Here, the standard orthography misleadingly suggests a constructional analysis in terms of segmental affixation ([verb-/ə/]), whereas a phonological transcription of the non-standard form exposes the morphological structure as suprafixation ([verb+/ː/]). Phonological form can be filled (as in lexemes or affixes, including suprafixes), but can be also quite schematic. Many clause-level constructions include schematic phonological patterns. Constructions in European languages signalling illocutionary force are typical examples, such as the German Polar Question in (11), which combines verb-initial word order with particular types of intonation, or the French Intonation Question in (12), which is solely expressed by specific intonational patterns. (11) German Polar Question a. [vfin₁ …, intonational patterns7] b. Sprecht ihr Deutsch? speak.ind.prs-2pl 2pl.nom German ‘Do you speak German?’ 7.  These can be formalised in various ways. The prototypical pattern includes an utterance-final f0 rise which can be interpreted in formal terms as an H-^H% boundary tone (for the GToBI system of intonational notation, see Grice, Baumann & Benzmüller 2005).

75

76

Steffen Höder

(12) French Intonation Question a. [intonational patterns] b. Vous parlez français? 2pl speak-ind.prs.2pl French ‘Do you speak French?’

An example from the domain of information structure is found in the English Contrastive Focus construction in (13) (also cf. the remarks on prosodic patterns in Fried & Östman’s [2005] discussion of pragmatic particles in Czech and Finland Swedish dialects). (13) English Contrastive Focus a. [stress on contrastive element] b. He’s done ˈthat. ‘He’s done that, whereas he hasn’t done other things.’

It is, however, remarkable that phonologically schematic forms can also encode quite specific grammatical functions. For instance, there are several ways of nominal compounding in Standard Swedish, all of which combine two nouns, either with our without one of different available (segmental) interfixes. However, all of them also include a schematic phonological pattern which (iconically) integrates both components into one phonological word and assigns accent 2 to the whole compound, regardless of the tonal accents of its components (Riad 2009: 215ff.). This is illustrated in the examples in (14). (14) Standard Swedish nominal compounds a. bokmanus /²buːkmaːnʉs/ book-manuscript