Atypical predicate-argument relations [1 ed.] 9789027266255, 9789027231437

Citation preview

Lingvisticæ Investigationes Supplementa 33

Atypical predicateargument relations edited by

Thierry Ruchot Pascale Van Praet

J OHN B ENJAMINS P UB LISHING COMPANY

ATYPICAL PREDICATE-ARGUMENT RELATIONS

LINGVISTICÆ INVESTIGATIONES SUPPLEMENTA

Studies in Linguistics and Language Resources issn 0165-7569 This series has been established as a companion series to the journal Lingvisticæ Investigationes A complete list of titles in this series can be found on http://benjamins.com/catalog/lis

Series Editors Éric Laporte (Université Paris-Est Marne-la-Vallée & CNRS) Annibale Elia (Università degli Studi di Salerno) Cédrick Fairon (Université catholique de Louvain) Duško Vitas (University of Belgrade)

Volume 33

Thierry Ruchot and Pascale Van Praet (eds.) Atypical predicate-argument relations

ATYPICAL PREDICATE-ARGUMENT RELATIONS

Edited by

THIERRY RUCHOT PASCALE VAN PRAET University of Caen, Normandy

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/lis.33 Cataloging-in-Publication Data available from Library of Congress: lccn 2016034637 (print) / 2016057619 (e-book) isbn 978 90 272 3143 7 (Hb) isbn 978 90 272 6625 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 Thierry Ruchot

vii

Part 1.  Atypical realization of the main arguments of the verb Verbs of pain and accusative subjects in Romanian Marleen Van Peteghem Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew Rivka Halevy IO realization variation in Spanish reverse psych verb sentences Chiyo Nishida Non-human agents as subjects in English and Dutch: A corpus-based translation study Steven Doms, Bernard De Clerck and Sonia Vandepitte

3

27 61

87

Part 2.  Valency-changing devices and non-finite verb forms The argument-structure configuration of English middle and related structures Casilda García de la Maza

115

Non-categorical categories: Aspect, Voice, Pred and the category of Participles Elena Soare

131

Part 3.  Variations in transitivity The semantic motivation of non-canonical predicative relations: The French transitive construction Meri Larjavaara

163

vi

Atypical Predicate-Argument Relations

Atypical argument structures in French: From metaphorical uses to atypical ones Geneviève Girard-Gillet Split intransitivity in Lamaholot (East Flores, Indonesia) Philippe Grangé

181 203

Part 4.  Norm variation in predicate-arguments relations Geographic variation in a non-canonical infinitive structure with the modal verb brauchen Pascale Van Praet and Gilbert Magnus

243

Verbal constructions in spoken language deviating from the norm: Reflections on the concept of atypicality Günter Schmale

265

Index of authors Index of subjects

283 287

Introduction Thierry Ruchot

Université de Caen, Basse Normandie

The papers presented here emerged from the proceedings of a conference held in Caen in November 2012 whose topic was: “Non-canonical predicative relations”. This issue is the result of a common study carried out by the Research Laboratory: CRISCO (University of Caen Basse-Normandie EA4255), and the Research Group CONTRAGRAM (Universiteit Gent). Traditional grammarians, as well as linguists, have long been essentially interested in the most typical predicate-argument constructions, leaving aside peripheral and less typical structures. Indeed, linguists used to focus, in single languages or cross-linguistically, on active, transitive constructions with verbs designating typical actions and taking a typical agent-like, animate, fully volitional first argument (subject) and a typical patient-like affected second argument (object) with uniform morphosyntactic marking, although all of these typical features remained implicit until Hopper and Thompson’s seminal work on typical transitivity. Nevertheless, typicality logically implies atypicality. Since then, linguists have become more and more aware of the necessity of investigating those atypical cases that offer a great number of challenges for current linguistic theories as well as for empirical studies. It may even appear that studying atypical cases sheds light by contrast on the more typical instances and therefore enables the development of more adequate theories. Although there is no agreement regarding what counts as typical or atypical in syntax, we can reasonably distinguish some instances of predicate-argument constructions that may be considered atypical in at least some linguistic theories, among them oblique case marking of central arguments, valency increasing, decreasing or reorganizing devices, the possibility of omitting an argument or, on the contrary, of expressing arguments not contained in the semantic structure of the predicate. Moreover certain types of atypical marking or atypical syntactic behavior can also occur with non finite forms of verbs or according to some factors like tense, modality, information structure, person, animacy, definiteness, specificity etc. All of the articles presented here focus on one or several of these aspects and display a great variety of perspectives, be they (morpho-)syntactic, semantic, doi 10.1075/lis.33.001int © 2016 John Benjamins Publishing Company

viii Thierry Ruchot

pragmatic or a combination thereof and are cast in different theoretical frameworks such as generative grammar, or construction grammar. Yet, some papers don’t adopt a particular framework and take a more descriptive standpoint. The originality of this issue is that most studies are based on a wide range of corpora for modern languages, such as translation corpora, written or oral medial corpora and conversation corpora. Most of the time, such corpus data illustrate brandnew or emerging language phenomena. The studies demonstrate also cases of diachronic or diatopic variation within one single language as well as cross-linguistic variation. The articles in the first part focus on factors that impinge on the realization of the main arguments of the verb. Marleen Van Peteghem’s article concentrates on the problem of pain verbs in Romanian whose Experiencer argument takes the accusative case instead of the dative case unlike other Romance languages. She tries to show that it is problematic for Burzio’s Unaccusative Hypothesis. Then she gives an explanation for this atypical morphosyntactic marking in the framework of construction grammar that shows that this atypical behavior is due to semantics factors, especially affectedness. Rivka Halevy focuses on the atypical marking of the single NP argument of thetic/presentational sentences in colloquial Modern Hebrew, that can receive a marking akin to object marking, which contradicts Burzio’s unaccusative generalization. She proposes that the boundary between subject and objects are blurred in focus position and that the subject category is a gradual one. Chiyo Nishida investigates the peculiarities of Psych verbs in Peninsular Spanish with reverse order. The normal word order is Indirect object Verb Subject with obligatory clitic doubling, but when the subject appears first, clitic doubling is not obligatory. Nishida studies syntactic and semantic factors that account for this variation. Working on parallel corpora of translations from English into Dutch and French, Steven Doms studies cases where English non-protypical agents in predicates which usually require prototypical agents are translated in the other languages by other constructions, like passive or impersonal constructions in order to restore a more prototypical predicative relation. He also shows cases where translators choose to maintain them in a more literal translation. The articles in the second part concentrate on valency decreasing devices as well as the relations between non-finite verb forms and voice. Casilda Garcia de la Maza investigates the constraints responsible for the use of the English middle constructions. Her study is based on data from novel or idiosyncratic usages, some of them still being rejected by some native speakers. She shows that, unlike the received view according to which middle constructions can

Introduction ix

be explained only by semantic factors, there is in fact a range of essentially pragmatic constraints, some of them stronger than others that concur to determine the degree of acceptability of the middle construction. Elena Soare’s paper is concerned with Romanian participle constructions. She assumes that participles can be defined as verbal constructions that have a truncated character. This allows to encompass constructions that are usually not included in the category of participles like supine, gerund and the like. The range of participial constructions available in each language depends on morphological devices displayed by each of them. Elena Soare also aims to explain why participial constructions display properties common to nouns, verbs and adjectives as well as the relations between participles and voice. The articles in the third part describe some cases of constructions where more arguments are expressed than usually implied by the verb meaning (semantic valency), a problem area especially popularized by works in construction grammars. Meri Larjavaara studies French verbs that are usually intransitive but allow in some quite frequent novel uses transitive constructions, like démissionner (to resign/occasionally to fire someone). She first examines functional and typological accounts of transitivity. Then, working in the framework of construction grammar, she discusses other possible factors that motivate these cases of non-­ canonical transitive constructions. Geneviève Girard-Gillet focuses on the possibility of French change-of-state verbs to have either a causative construction with faire (literally ‘to make’) or a transitive construction without faire of the type ils l’ont suicidé (literally ‘they suicided him’). She hypothesizes that French has means to differentiate external and internal causes. She also explains by this distinction the possibility of metaphorical uses. Philippe Grangé’s work deals with split intransitivity in lamaholot, an Austronesian language spoken in Indonesia. It investigates a fairly complex system of split intransitivity, where various factors are involved. The fourth and last part is dedicated to problems of synchronic norm variation according to regional factors, speech registers, or type of data taken into account (oral vs written for example). Based on a medial corpus made of written press articles, the study of Gilbert Magnus and Pascale Van Praet concentrates on the verb of “necessity” brauchen (‘to need’) in Modern German. This verb displays two alternating constructions one with the infinitive marker zu and one without it. The second construction is an atypical construction that tends to become more like typical modal verbs that also don’t take zu. The analysis shows that this phenomenon may have something to do with oral conceptualized contexts and with diatopic variation within the German-­speaking area.

x

Thierry Ruchot

Günter Schmale studies non-canonical predicative relations in idiomatic expressions and occurrences issued from the authentic everyday spoken German language. Taking on alternatively a morphosyntatic, a syntactic or a semantic point of view, he shows that there is a very thin line between typical and atypical structures, and more generally reflects on the establishment of the norm itself. These contributions show the liveliness of discussions in modern linguistics upon the central question of argument structure and argument realization. They confront a great variety of vantage points and treatment of those relations. The main thrust of these articles is that the phenomena under consideration are best studied with a complex interplay of syntactic, semantic and probably pragmatic factors, which open a very large field of investigations for the future. We would like to thank all the persons who read the articles or helped us with advice: Petr Arkadiev, Valérie Amary, Michael C. Ewing, Jacques François, Csaba Földes, Isabelle Haïk, Axel Holvoet, Rolf Kailuweit, Pierre Larrivée, Silvia Luraghi, Annie Montaut, Estelle Moline, Lea Nash, Peter de Swart, Lucia Tovena, Johan Van den Auwera. We also thank the Crisco laboratory for ist support in the organization of the conference that was at the outset of the publication of this book.

Part 1

Atypical realization of the main arguments of the verb

Verbs of pain and accusative subjects in Romanian Marleen Van Peteghem

Ghent University – Contragram / GLIMS

Verbs of pain in Romanian such as durea ‘ache’, ustura ‘burn’, and furnica ‘itch’ assign the accusative to their experiencer arguments, unlike other Romance languages, where the experiencer is dative-marked. The use of the accusative raises interesting problems in that it gives rise to a mismatch between the hypothesis on the syntax of inalienability in Romance in Generative Grammar (Guéron 1985) on the one hand and Burzio’s (1986) Generalization on the other hand. This article shows that the inversed nominative NP denoting the body part does not show subject properties, and that the accusative experiencer in sentence initial position does not show object properties, but instead displays subject properties, just like the dative in similar constructions. However, the difference between accusative and dative subjects in this construction is that the accusative is assigned to verb arguments and is a lexical case, whereas the dative is assigned to external possessors and is an inherent case. Surprisingly, the argument status of the accusative experiencer makes it even more subject-like than the dative experiencer, which is an adjunct and is dependent on the presence of an internal argument triggering verb agreement, whereas the accusative subject can also occur without an internal argument or with a locative PP.

0. Introduction1 The argument structure of verbs expressing pain in Romanian (e.g. durea ‘ache’, ustura ‘burn, irritate’, arde ‘burn’, etc.) raises interesting problems both from a comparative and a theoretical point of view. These verbs take two arguments: a theme denoting the aching body part and an experiencer denoting the person who feels the pain. As in other Romance languages, the theme is encoded as a grammatical subject triggering verb agreement in Romanian; however, the experiencer 1. I would like to express my gratitude to two anonymous reviewers for their insightful remarks. Needless to say, I am responsible for all remaining imperfections. doi 10.1075/lis.33.01van © 2016 John Benjamins Publishing Company

4

Marleen Van Peteghem

receives accusative case (1), whereas in languages such as Spanish, French and Italian it is dative-marked (2). (1) a. Îl doare cap-ul him.acc hurts head-the (He has a headache) b. Îl arde piele-a him.acc burns skin-the (His skin is burning) (2) a. Sp. Le duele la cabeza him.dat hurts the head (He has a headache) b. Fr. La peau lui brûle the skin him.dat burns (His skin is burning) c. It. Gli prude la pelle him.dat itches the skin (His skin is itching)

The use of the accusative instead of the dative in the Romanian examples in (1) is interesting because it results in a mismatch between the following two hypotheses put forward in Generative Grammar: a. the hypothesis on the syntax of inalienable possession, as it was formulated by J. Guéron (1983, 1985, 2003, 2007), J.-R. Vergnaud & M.L. Zubizaretta (1992) and J. Herschensohn (1992); b. the Unaccusativity Hypothesis, first formulated by D.M. Perlmutter (1978) within the framework of Relational Grammar, and later adopted by L. Burzio (1981, 1986) in Generative Grammar in what is generally called “Burzio’s Generalization”. According to hypothesis (a), the definite article of the body part NP is bound by the experiencer pronoun, which is only possible if the latter c-commands the former. Therefore the body part NP in subject position is to be analyzed as an internal argument of the verb, raised to subject position, and consequently the verb has to be unaccusative. However, according to Burzio’s Generalization, unaccusative verbs are unable to assign accusative case. Therefore, the experiencer pronoun should occur in the dative as in other Romance languages, and not in the accusative as in Romanian. The main focus of this article is to account for the atypical case assignment in the structure [măacc Vpain NPbody part] in Romanian. I will argue that the accusative nominal in (1) shows subject properties, just like the dative in similar constructions with other types of verbs of pain. The difference between accusative and dative subjects is that the accusative is assigned to a verb argument, regardless of the syntactic encoding of the theme argument, whereas the dative is not a verb argument, but an external possessor, hence an argument of the noun denoting the body part, and therefore highly dependent on the syntactic encoding of the body



Verbs of pain and accusative subjects in Romanian

part NP. This leads to the surprising conclusion that in the construction under investigation the accusative is less dependent on the morpho-syntactic encoding of other verb arguments than the dative and constitutes empirical evidence against the widely accepted idea that the accusative is basically a dependent case. The article is structured as follows. Section 1 discusses the mismatch between the two hypotheses mentioned above and shows that the Romanian data violate Burzio’s Generalization. Section 2 explores which verbs enter into the accusative construction and examines how this structure relates to the other argument structures of these verbs. Section 3 investigates the syntactic behavior of the two arguments of the construction in (1); more specifically, it addresses the question to what extent they display either object or subject behavior. Finally, Section 4 studies the competition between accusative and dative as case marking of quirky subjects of verbs of pain in Romanian. 1. The Inalienable Possession Hypothesis vs. the Unaccusativity Hypothesis 1.1

The syntax of inalienable possession

J. Guéron (1983, 1985, 2003), J.-R. Vergnaud & M.L. Zubizaretta (1992) and J. Herschensohn (1992), among others, showed that in French the main syntactic property of inalienable possession is the presence of a definite article functioning as a bound anaphor.2 The antecedent of this anaphor is the pronoun or the NP denoting the possessor, which can occur either as the indirect object, direct object or subject of the verb. In the corresponding structures in Romanian, the pronoun or NP can be either in the dative (3a), the accusative (3b), the nominative or be omitted (3c), since Romanian is a PRO-drop language. As for the anaphoric NP denoting the inalienable referent, it mostly occurs as a direct object (3a), but also within a PP (3b) or even as a subject (3d). (3) a. Maria ii- a mângâiat mânai Mary him.dat has stroked hand-the (Mary stroked his hand)

2. The use of the indefinite article is equally possible (cf. M.-O. Junker & F. Martineau 1987: 196): Edgar se lava soigneusement une oreille, puis l’autre. Edgar refl washed carefully one ear, then the other ‘Edgar carefully washed one ear, then the other’

5

6

Marleen Van Peteghem

b. Maria li- a luat de mânăi Mary him.acc has taken by hand (Mary took him by the hand) c. [Mariai] a ridicat mân-ai Maria has raised hand-the (Mary raised her hand) d. Îmii curge nas-uli me.dat flows nose-the (I have a runny nose)

This anaphora is not coreferential, but a kind of intraphrastic associative anaphor (cf. M. Van Peteghem 2006b).3 As argued by the authors mentioned above, it is syntactically bound, which means that it is c-commanded by the antecedent, like reflexives.4 In other words, the anaphoric NP denoting the body part must occur in a syntactic position that is locally c-commanded by the antecedent. If this hypothesis is correct, the anaphoric NP cannot occur in subject position, since the subject c-commands all other NPs of the clause. Therefore, a sentence such as (3d), where the anaphoric NP is the subject, should be ruled out. Nevertheless, both in Romanian and in other Romance languages, such examples do occur (cf. (1), (2)), although this structure is not productive in all of them. In French, for instance, it is restricted to a few particular verbs (cf. A.-M. Spanoghe 1995: 94–96; E. König & M. Haspelmath 1998: 539). (4) a. b.

La tête lui tourne vs. a′. *La tête lui craquait the head him.dat turns  the head him.dat cracked (He feels dizzy) Les yeux lui brûlent b′. *Les  yeux lui étincelaient the eyes him.dat burn  the  eyes him.dat sparkled (His eyes were burning)

Nevertheless, according to J. Guéron (1983, 1985), the examples (4a–b) do not violate the bound anaphora hypothesis, but can be explained by the fact that the subject is a deep object raised to subject position. In line with L. Burzio (1981), she argues that the structure in (4a–b) is unaccusative. Therefore the verb cannot assign the accusative case to the object, which raises into subject position in order 3. For more details on associative anaphors, see M. Riegel (1994), C. Schnedecker et al. (1994) and G. Kleiber (1999). 4. J.-R. Vergnaud & M.L. Zubizaretta (1992: Note 18) describe the structural relation between the dative and the possessee NP in terms of m-command, defined in the following way: “x m-­commands y iff x does not dominate y and every maximal projection z that dominates x dominates y.”



Verbs of pain and accusative subjects in Romanian

to receive nominative case (J. Guéron 1985: 51). As a consequence, these verbs have to resort to the dative case for the second internal argument (cf. M. Van Peteghem 2006a). This analysis accounts perfectly for the examples in (2), (3d), (4), in which the possessor pronoun is dative marked, but not for those of the type of mă doare capul in Romanian, in which an apparently unaccusative verb assigns accusative case to its second argument. In other words, the accusative structure in (1) gives rise to an inconsistency between the inalienable possession hypothesis as elaborated by J. Guéron (1983, 1985, 2003), J.-R. Vergnaud & M.L. Zubizaretta (1992) and J. Herschensohn (1992) on the one hand and Burzio’s Generalization on the other. 1.2

The Unaccusativity Hypothesis and Burzio’s (1986) Generalization

The Unaccusativity Hypothesis of D.M. Perlmutter (1978), reformulated in L.  Burzio’s Generalization (1981, 1986) states “that a verb that does not assign an external θ-role to its subject does not assign structural accusative Case to an object and conversely” (cf. E. Reuland 2000b: 1). In the absence of an external argument able to occupy the subject position, the internal argument moves into subject position, triggering verb agreement, and therefore the verb can no longer assign accusative case. However, as shown in several studies, counter-examples can be found in various languages. The most discussed counter-evidence comes from the impersonal construction, as in French (5a) or Italian (5b). In this construction, the internal argument does not move into subject position, which is filled by an expletive pronoun (or is empty in pro-drop languages). (5) a. Il est arrivé un accident It is arrived an accident (An accident has happened) b. Si leggerà volentieri alcuni articoli. (H. Haider 2000: 44) si.imp read.3.sg.fut voluntarily several articles (One will voluntarily read several articles)

In its original version, the Unaccusative Hypothesis was introduced to account for one-place predicates, such as intransitives or passives. However, in later studies it has been extended to two-place predicates lacking an external θ-role, such as psychological verbs of the piacere class, which take a theme and an experiencer argument (cf. A. Belletti & L. Rizzi 1988). However, as shown by various authors, one of the three classes of psychological verbs distinguished by A. Belletti & L. Rizzi, the preoccupare class, is also inconsistent with Burzio’s Generalization. Binding

7

8

Marleen Van Peteghem

facts indeed show that the subject arguments of such verbs are in a lower position than the objects (cf. A. Belletti & L. Rizzi 1988: 313). The subject is therefore to be analyzed as a derived subject, raised from a VP-internal position. Consequently, the constructions in (6) should be unaccusative and their verbs should not be able to assign accusative case, which they do. (6) a. Questi pettegolezzi su di se preoccupano Gianni these gossips on of himself worry Gianni più di ogni altra cosa more than each other thing (This gossip about himself worries Gianni more than anything else) b. Stories about herselfi rarely annoy heri profoundly. (H. Haider 2000: 36)

Various authors (cf. E. Reuland 2000b; A. Marantz 2000; H. Haider 2000) have shown that the problems originate from the fact that Burzio’s Generalization is “cross-modular” (cf. H. Haider 2000: 33) in that it relates case marking to argument structure. The assignment of an external semantic role to the subject is claimed to be a condition for the assignment of a structural accusative case and vice versa. Furthermore, as shown by A. Marantz (2000), Burzio’s Generalization appeals to two independently motivated principles: (i) the requirement that sentences have subjects (the “Extended Projection Principle, cf. N. Chomsky 1986: 4), according to which the internal argument should move into subject position in the absence of an external θ-role (cf. the “nominative first” condition, E. Reuland 2000b: 3), and (ii) the hypothesis of the dependency of the accusative on the nominative case. In fact, a language such as Russian, which has morphological cases, shows that neither of these principles is universal: internal arguments do not ne­ cessarily move into subject position and accusative case can be assigned without a nominative (cf. (7)). Furthermore, external arguments can be demoted from subject position and receive quirky case, as in the Russian example (8b), where the external argument of an unergative verb receives quirky dative case and the verb takes the reflexive impersonal form. (7) Menja tošnit. me.acc nauseate.3sg (I feel nauseous) (8) a. Ja zdes’ xorošo rabotaju I here well work (I always work well here) b. Mne zdes’ xorošo rabotaet-sja me.dat here well work-refl (I can work well here)

(Perlmutter & Moore 2002)



Verbs of pain and accusative subjects in Romanian

Within the framework of generative grammar, various solutions have been proposed to account for the above-mentioned counter-evidence (e.g. the papers collected in E. Reuland 2000a; and also E. Woolford 2003; H. Bennis 2004, etc.). There is a general consensus that the assignment of the nominative case to internal arguments is not related to the unability of verbs to license the accusative case, but to the requirement that an object gets the nominative case when there is no nominative subject (e.g. L. Burzio 2000; H. Haider 2000; E. Woolford 2003). However, this reformulation of Burzio’s Generalization does not account for the Russian dative subject construction in (8b), nor for the dative subject constructions in Faroese in (9), in which a dative subject co-occurs with an accusative object (cf. M. Barnes 1986; E. Woolford 2003). (9) Mær líkar henda filmin. Me.dat likes this film.acc (I like this film)

(Barnes 1986: 12)

In what follows we will show that it cannot account for the impersonal constructions of verbs of pain in Romanian either. 2. Verbs of pain and their argument structures Before examining the accusative structure in (1), I will first explore which verbs of pain can occur in it. These are not very numerous since, as will be shown in Section 4, most verbs of pain occur with a dative experiencer, as in the other Romance languages. A. Şerbănescu (1999) provides a list of the following nine verbs of pain assigning the accusative to their second arguments, to which I add one more verb: a pişca ‘pinch’. In my corpus research on the structure [mă Vpain NPbody part] I have not found any other verbs.5 (10) arde ‘burn’, durea ‘ache’, furnica ‘tingle’, frige ‘burn’, gâdila ‘tickle’, înţepa ‘sting’, mânca (here ‘itch’, primary meaning ‘eat’), strânge ‘tighten’, ustura ‘burn’

As in other languages, most of these verbs are drawn from other semantic fields, have various other argument structures, and their use as a verb of pain is the result of metaphorization (cf. A. Bonch-Osmolovskaya et al. 2007). Only few of them are verbs of pain in their primary use.

5. The corpus I have used is the Romanian version of Google (google.ro).

9

10

Marleen Van Peteghem

The most basic verb of pain is without any doubt durea ‘ache’, which can only be used with two arguments. Four other verbs belong to the semantic field of pain in their primary use: a ustura ‘burn’, gâdila ‘tickle’, înţepa ‘prick, sting’ and pişca ‘to pinch’. However, they differ from durea in that they are often used with an external argument, denoting either the person or another stimulus which causes the pain and occurs in subject position. The body part is then encoded as a direct object and the experiencer as a dative pronoun. This gives rise to an interesting alternation, illustrated in (11)–(12): the same NP denoting a body part can occur as a direct object in the structure with an external argument (11), or as a subject in the structure without an external argument (12). (11) a. Flăcări-le erau atât de aproape, încât simţea-m cum cenuş-a flames-the were so of close that felt-1 how ash-the îmi gâdila gât-ul me.dat tickled throat-the (The flames were so close that I felt how the ashes tickled my throat) b. Atunci când mâini-le voastre se întind pentru îmbrăţişare, nu then when hands-the your refl stretch for hugging not sunt altceva decât tentacule murdare care îmi înţeapă spate-le are other than tentacles dirty which me.dat sting back-the (When your hands stretch to hug, they are nothing but dirty tentacles which sting my back) c. Stropi-i de şampanie lunecă pe gât, iar bulele splashes-the of champagne are slipping by throat and bubbles-the îmi pişcă limb-a me.dat pinch tongue-the (Splashes of champagne are slipping down my throat and the bubbles tickle my tongue) (12) a. Dacă te gâdilă gât-ul, scarpină-ţi ureche-a! if you.acc tickles throat-the scratch-you.dat ear-the (If your throat tickles, scratch your ear!) b. De câteva zile mă înţeapă spate-le în parte-a stângă for several days me.acc sting back-the in side-the left (For several days my back has been hurting on the left side) c. Mă pişcă limb-a şi buze-le de la sare me.acc sting tongue-the and lips-the from salt (My tongue and lips sting because of the salt)

This alternation is close to the causative alternation. However, the difference with the latter alternation is that in the canonical causative alternation an intransitive



Verbs of pain and accusative subjects in Romanian

construction alternates with a transitive one, whereas the alternation illustrated in (11)–(12) involves two transitive constructions. The other five verbs are not directly linked to the field of pain in their primary meaning and express pain by metaphorization. One of them, furnica ‘tingle’, is basically an intransitive motion verb meaning ‘to swarm, to move around in a quick and chaotic way like ants’.6 Used as a verb of pain, it never takes an external argument and can only occur in the same argument structure as durea ‘ache’, with an experiencer and a theme (13b). (13) a. Sute de oameni furnicau în jurul biserici-i hundreds of people ant.v in round-the church.GEN-DEF (Hundreds of people were moving around the church) b. Mă furnică picioare-le me.acc itch feet-the (My feet are itching)

The other four come from completely different semantic fields: frige ‘grill, roast’ and arde ‘burn’ are typical verbs of physical change entering into the causative alternation, strânge means ‘pull together’ and can be included in the class of verbs of deformation (cf. A. Bonch-Osmolovskaya et al. 2007: 114), while mânca ‘eat’ turns into a verb of destruction when used as a verb of pain. (14) a. Căldur-a ta îmi frige piele-a warmth-the yours me.dat burns skin-the… (Your warmth burns my skin) vs. Mă frige piele-a me.acc burns skin-the (My skin is burning) b. Lumin-a îmi arde ochi-i light-the me.dat burn eyes-the… (The light burns my eyes) vs. Mă ard ochi-i me.acc burn eyes-the (My eyes are burning) c. Mesajul îmi strânge inim-a message-the me.dat pulls heart-the… (The message makes my heart shrink) vs. Mă strânge inim-a me.acc pulls heart-the (My heart shrinks) 6. The verb furnica derives from Lat. formicare (cf. o furnică ‘an ant’).

11

12

Marleen Van Peteghem

d. O bacterie îi mănâncă piele-a a bacteria him.dat eats skin-the (A bacteria is eroding his skin) vs. Mă mănâncă piele-a me.acc eats skin-the (My skin is itching)

In summary, all of these verbs (except for durea ‘ache’ and furnica ‘tingle’) display a causative alternation in which the body part NP can occur either as the subject or the object of the verb (cf. (11)–(13), (14)). The case of the experiencer argument is dependent on the syntactic function of the body part NP: dative when this NP is an object, accusative when it is a grammatical subject. 3. [Mă Vpain NPbody part]: Identifying subjects and objects This section will concentrate on the accusative structure in (1). More specifically it will address the question to what extent each of the two arguments displays subject or object properties. 3.1

Is the body part NP a real subject?

As noted before, the body part NP triggers verb agreement and is thus the grammatical subject of the clause. (15) a. Mă doare capul b. Mă dor picioare-le me.acc hurts head-the me.acc hurt feet-the (I have a headache) (My feet hurt)

Given its semantic role of theme, the question arises whether it behaves as a subject in other respects. The most obvious test comes from word order. Romanian is basically a SVO language, but its word order is quite flexible, so that the subject can be easily postposed to the right of the verb. Interestingly, a corpus investigation of each verb with its most frequent subject, both singular and plural (cf. Table 1), shows that the subject occurs far more frequently in postverbal position than in its canonical position before the verb. For all verbs, the inverted word order is the most frequent one, which suggests that the NP is a VP-internal argument (cf. also C. Cilianu-Lascu 2006). Nevertheless, frequencies vary according to the number of the subject, inversion being much more frequent in the singular (85.99%) than in the plural (67.86%).

13

Verbs of pain and accusative subjects in Romanian Table 1. The position of the subject Singular subject

%

S mă V

%

mă V S

%

32800

77.52

9510

22.48

mă ard ochi-i me burn eyes-the

8920

65.98

4600

34.02

mă doare cap-ul me hurts head-the

1250000

89.15

152000

10.85

mă dor ochi-i me hurt eyes-the

399000

64.88

216000

35.12

mă furnică piele-a me itches skin-the

8010

94.02

509

5.98

100

0

0

100

0

0

6

54.55

5

45.45

53

98.15

1

1.85

mă gâdilă degete-le me itch fingers-the

5250

99.98

1

0.02

4200

97.97

87

2.03

mă înţeapă plămâni-i my lungs sting

28

87.5

4

mă mănâncă ochi-i me itch eyes-the

16200

64.26

9010

35.74

mă ustură ochi-i me burns eyes-the

76500

87.23

11200

12.77

mă pişcă ochi-i me irritate eyes-the

2310

99.61

9

0.39

508474

67.86

240829

32.14

mă arde stomac-ul me burns stomach-the

mă frige stomac-ul me burns stomach-the mă gâdilă gât-ul me itches throat-the mă înţeapă inim-a ‘my heart stings’

mă V S

6

Plural subject

mă mănâncă piele-a me itches skin-the

55400

mă strânge inim-a me squeezes heart-the

4370

99.79

9

30000

76.3

9320

2810

92.37

232

7.63

1387649

85.99

226168

14.01

mă ustură piele-a me burns skin-the mă pişcă limb-a me irritates tongue-the Total

50.4

54500

49.6 0.21 23.3

mă furnică tălpi-le me itch feet-the mă frig ochi-i me burns eyes-the

260

S mă V

%

12.5

–

Total

14

Marleen Van Peteghem

This may be explained by the fact that, with most verbs, the agreement is morphologically more marked in the plural than in the singular. However, verbs of the first conjugation, such as ustura ‘burn’, înţepa ‘sting’, furnica ‘tingle’, and gâdila ‘tickle’, do not show a morphological difference between their singular and plural forms in the present tense, and yet furnica, gâdila, înţepa occur the most with an inverted subject. By contrast, mânca ‘eat’, the most metaphorical of the 10 verbs, occurs very easily with both word orders. The relatively high frequency of the non-­inverted word order could be explained in this case by the fact that the subject NP is more easily viewed as an external argument. Further investigation is required to evaluate the impact of the metaphorical use of these verbs on the word order of their arguments. Another fact which supports the analysis of the body part NP as an internal argument is that it can be encoded as a locative PP in an impersonal construction, as in (16). (16) a. Mă doare în piept / în partea stângă / în suflet / la inimă / înspate me.acc aches in chest / in part left / in soul / in heart / in back (I have pain in my chest / on the left side / in my soul / in my heart / in my back) b. Mă înţeapă la inimă / la buric / la plămâni me.acc stings in heart / in belly / in lungs (My heart / my belly stings / my lungs sting)

As shown in Table 2, the verbs under investigation allow the alternation between the personal structure [mă V NP] and the impersonal structure [mă V PP].7 The corpus investigation presented in Table 2 shows that the impersonal construction is used in almost a quarter of the examples (22.55%). Again its frequency varies with the verb, but it is possible with all verbs. Importantly, the experiencer argument bears accusative and not dative case. The accusative is thus assigned to the experiencer in the absence of a nominative, which goes against the hypothesis discussed above that sentences must have no­ minative case. Another fact that supports a VP-internal analysis of the body part NP has already been mentioned in Section 2. Most verbs display a causative alternation, with the body part NP occurring either as the subject of a structure without an 7. I excluded the verb strânge ‘tighten’ because its use is mostly personal when occurring with a PP. Mă strânge la piept şi-mi şopteşte la ureche: […]. me.acc pulls to chest and-me.dat whispers to ear ‘He pulled me to his chest and then whispers in my ear: […]’



Verbs of pain and accusative subjects in Romanian

Table 2.  Impersonal vs. personal use of verbs of pain Impersonal construction

%

Personal construction

%

mă doare în gât me hurts  in throat

72200

33.09

mă doare gât-ul me hurts throat-the

146000 66.91

mă arde  în piept me burns in chest

  110

10.52

mă arde piept-ul me burns chest-the

   936 89.48

mă furnică în tălpi me itches in feet

    3

01.22

mă furnică tălpi-le me itch feet-the

   242 98.78

mă frige în stomac me burns in stomach

    3

33.33

mă frige stomac-ul me burns stomach-the

     6 66.67

mă gâdilă în nas me itches  in nose

  170

45.70

mă gâdilă nas-ul me itches nose-the

   202 54.30

mă înţeapă la inimă me stings  to heart

 2910

06.47

mă înţeapă inim-a me stings heart-the

 42100 93.53

mă mănâncă în gât me irritates  in throat

 2600

97.12

mă mănâncă gât-ul me irritates throat-the

    77  2.88

mă ustură la ochi me burns to eyes

  241

 0.31

mă ustură ochi-i me burn eyes-the

 76500 99.69

mă pişcă la limbă me irritates to tongue

   63

 2.19

mă piscă limb-a me irritates tongue-the

  2810 97.81

78300

22.55

268873 77.45

external argument or the object of a structure with an external argument. In the latter case, the external argument expresses the external cause of the pain and the experiencer is then obligatorily dative marked (cf. also (12)). (17) a. Prea multe senzaţii îmi înţeapă inim-a, parcă ar too many sensations me.dat sting heart-the, as.if have.cond fi ace be needles (Too many sensations sting my heart, like needles) b. Fumul îmi ustură ochi-i smoke.the me.dat burns eyes-the (The smoke stung my eyes) c. Aşa că am lăsat deoparte « răutate-a » care îmi so that have.1 left to.a.side evil-the which me.dat furnică degete-le ant.v fingers-the (So I put aside the “evil” that made my fingers itch)

15

16

Marleen Van Peteghem

As mentioned earlier, the transitive construction with an external argument is possible with all verbs of our list, except for durea ‘hurt’.8 However, it is extremely rare with furnica ‘itch’ and ustura ‘burn’, although it is attested in examples (17b–c). The three facts just presented support the analysis of the body part NP as an internal argument, hence a derived subject, and consequently the analysis of the verbs as unaccusative. Nevertheless, the following unaccusativity test does not corroborate this analysis for all verbs. One of the most reliable criteria of unaccusativity is the use of the past participle as an adjectival modifier of the internal argument. This use is possible with the direct object of transitive verbs and with the subject of typically unaccusative verbs such as pleca ‘leave’, naşte ‘be born’, dispărea ‘disappear’, and sosi ‘arrive’ as in (18). (18) a. Scrisoare-a, scrisă de mână, dar nedatată, este semnată cu letter-the written of hand but undated is signed with iniţialele C.L. initials-the C.L. (The letter, handwritten, but undated, is signed with the initials C.L.) b. Copii-i născuţi la ţară nu sunt robotizaţi, au o lume children-the born at countryside not are robotized, have a world de basm of fairy-tale (The children born at the countryside are not robotized, they live in a fancy world) c. Iniţiativă pentru copii-i dispăruţi initiative for children-the vanished (Initiative for lost children)

As for the verbs of pain, although most of them seem to allow their past participle to modify the body part NP as in (20), this use is ruled out with the most prototypical verb durea ‘hurt’, and also with furnica ‘itch’ and ustura ‘burn’ as in (19), precisely the three verbs which are not or not commonly used with an external argument. This suggests that the past participle construction is only possible when an external argument is present – overtly as in (20a) or covertly as in (20b). 8. A durea can be used with a causative meaning, but its use is then mostly intransitive and the experiencer can never occur in the dative. (i) Adesea, pacienţi-i întreabă dacă implantul dentar doare sometimes patients-the ask whether implant-the dental hurts (Patients sometimes ask if a dental implant hurts)



Verbs of pain and accusative subjects in Romanian

(19) a. *Am plecat cu cap-ul durut have.1sg left with head-the hurt b. *A lucrat cu degete-le furnicate has worked with fingers-the itched c. *A lucrat la calculator cu ochi-i usturaţi de fum Has worked at computer with eyes-the burnt by smoke (20) a. Speriat, copilul cu ochi-i arşi de febră, a ridicat frightened child-the with eyes-the burnt by fever, has shrugged din umeri of shoulders (Frightened, the child with his feverish eyes, shrugged his shoulders) b. Temuţi-i strigoi, cu inima înţepată şi arsă feared-the ghost, with heart-the stung and burnt (The feared ghosts, whose heart was stung and burnt) c. Germani-i aşteaptă cu inim-a strânsă 100.000 de muncitori Germani-the wait with heart-the squeezed 100,000 of workers din est from east (The Germans are waiting with a fearful heart for 100,000 workers coming from the East)

In the same vein, passive is only possible with the verbs able to take external arguments, and not with those such as durea ‘ache’, ustura ‘burn’ and furnica ‘tingle’, which are rarely used with external arguments. (21) a. Ochi-i lui Remy erau arşi de febră eyes-the gen Remy were burnt of fever (Remy’s eyes were glazed with fever) b. Carnea îi era mâncată de lepră the flesh him.dat was eaten by leprosy (His flesh was eroded by leprosy) (22) a. *Cap-ul era durut head-the was ached b. *Ochi-i erau usturaţi eyes-the were burnt

This shows that the use of the past participle is dependent on the presence of an external argument and is not valid as a test for the construction under investigation, where no such external argument is implied. In other words, three of the four tests argue in favor of a VP-internal analysis, while one test argues against it or is at least inconclusive.

17

18

Marleen Van Peteghem

3.2

Is the experiencer argument a real object?

Let us now turn to the experiencer argument. In the construction under investigation, it bears morphological accusative case, which in Romanian is clearly distinct from the dative case in most pronouns. When the experiencer is a full NP, it is also clearly accusative marked; it is then preceded by the preposition pe, which marks animate definite object NPs in Romanian, and it is then mostly doubled by an accusative pronoun. (23) a. Îl doare burt-a pe bebe him.acc aches belly-the acc baby (The baby has a stomach ache) (24) b. L- am culcat pe bebe him.acc have laid acc baby (I laid the baby to sleep)

Morphologically it is thus clearly accusative-marked. Nevertheless, unlike other object NPs, it cannot become the subject of a passive construction (cf. (25)), nor can it be modified by the past participle corresponding to the verb (cf. (26)). (25) a. Mă doare cap-ul  *Sunt  durută de cap me.acc aches head-the am  hurt by head (I have a headache) b. Mă ustură ochi-i.  *Sunt  usturată de ochi me.acc burn eyes-the am  burnt by eyes (My eyes are burning) c. Mă gâdilă urechi-le  *Sunt  gâdilată de urechi me.acc itch ears-the am  itched by ears (My ears are itching) vs. d. Mă deranjează zgomot-ul Sunt deranjată de zgomot me.acc disturbs noise-the am disturbed by noise (The noise disturbs me) (I am disturbed by the noise) (26) a.  *Persoane-le durute [de cap]… persons-the hurt [by head] b.  *Persoane-le usturate [de ochi] persons burnt [by eyes] c.  *Persoane-le gâdilate [de urechi] persons itched [by ears] vs. d. Cărţi-le citite de către studenţi … books-the read by students (The books read by the students…)



Verbs of pain and accusative subjects in Romanian

The above-mentioned data show that, although marked with the accusative, the experiencer does not behave as a prototypical object. Neither does the nominative theme behave as a prototypical subject. Therefore it is not surprising that verbs of pain, cross-linguistically, are the most likely to occur in non-canonical constructions or case frames (cf. G. Bossong 1998; M. Haspelmath 2001; I.A. Seržant 2013). The theme denoting the body part can be interpreted either as a stimulus, a patient or even as a location of the pain, and hence can be encoded either as a nominative subject, an accusative object (as in Lithuanian, cf. I.A. Seržant 2013), or a locative PP. As for the experiencer, because of its animacy and discourse salience, it tends to occur in sentence initial position and take on subject properties. Therefore, it develops in many languages into a dative subject or even a nominative subject, as in English and French (cf. ‘I have a headache’). Only a few languages, mostly older languages (cf. Icelandic or German), assign the accusative to this experiencer argument, thus encoding it as a patient. In most of these languages, the accusative has been replaced with the dative (cf. old German Mich.acc schmerzt der Kopf vs. Mod. German Mir.dat schmerzt der Kopf), phenomenon called “dative sickness” by H. Smith (1994). In other words, the accusative experiencer found with verbs of pain in Romanian is archaic within the context of European languages. However, it does not trace back to Latin, which had a dative structure, just like Modern Spanish (cf. Latin Caput mihi.dat dolet9). Diachronic research is necessary to give a better insight into this very particular evolution of Romanian, where the dative experiencer has apparently been replaced by an accusative one with a few prototypical verbs of pain such as durea ‘ache’, ustura ‘burn’, and has also attracted into this pattern metaphorical verbs of pain, such as arde ‘burn’ and mânca ‘eat’ / ‘itch’. This evolution is surprising since Romanian is even more “dative-experiencer oriented” than other Romance languages (cf. G. Bossong 1998) and uses the dative case frame with a wide range of verbs of pain that we will comment on in the next section. 4. Verbs of pain and the opposition accusative vs. dative subjects Indeed, as mentioned above, verbs of pain without an external argument do not necessarily assign accusative case to the experiencer argument in Romanian. The dative is used obligatorily with all verbs of pain that are not included in the list in (10) (cf. (27)) and it is also mandatory in reflexive constructions, as in (28).

9. Note that LAT. dolet is the etymon of ROM. doare.

19

20 Marleen Van Peteghem

(27) a. Îmi / *Mă bate / tresare / vibrează inim-a me.dat / *me.acc beats / jumps / thrills heart-the (My heart is beating / is jumping / is thrilling) b. Îmi / *Mă merge / vâjâie / vuieşte / pocneşte cap-ul me.dat / *me.acc goes / hums / hums / explodes head-the (My head is working / humming / humming / exploding) (28) a. Mi se rupe / topeşte inim-a me.dat refl breaks / melts heart-the (My heart breaks / is melting) b. Mi se strânge stomac-ul me.dat refl squeezes stomach-the (My stomach is tightening)

Both constructions are productive in that they can be instantiated by a wide range of verbs. The construction in (27) is used with intransitive verbs or verbs used intransitively that express a process taking place within the body part and affect the experiencer only indirectly.10 As for the reflexive construction in (28), it is used with causative verbs in their intransitive use giving rise to an anticausative construction; from a semantic point of view these verbs express a change of state taking place within the body part.11 Both the accusative and the dative experiencer display subject properties in that they bind the definite article of the body part NP, as already noted earlier in Section 1.1, and can control the subjects of gerunds as in (29)–(30). This behavior argues in favor of an analysis of both accusative and dative nominals in this construction as quirky subjects. (29) a. Te doare suflet-ul văzând imagine-a!!! you.acc aches soul-the seeing image-the (Your heart breaks when you see this image) b. Mi se strânge inim-a de multe ori văzând ce se me.dat refl squeezes heart-the of many times seeing what REFL întâmplă în România happens in Romania (My heart often shrinks when I see what is happening in Romania)

10. The construction [îmi.dat V NPbody part] is used with verbs such as: ţiui ‘ring’, amorţi ‘numb’, bate ‘beat’, crăpa ‘split’, curge ‘run’, lăcrima ‘tear’, pocni ‘hit’, tremura ‘shake’, trosni ‘crackle’, etc. 11. Other verbs that enter in this construction are: a se înfunda ‘get blocked’, a se încinge ‘hot’, a se învineţi ‘bruise’, a se coji ‘peel’, a se topi ‘melt’, a se înroşi ‘redden’, a se mişca ‘move’, a se bloca ‘block’, a se usca ‘dry’, a se încălzi ‘warm’, etc.



Verbs of pain and accusative subjects in Romanian

(30) a. Îmi sângerează inim-a văzând cum este ignorată Iubire-a Mea me.dat bleeds heart-the seeing how is ignored love-the mine (My heart bleeds when I see how my love is ignored) b. Îmi vâjâie cap-ul doar citind despre aceste  nesfârşite me.dat howls head-the only reading about these  endless plimbări de la un ghișeu la altul. walks from a counter to other (My head spins when I read about about those endless walks from one counter to another)

What is important to note is that the accusative and the dative are almost always mutually exclusive. Although Table 3 seems to suggest that some of the accusative verbs in the list in (8) can assign the dative to the experiencer, a closer look at the examples shows that the dative case is only used when an external argument is present. The only exception is arde ‘burn’, one of the most typical verbs of the causative/anticausative alternation. I claim that the indeterminacy of arde as to the case Table 3.  Accusative vs. dative case + accusative

%

+ dative

%

mă doare cap-ul me.ACC hurts  head-the

1370000  99.99

îmi  doare cap-ul me.DAT hurts  head-the

   2

 0.01

mă arde piept-ul me.ACC burns chest-the

   1040  45.41

îmi arde piept-ul me burns chest-the

1250

54.59

mă furnică degete-le me.ACC itch feet-the

  31400  99.99

îmi furnică degete-le me.DAT itch feet-the

  16

 0.01

mă frige  stomac-ul    9850 100 me.ACC burns stomach-the

îmi frige stomac-ul    0 me.DAT burns stomach-the

 0

mă gâdilă nas-ul me.ACC itches nose-the

    178   2.16

îmi  gâdilă nas-ul me.DAT itches  nose-the

8060

97.84

mă înţeapă inim-a me.ACC stings heart-the

  46200  98.1

îmi înţeapă inim-a me.DAT stings heart-the

 893

 1.9

mă mănâncă piele-a me.ACC irritates skin-the

  60300  99.99

îmi  mănâncă piele-a me.DAT irritates skin-the

   2

 0.01

mă  ustură ochi-i me.ACC burn eyes-the

  81100  99.99

îmi  ustură ochi-i me.DAT burn  eyes-the

 499

 0.01

mă pişcă limb-a    2570  99.69 me.ACC irritates tongue-the

îmi pişcă  limb-a    8 me.DAT irritates tongue-the

 0.31

1561408  99.38

9762

 0.62

21

22

Marleen Van Peteghem

of the experiencer can be explained by the fact that, because of the causative/anticausative alternation, it is as frequent in intransitive as in transitive constructions and can therefore be used both in the accusative construction, which hosts transitive verbs, and in the dative construction illustrated in (31), which is typically used with intransitive verbs. (31) a. Am avut febră cu frisoane şi senzaţi-a că îmi arde have.1 had fever with chills and feeling-the that me.dat burns piept-ul acum vreo 5 zile de am dat fugă la urgenţe chest-the now about 5 days that have.1 given run to emergency (I had fever with chills and the feeling that my chest was burning about 5 days ago, so that I ran to the emergency) b. Dragul meu, te privesc şi simt că îmi arde dear-the mine you.acc watch and feel.1 that me.dat burns piept-ul chest-the (My dear, I look at you ant I feel that my chest is burning) (32) a. Simţea-m soarele cum îmi arde pieptul felt-1 sun-the how me.dat burns chest-the (I felt how the sun burned my chest) b. Aer-ul aspru îmi arde piept-ul air-the harsh me.dat burns chest-the (The harsh air is burning my chest)

An important difference between the accusative and the dative subject construction is that the accusative is assigned by a very limited set of verbs, whereas the dative is used with a broad set of intransitive and reflexive verbs. In other words, the dative subject structure is productive, whereas the accusative one is not. A possible explanation is that case is not assigned in the same way in these two structures. I will argue that the accusative is assigned to a verb argument and is a lexical case, whereas the dative is assigned to an external possessor and is an inherent case, in line with E. Woolford (2006: 113), who defines lexical case as “idiosyncratic, lexically selected case” and inherent case as “a case inherently associated with certain θ-roles/positions”: lexical case “may occur on themes/ internal arguments” and inherent case “on agentive/external arguments and on (shifted) DP goal argument, but not on themes/internal arguments”. Lexical case is licensed by lexical heads, whereas inherent case is licensed by “little/light v heads” to arguments occurring above the VP. This analysis is supported by the fact that, being an external possessor, the dative experiencer is highly dependent on the encoding of the body part NP, which has to be as an object or a derived subject, in other words a VP-internal NP. In



Verbs of pain and accusative subjects in Romanian

contrast, the accusative experiencer is not dependent on the syntactic encoding of the theme argument, which can be a subject or a locative PP, as shown above (cf. the examples in (16) above), or even an adverb in an impersonal construction. (33) a. Unde te / *îti doare? where you.acc / *you.dat aches (Where does it hurt?) b. Mă / *îmi ustură jos când urinez me.acc / me.dat stings down when urinate.1 (It stings down there when I urinate)

Moreover, an accusative subject allows the theme argument to be absent, as in the examples in (34), where the referent of the theme is difficult to identify, which shows that there is no zero anaphora. This structure is particularly frequent with the verbs durea ‘ache’ and ustura ‘burn’, in their impersonal form, but it is excluded with dative subjects. (34) a. Te / *îţi doare când naşti? you.acc / *you.dat aches when gives.birth (Does it hurt when you give birth?) b. Te / *îţi ustură când urinezi? you.acc / *you.dat burns when urinate.2 (Does it burn when you urinate?)

The fact that with accusative subjects the theme argument can easily alternate with a locative PP or with an adverb and can even be absent makes the theme more adjunct-like, presenting it as the location of the pain, while the experiencer is the most salient argument, directly affected by the verb. Therefore the verb has to express pain that directly affects the experiencer, giving rise to a metonymical interpretation according to which the whole person is affected by the process. By contrast, the dative subject can be used with processes occurring within the body part, but which cannot take the experiencer as a whole as an argument. This leads to the surprising conclusion that the accusative in the construction under investigation is even more subject-like than the dative, although dative subjects are much more common in Romanian: the accusative subject is a verb argument, whose encoding does not depend on the encoding of other arguments, and it can be used as the sole argument. As for the dative subject, like external possessors it is not an argument of the verb, but of the body part noun, and is therefore highly dependent on the encoding of the body part NP.

23

24

Marleen Van Peteghem

5. Conclusions The analysis of the accusative construction [mă Vpain NPbody part] in Romanian argues against the claim that the accusative is dependent on the licensing of the nominative, in that it shows that the accusative can be assigned in constructions lacking a nominative. It has also shown that the accusative displays subject behavior in this construction, just like the dative in similar constructions. The accusative is assigned to verbal arguments, whereas the dative is used with external possessors taking on subject behavior. The argument status of the argument experiencer makes it more subject-like than the dative in that the former is not dependent on the presence and the encoding of another argument, whereas the dative subject necessarily co-occurs with an internal argument denoting the body part encoded as a grammatical subject triggering verb agreement. Further investigation is needed to provide a more complete account of accusative and dative subjects in Romanian, accusative subjects being much rarer than dative subjects in Romanian and in Romance in general. Given the relatively higher frequency of accusative subjects in Slavic languages than in Romance, the question arises to what extent this Romanian construction is a result of the influence from Slavic languages or even German, which has similar accusative subjects (cf. mich/mir schmerzt der Kopf), a question which can only be addressed in a study with a much broader comparative and diachronic perspective.

References Barnes, Michael. 1986. Subject, nominative and oblique case in faroese. Scripta Islandica 37. 13–46. Belletti, Adriana & Luigi Rizzi. 1988. Psych-verbs and theta theory. Natural Language and Linguistic Theory 6. 291–352. doi: 10.1007/BF00133902 Bennis, Hans. 2004. Unergative adjectives and psych verbs. In A. Artemis, E. Anagnostopoulou, & M. Everaert (eds.), The unaccusativity puzzle. Explorations of the syntax-lexicon interface, 84–113. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199257652.003.0004 Bonch-Osmolovskaya, Anastassia, Ekaterina V. Rakhilina & Tatiana I. Reznikova. 2007. Conceptualization of pain: A database for lexical typology. In P. Bosch, D. Gabelaia, & J. Lang (eds.), TbiLLC 2007, LNAI, 110–123. Berlin/Heidelberg: Springer Verlag. Bossong, Georg. 1998. Le marquage de l’expérient dans les langues de l’Europe. In J. Feuillet (ed.), Actance et valence dans les langues de l’Europe, 259–294. Berlin/New York: Mouton/ de Gruyter. Burzio, Luigi. 1981. Intransitive verbs and italian auxiliaries. Ph.D. diss. MIT. Burzio, Luigi. 1986. Italian syntax: A Government binding approach. Dordrecht: Reidel.  doi:  10.1007/978-94-009-4522-7



Verbs of pain and accusative subjects in Romanian

Burzio, Luigi. 2000. Anatomy of a generalization. In E. Reuland Eric (ed.), Arguments and case. Explaining burzio’s generalization, 195–240. Amsterdam: John Benjamins.  doi:  10.1075/la.34.10bur

Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Cilianu-Lascu, Corina. 2006. O mănâncă limba/la langue lui démange. Quelques remarques sur la place du sujet dans les structures possessives en roumain et en français, Enonciation et syntaxe. Recherches ACLIF: Actes du Séminaire de Didactique Universitaire, Association des Chercheurs en Linguistique Française 3. 51–69. Guéron, Jacqueline. 1983. L’emploi possessif de l’article français. Langue française 58. 23–35.  doi:  10.3406/lfr.1983.6413

Guéron, Jacqueline. 1985. Inalienable possession, pro-inclusion and lexical chains. In J. Guéron, H. Obenauer, & J.-Y. Pollock (eds.), Grammatical representation, 43–86. Dordrecht: Foris. Guéron, Jacqueline. 2003. Inalienable possession and the interpretation of determiners. In M. Coene Martine & Y. D’hulst (eds.), The expression of possession in noun phrases. From NP to DP, Vol. II, 189–220. Amsterdam: John Benjamins. doi: 10.1075/la.56.13gue Guéron, Jacqueline. 2007. Inalienable possession. In M. Everaert & H. van Riemsdijk (eds.), The blackwell companion to syntax, 589–638. Malden: Blackwell Publishing. Haider, Hubert. 2000. The license to license: Licensing of structural case plus economy yields burzio’s generalization. In E. Reuland (ed.), Arguments and case. Explaining burzio’s generalization, 31–56. Amsterdam: John Benjamins. doi: 10.1075/la.34.05hai Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages. In A. Aikhenvald, Y. Alexandra, R. M. W. Dixon, & M. Onishi (eds.), Non-canonical marking of subjects and objects, 53–83. Amsterdam: John Benjamins. doi: 10.1075/tsl.46.04has Herschensohn, Julia. 1992. French inalienable binding. In C. Laeufer & T. A. Morgan (eds.), Theoretical analyses in romance linguistics, 367–384. Amsterdam: John Benjamins.  doi:  10.1075/cilt.74.24her

Junker, Marie-Odile, & France Martineau. 1987. Les possessions inaliénables dans les constructions objet. Revue romane 22. 194–209. Kleiber, Georges. 1999. Anaphore associative et relation partie-tout: condition d’aliénation et principe de congruence ontologique. Langue française 122. 70–100.  doi:  10.3406/lfr.1999.6288

König, Ekkehard, & Martin Haspelmath. 1998. Les constructions à possesseur externe dans les langues d’Europe. In J. Feuillet (éd.), Actance et valence dans les langues d’Europe, 525–606. Berlin: Mouton de Gruyter. Marantz, Alec. 2000. Case and licensing. In Reuland Eric (ed.), Arguments and case. Explaining Burzio’s generalization, 11–30. Amsterdam: John Benjamins. doi: 10.1075/la.34.04mar Perlmutter, David M. 1978. Impersonal passive and the unaccusative hypothesis. Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society (BLS 4), 159–189. Perlmutter, David M. & John Moore. 2002. Language-Internal explanation: The distribution of Russian impersonals. Language 78. 373–416. doi: 10.1353/lan.2003.0049 Reuland, Eric. (ed.). 2000a. Arguments and case. Explaining Burzio’s generalization. Amsterdam: John Benjamins. doi: 10.1075/la.34 Reuland, Eric. 2000b. Explaining Burzio’s generalization: Exploring the issues. In E. Reuland (ed.), Arguments and case. Explaining Burzio’s generalization, 1–10. Amsterdam: John Benjamins. doi: 10.1075/la.34.03reu

25

26 Marleen Van Peteghem

Riegel, Martin. 1994. Article défini, anaphore intra-phrastique et relations partie-tout. In C. Schnedecker, et al. (eds.), L’anaphore associative (Aspects linguistiques, psycholinguistiques et automatiques), 233–250. Paris: Klincksieck. Schnedecker, Catherine, Charolles Michel, Kleiber Georges & David Jean (eds.). 1994. L’anaphore associatiave (Aspects linguistiques, psycholinguistiques et automatiques). Paris: Klincksieck. Şerbănescu, Andra. 1999. Dativ posesiv, dativ experimentator. Studii şi cercetări lingvistice L, 1. 19–38. Seržant, Ilja A. 2013. Rise of canonical objecthood with the Lithuanian verbs of pain. Baltic Linguistics 4. 187–211. Smith, Henry. 1994. “Dative sickness” in Germanic. Natural Language and Linguistic Theory 12. 675–736. doi: 10.1007/BF00992930 Spanoghe, Anne-Marie. 1995. La syntaxe de l’appartenance inaliénable en français, en espagnol et en portugais. Frankfurt am Main: Peter Lang. Van Peteghem, Marleen. 2006a. Le datif en français: un cas structural. Journal of French Languages Studies 16. 93–110. doi: 10.1017/S0959269506002286 Van Peteghem, Marleen. 2006b. Anaphores associatives intra-phrastiques et inaliénabilité. In M. Riegel, C. Schnedecker, P. Swiggers, & I. Tamba (eds.), Aux carrefours du sens. Hommages offerts à Georges Kleiber pour son 60e anniversaire, 441–456. Leuven: Peeters. Vergnaud, Jean-Roger & María Luisa Zubizaretta. 1992. The definite determiner and the inalienable construction in French and English. Linguistic Inquiry 23. 595–652. Woolford, Ellen. 2003. Burzio’s generalization, markedness, and locality constraints on nominative objects. In E. Brandner & H. Zinsmeister (eds.), New perspectives on case theory, 299–327. Stanford: CSLI Publications. Woolford, Ellen. 2006. Lexical case, inherent case, and argument structure. Linguistic Inquiry 37(1). 111–130. doi: 10.1162/002438906775321175

Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew* Rivka Halevy

Hebrew University of Jerusalem 

In memory of Prof. Gideon Goldenberg

The paper deals with the non-typical structure and coding properties of ‘existential-like’ constructions in Colloquial Modern Hebrew (CMH), with reference to parallels in some major Indo-European languages. The construction explored consists of an invariable (neuter) predicate incorporating an empty referential subject (S) morpheme, plus an explicit postverbal NP representing the logic-­ semantic subject (S′) that is deficient in topicality and behaves like an O (though it is not a Patient argument). This construction exhibits inconsistency and instability in several aspects of its encoding. Taking the structure-based approach as its starting point, the paper’s main argument is that the construction under investigation is a special impersonal construction displaying a split between the grammatical S and semantic S′. Typologically, it proposes a unified account of the construction in both synthetic inflectional languages like Hebrew, which do not require an expletive/dummy-­ subject, and in analytic inflectional languages like Germanic languages and French that do require it. The paper disputes the assumption that the postverbal NP in this construction is an O or an S that became an O. The underlying assumption of the paper is that a construction is a form-­ meaning-function unit; accordingly, the construction at hand is examined not only from the structural and semantic viewpoint but also from the viewpoint of functional sentence perspective and the speaker’s perspectival choice with respect to the construal of the event.

* I am indebted to my colleagues, Prof. Tamar Zewi for sharing ideas with me in the initial stage of this research, and Prof. Shlomo Izre’el for some valuable comments on an early draft of this paper. This subject was also of great interest to Prof. Gideon Goldenberg, and we had interesting discussions about it. His spirit is everywhere in this paper. doi 10.1075/lis.33.02hal © 2016 John Benjamins Publishing Company

28

Rivka Halevy

1. Introduction This study deals with a marked P(redicate)-initial construction in Colloquial Modern Hebrew (CMH) that does not conform to the syntactic ‘default rules’ of the language, with some cross-linguistic reference to similar phenomena in European languages. The construction under investigation commonly appears in the beginning of a discourse, signals a break in information flow, or answers the question “what happened?”/“what’s going on?”. It shows behavioral and coding properties typical of existential-presentational sentences, though the range of situation types it denotes is broader, and its function is not necessarily to introduce a new entity into the discourse. The predicates participating in this category are classified into four major semantic groups: (a) verbs denoting assertion of existence or non-existence, i.e., verbs of occurrence or disappearance, commencement or ending; (b) “informationally light” predicates (B. Birner 1994) in impersonal passive constructions (e.g., be said, be written and the like); (c) a lexically and morphologically mixed group of predicates with modal-evaluative function, and (d) a subclass of psych-­perception expressions and verbs denoting physical conditions with an S′ representing the stimulus, and another argument representing the undergoer or ‘experiencer’ usually represented by an overt constituent preceded by a dative clitic. At first glance, the classes listed above may seem like a mixed category in terms of their syntactic and semantic properties. But, as will be discussed and elucidated, the predicates appearing in this construction are all “suppressed” by the existential construction and thus conceived as statements of existence (or denial of existence), including evaluative assertions denoting a desired or (undesired) state of affairs. The semantic and syntactic properties of the verbs belonging to this category overlap in many respects with unaccusatives (and some unergatives that conform to the meaning-function of the construction). Like unaccusatives, these verbs occur in V-initial construction. These constructions often have a plain (‘categorical’) counterpart with similar propositional content. The reverse, however, is not necessarily true, as not all plain sentences can be transformed into this type of construction. Existential and presentational sentences are special cases of a broader category traditionally known as thetic.1 A ‘thetic sentence’ is a construction that is 1. For the linguistic aspects of the thetic-categorical distinction, see Ulrich (1985) and references therein. It should be noted that, from a discourse point of view, the term sentence-focus (Lambrecht 2000: 617) seems to be more adequate than ‘thetic sentence’. Nevertheless, as the term ‘thetic’



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

indivisible at the information level and which is therefore regarded as one-­ membered. More specifically, it lacks an overt referential constituent bearing the grammatical and functional properties of a Subject. The hallmark of such sentences is then the absence of the basic predicative relation between subject and predicate (or Topic and Focus). Constructions of this type have been recognized as a category crosslinguistically, either on the basis of their semantic-­pragmatic properties, such as asserting existence of some entity or introducing an entity into the discourse, or on the basis of their morphosyntactic characteristics. The construction under consideration is a special case of ‘thetic’ or sentence focus constructions. Unlike meteo-statements, it features a semantically contentful, referential NP that qualifies as the semantic subject (henceforth S′),2 though this constituent does not display the characteristics of a typical grammatical Subject (henceforth S), or of an independent semantic argument. In SV languages of the configurational type the common strategy for uttering such sentences is through word-order inversion. The non-topic status of S′ tends to be signaled via morphosyntactic, prosodic or behavioral features that are normally associated with a focal Object (henceforth O). Cross-linguistically, such sentences are unstable in terms of Case marking, verb agreement and constituent order, and in a number of languages also in terms of pronominalization. In many languages, existential/presentational sentences are encoded in a special formula consisting of an existential predicate, often in an invariable form, and an expletive/dummy-S. Among Germanic and Romance languages, the ones that use an expletive in existential formulas are, notably, English (there is), German (es gibt), Scandinavian, e.g. Swedish (det är), and French (il y a); others, like Spanish (hay), do not require an expletive but employ a V-initial construction with an impersonal existential predicate.3 As for Hebrew, it employs the existential (EXS) particle (or “verboid” in Rosén 1977) yeš ‘[there]-is/are’ (negative: 'en) without an is widely known in linguistic literature, I will occasionally use it, though as a neutral term, stripped of its philosophical implications. Other adequate terms are ‘ostension dispersonnelle’ (J.  Damourette & É. Pichon 1952, vol. 4: 464–465) and ‘covert pragmatic impersonals’ (D.  Creissels 2006, 2008); some other terms used are presentative sentences (Bolinger 1977), neutral descriptions (Kuno 1972), news sentence (Schmerling 1976), and event-reporting sentences (Lambrecht 1994; Sasse 1987). 2. In French grammar there is traditionally a distinction between le sujet reél and le sujet apparent. For example, in Grammaire de l’Académie française (Paris 1932: 186–187): “Quand le sujet apparent est il, le verbe employé impersonnellement reste au singulier, même si le sujet réel est au pluriel: Il y a des fruits. Il tombe de grosses gouttes. Il est venu deux personnes”. Cf. also Lazard (1994: 2–3). 3. See Gast and Haas (2011) for a distinction between languages allowing V-initial order in thetic sentences (‘thetic-V1 languages’) and languages disallowing such an order (‘thetic-XV

29

30

Rivka Halevy

overt expletive S. In the past and future tenses, the invariable yeš is replaced by the fully-declinable existential verb lihyot (‘to be’). In spoken language, the existential predicate often occurs in the default form of third person masculine singular4 regardless of the number and gender of the following S′ argument, namely as a verb-form incorporating a zero (0) S morpheme (i.e., a semantically empty morpheme in 3SG.M), as will be explained below. In assertions of possession, Hebrew uses a special type of the existential construction in which the existence of the possessed argument is in focus. To mark possession, Hebrew employs a construction equivalent to French être à (i.e., note that Hebrew is a non-habere language-type), featuring yeš (neg. 'en) or lihyot + the possessor preceded by a datival clitic + an NP denoting the possessee, e.g., yeš/haya/yihye la-hem bayit (they have/had/will have a house; lit. 0-is/was/will-be to-them a house).5 Unlike in some European languages (e.g. English), the Hebrew existential construction does not exclude a definite NP or pronoun representing the S′ NP.6 In MH, such a definite S′ is preceded by a differential Object marker (DOM),7 e.g., haya 'et ha-sefer ba-sifriya (The book was in the library; lit. 0-was DOM the-book in-the-library) / haya 'oto ba-sifriya (It was in the library; lit. 0-wasPRO.ACC.3.M.SG in the library). According to the Unaccusative Hypothesis (D. Perlmutter 1978), widespread in generative grammar, the sole dependent of unergatives is an external argument, which patterns with the Subject of transitive verbs, while the sole dependent of unaccusatives is an internal argument, which patterns with the Object. According to Burzio’s generalization (1986, 2000), a verb lacking an external argument fails to languages’). Compare also to Russian, where there is no need for a dummy, e.g. prišlo tri čeloveka (there came3SG three manPL) (Malchukov & Ogawa 2011: 30). 4. Benveniste (1966: 228): “La forme dite de 3e personne comporte bien une indication d’énoncé sur qulqu’un ou quelque chose, mais non rapporté à une ‘personne’ spécifique […] il ne présente que l’invariant inhérent à toute forme d’une conjugaison […] c’est même la forme verbale qui a pour function d’experimer la non-personne”. 5. For further discussion and examples, see Berman (1980: 761–762), Glinert (1989: 179– 181), Kuzar (2012: 87–95), Halevy (2013: §5 and references therein). Note that Hebrew is a nonhabere language-type; therefore, the possessor, preceded by a datival clitic, cannot be interpreted as the grammatical S. 6. Unlike Modern Hebrew, in Biblical Hebrew existential yeš occurs only with indefinite subjects. 7. Morphological Case-marking has disappeared in Hebrew, as in many languages elsewhere, or in some cases replaced by adpositional markers, e.g. definite nouns in O-position are preceded by the DOM 'et.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

assign accusative Case. Based on a structural analysis of the construction at hand, and in line with CG approach, I aim to show that, contra to what is entailed by Burzio’s generalization, the postverbal NP in our construction is not interpreted as an Object (Patient/Theme) and does not function as one; rather, as will be explained, its status is construction-specific or part of the grammaticalization of the clause-type construction. Across languages, Case, agreement (or lack thereof) and constituent order generally serve to mark grammatical relations. However, as will be demonstrated below, these features also have much to do with information structure (Givón 1976b; Siewierska 2004; and others). In many languages defined as SV, existential/presentational sentences place the S′ in a postposed position (conventionally in Object position). However, S′ may apparently play a role in verb agreement – though, as will be further argued, contrary to a canonical Subject, it does not govern agreement but attracts it – as in the case of English and German, where the verb generally agrees with the S′, as opposed to French il y a sentences, where it does not.8 So far these constructions have received little attention in the study of contemporary Hebrew.9 The phenomenon whereby a subset of unaccusatives (and some unergatives) appear with a S′ that is deficient in topicality and behaves like an O, yet in some cases attracts agreement (as will be argued below), is commonly described in the literature as a case of mismatch between morphological and syntactic alignment. Constructions of this kind feature in recent literature about grammatical asymmetry (A. Malchukov 2008, among others), which has been widely discussed in the literature on impersonals and unaccusativity, most notably in Romance languages (D. Perlmutter 1983; B. Levin & M. Rappaport-Hovav 1995). It is unsurprising that these constructions exhibit inconsistency and instability, both synchronically and diachronically, across many languages. For example, in presentational constructions in colloquial English, the S′ can appear in the accusative case, e.g. There’s him, and there’s you and me (D. Bolinger 1977: 116), and it does not always control agreement with the existential verb, e.g. There’s some difficulties (T. Givón 1990[1984]: I, 380, II, 743–744). In CMH, these constructions likewise exhibit indifference to S-P agreement in many cases (i.e., the predicate does not agree with the S′ but is realized in the neuter invariable form), though in standard ‘normative’ Hebrew, agreement with S′ is consistently maintained. However, both standard and colloquial MH conventionally mark the S′ as an O

8. In French the structural changes involve word order and agreement (but not Case). 9. But see Kuzar (2012: Ch. 3) and references therein to his two previous papers (in Hebrew) on this issue.

31

32

Rivka Halevy

when definite, which contrasts with the default nominative-accusative alignment in the language. Nonetheless, unlike in “pure” existentials and statements of possession, where this atypical Case-marking is conventionalized, the particular existential-like construction at issue shows variability in both Case-marking and agreement, as shown in (1). (1) a. niš'ar 'et ha-'uga me-'etmol remained3M.SG DOM the-cakeF.SG from-yesterday (There’s the cake left from yesterday; lit. 0-remained the cake…) b. niš'ara ha-'uga me-'etmol remained3F.SG the-cakeF.SG from-yesterday (There’s the cake left from yesterday; lit. remained the cake …) c. niš'ar ha-'uga me-'etmol remained3M.SG the-cakeF.SG from-yesterday (There’s the cake left from yesterday; lit. remained the cake …) d. *niš'ara 'et ha-'uga me-'etmol remained3F.SG DOM the-cakeF.SG from-yesterday (There’s the cake left from yesterday; lit. remained the cake …)

In some cases, S′ is marked for accusative case, as in (1a) for example, and in other cases it is not, as in (1b) and (1c). Agreement shows variability as well: the predicate may display agreement with the S′, as in (1b), or fail to agree with it, as in (1a) and (1c). The choice between the two options depends on register (non-­ agreement and inclusion of -'et being typical of colloquial Hebrew), on the speaker’s idiolect, and to some extent on the semantics/pragmatics of the predicate and the postverbal NP (as will be discussed below). It is important to note that there is partial correlation between the two parameters: an accusative marking of S′ precludes verb agreement with it, as shown in (1d). This kind of instability and inconsistency in Case-marking and agreement are a prominent characteristic of the construction under investigation. A. Malchukov & A. Ogawa (2011: 39) observe that the loss of S coding properties proceeds in the following order: position (constituent order) > agreement > Case. And indeed, as will be shown, these three forms of marking interact with each other in our construction. The main questions addressed in this study pertain to: (i) the meaning and structure of the construction at hand; (ii) the coercive force of the construction as a whole; (iii) the speaker’s perspective and the communicative function of this construal. This study is based on and substantiated with data attested in the spontaneous speech of native speakers: conversations in which I participated or which I over-



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

heard, and chats I collected from the internet.10 However, my data do not amount to a corpus-base, and therefore it is impossible at this point to draw definite conclusions about the degree of the phenomenon’s entrenchment in the language. The paper is structured as follows: Section 1 introduced the existential-like construction in CMH with cross-linguistic reference to its equivalents in prominent European languages, and set out the main questions of the study and the data used; Section 2 provides a semantic classification of the situation types and predicates appearing in the construction at issue; Section 3 introduces the ‘coercive force’ of the (existential) construction and deals with its semantic scope; Section 4 reviews and discusses the morphosyntactic properties of the construction in CMH; Section 5 suggests a structural analysis of the construction and discusses the typological significance of this analysis for analytic inflectional languages of the SV type; Section 6 addresses the construction from the functional sentence perspective and the speaker’s perspectival choice with respect to the construal of the event and state of affairs, with emphasis on the controversial status of the postverbal NP; Section 7 concludes and summarizes the main arguments of the paper. 2. A semantic classification It is possible to delineate four subclasses of situation types and lexical predicates participating in the existential-like construction under investigation, though there are also grey areas where the distinction between the subclasses is not clear-cut: Class I: statements of occurrence and disappearance, (in)sufficiency, commencement and ending (and related “negative presentationals”), e.g.,11 (2) qara l-o bediyuq 'et ma še-xašavti  (Fr. il lui est arrivé) happened3M.SG to-him exactly DOM what that-I-thought (Exactly what I thought would happen to him, happened to him; lit. 0-happened to-him exactly what I thought) (3) hispiq l-i ha-be'ay-ot ba-'avoda sufficed3.M.SG to-me the-problemF.PL at-the-work (The problems at work were enough for me; lit. 0-sufficed to me the problems at work)

10. In order to simplify matters for non-speakers of Hebrew, the context in which the examples were uttered is presented only in brief. For reasons of space, I also decided to omit reference to the discourse in which these utterances appeared. 11. See Kuzar (2012: 92–103).

33

34

Rivka Halevy

Class II: “informationally light” predicates in impersonal passive form, such as ‘be said’, ‘be written’, ‘be published’, ‘be determined’, ‘be informed’; ‘be accustomed’, ‘become known’ etc., e.g.: (4) katuv 'et kol ha-šlab-im ba-“help” written3M.SG DOM all (of) the-stageM.PL in-the-“help” (All the stages are written in the ‘help’ [section] lit. 0-is-written all the-stages…) (5) lo yadua' 'et kol ha-prat-im not known3M.SG DOM all the-detailM.PL (All the details are not known lit. 0-is-known all the-details)

Class III: a mixed lexico-syntactic group of predicates in modal-evaluative function,12 e.g.: (6) 'adif 'et ha-'avoda ha-zo't preferableM.SG DOM the-jobF.SG the-thisF.SG (This job is preferable; lit. 0-preferable this job) (7) mat'im et ha-tsva'im ha-'ele ka'n suit3M.SG DOM the-colorM.PL the-these here (These colors suit here; lit. 0-suits these colors here)

Class IV: statements involving psych-perception and physical condition predicates affecting sentient animate beings, usually followed by a facultative dative ‘experiencer’, e.g., ‘understood’, ‘seem’, ‘be remembered’, ‘be familiar’; ‘feel like’, ‘ache/hurt’ etc. (8) nidme (l-o) 'et ze /'et ha-dvar-im ha-'ele appearPTCP.3M.SG (to-him) DOM this/DOM the-thingM.PL the-these (He’s imagining it/ these things; lit. 0-appears (to-him) this/these things)) (9) ba' l-i 'et ha-'agil-im ha-'ele come3M.SG to-me DOM the-earingM.PL the-this. PL (I feel like [having] these earrings; lit. 0-feels to-me like these earrings) (10) ko'ev l-o kol ha-guf / 'et ze 'axare 'imun hurt3M.SG to-him all-the-bodyM.SG / DOM this after training (All his body hurts/ it hurts after a training session; lit. 0-hurts to-him all thebody/it hurts him after a training session)

Semantically, the last subclass constitutes an intermediate case between personal and impersonal constructions, as it involves an (overt or covert) animate (but 12. See Kuzar (2012: 103–113).



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

non-­agentive) participant. That is, although the Predicate appears in the neuter invariable (impersonal) form, there is an animate entity conceived as experiencing or as affected by a feeling/perception in his/her body or mind. However, as in the other subclasses listed above, the NP denoting the sentient ‘experiencer’ appears in the dative case and does not function as the grammatical S, so (synchronically, and perhaps also diachronically) it is misleading to call it a ‘dative-subject’ (or ‘oblique subject’) of the Predicate. In fact, Hebrew can, though infrequently, form an active (transitive) VS construction in which the human participant is an agent, yet the predicate appears in the neuter invariable form, exhibiting a tension between the impersonal grammatical S incorporated in the inflected verb and the animate pronominal postverbal Subject, e.g. domani [dom(e)-ani], meaning ‘I imagine’, lit. ‘imagine-0 (invariable SG.M PTCP verb)-I’. In what follows, it will be argued that the common denominator of all instances of this construction is their existential-like meaning. 3. The ‘coercive force’ of the construction and its semantic scope Construction Grammar (CG) is a functional approach that posits only one level of syntactic representation and assigns meaning not only to words but also to the constructions in which they occur (Ch. Fillmore 1988 and W. Croft 2001, 2003, A. Goldberg 1995, 2003, 2006, among others). According to this approach, language is a gradient and multi-dimensional system. A construction may share properties with another construction in one dimension, and with a different construction in another dimension. Specifically, an S or O can display more of the properties typical to its category, or less. At first glance, the behavioral and coding properties of our construction may seem to be lexical in nature, related to the verb-classes involved. But on closer inspection, it is clear that they have to do with the nature of the construction as a whole. The idea that the meaning of a sentence depends not only on the meanings of the individual words of which it is made, but also on the meaning of the construction as a form-meaning-function unit, is an essential concept in the constructionist approach. In this vein, it is assumed that the grammatical class of elements that can realize functions in various constructions, and the restrictions applying to them, are “derivative from the constructions that define them” (W. Croft 2001: 85). Accordingly, the meaning of our construction is a property of the construction as a whole. More than this, the construction has a coercive force (P. Lauwers & D. Willems 2011): its meaning is forced upon the predicates it includes, suppressing any polysemous meaning. In our case, the existential

35

36

Rivka Halevy

construction as a form-­meaning-function unit imposes its properties on the sentences at hand, which are all construed from a similar perspective. As shown above, our construction is patterned primarily around verbs expressing existence, appearance, disappearance, or the beginning or termination of a state of affairs. Such predicates center on the process itself and are thus compatible with the existential meaning of the construction. ‘Informationally light’ verbs, whose core meaning is depleted to yield a more general meaning, are also potential members of this category. All the same, as has already been pointed out, not only unaccusatives can appear in our construction but also some unergative verbs that are highly compatible with the meaning and function of the construction. In fact, it seems that there is no semantic subclass of intransitive verbs completely excluded from our construction, as long as the verb’s lexical meaning is capable of being suppressed by the ‘coercive power’ of the construction as a single form-meaning-function unit.13 More specifically, predicates occurring in this construction are part of a semantic network, or ‘radial category’ (G. Lakoff 1987), structured around the core meaning of assertion of existence, assertion that ‘such and such actually occurred’ or a modal-evaluative assertion of an (un)expected or (un)desired state of affairs. On the fuzzy border of this category one finds polysemous verbs whose meaning is suppressed by the overall meaning of the construction. To illustrate this, compare the following pair of examples. In the first, the lexeme nigmar displays its core meaning of completion, whereas in the second it takes on a non-existence meaning which is imposed by the construction: (11) a. ha-šiputs nigmar sof sof the-renovationM.SG endedM.SG finally (The renovation process has finally been completed; i.e., started and finished) b. nigmer-u ha-teruts-im finishedPL the-excuseM.PL (No more excuses, i.e., further excuses do not exist)

Likewise in the following pair, where in the first the lexeme ba (‘come’) displays its core meaning of arrival, while in the second it is lexicalized14 in a modal function of describing a desired state of affairs, e.g.,

13. Lambrecht (2000: 623) notes that the class of predicates participating in ‘sentence focus’ sentences is an open class. For a similar observation regarding French, see references in Creissels (2008: 36). 14. Compared to the plain construction the existential-like construction frequently triggers a metaphoric meaning.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

(12) a. ba l-i 'oreax me-xul come3.M.SG to-me guestM.SG from-abroad (A guest from abroad comes to me) b. ba l-i 'et ha-'uga ha-zo't come3.M.SG to-me DOM the-cakeF.SG the-thisF.SG (I feel like having this cake; lit. 0-comes to-me this cake)

As for the S′ constituent, in spite of being referential (generic nouns are commonly overridden by the existential construction),15 it does not control the event, morphologically or semantically. More particularly, the predicate is indifferent to the involvement of S′ in the event, as evident from its invariable form, which excludes any reference to the S′ constituent. 4. Behavioral and coding properties of the construction in CMH16 4.1

Overview

This section will discuss the following properties of the construction: a. P-initial constituent order. b. Tendency towards cohesion between P and the postverbal NP (S′) similar to the cohesion between a verb and an internal argument and to some extent also to the morphosyntactic and lexical cohesion in phraseological units. c. Tendency towards reduced control of verb agreement (especially in colloquial speech), where P occurs in the invariable-neuter form, yielding a morphosyntactic and logical tension between the semantic S′ and the grammatical S. d. An accusative particle is occasionally inserted before a definite, pronominal anaphoric/deictic S′ constituent (sometimes even after invariable nominal predicates, typically adjectival predicates17). e. While a plain S constituent bound by a matrix antecedent can undergo prodrop (‘equi omission’), the ‘hybrid’ S′ in the utterances at issue is not omitted but tends to recur in a pronominal accusative form.

15. Cf. Givón (1976: 173): “many languages can introduce indefinite non-generic nouns as subjects only in existential constructions”. 16. See Onishi (2001) for the parameters and properties of “non-canonically marked core arguments”, in languages from different families. 17. Historically, Hebrew adjectives are participial verb forms inflected for past and future tense (in classical Hebrew there is no separate adjective category).

37

38

Rivka Halevy

f. Even when S′ is fronted, playing a role of contrastive-Subject focus, it frequently retains its irregular Object-like marking. g. On a par with the backgrounding of S′ human participants occurring in this construction are often conceived as undergoers or ‘experiencers’, frequently preceded by a datival clitic and infrequently by the accusative marker (though, as noted before, it is misleading to analyze such NPs as “oblique-Subject”). 4.2

Discussion

4.2.1 Constituent order Like classical Arabic, Hebrew is historically a VS(O) language. However, in present-­day Hebrew, as in modern spoken Arabic, the default order is SV(O), apparently due to the influence of European languages.18 Nevertheless, present-­day Hebrew retains some flexibility in constituent order. As noted above, unlike in European languages such as English, German and French, P-initial order in MH is not limited to existential-presentational sentences, nor to passives and unaccusatives as opposed to transitive and unergative verbs, as is sometimes suggested, especially by generativists (e.g., U. Shlonsky 1997: 163).19 For instance, note the following ‘existential-like’ utterance with an unaccusative (or precisely anticausative) predicate, that is excluded in the English ‘subject inversion’ construction:20 (13) nišpax l-i mayim 'al ha-maxšev spilled3M.SG to-me waterM.PL on the-computer (Water spilled on my computer; lit. 0-spilled to me water on the computer)

P-initial order in present-day Hebrew is conventionally found in collocations and various phraseological units, as well as in narrative referring to entities that are not known, given or previously mentioned. In languages of the SV type, subject inversion is considered a marked construction limited to intransitive verbs, notably unaccusatives and presentational sentences, which signals the detopicalization of S (K. Lambrecht 2000; D. Creissels 2006, 2008). This analysis presupposes that, in the logical or ‘canonical’ order, S 18. The disappearance of the classical verb forms with waw-consecutive (e.g. way-yo’mer ‘he said’; lit. and-he-said), which are initial verb-forms, may also be responsible for this change in constituent order. 19. For examples of VS order in ‘existential-like’ construction with an unergative verb, see e.g. (17a), (17b) below. 20. Note that in this case the postverbal NP is equivalent to the internal argument (Patient/ Theme) in the causative construction.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

precedes V (or P), while the postverbal position is occupied by the accusative Patient. However, since Hebrew is not a rigid SV language, there is no reason to assume that it employs ‘subject inversion’ to signal detopicalization.21 4.2.2 Invariability and attraction of agreement Typically, the S-P agreement depends on the semantic feature of control, i.e., on the degree to which the referent of the NP controls the activity in the particular event referred to. Accordingly, S typically controls agreement when it is the Topic,22 while lack of agreement tends to be associated with a non-topical S, predominantly in spoken language.23 Agreement markers necessarily involve a semantic reference to the agreed-with nominal (in synthetic inflectional languages like Hebrew agreement involves morphosyntactic properties of the agreed-with constituent). Indifference to grammatical agreement is frequently found in P-initial sentences.24 K. Lambrecht (2000) notes that lack of agreement is a cross-linguistic characteristic of ‘sentence focus’ utterances. And indeed, it is ubiquitous in the CMH construction under discussion, e.g., (14) nigmar l-i ha-solela maher. kol xameš daqot finished3M.SG to-me the-batteryF.SG fast. Every five minutes yored l-i kama 'axuzim mi-mena go-down3M.SG to-me a-few percentM.PL of-it (My battery runs out quickly. Every five minutes it goes down by a few percent lit. 0- finished to-me the battery…)

CMH even exhibits indifference to agreement in sentences like the following:

21. For word order in MH see Berman (1980, 1992); Givón (1976); Halevy EHLL (2013: §13), among others. 22. Givón (1976b) shows that agreement arises via topic-shifting constructions in which the topicalized NP is coreferential to one argument of the verb. He assumes that the diachronic source of grammatical agreement is some form of ‘topic-agreement’. From a different perspective Corbett (2006), distinguishes between two types of structural agreement: morphosyntactic agreement, and index agreement (i.e., semantic agreement, determined specifically by reference). He also shows that agreement can be a matter of pragmatics. Jespersen (1913: 169) states that “[o]ccassionally […] the verb will be put in the sg., even if the subject is plural; this will especially happen when the verb precedes the subject, because the speaker has not made up his mind when pronouncing the verb, what words are to follow.” 23. See Givón (1976b); Lazard (1998: 17), among others. 24. In Classical and Standard Arabic, a verb form that precedes a nominal subject generally appears in an invariable form of unmarked personal morpheme.

39

40 Rivka Halevy

(15) ba-sof nolad l-a bat eventually born3M.SG to-her a-girl (Eventually she had a girl; lit. 0-born to-her a-girl)

Lack of agreement is observed even when predicates belonging to the relevant semantic subclasses appear in the marked SV order, namely when S′ is fronted as a contrastive subject-focus, for example in impersonal modal-evaluative statements, e.g.,25 (16) toda magia' l-o lo 'oneš gratitudeF.SG arrive (deserve)3.M.SG to-him not punishment (It is gratitude (thanks) he deserves, not punishment; lit. gratitude 0-deserves to him…)

Other languages too exhibit a lack of agreement between V and S′ in colloquial instances of this type of construction, e.g. English there’s a lot of people in the room, and Spoken Italian (P. Koch 2003: 158) c’era dei contadini, (there were lit. was farmers).26 In French, sentences with ‘dummy’ il and an invariable verb that does not agree with S′ are the rule, e.g., il est venu deux soldats (there came two soldiers; lit. it came3.SG three soldierM.PL). There are, however, French ‘existential-­ like’ sentences in which the verb agrees with the postverbal NP (S′), not with the dummy il (SG), e.g., Il arrivent des cars jusque là (there are cars arriving there; lit. it arrivePL carPL till there). Lack of agreement between the expletive and the inflected verb is the rule in the parallel construction in German, e.g., Es kommen so viele Menschen (so many people are coming; lit. it comePL so many people). As pointed out by D. Creissels (2008) regarding French, this ‘ergative’ alignment is not triggered by the semantic feature [−control], but rather has the pragmatic function of expressing an existential/presentational (or ‘thetic’) organization of predication. As for MH, agreement between P and postverbal S′ is the rule in the standard “educated” language (which, accordingly, excludes an accusative marking of the postverbal S′). Thus, one might suggest that the relevant utterances in standard 25. The unmarked order is:

magia' l-o toda lo 'oneš arrive3M.SG (deserves) to-him gratitudeF.SG not punishment (He deserves gratitude not punishment; lit. 0-deserves to-him gratitude not punishment) 26. Compare also to German, where the construction with the dummy marker es + an inflected verb (which does not agree with es) yields VS sentences which differ from the common inversion and may likewise appear with a definite S′, e.g. es schlafen die Menschen in ihren Betten (‘there are people (lit. the people) sleeping in your beds’; lit. it sleepPL the people in your beds). Taken from Goldenberg (2013: 180).



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

MH and in CMH constitute two different constructions (i.e., a personal vs. an impersonal). However, following G. Goldenberg (2013: 180), I am reluctant to adopt such an analysis. Instead, I assume that both are impersonal constructions and have an impersonal zero S incorporated into the predicate. As for the agreement between the P and the postverbal NP in standard MH, most crucially in in prototypical yeš/’en existentials (and in the equivalent constructions in several other languages), as suggested by Goldenberg, it is a case of attraction of agreement where a non-Subject morpheme (which is not the verbal suffix) attracts agreement. Alternatively, to quote J. Bybee et al. (1994), it can be viewed as a case of “harmonic occurrence of grams [grammatical constituents] in context” [emphasis mine].27 That is, because the incorporated S is semantically empty, whereas the postverbal NP is the S′, the latter, rather than the former, “attracts” the agreement, even though it is not the grammatical S. The result of the tension in our construction between the empty person-morpheme incorporated in the finite verb-­form and the prototypical function of a Topical referential S is thus evoked in the attraction of agreement (as evident in formal-normative Hebrew as well as in some other languages worldwide). Evidently, in order to fully understand the instability and shifts that characterize the structure of these sentences, various sociolinguistic aspects of the relationship between “educated” and colloquial language should be considered. Also, in the case of MH, it seems useful to investigate the diachronic development of these constructions, while recognizing the role played by the language policy of the revivers who directed the development of MH. However, since the focus of this study is on the synchronic aspects of the construction, this issue is set aside for future research. 4.2.3 Object-like marking of postverbal S′ It has been claimed in the literature that unaccusatives do not assign accusative Case to their internal arguments. This claim is not without exceptions. In Hebrew both unergative and unaccusative verbs can, in special cases, be followed by NPs in the accusative Case, even when this NP is deficient in Patient properties, as elucidated, e.g. above, in (1), (2), (4), (5), (8), (10) with unaccusatives and in (18) below with the unergative verb equivalent to English ‘came out’. Furthermore, accusative marking of S′ is also common in CMH impersonal modal-­evaluative statements with an adjectival predicate, as evidenced in (6) and (7) above.

27. Bybee et al. (1994: 294): “Sequence of tense phenomena may be attributed to harmony, and other types of agreement, such as negative agreement, or even number and gender agreement, may result from the harmonic occurrence of grams in context.”

41

42

Rivka Halevy

As already pointed out in ‘pure’ existentials and in assertions of possession in MH, definite postverbal NPs conventionally appear in the accusative case. The existential-like utterances at issue, on the other hand, exhibit inconsistency in this respect, as they are still undergoing an active process of grammaticalization.28 As an existential-like construction, our construction typically involves postverbal NPs with “weak” referents, i.e., referents that are non-salient and non-­ affected. This is exemplified in the utterances in (17), and (18) with motion verbs. The DOM can occur only with the inanimate NP in (17a) and (18a), and is excluded in (17b) and (18c), where the postverbal NP is animate. (17) a. higia' 'et ha-xavila/ 'et ze ha-yom arrived3M.SG DOM the-packageF.SG / DOM this today (The package/ it arrived today; lit. 0-arrived the package/this today) b. higia' (*'et) ha-menahel arrived3M.SG (*DOM) the-manager (There/here is the manager (he arrived)) (18) a. yatsa 'et ha-misxaq-im ha-'ele la-facebook lo went-out3M.SG DOM the-gameM.PL the-these to-the-facebook not mizman from-time (These games came up onto facebook recently; lit. 0-came-out these games to-facebook …) b. yatsa lanu 'et ha-tisa ha-zo't/'ota came-out3M.SG to-us DOM the-flightF.SG the-thisF.SG/ACC.PRO.F.SG be-ta'ut by-mistake (We got this flight by mistake; lit. 0-came-out to-us this flight …) c. hinne yotse (*'et) ha-menahel here go-out3M.SG (*DOM) the-managerM.SG (There is the manager coming out; lit. here comes-out the-manager)

However, unlike in many European languages that commonly avoid encoding pronominalized postverbal NPs in the equivalent constructions,29 since, as is argued, weak pronouns cannot be used to introduce new referents, in Hebrew this constraint is not evident, as shown in (17a), where the postverbal NP can be a deictic pronoun marked with accusative Case, and likewise pronominalized in 28. Object marking of S arguments is known in Biblical Hebrew, especially ‘incipient’ DOM (see Khan 1984; Zewi 1997); however, this is not the pragmatic-driven phenomenon of MH. 29. In French such encoding would yield an ungrammatical utterance, e.g. Il tombe une goutte (‘there fall a drop’) vs. il *la tombe (lit. itACC.PRO drop) (Lambrecht 2000: 636).



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

the accusative in (18b). In fact, in CMH, it is sometimes preferred or even obligatory to explicitly realize an accusative pronominal S′, as seen in the following coordinate proposition: (19) mitsta'er, qarta i-havana, ve-lo iqre sorry, happened misunderstanding, and-not will-happen3M.SG 'ot-a šuv ACC-PRO.3.F.SG again (Sorry, there was a misunderstanding, and it will not happen again)

In many languages pro-drop in conjoined coordinate propositions is restricted to topical Subjects, probably because a referent that is cognitively salient can more easily be ‘taken for granted’ and therefore left unexpressed (K. Lambrecht 2000: 661–662).30 Evidently, our non-topical S′ cannot be ellipted this way. MH also allows accusative postverbal NPs in impersonal passive constructions,31 particularly with verbs of informative value, as demonstrated in (4) and (5) above. Furthermore, even when S′ is fronted as a contrastive subject-focus, it can still retain this non-typical accusative marker, as in e.g., (20) raq 'et ha-ma'amar-im xaser 'adayin only DOM the-articleM.PL lack3M.SG still (Only the articles are still missing; roughly: there is only the articles still missing)

It thus seems uncontroversial to postulate that: S′ can be assigned an Object-like marking if and only if S′ is focalized. At any rate, it is clear that the accusative marker in our construction does not mark the NP as a Patient, or in the terms of Case Grammar (Ch. Fillmore 1968), the postverbal S′ is not understood as having the case role of Theme (viz. Patient). In line with the structural analysis that will be proposed here, it will be argued that this accusative marking of S′ derives from information packaging structure rather than from verb assignment (i.e., the semantic valence of the verb). I will return to this point in Section 6.

30. But compare to German: Es war ein Schaf im Garten, und es fraß Heu, and French: il y avait un mouton dans le jardin, et il mangeait du foin. 31. Impersonal passives governing a direct accusative object are well-known in Arabic and Biblical Hebrew. See Blau (1996: 114–121); Goldenberg (2013: 170–174), among others. Some scholars have inappropriately defined such cases as split passive-active construction.

43

44 Rivka Halevy

5. Structural analysis Before proceeding to the analysis of our construction, it is essential to present the underlying hypothesis of this study and its corollaries for its structural analysis. 5.1

The finite verb as a dimorphemic complex

Following the theory elaborated by leading Semitic scholar G. Goldenberg (1998, 2006, 2013 and elsewhere), this study presupposes that a finite verb-form is a dimorphemic complex consisting of the verbal lexeme and a person morpheme representing the grammatical Subject (S). The suffixed φ-features (of person, number, and gender) represent the person morpheme incorporated into the finite verb-­form.32 Between these members there holds a predicative relation known in the literature as ‘nexus’ (a term coined by O. Jespersen). The recognition of such a predicative link is traced back to Aristotle, who conceived the finite verbform as equivalent to the present form of the verb ‘to be’ plus a verbal attribute. The former, devoid of its lexical value, represents the predicative link, whereas the latter is a deverbal nominal that carries the predicative value. For example, the English periphrastic verbal form I was writing is equivalent to the Hebrew synthetic form katavti. In both languages, the predicate expression is comprised of an S pronominal index plus a predicate-specifying lexeme (e.g., writing, or its Hebrew equivalent, the lexical root k-t-b interwoven into the verb pattern). The ‘nexus’ is the built-in predicative bond (represented in the English example by the past tense of the copulative verb ‘to be’). The notion that the pronominal S is an integral part of the finite verb was recognized by Medieval Arab grammarians as well as by the authors of Grammaire de Port-Royal. As shown by G. Goldenberg (1998), Arab grammarians conceived the Semitic verb-form as the agglutination of a sentence into one synthetic word.33 In well-known European languages like English, French and German, in 32. Present-tense predicates, considered to be nominal participles rather than verbal forms, do not feature pronominal suffixes but obligatorily appear with an independent pronoun or noun representing the Subject. Nevertheless, like adjectives, they incorporate a S morpheme (representated by the inflected number and gender morphemes) + a base lexeme expressing the nominal attribute and a predicative-attributive relation as the linking feature (Halevy 1992; Goldenberg 1995/1998). 33. See Benveniste (1966: 231): “volat avis ne signifie pas ‘l’oiseau vole’, mais ‘il vole, (scil.) l’oiseau’. La forme volat se suffit à elle-même et, quoique non personelle, inclut la notion grammatical de sujet”; Cohen (1975/1984): [le verb] peut constituer par lui-même un énoncé assertif fini du fait de sa nature dimorphématique: base verbale + marque personnelle représentant



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

which the finite verb evolved into an analytic/periphrastic expression, the person pronominal suffixes became independent grammatical markers, equal in status to inflectional morphemes.34 Thus, as the predicate in itself is already the determiner of S, it would be reasonable to assume that, at least in some languages, the presence of another S index next to the predicate would not be indispensable.35 For a clearer insight into the structure of V (or P)-initial sentences and its relevance to the construction at issue, it is illuminating to quote Goldenberg: Finite verbs being intrinsically nexus-complexes, we can probably say that the basic contrast is between (a) sentences centred on the “procès pour lui même” […] without necessarily referring to a noun phrase indicating a thematic subject, and (b) sentences formed of a topical subject noun-phrase and a predicate (whether verbal or not) said, asserted, affirmed or denied concerning that subject. Verbal predicative complexes, inflected or phrasal, which constitutionally incorporate a person morpheme, are involved in both types of sentences. In type (a) the personal morpheme will ideally be neutral and unmarked, since it is there just for filling the formal requirement of the verbal complex; in type (b) it represents the topical subject within the verbal complex. […] In French and some other “SV-languages”, (b) might be regarded as the plain construction, whereas (a) has been described as marking the detopicalization of S.  (G. Goldenberg 2006: 334–335)

5.2

The ‘split-subject’ assumption

As a corollary of the finite verb dimorphemic complex hypothesis, it is plausible to assume that a Subject36 is capable in some cases of being stripped of its referential meaning. This may lead to a discrepancy between the linguistic form (the le sujet”. See specifically Goldenberg’s (1998 [1985]) comprehensive and thorough study “On Verbal Structure and the Hebrew Verb”, as well as Goldenberg (2013: §6.5). 34. Meillet (1921: 177) discusses the French person pronominals je, tu, il, ils, which evolved into simple grammatical markers equal in status to inflectional morphemes and today regarded as part of the verb. 35. Benveniste (1966: 128): “On sait qu’un seul signe suffit à constituer un prédicat. De même la presence d’un ‘sujet’ auprès d’un prédicat n’est pas indispensable: le terme predicative de la proposition se suffit à lui- même puisqu’il est en réalité le determinant du ‘sujet’”. 36. “Subject” is to be understood here as a Theme/Topic, that is, an independent constituent (‘external argument’ in Generative parlance) singled out from the event itself. Such a constituent conventionally controls verb agreement and exhibits the ‘behavioral’ properties of an S. Cf. Li & Thompson (1976: 484): “Subjects are essentially grammaticalized topics” (cf. also Hopper & Traugott 2003: 28–29).

45

46 Rivka Halevy

usual dimorphemic structure of a predicative statement) and the logical contents on the information level (G. Goldenberg 2013: 158, 300). That is, the Subject can split into (a) a formal obligatory S that is referentially empty (“sujet apparent”), and (b) a thematic and referential S′ (“sujet réel”). Insofar as a finite verb requires the actantial presence of a person morpheme, the third masculine singular is the natural candidate for providing the required formal, non-­referential, person-­morpheme, as third person singular is essentially a non-personal form that does not imply any person (note that in Semitic, unlike first and second person singular it is a zero-suffix, not marked morphologically). Such a formal, non-­ referential indexation of the Subject is treated in the literature as ‘zero subject’ (O. Jespersen 1937: §34.3), ‘non person’ (É. Benveniste 1966: 228–231]) or as a pronoun which has weakened and evolved in turn into an agreement marker (T. Givón 1976b: 154). In synthetic inflectional languages like Hebrew, an impersonal verb may occur in the default invariable 3.M.SG, with no need to explicitly encode a separate ‘dummy Subject.37 In contrast, in languages of analytic inflexion, which require a separate indexation of some preceding nominal or pronominal S, a ‘dummy’/‘expletive’/ ‘null Subject’ fulfills this formal requirement, as is shown in the following utterance in Hebrew vs. the equivalent construction in the English translation: (21) laqax l-i ša'a le-hagia' took3.M.SG to-me an-hourF.SG to-arrive (It took me an hour to arrive; lit. 0-took me an hour to arrive)

In such utterances, the split between the formal zero S and the overt semantic S′ is manifested first and foremost by indifference to agreement control. As already indicated, in statements of existence or deficiency, S′ is conventionally marked as an Object-like accusative constituent, e.g., (22) xaser la-nu 'et ha-sfarim šel Benveniste lackM.SG to-us DOM the-books of Benveniste (We are out of Benveniste’s books; lit. 0-lacks to us Benveniste’s books)

The acceptability of such patterning differs greatly from speaker to speaker, and also depends on the register and particular Predicate and semantic S′. Some CMH speakers insert the accusative marker 'et to mark S′ while ignoring agreement control; others don’t. Nonetheless, it should be stressed that this kind of

37. In CMH, invariable verbs in initial position are especially prevalent in “unplanned discourses” (Halevy 2013).



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

variability is not observed in MH in sentences with ‘true’ Objects, where 'et obligatory precedes a definite Object. 5.3

A special type of impersonal construction

An impersonal construction is by definition a construction in which the person index incorporated into the finite verb-form is a semantically empty and does not refer to a topical participant. Thetic statements, which are not built upon a ‘predication base’ (viz., Topic-Subject),38 are typically impersonal. According to our analysis, the existential-like construction under investigation is impersonal, as it lacks a constituent that exhibits the behavioral and/or coding properties considered typical of a referential and topical S. A. Siewierska (2008) distinguishes between semantic, syntactic, morphological, and communicative-functional perspectives of impersonality. From the structural viewpoint, she associates it with the lack of a canonical subject; from the communicative-functional perspective, she defines it in terms of agent defocusing/ backgrounding (comparable to ‘sentence-focus’ as characterized by K. Lambrecht 2000),39 and from the morphological perspective, she associates it with an invariant marking of the verb for person (in Hebrew invariability is also in terms of gender and number). A. Malchukov & A. Ogawa (2011) distinguish three types of impersonals: (a) impersonals with experiential predicates and an S lacking in agentivity (e.g., psych verbs + an ‘experiencer’); (b) impersonal constructions with a notional S deficient in topicality (e.g., presentationals); and (c) impersonals with a notional S (null S) deficient in referential properties (e.g., meteo-statements). According to this classification, our construction belongs to type b. D. Creissels (2008) distinguishes between ‘simple’, ‘special’ and ‘covert’ impersonal constructions. ‘Simple’ impersonals are ‘canonical’ constructions that display no unusual characteristics in terms of their argument structure or coding properties. Their only distinguishing feature is that the subject role is filled by something other than a canonical referential NP, thereby triggering an arbitrary interpretation of the subject. By contrast, special and covert impersonal constructions – like our construction – may include non-canonical subjects but also imply formal changes to the rest of the construction or a modification of argument structure. 38. In Sasse’s terminology (1987: 555ff.), following Aristotle. 39. From a discourse point of view, ‘sentence-focus’ seems to be a more adequate term than ‘thetic sentence’. Nevertheless, I am also using the term ‘thetic’ as a neutral term, stripped of the philosophical implications usually associated with it in linguistic parlance.

47

48 Rivka Halevy

As to the typology of ‘thetic’ constructions, two types are recognized cross-­ linguistically: (i) thetic-V1, and (ii) thetic-XV (V. Gast & F. Haas 2011). Typical examples of European ‘thetic-V1’ languages are Spanish and Italian, which do not have an obligatory preverbal slot in their structural clause templates and thus do not require a dummy pro (or, according to a different analysis, allow a phonologically silent pro). Examples of ‘thetic-XV’ are Germanic languages and French, which have an obligatory preverbal slot that must be filled by an overt expletive. According to this classification, Hebrew is a ‘thetic-V1’ language-type. In such thetic sentences, the S incorporated into the synthetic inflectional verb-form is semantically empty, represented by an invariable 3rd-person singular morphology, yielding an impersonal construction (G. Goldenberg 2013: 177–179). That is, Hebrew uses a morphosyntactic operation for impersonalizing intransitive verbs. This happens especially (but not exclusively) with intransitive verbs that have an unspecific inanimate argument. By contrast, in ‘thetic-XV’ languages of analytic inflection, which have an obligatory S-slot external to the predicate, a ‘dummy’, ‘expletive’ or ‘null’ Subject fulfills this formal requirement. From this perspective, the Germanic languages and French equivalents of the construction under investigation are also impersonal constructions, or ‘covert impersonasls’, to use Creissels’ term (2008). It is noteworthy, however, that although Hebrew does not require an expletive constituent, a neuter (NEUT) demonstrative pronoun can occasionally be inserted before a sentence headed by an invariable predicate, especially before modal-­ evaluative predicates that have not yet been conventionalized as impersonals. (23) ze hirgiš muzar li-hyot šam thisNEUT felt3.M.SG strange to-be there (It felt strange to be there)

In this utterance, with the depersonalized predicate hirgiš (‘felt’) conveying a modal-­evaluative meaning (probably English-induced), the demonstrative pronoun plays the role of a focus marker (viz., pointing at the rhematicized nominal complement). However, it is important to note that it is not an expletive, as indicated by its translation into ‘this’ instead of ‘it’.40

40. Kuzar (2012: 206–209) characterizes the usage of invariable ze (‘this’) in MH impersonal constructions as “expletive addition in Hebrew”. In light of the structural analysis suggested above, explaining the invariable demonstrative in the constructions at issue as an “expletive” (or dummy subject) is erroneous. Such interpretation seems to reflect the tacit assumption that all sentences require an “external” S quite widespread among linguists dealing with impersonal constructions even when confronted with syntactical inflectional languages like Hebrew.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

V. Gast & F. Haas (2011: 177–178ff.) state that ‘thetic-XV’ languages, which require an expletive (e.g., the Germanic languages and French) are impersonal, as S properties have been transferred to the expletive, whereas ‘thetic-V1’ languages, which do not require an expletive (e.g., all Romance languages except French) are not impersonal, because the postverbal NP retains its status as an S. The unified analysis suggested here – which treats these constructions as impersonal in all these languages, and does not regard the postverbal NP as the grammatical S in any of them – seems to provide a more plausible account for the alignment of these constructions. 6. The functional sentence perspective and the speaker’s perspectival choice 6.1

“Block predication”

Traditionally, the absence of a bipartite division of the proposition into S-P or Topic-Comment is regarded as the criterion for distinguishing between ‘thetic’ and ‘categorical’ sentences.41 The differences between these two kinds of sentences is at the level of information structure and of pespectival structure (V. Borschev & B. Partee 2002). In a ‘categorical’ (plain) sentence, P is in focus and S is within the presupposition, while a ‘thetic’ assertion is a sentence that consists solely of a rheme (viz., is a ‘sentence focus’). In such a sentence, the semantic S′ is not conceptualized as actively involved in the event but as merely appearing ‘on the scene’. In other words, the perspectival center in these sentences is the ‘ground’ (the event or the state of affairs), whereas the ‘figure’ (represented by the S′ constituent) is backgrounded and perceived as part of the ‘rheme-only’ sentence.42 The event or state of affairs is viewed from the outside, as unit: while a plain sentence is subject-­ oriented, a thetic one is event-central (H.-J. Sasse 1987). Since a thetic sentence is a single block of predication, there is a strong tendency to avoid the insertion of lexical material between P and S′. In this respect, our construction shows some affinity to collocations and phraseological units, which disallow the separation of their lexical constituents and are likewise V-­initial 41. See Kuroda (1972), Sasse (1987), Lambrecht (1987a), Matras & Sasse (1995), and references therein. 42. Borschev & Partee (2002) describe Perspectival Structure as metaphorically similar to making a choice of what to track with a video camera: to follow the ‘figure’, or to stay fixed on the ‘ground’. Perspectival structure is not synonymous with information structure, although it bears some similarity to it.

49

50

Rivka Halevy

(R. Halevy 1998). However, it should be borne in mind that the “one-­membered” articulation of the utterances at issue is a necessary, but not a sufficient, condition for their grammatical status. 6.2 S′ is part of the new information It has long been claimed that existential sentences disallow a lexically definite postverbal NP. This has come to be known as the ‘Definiteness Effect’ (or ‘Definiteness Restriction’). According to this view, sentences of this type introduce only ‘weak’ referents, such as indefinite NPs or indefinite pronouns.43 This restriction, proposed by G. Milsark (1977) and others, seems to apply only to a subset of presentational sentences, namely there-sentences in standard English, hay-­sentences in standard Spanish, and to a lesser degree, their equivalents in German and French.44 Elsewhere, the restriction does not necessarily apply, as already observed by many. As for Hebrew, definite NPs, including proper names, are not ruled out in existential (and in existence of possession) constructions, particularly in the spoken language, e.g., (24) ('al tid'ag), yeš 'et Roni še-yavo lehašqot 'et ha-gina (don’t worry), isEXS DOM Roni that-will-come to-water DOM the-garden (don’t worry), we have Roni to come water the garden; lit. 0-is Roni who will come …)

It is worth stressing that notions such as identifiability or familiarity (or lack thereof) are neither a necessary nor a sufficient condition for the felicitous use of the definite or indefinite article. The interpretation of a given definite description arises in a particular context through a combination of lexical, grammatical and pragmatic factors.

43. According to Milsark (1977), “weak” NPs are an indefinite referential expressions (such as indefinite pronouns) or NPs with a cardinal determiner (a/an, some, many etc.). “Strong” NPs are definite referential expression (such as proper names and definite pronouns), NPs determined by a universal quantifier (every, all, etc.), possessives or demonstratives. 44. German and French, unlike English, allow definite NPs in there-constructions, e.g. Ger. es hat die Nachtingall die ganze Nacht gesungen; Fr. Il y a le rossignol qui a chanté toute la nuit; English #there was the nightingale singing all night (cf. Lazard 1994: 14). In spoken French proper names may also occur, e.g. il viendra Louise (‘there will be Louise coming’) (Damourette & Pichon 1930–1952, IV: 464–465); similarly in North Italian dialects, gli è venuto la Maria ‘There came Maria’ (Brandi & Cordin 1989: 121), examples quoted by Goldenberg (2013: 179).



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

Rhematicity is naturally associated with indefiniteness, since it is natural for information flow to move towards a new referent that is introduced into the discourse. However, information flow can also move towards a known referent that is being reintroduced. Furthermore, identifiability is not only achieved through previous mention in the text, but can also be situationally evoked (E. Prince 1981, 1992). In such cases, the NP is identifiable by virtue of its salience in the extra-­ textual context. Also, a definite S′ may represent the type of the entity rather than one of its ‘tokens’ (i.e., a specific/unique referent),45 as exemplified in (22) above, where ‘Benveniste’s books’ stands for a ‘type’ rather than a definite individual entity. In fact, predicates in statements of absence or disappearance necessarily refer to entities whose existence is known to the interlocutor. The existential-like construction under investigation can likewise include definite postverbal NPs, as demonstrated in the following: (25) mofia' l-i raq 'et ha-ktovet šel Roni appear3M.SG to-me only DOM the-addressF.SG of Roni (I have only got Roni’s address; lit. 0-appears to-me only Roni’s address)

This utterance, like the existential utterance in (24) above, is not about Roni or Roni’s address. They present events (or states) that are introduced as new. As already indicated, in our construction, S′ is not conceived as a Topic-Subject,46 but rather as an integral part of the ‘block predication’ as a “rheme-only” sentence. Therefore, from an information structure perspective, the relation of S′ to the rest of the sentence is new, and it does not matter if the referent of S′ is definite or not. To conclude, as evidenced in Hebrew, the NP representing S′ does not necessarily have to be new in itself. It may refer to a familiar entity – whether previously known or situationally inferable – and thus occur as a pronoun, a proper name, or as an NP with the definite article. This does not mean, however, that it does not convey new information, for it is not the speaker’s intent to present it as a Topic but rather as part of the new information conveyed by the sentence as a whole. In other words, S′ conveys new information insofar as the entire sentence conveys new information.

45. Ward & Birner (1995: 731–734); regarding Hebrew see Ziv (1982). 46. Traditionally, the basic characterizations of a Theme/Topic are: (a) the constituent that establishes a relationship of aboutness (Kuno 1972; Lambrecht 1994; Reinhart 1981; among others), i.e., the element about whose referent the proposition is construed, or about which the hearer’s knowledge is increased; (b) the given information; and (c) the constituent that is the predication base (Gr. Hypokeimenon, Lat. sub-jectum). Li & Thompson (1976: 484): “Subjects are essentially grammaticalized topics” (see also Hopper & Traugott 2003: 28–29).

51

52

Rivka Halevy

6.2.1 The construction-specific status of the postverbal NP The neutralization of the S-O opposition is perhaps the most important grammatical feature of existential-like constructions across languages. This might be explained in terms of grammatical change, as is sometimes claimed, especially regarding European languages. In French and in some other European languages,47 the postverbal NP is traditionally viewed as an O in the ‘canonical’ Patient position.48 In his famous essay titled “l’actant H: sujet ou objet?”, G. Lazard (1994) chose to arbitrarily name the postverbal NP “actant H”. He states that il in the French construction can be analyzed as an S – perhaps a “reduced” S, but nevertheless an S, as evident from its agreement relations with the verb avoir – whereas “[l]’actant H n’a donc rien d’un sujet” (ibid: 9). In his typological survey, P. Koch (2003) takes a similar approach. According to this view the diachronic frequency of the so-called “actant H” indicates a continuum between prototypical S participants and prototypical O participants. K.  Lambrecht (2000) argues that, diachronically, the postverbal NP is a detopicalized S devoid of (some or all of) its S properties; consequently, it has a status somewhere between subjecthood and objecthood. Along similar lines, A. Malchukov & A. Siewierska (2011) argue that these constructions are an intermediate stage in a fundamental diachronic change from a transitive to an intransitive construction, or from a participant-centered to an event-centered one. These views presuppose that the NP at issue had S properties at some point in its historical development. However, such accounts cannot be applied universally, and certainly not to Hebrew. This, because Hebrew, which now favors SV as the default order, was originally a VS(O) language, and it still exhibits many VS constructions, which are even obligatory in certain contexts. Moreover, V-initial sentences did not evolve into an existential-like construction via a process of grammaticalization. Therefore, unlike in SV languages, in Hebrew the postverbal NP at issue is not necessarily a ‘postposed’ or ‘inverted’ S, so terms like ‘detopicalization’ and ‘neutralization’, employed from a diachronic viewpoint, are inapplicable. In other words, saying that postverbal S is in the position of a ‘canonical’ O – which implies that in the ‘canonical’ word order the Subject necessarily precedes the verb and the postverbal position is of an O (Patient argument) – does not seem correct for a language-type like Hebrew. 47. See Faarlund (2001: 1158) regarding Norwegian: “The NP argument is now in object position, following the non-finite verb. The only possible subject in the sentence is therefore det [equivalent to dummy ‘there’ in English. RH]”. And see Gast & Haas (2011: 155–157) for a survey on Scandinavian languages and Dutch. 48. Creissels (2008: 15): “Pragmatic impersonals concern intransitive verbs whose S argument takes characteristics of the patient of prototypical transitive verbs”.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

What we are dealing with here is a thematic argument which does not exhibit the coding and behavioral properties of an S but at the same time is not a semantic Patient. That is, it is a “hybrid” construal where the postverbal NP, occurring in an impersonal construction involving intransitive verbs, has the syntactic role of an O, but is identical in its semantic role to the S of the same verb in the plain (‘categorical’) construction. From the perspective of propositional content, this unexpected alignment seems to contribute nothing. However, from an informational viewpoint it does alter the interpretation process.49 For example, when used for rhetoric and stylistic reasons, it may reflect a deliberate break in discourse coherence, for instance by presenting an unexpected event. As indicated, the existential-like construal maps the S′ argument as part of the ‘block predication’. That is, this irregular construal modifies the information structure (or logical structure) of the sentence, canceling the default Topic status of S′ and yielding an existential-like meaning. In line with the structural analysis suggested above, it is evident that, since S′ does not have a Topic role, there is no functional motivation for a semantic and syntactic reference to it; hence there is no anaphoric/deictic index of it marked on the inflected verb. Thus, as a corollary of its non-Topic status, it is frequently marked with a non-­nominative Case. Discussing the parallel construction in Scandinavian languages H. Lødrup (1999: 206–208) suggests to define this Object-like NP in negative terms, as a ‘non-­Subject’ rather than an O. This suggestion seems to me more convincing than equating the postverbal NP with the O position in transitive sentences. Along similar lines, I suggest to treat the postverbal NP in our construction as a constituent neutral of both S and O properties, and thus as a non-independent constituent which is syntactically and functionally an integral part of the thetic predication as a whole. More specifically, in line with the structural analysis developed above, I maintain that the exceptional accusative marking of S′ (alongside neutralization of agreement) serve to distinguish it from the impersonal S index incorporated in the Predicate. In other words, its function is to pose (or delimit) the referent of S′ in the discourse in such a way that it does not coincide with the grammatical impersonal S.50 49. Based on ‘relevance theory’, Kempson (1988: 204) states that “whatever information the speaker presents the hearer ab initio will be immediately considered as accessible for this process. The effect of a dummy pronoun in subject position inhibits this process, for the information associated with the argument of the predicate is withheld”. 50. Lambrecht (2000:623), maintains that the non-nominative marking of S′ is also paradigmatically motivated, by the need to minimally distinguish the ‘sentence focus’ construction

53

54

Rivka Halevy

Thus, in my view, saying that the postverbal NP takes on syntactic properties normally associated with O does not mean that this constituent undergoes a change of grammatical relation, nor does it imply that, diachronically, this patterning represents a continuum between S and O or an intermediate stage between transitive and intransitive (or ‘fluid intransitivity’, as suggested by D.  Creissels 2006, 2008). 7. Summary and conclusions In the framework of Construction Grammar, this study presupposes that the existential construction is the ‘coercive force’ behind the form-meaning-­function of the utterances under investigation. From the semantic viewpoint, it was shown that the ‘existential’ meaning can be imposed upon unaccusative and some unergative verbs of various lexical classes, not only on purely ‘existential’ predicates. The main argument of this study was that the ‘non-canonical’ alignment of these utterances in CMH is not a result of ‘subject-verb inversion’ performed as a strategy of detopicalization, for MH is not a grammaticalized SV language (although, under the influence of European languages, it generally favors SV order as its default order). The factors motivating the irregular patterning discussed here remain largely pragmatic, related to information packaging structure and to the perspective from which the speaker views the scene. It was argued that both the information structure and perspectival structure of the given utterances trigger a ‘thetic-like’ interpretation whereby, instead of foregrounding the S′ as the ‘figure’ by marking it as the Topic-S, this constituent is conceived as an integral part of the ‘ground’, i.e., the event-centered or “rheme- only” sentence. Accordingly, it was demonstrated that, in the Hebrew construction, the semantic S′ itself may be either old or new to the hearer and in the discourse, and may therefore be represented not only by an indefinite NP but by a definite one (including a pronoun or proper name). In this analysis of the existential-like sentences, I took the structural approach as the starting point. The controversial debate regarding the status of the ‘dummy-­subject’ and of the postverbal NP were resolved based on two principles: (a) the finite verb-form is a dimorphemic complex, and, following from this, (b) the principle of Split-Subject, i.e., a split between the formal indexation of S and the

from its corresponding plain construction. All the same, he emphasizes that ‘detopicalization’ / ‘objectivization’ of S′ is a pragmatically driven phenomenon, involving primarily the expression of the pragmatic relations of topic and focus.



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

semantic indexation of S′. Along these lines, I described the construction at hand as a special type of impersonal construction. The unified account of the existential-like construction proposed here which is applied both to V-initial synthetic inflectional languages that do not require an expletive and to SV languages that do require it, is, to my best knowledge, novel in the literature. As regards the postverbal NP, this study challenged the view that it is an O or an S that has become an O. In my view, the fact that this NP behaves like, and even has coding properties of, an O (albeit inconsistently), does not imply that it has become an O (“Patient”). More specifically, I maintain that the status of this postverbal NP, including its exceptional accusative marking and the neutralization of agreement, are construction specific. Its Object-like encoding is a morphosyntactic device that indicates its exceptional pragmatic function (i.e., as a non-Topic). In other words, this Object-like marking is a device for disambiguating its position as a logic-semantic S′ that does not coincide with the formal empty S morpheme incorporated into the verb. As postulated above, S′ can be assigned an Object-­ like marking if and only if S′ is focalized. Thus, I propose to define the accusative marking of S′ as a special type of ‘pragmatically driven’ Case marking that marks constituents (most commonly S′) as ‘Non-Subjects’. As is well-known, there are different characterizations and types of ‘non-­ canonical’ Subjects, including non-agents, non-topics, Subjects lacking prototypical S properties, partially referential Subjects, and inherently non-referential and null Subjects. Moreover, there is no one-to-one correlation between a syntactic category and a category based on information packaging. This view is based on the assumption that prototypical categories like S and O are not absolute, and that language in general is a gradient system. Generally, following E. Keenan’s (1976) prototype-­based approach, I contend that subjecthood should be regarded as a gradient category, as was argued regarding the construction at issue, where it is a pragmatically-­driven phenomenon. In this vein, we can be assume that not every clause requires determining which argument is which, once we recognize that the foundations on which the given grammatical element is based are pertinent to different areas of language structure (Mithun & Chafe 1999: 591–592). Thus, assuming that language is a gradient system, it could be expected to produce fluid categories, such as a ‘more-or-less’ Subject, as evidenced in the case of the postverbal NP in the construction at issue. The compelling arguments that favor the account proposed here are: (i) P is semantically and syntactically indifferent to the thematic role of the postverbal NP (manifested in morphosyntactic invariability); (ii) It is no coincidence that uncertainty, inconsistency and instability characterize the encoding of the

55

56

Rivka Halevy

construction at issue. That is, in some cases Hebrew speakers insert the accusative marker, while in other cases they avoid it. Likewise regarding agreement: in some cases speakers follow the grammatical rule of agreement (in fact attraction of agreement), and in other cases they do not. There are cases where the “hybrid” nature of the construction is even more striking in MH, notably with ‘pure’ existentials and statements of existence of possession; (iii) This non-­canonical construction (crucially with an accusative-marked S′) is tolerated in Hebrew even with nominal predicate forms. This seems to dispute assumptions made by some linguists (e.g., Haspelmath 2001) regarding passive derivations and agentivity as a possible origin of this kind of construction. The uncertainty, instability and inconsistency in uttering the “hybrid” sentences discussed here prove that the speakers themselves conceive this construction as cognitively and pragmatically non-­typical and divergent from the plain ‘canonical’ construction. Finally, it is interesting to note that, counter to the common intuition, in spontaneous speech speakers of MH prefer to employ the existential-like construction as the unmarked option, while the plain construction is considered as the marked option.

References Aikhenvald, Alexandra, Robert M. W. Dixon & Masayuki Onishi (eds.). 2001. Non-Canonical marking of subjects and objects. Amsterdam: John Benjamins. doi: 10.1075/tsl.46 Benveniste, Émile. 1966. A phrase nominale. In Problèmes de linguistique générale, vol. I, 151– 167. Paris: Gallimard. Berman, Ruth A. 1980. The case of an (S)VO language: Subjectless constructions in Modern Hebrew. Language 56. 759–776. doi: 10.2307/413487 Birner, Betty J. 1994. Information status and word order: An analysis of English inversion. Language 70. 233–259. doi: 10.2307/415828 Blau, Joshua. 1996. On the impersonal passive in the Bible: A comparative study against the background of the impersonal passive in Classical Arabic. Studies in Hebrew Linguistics (in Hebrew), 114–121. Jerusalem: Magness Press. Borschev, Vladimir & Barbara H. Partee. 2002. The Russian genitive of negation in existential sentences: The role of Theme-Rheme structure reconsidered. In Eva Hajicová, Petr Sgall, Jirí Hana & Tomáš Hoskovec (eds.), Travaux du Cercle Linguistique de Prague (nouvelle série), 185–250. Amsterdam: John Benjamins. Bolinger, Dwight. 1977. Meaning and Form. London & New York: Longman. Brandi, Luciana & Patrizia Cordin. 1989. Two Italian dialects and the null subject parameter. In Jaeggli Osvaldo & Kenneth J. Safir (eds.), The Null Subject Parameter, 111–142. Dordrecht: Kluwer. doi: 10.1007/978-94-009-2540-3_4 Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. doi: 10.1007/978-94-009-4522-7



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

Burzio, Luigi. 2000. Anatomy of a generalization. In Eric Reuland (ed.), Arguments and case: Explaining burzio’s generalization, 195–240. Amsterdam: John Benjamins.  doi:  10.1075/la.34.10bur

Bybee, Joan, Perkins Revere, Pagliuca William. 1994. The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago & London: University of Chicago Press. Corbett, Greville. 2006. Agreement (Cambridge textbooks in linguistics). Cambridge: Cambridge University Press. Creissels, Denis. 2006. Syntaxe générale une introduction typologique, 2 vols. Paris: Hermes Science Lavoisier. Creissels, Denis. 2008. Impersonal and related constructions: A typological approach. Text of a series of 3 lectures given at the University of Tartu on June 02–03 2008. 1–52. Croft, William. 2001. Radical construction grammar, syntactic theory in typological perspective, Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780198299554.001.0001 Damourette, Jacques & Éduard Pichon. [1930] 1952. Des mots à la pensée. Essai de grammaire de la langue française, vol. 4. Paris: d’Artrey. EHLL – Encyclopedia of Hebrew language and linguistics. 2013. Geoffrey, Khan (general ed.). Leiden: Brill. Also available online. Faarlund, Jan-Terje. 1998. L’actance des langues germaniques. In Feuillet Jack (ed.), Actance et Valence dans les Langues de l’Europe, 789–809. Berlin: Mouton de Gruyter. Fillmore, Charles. J. 1968. The case for case. In E. Bach & R. T. Harms (eds.), Universals in linguistic theory, 1–88. New York: Holt, Rinehart, & Winston. Fillmore, Charles J. 1988. The mechanisms of construction grammar. Berkeley Linguistics Society 14. 35–55. doi: 10.3765/bls.v14i0.1794 Gast, Volker & Florian Haas. 2011. On the distribution of subject properties in formulaic presentationals of Germanic and Romance: A diachronic-typological approach. In A. Malchukov & A. Siewierska (eds.), 127–166. Givón, Talmy. 1976a. On the VS word order in Israeli Hebrew: Pragmatics and typological change. In Peter Cole (ed.), Studies in modern Hebrew syntax and semantics, 153–181. Amsterdam: North-Holland. Givón, Talmy. 1976b. Topic, pronoun, and grammatical agreement. In Charles N. Li (ed.), Subject and topic, 149–188. New York: Academic Press. Givón, Talmy. 1990 [1984]. Syntax: A functional-typological introduction, vol. 2. Amsterdam: John Benjamins. doi: 10.1075/z.50 Glinert, Lewis. 1989. The Grammar of modern Hebrew. Cambridge: Cambridge University Press. Goldberg, Adele E. 1995. Constructions. Chicago: University of Chicago Press. Goldenberg, Gideon. 1995. Attribution in semitic languages. Langues Orientales Anciennes: Philologie et Linguistique 5–6. 1–20. [= Studies in Semitic Linguistics: Selected Writings. Jerusalem: Magnes. 1998. 46–66.] Goldenberg, Gideon. 1998. On verbal structure and the Hebrew verb. In Studies in Semitic Linguistics: Selected writings, 148–196. Jerusalem: Magness Press. Goldenberg, Gideon. 2006. On grammatical agreement and verb-initial sentences. In Pier-­ Giorgio Borbone, Alessandro Mengozzi, & Mauro Tosco (eds.), Loquentes linguis: Studi linguistic e oriental in onore di Fabrizio A. Pennacchetti, 329–335. Wiesbaden: Harrassowitz. Goldenberg, Gideon. 2013. Semitic languages: Features, structures, relations, processes. Oxford: Oxford University Press.

57

58

Rivka Halevy

Grammaire de l’Académie française. 1932. Paris: Firmin-Didot et cie Halevy, Rivka. 1992. Free and restricted adjectives in contemporary Hebrew (in Hebrew). In M. Bar-Asher (ed.), Language Studies 5–6, 521–536. Jerusalem: Magness. Halevy, Rivka. 1998. Between syntax and lexicon: Restricted collocations in contemporary hebrew (in Hebrew). Jerusalem: Magness Press. Halevy, Rivka. 2006. The functions of the non-lexical ze in contemporary Hebrew. (in Hebrew). Lešonenu 67 (The Academy of Hebrew Language: Jerusalem). 283–307. Halevy, Rivka. 2013. Syntax: Modern Hebrew. EHLL 3. 707–723. Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages. In Alexandra Aikhenvald, Robert. M. W. Dixon, & Masayuki Onishi (eds.), 53–83.  doi:  10.1075/tsl.46.04has

Hoop, Helen de & Andrej Malchukov. 2007. On fluid differential case marking: A bidirectional OT approach. Lingua 117. 1636–1656. doi: 10.1016/j.lingua.2006.06.010 Hopper, Paul & Elizabeth C. Traugott. 2003. Grammaticalization. Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139165525 Jespersen, Otto.1937. Analytic syntax. London: Allen & Unwin. Keenan, Edward. 1976. Towards a universal definition of subject. In Charles N. Li (ed.), Subject and topic, 305–334. New York: Academic Press. Kempson, Ruth M. 1988. On the grammar-cognition interface: The principle of full interpretation. In Ruth M. Kempson (ed.), Mental representations – the interface between language and reality, 199–224. Cambridge: Cambridge University Press. Khan, Geoffrey. 1984. Object markers and agreement pronouns in Semitic languages. Bulletin of the School of Oriental and African Studies 47. 468–500. doi: 10.1017/S0041977X00113709 Koch, Peter. 2003. From subject to object and from object to subject: (De)personalization, floating and reanalysis in presentative verbs. In Giuliana Fiorentino (ed.), Romance Objects: Transitivity in romance languages, 153–185. Berlin – New York: Mouton de gruyter. Kuno, Susumu. 1972. Functional sentence perspective. A case study from Japanese and English. Linguistic Inquiry 3. 269–320. Kuroda, S.-Y. 1972. Categorical and thetic judgments: Evidence from Japanese syntax. Foundations of Language 9. 1–37. [French translation 1973: Jugements catégoriques et jugements thétiques. Languages 30. 82–110]. Kuzar, Ron. 2012. Sentence Patterns in English and Hebrew. Amsterdam: John Benjamins.  doi:  10.1075/cal.12

Lakoff, George. 1987. Women, fire, and dangerous things. Chicago: University of Chicago.  doi:  10.7208/chicago/9780226471013.001.0001

Lambrecht, Knud. 1994. Information structure and sentence form: Topic, focus, and the mental representation of discourse referents. Cambridge: Cambridge University Press.  doi:  10.1017/CBO9780511620607

Lambrecht, Knud. 2000. When subjects behave like objects: An analysis of the merging of S and O in Sentence-Focus constructions across languages. Studies in Language 24(3). 611–682.  doi:  10.1075/sl.24.3.06lam

Lauwers, Peter & Dominique Willems. 2011. Coercion: Definitions and challenges, current approaches, and new trends. Linguistics 49(6). 1219–1235. doi: 10.1515/ling.2011.034 Lazard, Gilbert. 1994. L’actant H: sujet ou objet?. Bulletin de la Société de linguistique de Paris 89(1). 1–28. doi: 10.2143/BSL.89.1.2013024 Lazard, Gilbert. 1998. Actancy. Berlin and New York: Walter de Gruyter [translated from L’actance, Paris 1994]. doi: 10.1515/9783110808100



Non-canonical ‘existential-like’ constructions in colloquial Modern Hebrew

Levi, Yaakov. 2013. Agreement: Biblical Hebrew. In EHLL, vol. 1 (Brillonline). Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the syntax-lexical semantics interface. Cambridge, MA: MIT Press. Li, Charles. N. & Sandra A. Thompson. 1976. Subject and topicSubject and topic: A new typology of language. In Charles N. Li (ed.), Subject and topic, 457–489. New York: Academic Press. Lødrup, Helge. 1999. Linking and optimality in the Norwegian presentational focus construction. Journal of Nordic Linguistics 22. 205–230. doi: 10.1080/03325860050179254 Malchukov, Andrej L. 2008. Animacy and asymmetries in differential case marking. Lingua 118. 203–221. doi: 10.1016/j.lingua.2007.02.005 Malchukov, Andrej L. & Ogawa Akio. 2011. Towards a typology of impersonal constructions. In Andrej Malchukov & Anna Siewierska (eds.), 19–56. Malchukov, Andrej & Siewierska Anna (eds.). 2011. Impersonal constructions. Amsterdam: John Benjamins. doi: 10.1075/slcs.124 Matras, Yaron & Hans-Jürgen Sasse (eds.). 1995. Verb-Subject order and theticity in European languages. Sprachtypologie und Universalienforschung 48 [special issue]. Milsark, Gary. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3. 1–29. Mithun, Marianne & Wallace, Chafe. 1999. What are S, A, and O. Studies in Language 23(3). 569–596. doi: 10.1075/sl.23.3.05mit Onishi, Masayuki. 2001. Non-canonically marked subjects and objects: Parameters and properties. In Alexandra Aikhenvald, Robert M. W. Dixon, & Masayuki Onishi (eds.), 1–51. doi: 10.1075/tsl.46.03oni Perlmutter, David. 1978. Impersonal passives and the unaccusative hypothesis. In Proceedings of the Fourth Annual Meeting of the Berkeley Linguistic Society, 159–189. Berkley, CA: Berkeley Linguistics Society. Perlmutter, David. 1983. Personal vs. impersonal constructions. Natural Language and Linguistic Theory 1. 141–198. doi: 10.1007/BF00210379 Prince, Ellen F. 1981. Towards a taxonomy of given/new information. In Peter Cole (ed.), Radical Pragmatics, 223–244. New York: Academic Press. Prince, Ellen F. 1992. The ZPG letter: Subjects, definiteness, and information-status. In Sandra Thompson & William Mann (eds.), Discourse description: Diverse analyses of a fundraising text, 295–325. Amsterdam: John Benjamins. doi: 10.1075/pbns.16.12pri Rosén, Chaim B. 1977. Contemporary Hebrew. Mouton: The Hague. doi: 10.1515/9783110804836 Reinhart, Tania. 1981. Pragmatics and linguistics. An analysis of sentence topics. Philosophica 27. 53–94. Sasse, Hans-Jürgen. 1987. The thetic/categorical distinction revisited. Linguistics 25. 511–580.  doi:  10.1515/ling.1987.25.3.511

Schmerling, Susan F. 1976. Aspects of English sentence stress. Austin: University of Texas Press. Shlonsky, Ur. 1997. Clause structure and word order in Hebrew and Arabic: An essay in comparative semitic syntax. New York: Oxford University Press. Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press.  doi:  10.1017/CBO9780511812729

Siewierska, Anna. 2008. Introduction: Impersonalization from a subject vs. agent-centered perspective. Transactions of the Philological Society 106. 1–23.  doi:  10.1111/j.1467-968X.2008.00211.x

59

60 Rivka Halevy

Ulrich, Miorita. 1985. Thetisch und Kategorisch: Funktion der Anordnung von Satzkonstituenten am Beispiel des Rumänischen und anderer Sprachen. Tübingen: Gunter Narr. Ward, Gregory & Betty Birner. 1995. Definiteness and the English existential. Language 17. 722–742. doi: 10.2307/415742 Zewi, Tamar. 1997. Subject preceded by ’et in Biblical Hebrew. In Andreas Wagner (ed.), Studien zur Hebräischen Grammatik, 171–183. Freiburg/Göttingen: Universitätsverlag/ Vandenhoeck & Ruprecht. Ziv, Yael. 1982. Another look at definites in existentials. Journal of Linguistics 18. 73–88.  doi:  10.1017/S0022226700007246

IO realization variation in Spanish reverse psych verb sentences* Chiyo Nishida

University of Texas, Austin

This paper examines the little-known morphosyntactic variation involving Spanish psych verbs that take an experiencer IO and a stimulus subject. With these verbs, a dative clitic duplicating the IO is widely assumed to be obligatory in the canonical [IO-V-S] order (A la gente joven LE/*Ø gustan los deportes ‘Young people like sports’). However, naturally occurring data from corpora show that clitic doubling is not obligatory in the non canonical [S-V-IO] order, yielding two variant constructions (Los deportes LE/Ø gustan a la gente joven ‘Sports appeal to young people’). Using written corpus data from Peninsular Spanish, the paper investigates two issues: (a) what is the overall distribution of clitic doubling in [S-V-IO] psych verb sentences?; (b) are there any systematic distributional differences between the two variants? With respect to (a), we found that for none of the 10 psych verbs surveyed was clitic doubling obligatory. With respect to (b), we found the presence of a clitic tends to restrict the referential properties of the lexical IO in terms of animacy, pronominality, individuality, and number. The findings of this study indicate that dative clitics, which are commonly analyzed as IO-V agreement markers, actually make a substantive contribution to the semantics of psych verb sentences.

1. Introduction Spanish reverse psych verbs (psych verbs, hereafter) like gustar ‘to appeal/to like’ take an experiencer IO and a stimulus (or theme) subject, and the canonical word order for psych verb sentences is [IO-V-S], as shown in (1) below. Functionally, sentences like (1) will be felicitous as responses to questions like “What’s happened?” or “What’s new?”, with the focus falling on the entire sentence (Masullo * My special thanks go to Thierry Ruchot, volume editor, and two anonymous reviewers for their helpful comments. I am also indebted to Elisenda Grigsby and Robyn Wright for their help in the preparation of the manuscript. However, I am solely responsible for any errors. doi 10.1075/lis.33.03nis © 2016 John Benjamins Publishing Company

62

Chiyo Nishida

1992; Gutiérrez-Bravo 2007; inter alia). Note that the IO is introduced by a, which is homophonous with the directional preposition, and duplicated by a coreferent dative clitic. The verb agrees with the subject in person and number. (1) A la gente joven LE gustan los deportes.1 to the people-SG young clitic-dat.3.SG appeal the sports ‘Young people like sports.’

It is widely assumed in previous syntax studies that the use of a dative clitic is obligatory in psych verb sentences with an [IO-V-S] order like (1) (Jaeggli 1982; Cuervo 2003/2010; inter alia); thus, the lack of a clitic would yield an ill-formed string, as in (2). (2) *A la gente joven Ø gustan los deportes.

Here we have only used gustar to show the basic morphosyntactic properties associated with psych verbs. However, these properties apply to all other psych verbs, which include asustar ‘to frighten’, encantar ‘to delight’, fascinar ‘to fascinate’, importar ‘to matter’, interesar ‘to interest’, etc., when appearing in sentences with an [IO-V-S] order.2 Psych verb sentences may occur in the non canonical [S-V-IO] order, as shown in (3). (3) Los deportes LE gustan a la gente joven. ‘Sports appeal to young people.’

This sentence has information structure different than (1) and would be felicitous as a response to both “What about sports?” and “Who do sports appeal to?”. In either case, los deportes functions as a discourse-pragmatic relation topic (cf. Lambrecht 1994). However, corpus data show that verbs like gustar may not have a dative clitic duplicating the postverbal IO, as illustrated in (4) and (5), examples extracted from Royal Spanish Academy’s Corpus de Referencia de Español Actual (CREA).3

1. This sentence may also be interpreted as a response to questions like “What do young people like?” or “What about young people?”. In either case, la gente joven functions as the discourse-pragmatic relation topic (cf. Lambrecht 1994). 2. However, many of these verbs are known to be compatible with an accusative as well as a dative clitic, indicating that they may take a DO or an IO. See 2.3 below for a discussion on this point. 3. CREA contains 200 million words. Samples stored are dated 1974–2004 and organized according to sources (Newspaper, Magazines, Books, Miscellaneous, and Oral), fields (Science &



IO realization variation in Spanish reverse psych verb sentences

(4) Les decía, además, que el fútbol Ø gusta a los grandes tenores como Pavarotti, Plácido Domingo y Josep Carreras, … ‘He would tell them, besides, that football appeals to great tenors like Pavarotti, Placido Domingo, and Josep Carreras, … (La Vanguardia, 17/06/1994) (5) Ella Ø gustaba a los hombres, pero sufría muchas decepciones. ‘She appealed to men, but she suffered a lot of disappointments.’  (Llongueras, Lluís. 2001. Llongueras tal cual.  Anécdotas y recuerdos de una vida)

In both sentences, the subject functions as the topic, judging from the contexts in which these sentences appear. In Spanish, the topic is typically a null pronoun (Casielles-Suárez 2004) – pro in the generative tradition –; indeed a null subject may serve as the topic to a predicate not containing a clitic, as shown in (6). (6) … ese faraón tan dulce y simpático no gusta a todo el mundo. Te diré un secreto: ___ no Ø gusta a casi nadie. ‘… that farah so sweet and likable is not appealing to everyone. I will tell you a secret: he is not appealing to almost anyone.’  (Moix, Trenci. 2002.  El arpista ciego. Una fantasia del reinado de Tutankamón)

Data from CREA also reveal that the postverbal non clitic-doubled IO associated with gustar is particularly frequent in relative clause contexts, as illustrated in (7) and (8). (7) era un rubio que Ø gustaba a las chicas … ‘he was a blond boy who appealed to girls …’ 

(El Mundo, 13/04/1995)

(8) … porque ellos saben qué es lo que Ø gusta a la gente. ‘… because they know what (is that which) appeals to people.’ (El Mundo, 28/07/1994)

Following Kuno (1972), Lambrecht (1994) claims that a relative pronoun may also constitute a topic expression just like a regular pronoun. What is more interesting is that gustar may appear with a postverbal clitic-­ doubled IO as frequently as a non-clitic-doubled counterpart, exhibiting apparent IO realization variation. Observe the two sentences appearing with postverbal clitic-­doubled IOs in (9) and (10) below to be contrasted with similar sentences with non-clitic-doubled IOs in (7) and (6), respectively, seen above.

Technology, Arts, Health, Politics, etc.) as well as countries (Spain, Hispanic American countries, the US, and the Phillippines).

63

64 Chiyo Nishida

(9) … seguramente oíste hablar de él cuando estabas en casa, Josefa habla de él todo el tiempo, un tipo que LES gusta a las mujeres … (cf. ex. 7) ‘… you surely heard people talk about him when you were at the house, Josefa talks about him all the time, a guy who appeals to women …’ (Cambio 16, nº 959, 09/04/1990) (10) A Andrés nunca le gustó Bill. Sabía que no LE gustaba a nadie. (cf. ex. 6) ‘Andrés never liked Bill. He knew that he was not likable to anyone.’  (Grandes, Almuden. 2002. Los aires difíciles)

The existence of two IO variants in ditransitive sentences, as shown in (11) and (12), is well known and has been widely studied from both the syntactic and functional points of view. (11) El padre Michel LE entregó a Sole una pequeña campana de bronce. ‘Father Michel gave Sole a little bronze bell.’  (González, Elías. 1999. Quien, como Dios) (12) … a su vez, Ø entregó a Cortés un collar de caracoles, … ‘… in return, he gave a necklace of shells to Cortés’  (Solares, Ignaico. 1994. Nen, la inútil)

However, the IO realization variation involving reverse psych verbs as seen so far is largely unrecognized in previous studies. To the best of our knowledge, there have been only two studies, Demonte (1994) and Vanhoe (2002), that make reference to two IO variants in psych verb sentences in some capacity. First, Demonte looks at several pairs of sentences like the one in (13) below (translation ours). (13) a. b. 

La foto de sí mismo LE gustó mucho a Juan. The photograph of himself clitic-dat.3sg appealed much to John ‘John liked the photograph of himself.’  ?La foto de sí mismo Ø gustó mucho a Juan. (Demonte’s exx. 34a&b, p. 187)

With this pair of sentences, Demonte is concerned with how the presence or the absence of a dative clitic may affect the binding possibilities of the reflexive sí mismo ‘himself ’, which precedes its binder. She observes that with a clitic the sentence sounds fine, whereas without one, the degree of acceptability diminishes. However, she does not engage in a discussion as to the distribution and the nature of the variation at issue. It should be noted, nonetheless, that she does not seem to think that the presence or the absence of the clitic has any effect on the propositional content since she postulates that the two sentences in (13) share the same underlying structure.



IO realization variation in Spanish reverse psych verb sentences

Second, Vanhoe (2002) makes a brief observation that certain psych verbs may occur without a dative clitic. While he lists interesar ‘to interest’ as one such verb, he doubts that gustar may be used without a dative clitic, providing the pair of sentences in (14). (14) a. El teatro *?(le) gusta a Juan. Theater appeals to John b. El teatro (le) interesa a Juan. ‘Theater interests John.’

(Vanhoe’s ex. 34, p. 386) (Vanhoe’s ex. 35, p. 386)

However, Vanhoe states in a footnote that the “rule” (requirement of a dative clitic for gustar) can be easily overridden in certain contexts but admits that “the exact nature still has to be determined” (p. 386). Also, he does not take the word order factor into consideration at all. The scarcity of previous work on the topic calls for a need for a thorough and systematic examination and analysis of the data pertaining to the variation at issue. This paper takes the first step to fulfill this objective by addressing the following questions: Question 1: In what form may the IO of psych verbs like gustar be realized in sentences with a [S-V-IO] word order, mostly with or without a clitic, or equally in both forms? Question 2: Do the two IO variants show any systematic distributional differences?

In order to provide responses to these questions, a corpus study was conducted using Corpus de Referencia de Español Actual (CREA, hereafter) compiled by Spanish Royal Academy. Section 2 presents an answer to Question 1. The findings based on 1,331 tokens of the IO occurring with 10 reverse psych verbs show that unlike in sentences with an [IO-V-S] word order, clitic doubling is not obligatory in those with a [S-V-IO] order. Section 3 deals with Question 2, where we show that there are systematic distributional differences between the two IO variants in terms of certain referential properties. In this section, we also show that the two variant sentences may differ in meaning in certain cases. Section 4 concludes and discusses some implications of this study.

65

66 Chiyo Nishida

2. Distribution of clitic doubling in psych verb sentences with a post verbal IO The goal of this section is to describe the overall patterns of IO realization in psych verb sentences with a [S-V-IO] order. However, we first describe how we selected the verbs used for our frequency study. 2.1

Psych verbs as a heterogeneous verb class

In section one, we defined the fundamental morphosyntactic properties shared by the group of verbs commonly referred to as reverse psych verbs. However, some clarifications must be made before we select the verbs used for the survey. In the first place, the so-called psych verbs do not constitute a homogeneous class semantically, but they are made up of at least two subclasses. Some verbs like gustar ‘to appeal’, encantar ‘to delight’, etc. denote psychological state-­ evoking properties always involving a experiencer, i.e., a sentient IO, whereas others, though commonly classified as psych verbs, may be used with or without a psychological meaning, as illustrated in (15) and (16) below with interesar ‘to interest/to involve’. (15) Al pintor barroco no le interesa la aplicación estricta de fórmulas matemáticas … ‘The Barroque painter is not interested in the strict application of mathematical formulas …’  (Colorado Castellary, Arturo. 1991.  Introducción a la pintura. De Altamira a Guernica) (16) A la lógica no le interesa en absoluto el razonamiento como actividad mental. ‘Logic is not involved at all in reasoning as a mental activity.’  (VV.AA. 1998. Filosofía. 1º bachillerato. Anaya)

Here, whether the sentence has a psychological sense or not depends on whether or not the sentence includes a sentient IO, i.e. an experiencer. Other verbs with a similar property include convenir ‘to suit’ and importar ‘to matter’; verbs like convenir, interesar, and importar may denote certain judgments in place of emotive meanings. With these verbs, although a non-sentient IO, as shown in (16), is rare in sentences with a preverbal IO, it is commonly found in sentences with a postverbal IO. For our survey, we included both subtypes of psych verbs. In the second place, many psych verbs, such as sorprender ‘to surprise’, asustar ‘to scare/to frighten’, etc., can be used not only with a dative but also with an



IO realization variation in Spanish reverse psych verb sentences

accusative clitic, as shown in (17) and (18), respectively, indicating that they have two argument structures associated with them.4 (17) A Bárbara LEDAT sorprendió la cantidad y la naturaleza de la información que él manejaba. ‘Barbara was surprised at the quantity and the nature of the information that he was handling.’  (Rojo, Alfonso. 2002. Matar para vivir) (18) Volvía del pueblo cuando LAACC sorprendió el bombardeo y no logró encontrarla. ‘She was returning from the village when she was caught by surprise by the bombing and couldn’t find it.’  (Regás, Rosa. 1994. Azul)

Note that with verbs like sorprender, it is not possible to tell whether the postverbal, non-clitic-doubled complement, as shown in (19), is indeed an IO or simply a DO accompanied by a differential object marker.5 (19) Su llegada a las playas no es un fenómeno nuevo, ni tampoco sorprende a los científicos, … ‘Their arrival in the beaches is not a new phenomenon, nor does it surprise scientists, …’  (La Vanguardia, 16/07/1995)

Due to this ambiguity, we did not include any of the verbs that are known to take both a dative and an accusative clitic. Finally, those denoting (non)existence, such as faltar ‘to lack’, sobrar ‘to remain in excess’, etc., as in (20), are often included in the group of verbs called gustar-­type verbs (Whiteley 1995; Vázquez Rozas 2006; inter alia). (20) A este niño le falta una oreja, a este caballo le to this boy clitic-dat.3sg lacks an ear, to this horse clitic sobra una pata … is.in.excess a leg ‘This boy is missing an ear, this horse has one leg too many; …’  (Luca de Tena, Torcuatro. 1979. Los renglones torcidos de Dios)

Cuervo (2003) calls these verbs unaccusative existentials and analyzes them differently from psych verbs like gustar ‘to appeal’. She claims that the IO of these

4. In this paper, we do not deal with the question of what factors condition the accusative-­ dative clitic choice. See Parodi and Luján (2000); Ackerman and Moore (2001); Ganenshan (2015); inter alia for different analyses of the said alternation. 5. In Spanish, the so-called a-personal is used to mark the DO with an animate and specific referent (cf. Aissen 2003; von Heusinger 2008; inter alia).

67

68 Chiyo Nishida

verbs is a possessor6 rather than an experiencer. We excluded these verbs from our survey to maintain the semantic uniformity.7 To recapitulate, for our survey, we chose 10 verbs that are known to take only the IO (and not the DO) and fall within the category of psych verbs, i.e., those denoting psychological states or judgements. The final list consists of agradar ‘to please’, convenir ‘to suit’, desagradar ‘to displease’, doler ‘to hurt’, encantar ‘to delight’, extrañar ‘to seem strange’, gustar ‘to appeal’, importar ‘to matter’, interesar ‘to interest’, and repugnar ‘to disgust’.8 2.2

What is the overall distribution of clitic doubling in [S-V-IO] psych verb sentences?

In order to answer this question, we conducted a frequency test using corpus data. Using an on-line corpus, Corpus de Referencia de Español Actual (CREA), we collected tokens of sentences containing the 10 verbs selected that exhibited two structures, as shown in (21) below. Both structures have a preverbal subject and a postverbal IO introduced by a, either with or without a clitic. (21) a. Subject (clitic) V IO … where Subject is overt or null. b. NP querel (clitic) V IO … where querel is a relative pronoun;

The verb may be inflected in any tense of the indicative or subjunctive mood. In addition, we collected tokens in which the verb was in the infinitive (occurring with modal, aspectual, and raising verbs). We only considered 3rd person IOs because they are the only ones that show the variation at issue. 1st and 2nd person IOs do not show the variation at issue because the lack of clitic doubling would yield an ungrammatical sentence, as shown in (22b) below.

6. The IO for these verbs is not always strictly animate, as can be seen in the following example. (i) No es relevante lo que LE falta a este libro. neg is relevant that which clitic-dat.3sg lacks to this book ‘What is the book is lacking is not relevant.’  (La Vanguardia, 02/09/1995) 7. Our preliminary analysis demonstrated that unlike psych verbs, they showed little variation, almost always requiring a clitic in any word order and even for an inanimate postverbal IO, as shown in the Note 6 above. These properties were critically different from those shown by truly psych verbs like gustar, etc. 8. This list follows Vázquez Rozas (2006); however, her list includes those that do not denote psychological states or judgements like suceder/pasar ‘to happen’, tocar ‘to be someone’s turn’, incumbir ‘to concern’, etc., as well as verbs like faltar and sobrar, as shown in (20).



IO realization variation in Spanish reverse psych verb sentences

(22) a. El futbol me gusta a mí. ‘Football appeals to me.’ b. *El futbol Ø gusta a mí.

CREA catalogues its data according to the countries. We chose European Spanish because it provides the largest data banked in CREA. The data collected consisted solely of written sources (newspapers, magazines, literary books, and non literary books) for the sake of uniformity. CREA includes spoken data; however, we did not choose spoken data because they are extremely small. Next, for each verb, we counted how many tokens appeared clitic-doubled (CLD, hereafter) and how many with a non-clitic-doubled (NCLD, hereafter) postverbal IO. Table 1 provides the results for all 10 verbs in raw numbers. The most striking result observed in Table 1 is that for none of the 10 verbs is clitic doubling obligatory in psych verb sentences with a postverbal IO. Overall, the NCLD IO is more frequent than the CLD counterpart, suggesting that the former variant has a broader distribution. The reason for this will become clear in the next section. Also, a great degree of variance is observed among the 10 verbs. We can observe that while three verbs, gustar, doler, and importar, occur with the CLD and the NCLD IO at an almost equal frequency, most verbs appear with the latter IO variant far more frequently than with the former. Here, we will not pursue why this lexical difference in terms of the frequency of clitic doubling exists but simply mention that a similar variance is observed among ditransitive verbs (Nishida 2012). Table 1.  Frequency of postverbal CLD and NCLD IO tokens according to verbs Verbs  1.  2.  3.  4.  5.  6.  7.  8.  9. 10.

agradar desagradar gustar importar convenir doler encantar extrañar interesar repugnar

‘to please’ ‘to displease’ ‘to appeal’ ‘to matter’ ‘to suit’ ‘to hurt’ ‘to delight’ ‘to make.feel.strange’ ‘to interest’ ‘to disgust’

CLD IO

NCLD IO

Total

  6   1 180  28  54  11   8   4  53   0 345 (26%)

 77  28 191  29 219  10  31  31 336  34 996 (74%)

   83    29    371    57    273    21    39    35    389    34 1,331

69

70 Chiyo Nishida

3. Any systematic distributional differences between the two variant constructions? 3.1

Previous analyses on IO realization variation

The co-existence of two forms instatiating the recipient IO in ditransitives, as shown in (11) and (12) above, has prompted some scholars to postulate that the selection between the two variants is conditioned by some discourse considerations. In this regard, Weissenrieder (1995), taking insight from Givón (1976) and Croft (1990), claims that the CLD IO tends to be human and definite, for which it is likely to function as the discourse topic. Koons-Garboden (2002), however, found, through his corpus study, that neither definiteness nor animacy defined the postverbal CLD IO entirely. Instead, he found that it was the preverbal IO that was always duplicated by a clitic when animate and definite.9 Belloro (2007) argues that the three-way IO encodings (by a clitic alone, a CLD IO, and a NCLD IO) directly correlate with three different cognitive states of discourse referents. She proposes that the CLD option is used for referents that are identifiable (i.e. definite) but less accessible in the sense that the encoding by a clitic alone does not suffice. The NCLD option, according to Belloro, is for referents that are neither identifiable (i.e. indefinite) nor accessible. Turning to psych verb sentences with a [S-V-IO] order, the discourse perspectives proposed by Weissenrieder and Belloro do not seem to serve as good instruments for accounting for IO realization variation. First, in these sentences, it is the subject that always functions as the topic; therefore, the IO is either part of the predicate focus or constitutes the focus by itself. This situation holds regardless of whether the postverbal IO is duplicated by a clitic or not. The IO, as a focus expression (or as part of a focus expression), is not categorically identifiable (=definite) or non-identifiable (=indefinite).10 Without taking a discourse perspective like Weissenrieder or Belloro, some have investigated possible links between referential properties and clitic doubling in ditransitive sentences. Based on data gathered from an Argentine newspaper, Aranovich (2012), for instance, observes the definiteness effect on the CLD IO. However, examining both Peninsular and Latin American Spanish data, Aranovich (2011) states that definiteness is not but animacy is a significant 9. This is the construction commonly known as clitic left-dislocatiton (CLLD), where the object dislocated to the left periphery functions as a topic, as in Al padre LE enviaron un libro ‘To the father they sent a book.’ CLLD is not considered as an instance of clitic doubling. 10. Overall, definite IOs are more common in psych verb sentences. However, as we shall see below in 3.2.2, definiteness has no effect on IO realization in psych verb with a postverbal IO.



IO realization variation in Spanish reverse psych verb sentences

predictor for the CLD recipient IO. Nishida (2012) reports that the CLD recipient IO tends to be animate, definite and singular in her corpus study based on Mexican Spanish written data. In Maldonado (2002), we find a brief comment, “the clitic doubling option is not commonly selected when the dative comprises an institution or a mass (p. 18).”11 3.2

Referential effects

Outside of Spanish, referential effects on argument realization have been widely recognized in typological studies (cf. van Lier 2012). Such effects are generally accounted for by way of a set of referential properties (such as animacy, person, number, referentiality, etc.) each made up of two hierarchically ordered values. For instance, animate is higher than inanimate in the animacy hierarchy, whereas singular is higher than plural in the number hierarchy, etc. We postulate that certain referential properties affect the IO realization in psych verb sentences with a postverbal IO and predict that the higher or the lower value in a given hierarchy tends to be correlated with either a CLD or a NCLD IO, respectively. In determining which referential properties are involved in IO realization at issue, we take two steps. As Step 1, we classify the nominal expressions12 standing as the IO of two prototypical psych verbs, gustar ‘to appeal’ and interesar ‘to interest’, into classes. We chose these two verbs because a preliminary search proved that they would provide a broad spectrum of possible nominal classes. Next, based on this classification, we propose a set of referential hierarchies and examine whether we can establish any statistically significant correlation between a particular value of each one of these hierarchies and one of the IO variants. 3.2.1 Step 1: Classification of nominal expressions functioning as IOs First observe the classification in (23) below, where each class is given a descriptive name and accompanied by examples from the corpus.

11. Maldonado’s main proposal, however, is that clitic doubling is a marker of evidentiality in that a speaker uses a clitic to “validate a giving event” (2002: 19), for instance, by witnessing it. However, as reviewed in Nishida (2012), this claim is not supported by her corpus data. 12. In “nominal expressions”, we include the so-called DPs (el niño ‘the boy’), NPs (niños ‘boys’), QPs (varios niños ‘several boys’) as well as non-referential expressions like todo el mundo ‘everyone’, cualquier niño ‘any boy’, etc., which quantify over variables.

71

72

Chiyo Nishida

(23) Classification of nominals standing for IOs: gustar and interesar Class I: Animate (human) individuals: Einstein, los abogados de Conde ‘Conde’s lawyers’ Class II: Human collectives: la gente ‘(the) people’, generaciones futuras ‘future generations’ Class III: Organizations/Institutions: una empresa ‘a corporation’, los sindicatos ‘unions’ Class IV: Locations: nuestro mundo ‘our world’, los dos países ‘the two countries’ Class V: Abstract concept pertinent to humans: la ciudadanía ‘citizenship’, la propiedad ‘ownership’, la dirección ‘the management’ Class VI: (purely) Inanimate objects/concepts: un pensamiento convencional ‘a conventional thourght’, la geografía ‘geography (as a discipline)’ Class VII: Objects possessed by humans/Personalized object: los oídos ‘the ears’, el esqueleto ‘skeleton’, la opinión pública ‘public opinion’

Class I represents the most common type of nominals to serve as the IO for psych verbs; its members are characterized as [+animate] and [+individual]. In this paper, we simply use the property [+animate] to refer to humans since there was little need to make the distinction between humans and animals for the quantitative analyses conducted. The feature [±individual] is used to differentiate individuals and collectives. Class I members may be further characterized with other properties like [±referential], [±pronominal], [±definite], and [±singular]. The different combinations of these features create six subclasses of Class I, as shown in (24). (24) Class I [+animate, +individual]: Subclasses a. [+referential, +pronominal]: Ud. ‘you’ (sg, formal), ellos ‘they’13 b. [+referential, –pronominal, +definite, +singular]: mi hermano ‘my brother’ c. [+referential, –pronominal, +definite, –singular]:14 sus mujeres ‘their wives’ d. [+referential, –pronominal, –definite, +singular]: un trabajador ‘a worker’

13. Pronouns can be [±singular], but this contrast proves to be insignificant for the quantitative analysis conducted later; thus, we disregard this distinction. 14. We define [+definite] those accompanied by a definite article, a possessor or a demonstrative. Therefore, generics in Spanish are also classified as [+definite] because in contrast to English, they are accompanied by a definite article.



IO realization variation in Spanish reverse psych verb sentences

e. f.

[+referential, –pronominal, –definite, –singular]: algunos niños ‘some children/boys’ [–referential]: todo el mundo ‘everyone’, nadie ‘no one’

Class II is also inherently animate but its members contain collective nouns. They are characterized as [+animate, –individual] for our analysis below. Class III, IV, and V are not inherently animate in that they do not directly denote human individuals but may be construed as groups of humans when used as the IO of psych verbs like gustar; thus, they are only implicitly human. These classes are characterized as [–animate] for the quantitative analysis below. Class VI is made up of purely inanimate nominals. These differ from Class III–V because they do not imply animacy at all. Verbs like importar ‘to matter’, interesar ‘to interest’, and convenir ‘to suit’ are compatible with Class VI IOs, as shown in (25) below. (25) Realismo es la palabra que conviene a esta pieza y su representación. ‘Realism is the word that suits this piece and its representation.’ (ABC, 19/04/1986)

We also characterize Class VI as [–animate]. Finally, Class VII includes nominals that refer to inalienable parts of humans like sus oídos ‘his ears’ or personalized objects like opiniones públicas ‘public opinions’. These by themselves do not denote human beings; they are construed as if they were, when appearing as the IOs of psych verbs. We also classified these as [–animate]. 3.2.2 Step 2: Referential effects on IO realization Next, we examine whether any particular referential properties show effects on IO realization. For this purpose, we use the referential hierarchies given in (26) below, which are modified from those proposed in van Lier (2012). As is evident, each hierarchy consists of two opposing values. The value on the left is considered to be hierarchically higher than the one on the right. Next to each hierarchy we have defined the group of tokens for which the hierarchy is examined. For instance, for the pair of values [±pronominal] we only look at those tokens that are [+animate, +referential, +individual] because this contrast is not relevant for [–animate], [–referential] or [–individual] nominals. On the other hand, for the pair [±animate], all tokens are included.

73

74

Chiyo Nishida

(26) Referential hierarchies & domains of application a. [+animate] > [–animate]; All tokens b. [+referential] > [–referential]; [+animate] tokens c. [+individual] > [–individual]; [(+anim), +referential] tokens d. [+pronominal] > [–pronominal]; [(+anim, +ref), +individual] tokens e. [+proper] > [–proper];  [(+anim, +ref, +ind), –pronominal] tokens f. [+definite] > [–definite];  [(+anim, +ref, +ind, –pronom), –proper] tokens g. [+singular] > [–singular];15  [(+anim, +ref, +ind, –pronom), –proper] tokens

3.3

Referential hierarchies and IO realization: Findings

Let us now discuss how each one of the seven pairs of values in the hierarchies in (26) may interact with IO realization in psych verb sentences with a [S-V-IO] order. We start with the pair of values [±animate]. For the [+animate] group, we included the tokens pertinent to both Class I and Class II; for the [–animate] group, we used the tokens belonging to Class III–VII. The survey results are shown in Table 2 below. This table shows that overall, both [+animate] and [–animate] tokens appear without a clitic (NCLD) more frequently than with one (CLD), as already suggested in Table 1. Nonetheless, we see significant distributional differences between Table 2.  [±animate] and IO realization CLD [+animate] [–animate]

NCLD

310 (36%) 673 (64%)   35 (10%) 313 (90%) χ2 = 61.75, p < 0.001

Total 983 34816

15. Generally, referential hierarchies include the property person, where 1st person is higher than 2nd person and the latter is higher than 3rd person. This may appear to be relevant for the IO realization since 1st and 2nd person IOs require clitic doubling, as seen in (23a) above, whereas 3rd person IOs may not, as seen so far. However, 3rd person pronominals require clitic doubling, as shall be seen below. Therefore, it is pronominality, not person that is critical for IO realization because 1st and 2nd person are pronominal by nature. 16. Of the referential hieararchies (b)–(g) in (26) above, only number and definiteness are relevant to [–animate] tokens. Overall, neither property was found significant for IO realization. Similar results were obtained for [animate] tokens. See Tables 7A–8B below.



IO realization variation in Spanish reverse psych verb sentences

the two classes of tokens: [–animate] tokens are more inclined to occur without a clitic than [+animate] ones. This means that the dative clitic tends to avoid co-occurrence with a [–animate] IO. On the other hand, if there is no clitic in the sentence, no such restriction seems to apply. Next, we take all [+animate] tokens and examine how referentiality interacts with IO realization. The survey results are shown in Table 3. Table 3.  [±referential] and IO realization CLD [+referential] [–referential]

NCLD

252(32%) 528 (68%)   58 (29%) 145 (71%) χ2 = 1.09, p = 0.296471 (p > 0.05)

Total 780 203

Unlike animacy, referentiality is not a critical factor for IO realization. In fact, our data shows that the two quantifiers todo el mundo ‘everyone’ and nadie ‘no one’ may appear equally with and without a clitic. Next, we consider the pair of values [±individual], for which we compare [+referential] tokens of Class I and all tokens of Class II. The survey results are shown in Table 4. Table 4.  [±individual] and IO realization CLD [+individual] [–individual]

NCLD

229 (36%) 410 (64%)   23 (16%) 118 (84%) χ2 = 20.14, p < 0.001

Total 639 141

Here we see patterns similar to those for animacy. Overall, the NCLD variant is more frequent regardless of the value for individuality. Nevertheless, the statistical analysis shows that [–individual] tokens are more inclined to appear without a clitic than their [+individual] counterparts. The results here indicate that the dative clitic has a fairly strong tendency to avoid co-occurrence with a collective IO. On the other hand, if there is no clitic in the sentence, no such restriction is observed. Next, we consider the possible interaction between [±pronominal] and IO realization. We looked at all tokens of Class I, after eliminating non-referential expressions since pronominality is a characteristic of referential expressions. The survey results are shown in Table 5 below. Here we see patterns not observed before: [+pronominal] tokens are unanimously realized with a clitic whereas [–pronominal] ones may go either with or without a clitic, the latter option being more frequent. This confirms the widely

75

76

Chiyo Nishida

Table 5.  [±pronominal] and IO realization CLD [+pronominal] [–pronominal]

NCLD

Total

  41 (100%)   0 (0%) 188 (31%) 410 (69%) χ2 = 123.49, p < 0.001

 41 598

accepted claim (Jaeggli 1986; Suñer 1988; inter alia) that the pronominal IO (or DO) must be clitic-doubled,17 whereas the non-pronominal IO is not subject to such a requirement. Next, we examined the pair of values [±proper] for all [–pronominal] tokens, whose result is shown in Table 6. Table 6.  [±proper] and IO realization CLD [+proper] [–proper]

NCLD

Total

98 (48%) 108 (52%) 90 (23%) 302 (77%) χ2 = 37.95, p < 0.001

206 392

Note that [+proper] tokens are divided almost equally between the two variants, whereas [–proper] tokens are significantly more inclined to occur without a clitic. However, recall that [–proper] tokens may be [±definite] as well as [±singular]. We will see whether these features may be affecting the outcome of [–proper] tokens. For this purpose, we divided [–proper] tokens into four groups, which bear different combinations of the two sets of features, and examined how each group is distributed between the two variants. First see the results in Table 7A and 7B below, where the two values for number are compared for both [+definite] and [–definite] tokens, respectively. Table 7A.  [+definite, ±singular] and IO realization CLD [+definite, +singular] [+definite, –singular]

NCLD

35 (52%)   32 (48%) 46 (17%) 220 (83%) χ2 = 35.51, p < 0.001

Total  67 266

17. Given that the [+pronominal] IO must be clitic-doubled, we have to confirm that the statistical results for Tables 2–4 are not skewed due to the inclusion of [+proniminal] tokens. So we re-calculated chi-square values for each table excluding these tokens. The new statistical results remained the same in terms of significance.



IO realization variation in Spanish reverse psych verb sentences

Table 7B.  [–definite, ±singular] and IO realization CLD [–definite, +singular] [–definite, –singular]

NCLD

Total

9 (33%) 18 (67%) 27 1 (3%) 31 (97%) 32 Fisher Probability Test: p = 0.002521388 (p < 0.01)

These tables show that regardless of the value for definiteness, [–singular] tokens are strongly linked with the NCLD IO, whereas [+singular] show no such strong relation with either one of the variants. Let us see if definiteness shows a similar pattern. This time, we kept the number value constant and compared [+definite] and [–definite] tokens. The results are shown in Table 8A and 8B, where the two values for definiteness are compared for both [+singular] and [–singular] tokens, respectively. Table 8A.  [+singular, ±definite] and IO realization CLD [+singular, +definite] [+singular, –definite]

NCLD

35 (52%) 32 (48%)   9 (33%) 18 (67%) χ2 = 2.76, p = 0.096648 (p > .05)

Total 67 27

Table 8B.  [–singular, ±definite] and IO realization CLD [–singular, +definite] [–singular, –definite]

NCLD

46 (17%) 220 (83%)   1 (3%)   31 (97%) χ2 = 4.32, p = 0.037667 (p < .05)

Total 266  32

Table 8A shows that as far as [+singular] tokens are concerned, the two sets of tokens, i.e., [+definite] and [–definite], do not show significant distributional differences. Table 8B shows that when we limited to [–singular] tokens, the two sets of tokens did show significant distributional differences. Nonetheless, the p value here is very high, already approaching non-significant (0.05). From these results, we conclude that definiteness is not an entirely critical factor for IO realization. This result concurs with Koontz-Garboden (2005) and Aranovich (2011), but directly clashes with Weissenrieder (1995), Belloro (2007), and Aranovich (2012), all of whom examined the recipient IO in ditranstive sentences using corpora data. To recapitulate this section, we have found evidence that certain referential properties significantly affect IO realization. As illustrated in (27), one of the values in four hierarchies (given as the last one in each feature complex) is obligatorily or strongly linked with one of the IO variants (CLD or NCLD).

77

78

Chiyo Nishida

(27) Types of objects Examples IO realization a. [–anim] al país ‘to the country’ ←→ NCLD b. [+anim, +ref, –ind] a la familia ‘to the family ←→ NCLD c. [+anim, +ref, +ind, +pron] a ella ‘to her’ === CLD d. [+anim, +ref, +ind, –pron, –prop, ±def, –sg] a los hombres ‘to the men’ ←→ NCLD where ←→ means “is strongly linked to”; === means “must be realized as”.

In contrast, objects with values not mentioned above, as shown in (28), are not significantly linked with either one of the two IO variants; thus, they are equally compatible with both variants. (28) Types of objects Examples IO realization a. [+anim, –ref] a nadie ‘to no one’ ==> (N)CLD b. [+anim, +ref, +ind, –pron, +prop] a José ‘to Joe’ ==> (N)CLD c. [+anim, +ref, +ind, –pron, –prop, ±def, +sg] al niño ‘to the boy’ ==> (N)CLD where ==> (N)CLD means “compatible with both CLD and NCLD IO”.

These results may be generalized to conclude that the CLD IO is much more restricted in terms of referential properties than its NCLD counterpart. First, the CLD IO may be [+pronominal]; in fact, a pronominal IO must be duplicated by a clitic. This is a general rule across both the DO and the IO. Second, if it is [–pronominal], the CLD IO is strongly inclined to be [+animate, +referential, +individual, +proper] or [+animate, +referential, +individual, –proper, +singular].18, 19 In contrast, the NCLD IO is subject to no such constraints. To summarize, the two IO variants are not in complementary distribution as some previous studies have claimed for the recipient IO (cf. Belloro 2007, in particular); there are areas of mutual exclusion as well as those of overlap, where the new and the old IO forms co-exist. In general, the NCLD IO enjoys broader distribution than its CLD counterpart, which partially explains the much higher frequency of NCLD IO tokens in the corpus.

18. Actually, a proper noun is inherently singular. Therefore, these two sets can be generalized as [+animate, +individual, +referential, +singular]. 19. A clitic-doubled IO may consist of a [–referential] expression like a quantifier; however, this seems to be true only if the speaker intends a particular interpretation for the set of bound variables. See 3.4.2 below for a discussion.



IO realization variation in Spanish reverse psych verb sentences

3.4

Do the two IO variant constructions differ semantically?

3.4.1 Event types In recent syntactic studies within the minimalist framework, the IO realization variation, as shown in (11) and (12) above, is assimilated to the English dative alternation (Demonte 1995; Bleam 2001; Cuervo 2003/2010; inter alia). In such studies, it is claimed that all ditransitive verbs like entregar ‘to give/to hand’, en­ viar ‘to send’, etc. uniformly have two meanings associated with them: (a) caused possession represented as [X cause [Y possess Z]] and (b) caused motion represented as [X cause [Z go to Y]], where X is the subject, Y the IO, and the Z the DO. It is claimed that the first meaning is mapped into syntax with a CLD IO and the second with a NCLD IO. Can the uniform polysemous analysis be extended to psych verb sentences? Observe the following pair of sentences, one with a clitic and the other without one. (29) Aquello no LE gustó a Anás. ‘That did not appeal to Anás.’  (Benítez, Juan José. 1980. Caballo to Troya 1) (30) La ironía del candidato nacionalista no Ø gustó a Pérez Touriño. ‘The irony of the nationalist did not appeal to Pérez Touriño.’ (Faro de Vigo, 26/06/2001)

The two clauses do not seem to differ, if we postulate the meaning of psych verb sentences taking an IO experiencer to be as schematized in (31) below. (31) Meaning of psych verb sentences taking an IO experiencer: X evokes a certain psychological state in Y, where X is the entity standing as the subject and Y as the IO; Y is sentient.

See the additional pair of similar strings in (32) and (33) with another psych verb encantar ‘to delight’, both of which equally fit the event scheme in (31). (32) Creo que Alicia en el País de las Maravillas. Siento adoración por este libro, que también LE encantaba a Borges. ‘I believe that (it is) Alice in the Wonderland. I feel adoration for this book, which also used to delight Borges.’  (Cambio 16, nº 974, 23/07/1990) (33) Su humor, cambiante, alegre y juvenil, Ø encantaba a doña Mariana, que fue también alegre en sus primeros años, … ‘His humor, (which was) quick, cheerful, and youthful, used to delight doña Mariana, who was also cheerful in her early years, …’  (Luján, Nestor. 1991. Los espejos paralelos)

79

80 Chiyo Nishida

3.4.2 Quantifier IO Is there any other way that the two variant sentences may differ in meaning? We look at one instance where the presence of a clitic has an effect on the interpretation of psych verb sentenes. Observe a passage from a theater piece, as shown in (34). (34) Daniel: (Cruzando la escena, en dirección a la derecha. Al pasar tras el grupo que forman, junto a los ventanales, Paolo, María Teresa y Aurora, habla en voz alta, a todos a la vez (sic) (Bajando la voz, hablando como para sí mismo.) Digo que hace mucho calor. Si no LE importa a nadie, voy a quitarme la chaqueta. (Pausa. Nadie da la impresión de haberse dado cuenta aún de su presencia.) Pues no. No parece que LE importe a nadie … nada. ‘Daniel: (Crossing the scene, in the direction towards the right. As he passes behind the group (that is) formed by Paolo, María Teresa and Aurora, by the big windows, he speaks in loud voice, to all at the same time (sic) (Lowering his voice, like he’s speaking to himself.) I say it is very hot. If it does not matter to anyone, I am going to take off my jacket. (None gives the impression of having noticed his presence yet.) Well no. It does not seem to matter to anyone … nothing.’  (Sierra, Rodolfo. 1991. Indian Summer)

Note that in the two sentences appearing towards the end with the verb importar ‘to matter’, the IO consists of a negative quantifier nadie ‘no one’ doubled by a clitic. Generally, nadie negates the existence of any human in the universe of discourse that fulfills the condition stated in the sentence. However, in these particular sentences, the context indicates that the set of individuals that nadie quantifies over is a specific set of pepeole, i.e., {Paolo, María Teresa, Aurora}, who have been mentioned earlier. Now observe the passage in (35), focusing on the last sentence, which has an IO consisting of nadie but not doubled by a clitic. (35) Apenas sabía nada de Maica. Ella se negaba sistemáticamente a visitarle, alegando pretextos fáciles. Sus cartas llegaban de tarde en tarde. El abogado había olvidado sus promesas. Su encierro no Ø importaba a nadie en este mundo. ‘I knew almost nothing about Maica. She would systematically refuse to visit him, claiming easy pretexts. His letters would arrive from time to time. The lawyer had forgotten his promises. His reclusion did not matter to anyone in this world.’  (Tomás García, José Luis. 1984. La otra orilla de la droga)

In the sentence at issue, the IO a nadie does not quantify over a specific set of people but all humans in the universe of discourse. This is clearly indicated by the modifier en este mundo ‘in this world.’



IO realization variation in Spanish reverse psych verb sentences

A similar contrast is also observed between (36) and (37), where we find the universal quantifier todo el mundo ‘everyone’. (36) El parque LE gustó a todo el mundo, incluso a aquellos contra los que iba dirigida. Era una novela materialista que le alabaron mucho los partidarios del espíritu. ‘The park appealed to everyone, including those against whom it was directed. It was a materialist novel that the advocates of spirit praised a lot.’  (Millás, Juan José. 2000. Dos mujeres en Praga) (37) En general, ya no se trata de que un anuncio Ø guste a todo el mundo, basta con que consiga atraer la atención y destacar entre los posibles compradores del producto. ‘In general, it is no longer a question of a commercial appealing to everyone, it suffices with the commercial managing to attract attention and to stand out among the possible buyers of the product.’  (El Mundo, 25/05/1995)

In (36), note the conjoined clause incluso a aquellos contra los que iba dirigida, which follows the psych verb sentence, suggests that todo el mundo quantifies over a specific set of people in lieu of everyone in the world. On the other hand in (37), todo el mundo quantifies over all possible humans in the universe of discourse, who are contrasted with a restricted set of people, i.e., los posibles compradores del producto, found in the clause to follow. To summarize this section, a qualitative analysis of the data involving quantifiers has provided evidence that the CLD and the NCLD IO may yield different interpretations from one another. A dative clitic is used by the writer (or a speaker) when he intends to have a quantifier IO to apply to a context-dependent restricted set rather than an open set. 4. Implications In this paper we took a close look at the little known IO realization variation – with or without a clitic – in reverse psych verb sentences with the non-­canonical [S-V-IO] order. Through a frequency study of naturally-occurring Peninsular written Spanish data, we have found that for none of the verbs surveyed is clitic doubling obligatory; in fact, the non-clitic-doubled IO is much more frequent than the clitic-doubled counterpart. We have shown that there are systematic distributional differences between the two IO variants with regard to certain referential properties. In short, the clitic restricts its co-occurring IO to have more individuated properties: [+pronominal] or [+animate, +individual, +singular] if it is [–pronominal]. On the other hand, if the [–pronominal] IO is not duplicated

81

82

Chiyo Nishida

by a clitic, it is free of constraints for animacy, individuality, and number. We have also illustrated through a qualitative analysis of some passages that the presence of a clitic in a psych verb sentence may also affect the interpretation of the IO, in particular, when this IO constitutes a quantifier expression like nadie ‘no one’ or todo el mundo ‘every one’. It forces these expressions to quantify over a restricted, specific set of humans instead of an open, generic set. The findings of this paper shed new light on the current stage and the nature of dative clitic doubling in modern European Spanish. According to Company Company (2001, 2003), dative clitic doubling in Spanish emerges historically as part of the grammaticalization process in which “the IO strengthens its status as object by reinforcing its marking devices” (2003: 225). She explains that the need for this process arose when the preposition a spread to mark the animate DO and later the inanimate DO, making the DO indistinguishable from the IO. Company Company argues that as dative clitics are used as IO markers, they lose their pronoun status and transform into agreement markers. Vázquez Rozas and García Salido (2012) also argue that dative clitics in the present day Spanish are agreement marking verbal affixes, though they postulate for a different scenario as to the factor that pushed dative clitics to transform into verbal affixes. Givón (1984) discusses the shift that Spanish dative clitics have undergone from a more general grammaticalization perspective. The analysis of dative clitics as IO-V agreement markers is broadly supported in the generative camp (Suñer 1988; Franco 2000; inter alia) as well.20 Nevertheless, the findings of our corpus study show that dative clitics play a much more substantive role in psych verb sentences than simply marking grammatical agreement in person and number. It has been widely claimed that dative clitic doubling is free of constraints (cf. Suñer 1988). However, our findings show that dative clitic doubling is highly selective, and a clitic-doubled IO has a strong tendency to be made up of a highly individuated object ([+animate, +pronominal] or [+animate, –pronominal, +individual, +singular]).21 Based on these findings, we propose that a dative clitic in psych verb sentences functions as an explicit marker of a prototypical experiencer IO in modern Peninsular Spanish, on top of partially marking grammatical agreement. Because of this particular role the clitic may play, clitic doubling has not spread to all instances of the IO. As 20. However, Cuervo (2003/2010) proposes that the dative clitic is the head of the functional projection Applicative Phrase, following the analysis of Bantu applicatives proposed by Pylkkänen (2001). 21. Company seems to include in her count both preverbal and postverbal clitic-doubled IOs as well as 1st and 2nd person IOs. It is not clear what the percentage would be if her survey were limited to 3rd person non pronominal postverbal IOs.



IO realization variation in Spanish reverse psych verb sentences

we have seen, the resistance of clitic doubling to inanimate objects (which cannot be interpreted as experiencers) is particularly strong. Also, since it is an explicit marker of certain referential properties, not all IOs with such properties may be necessarily clitic-doubled. Independent from the fact that dative clitic doubling is selective, our corpus data also show that the rate of clitic doubling in psych verb sentences with a post verbal IO is much lower than how dative clitic doubling is commonly portrayed. We have seen that with the [–pronominal] IO, clitic doubling in [S-V-IO] psych verb sentences is extremely restricted and the non-clitic-doubled IO is still the more common form with the majority of psych verbs. Company Company (2003), in her corpus study, gives the rate of IO clitic doubling for the 20th century to be as high as 83% in comparison to 40% in the 18th century (cf. Table 4, p. 23).22 However, with the newly identified context of optional clitic doubling as well as other contexts she may not have considered, it can be predicted that the current use of clitic doubling may well be much more conservative, in particular for 3rd person non-pronominal IOs. In fact, the rate of clitic doubling on the basis of the ten psych verbs surveyed in our study reaches only 26%. Even for the animate IO, which needed to be distinguished from the a-marked DO according to Company Company (2006), the clitic doubling rate is 36% (cf. Table 2), indicating that the grammaticalization process to transform dative clitics into “IO marking verbal affixes” has been much slower to spread in certain contexts, contrary to those studies that depict dative clitic doubling as being almost obligatory in modern Spanish (cf. Parodi 1998). The current study provides evidence from naturally occurring data that the distribution of dative clitic doubling is no way uniform across all contexts but a great deal of variation exists across constructions and within the same construction. Nevertheless, we have shown that the variation observed in one particular construction can be systematically explicated. In order to come to a better understanding of the nature and role of clitic doubling in modern Spanish, it is important to examine where variation exists, what the variation consists of, and what role the variation may be playing in the language. There is one variation issue involving psych verb sentences that we were not able to address in this paper due 22. Bresnan (2001) points out that it is common in many languages that bound pronouns, before they complete a grammaticalization process to turn into agreement markers, may often show sensitivities to certain semantic properties of their accompanying syntactic argument, like pronominality, animacy, specificity, etc., resulting in what she calls ‘finer transitional states’ (146). Bresnan mentions a number of such cases attested in a variety of languages. For instance, the Bantu language Kechaga verbal object inflection is obligatory with object pronouns (Bresnan & Moshi 1994), just like the obligatory clitic doubling with a pronominal object in Spanish.

83

84

Chiyo Nishida

to a space constraint: Why is clitic doubling obligatory in the canonical [IO-V-S] word order but optional in the non-canonical [S-V-IO] order? We leave this question for future work.

References Ackerman, Farrell & John Moore. 2001. Proto-properties and grammatical encoding: A correspondence theory of argument selection. Stanford: CSLI Publications. Aissen, Judith. 2003. Differential object marking: Iconicity vs. Economy. Natural Language and Linguistic Theory 21. 435–483. doi: 10.1023/A:1024109008573 Aranovich, Roberto. 2011. Optional agreement and grammatical fucntions: A corpus study of clitic doubling in Spanish. Doctoral thesis, University of Pittsburg. Aranovich, Raúl. 2012. A lexical functional account of Spanish dative usage. In P. de Swart & M. Lamers (eds.), Case, word order, and prominence, 17–41. Berlin: Springer.  doi:  10.1007/978-94-007-1463-2_2

Belloro, Valeria A. 2007. Spanish clitic doubling: A study of the syntax-pragmatics interface. Doctoral thesis, State University of New York at Buffalo. Bleam, Tonia. 2003. Properties of the double object constructions in Spanish. In R. NúñezCedeño et al. (eds.), A Romance perspective of language knowledge and use, 233–252. Amsterdam: John Benjamins. doi: 10.1075/cilt.238.19ble Casielles-Suárez, Eugenia. 2004. Syntax-Information structure interface. London: Routledge. Company Company, Concepción. 2001. Multiple dative-marking grammaticalization: Spanish as a special kind of primary object language. Studies in Language 25. 1–47.  doi:  10.1075/sl.25.1.02com

Company Company, Concepción. 2003. Transitivity and grammaticalization of object. In G. Fiolentino (ed.), Romance objects, 216–260. Berlin: Walter de Gruyter. Company Company, Concepción. 2006. El objeto indirecto. In C. Company Company (ed.), Sintaxis histórica de la lengua española. Primera parte: la frase verbal, 477–572. Mexico: UNAM & FCE. Croft, William. 1990. Typology and universals. Cambridge, UK: Cambridge University Press. Cuervo, María Cristina. 2003. The dative at large. Doctoral thesis, MIT. Cuervo, María Cristina. 2010. Against Ditransitivity. Probus 22. 151–180.  doi:  10.1515/prbs.2010.006

Demonte, Violeta. 1994. Teoría sintáctica. Madrid: Síntesis. Demonte, Violeta. 1995. Dative Alternation in Spanish. Probus 7. 5–30.  doi:  10.1515/prbs.1995.7.1.5

Franco, Jon. 2000. Agreement as a continuum: The case of Spanish pronominal clitics. In F. Beukema & M. den Dikken (eds.), Clitic phenomena in European languages, 147–189. Amsterdam: John Benjamins. doi: 10.1075/la.30.07fra Ganeshan, Ashwini. 2015. Case marking in Spanish reverse psychological verbs: A lexical semantic perspective. Unpublished doctoral thesis, University of Texas, Austin. Givón, Talmy. 1976. Topic, pronoun and grammatical agreement. In C. Li & S. Thompson (eds.), Subject and topic, 149–188. London: Academic Press. Givón, Talmy. 1984. Syntax. A functional-typological introduction, vol. 1. Amsterdam: John Benjamins. doi: 10.1075/z.17



IO realization variation in Spanish reverse psych verb sentences

Guitiérrez-Bravo, Rodridrigo. 2007. Prominence scales and unmarked word order in Spanish. Natural Language and Linguistic Theory 25. 235–271. doi: 10.1007/s11049-006-9012-7 Koontz-Garboden, Andrew. 2002. A quantitative analysis of Spanish indirect object doubling. In J. F. Lee et al. (eds.), Papers from the 4th hispanic linguistics symposium, 193–211. Somerville: Cascadilla. Kuno, Sususmu. 1972. Functional sentence perspective: A case study from Japanese and English. Linguistic Inquiry 3. 269–320. Laca, Brenda. 2006. El objeto directo: la marcación preposional. In C. Company Company (ed.), Sintaxis histórica del español. Primera parte: La frase verba, 423–475. Mexico: UNAM & FCE. Lambrecht, Knud. 1994. Information structure and sentence form. Cambridge, UK: Cambridge University Press. doi: 10.1017/CBO9780511620607 Maldonado, Ricardo. 2002. Objective and subjective datives. Cognitive Linguistics 13(1). 1–65.  doi:  10.1515/cogl.2002.010

Masullo, Pascual. 1992. Incorporation and case theory in Spanish: A crosslinguistic perspective. Doctoral thesis, University of Washington. Nishida, Chiyo. 2012. A corpus study of Mexican Spanish three-participant constructions with and without clitic doubling. In E. van Lier (ed.), Referential effects on the expression of three-participant events across languages. Linguistic discovery 10(3). 208–240. Pylkkänen, Maliina. 2001. Introducing arguments. Doctoral thesis, MIT. Parodi, Teresa. 1998. Aspects of clitic doubling and clitic clusters in spanish. In R. Fabri (ed.), Models of Inflection, 85–102. Tübingen: Max Niemeyer Verlag GmbH.  doi:  10.1515/9783110919745

Parodi, Claudia & Marta Luján. 2000. Aspect in Spanish psych verbs. In H. Campos (ed.), Hispanic linguistics at the turn of the century, 210–221. Somerville: Cascadilla Press. Suñer, Margarita. 1988. The role of agreement in clitic-doubled constructions. Natural Language and Linguistic Theory 6(3). 391–433. doi: 10.1007/BF00133904 Vanhoe, Henk. 2002. Aspects of the syntax of psychological verbs in Spanish. In M. Butt & T. H. King (eds.), Proceedings of the LFG Conference, 378–389. Stanford: CSLI Publication. van Lier, Eva. 2012. Referential effects on the expression of three-participant events across languages: An introduction in memory of anna sierwierska. Linguistics Discovery 10(3). 1–16. Vázquez Rozas, Victoria. 2006. Gustar-type verbs. In J. C. Clements & J. Yoon (eds.), Functional approaches to spanish syntax. Lexical semantics, discourse and transitivity, 80–114. Hampshire/New York: Palgrave MacMillan. doi: 10.1057/9780230522688_4 Vázquez Rozas, Victoria & Miguel García Salido. 2012. A discourse-based analysis of object clitic doubling in Spanish. In K. Davidse et al. (eds.), Grammaticalization and language change: New reflections, 271–297. Amsterdam: John Benjamins. doi: 10.1075/slcs.130.11vaz von Heusinger, Klaus. 2008. Verbal semantics and the diachronic development of differential object marking in Spanish. Probus 20. 1–31. doi: 10.1515/PROBUS.2008.001 Weissenrieder, Maureen. 1995. Indirect object doubling: Saying things twice in Spanish. Hispania 78. 169–177. doi: 10.2307/345242

Corpus Corpus de Referencia de Español Actual (CREA). Spanish Royal Academy. http://corpus.rae.es/ creanet.html

85

Non-human agents as subjects in English and Dutch A corpus-based translation study Steven Doms, Bernard De Clerck and Sonia Vandepitte Ghent University

In English sentences with a verb denoting an action like give, the subject usually plays the semantic role of agent. While in English non-human agents such as this manual in This manual gives instructions on the correct assembly occur quite frequently, Dutch seems to apply more restrictions, as illustrated in Dit handboek bevat voorschriften over de juiste montage in which the Dutch subject dit handboek is not an agent but rather a possessor (see e.g. Delsoir 2011; Vandepitte & Hartsuiker 2011). This article investigates how a set of 154 English sentences from the Dutch Parallel Corpus with non-human agents as subjects of give are translated into Dutch. The lower number of Dutch non-human agents are discussed with regard to translation tactics and explained in terms of differences in verb meanings between English give and its Dutch cognate geven and the lexico-­semantics of the non-human agents . The lexical choices translators made lead to valency loss in Dutch.

Introduction Ever since N. Chomsky’s (1957) Syntactic structures, a sentence can typically be divided into two main parts: a subject, e.g. the inhabitants of the city in (1) and I in (2) and a predicate, e.g. only pay direct tax in their place of residence in (1) and two years ago at Christmas time sent you my book in (2). The subject corresponds to ‘what is being talked about’, while the predicate typically denotes ‘what is said about it’. Apart from its grammatical function, the subject also fulfills what F. Palmer calls a “grammatical role” like Agent, Theme or Recipient (1994: 4). If the predicate contains an action verb (e.g. pay in (1) or send in (2)), the subject is usually an agent. If the predicate contains a state or a reception verb, it is a theme or a recipient, as in (3) and (4); see Section 4 for a more elaborate discussion of doi 10.1075/lis.33.04dom © 2016 John Benjamins Publishing Company

88

Steven Doms, Bernard De Clerck and Sonia Vandepitte

these roles. These meaning-based participants in a sentence have also been referred to as ‘thematic roles’ (see e.g. A. Sanfilippo 1990; D. Dowty 1991; B. Primus 1999; A. Tamm 2012), while others use ‘semantic roles’ (see e.g. C. Fillmore 1977a; M. Klaiman 1991; T. Givón 1993; S. Fauconnier 2012), a term which we will adopt in this study. (1) The inhabitants of the city only pay direct tax in their place of residence (2) Dear sir, two years ago at Christmas time I sent you my book (3) The nineteen municipalities of Brussels constitute the Brussels Capital Region (4) New immigrants receive substantial assistance in their compulsory integration

While agents are typically human beings performing the action denoted by a particular verb, instances of non-human agents occurring as the subjects of action verbs can also be attested in English, as is illustrated in examples (5a) and (6a). The occurrence of non-human agents has led several authors (e.g. B. Levin 1993; D. Biber, et al. 1999; R. Huddleston & G. Pullum 2002) to suggest that these subjects perform semantic roles other than that of agent (see §1.2), while I. Schlesinger (1989), and D. Dowty (1991), inter alia, still consider them as agents, albeit as less prototypical instantiations due to the absence of some important proto-­agent properties. For the sake of clarity, we will limit ourselves to the term ‘non-­human agents’ in the paper, but the concept will nevertheless be framed in and described in terms of existing theoretical concepts in §1.3. (5) a. The constitutional blueprint proposes a legal personality for the Union. b.  ?Het ontwerp van de grondwet stelt een rechtspersoonlijkheid voor de Unie voor. c. Er wordt rechtspersoonlijkheid voor de Unie voorgesteld. (Legal personality for the Union is proposed) (6) a. This sends a message: ‘I have forgiven you’ b.  ?Dit zendt een boodschap: ‘Ik heb je vergeven’ c. Dat houdt meteen ook een boodschap in: ‘Ik heb je vergeven’ (That also immediately implies a message: ‘I have forgiven you’)

What the examples (5) and (6) above demonstrate, however, is that although these non-human agents do not present specific problems in the English examples (5a) and (6a), literal translations in (5b) and (6b) will probably be considered unidiomatic by many native speakers of Dutch, suggesting that the Dutch language imposes more restrictions on the use of (these) non-human agents with action verbs, as is demonstrated by S. Vandepitte & R. Hartsuiker (2011) (see also J. Hawkins 1986 for an English-German comparison). The Dutch translations in (5c) and (6c), on the other hand, are more idiomatic, but do no longer contain a non-­



Non-human agents as subjects in English and Dutch

human agent. In (5c), for instance, there is no rendering of the source-text non-­ human agent the constitutional blueprint in the passive Dutch translation. And in (6c), the source-text non-human agent was translated literally, but the Dutch verb is not the cognate of send, but rather a verb denoting a state of affairs instead of an action (inhouden or imply). In this paper, we will further explore the similarities and differences in the use of non-human agents as subjects in a selection of English source-text sentences and their Dutch translations. We will try to get an insight into the kinds of non-­ human subjects which are accepted as agents in translated Dutch (and which are not) and describe the solutions that the translators have opted for instead of a non-­human agent in subject position. We will limit ourselves to translational data from utterances that contain the English action verb to give, which typically calls for a human agent, but which also appears with three verb meanings that allow for non-­human agents as subjects (see §2). Before analyzing the data, we will first of all provide a theoretical framework in which human and non-human agents are seen as instances of the agent role, regardless of prototypicality. Data and examples have been taken from the Dutch Parallel Corpus (DPC) (see L. Rura, et al. 2008) unless indicated otherwise and subjects are underlined. 1. Theoretical preliminaries In this section, we will zoom in on the definitions and features that have been proposed for agents in the literature. First, listed features will be discussed and compared to non-human instances of subjects in §1.1 In §1.2 attention will be paid to approaches that frame these non-human instances as instruments or causers and which, therefore, have not always been considered non-human realizations as a part of the agent category. In §1.3, however, we propose our own concept of agent which is based on only one agentive feature and which considers human and non-human agents as two distinct instances of the agent role. 1.1

Definitions and features of agents

One of the first definitions of agent is found in C. Fillmore’s case grammar of The Case for Case, in which he describes the agentive case as “the case of the perceived instigator of the action identified by the verb, typically animate” (1968: 46). In this definition, two features which appear to be typical of agents stand out: instigation and animacy. Later definitions most often highlight similar agentive features. D. Biber et al. (1999: 75), for instance, also emphasize instigation as an

89

90 Steven Doms, Bernard De Clerck and Sonia Vandepitte

inherent property of agents alongside willfulness: “the willful initiator of the action”. B. Comrie sees the agent as a “conscious initiator” (1981: 53) and introduces control as the distinctive characteristic to distinguish agents from other semantic roles like experiencer and patient. More recently, control as an agentive property is also found in V. Velupillai who states that the prototypical agent is “the entity that performs an action and has a high degree of control” (2012: 231). B. Primus (1999) argues that the agent is able to start and stop the event at will, which corresponds to the agentive feature control, but also adds responsibility as an agentive property. Finally, T. Givón (2001) also emphasizes responsibility and even includes deliberateness in saying that an agent is “the participant, typically animate, who acts deliberately to initiate the event, and thus bears the responsibility for it” (2001: 107). In his Thematic Proto-Roles and Argument Selection, D. Dowty (1991) does not give a definition of agent, rather he sees a prototypical agent as a cluster concept consisting of five so-called proto-agent properties: – – – – –

“volitional involvement in the event or state”, “sentience (and/or perception)”, “causing an event or change of state in another participant”, “movement (relative to the position of another participant)” and “exists independently of the event named by the verb” (1999: 572).

An agent argument which contains all five features is considered a prototypical agent, like John in Dowty’s example (7a) on the condition that we assume that John performs the action volitionally. The inanimate subject the storm in (7b), on the other hand, does not exert volition and is, hence, a less prototypical agent than John in a similar sentence. In this way, following D. Dowty (1991), less prototypical agents can be regarded as agents containing fewer than five agentive features. According to Dowty, one of the five features, i.e. the fifth and last feature that he lists, to wit independent existence, is always entailed if one of the four other features is present, so that at least two proto-agent properties can always be attested. In (7c), sentience is the second proto-agent property, while in (7d) independent existence is entailed by causation only. Movement is attested in (7e). (7) a. John broke the twig b. The storm broke the twig c. John knows the statement d. His loneliness caused his unhappiness e. The rolling tumbleweed passed the rock

At the same time, Dowty illustrates in these examples that less prototypical agents are very often non-human agents like the storm in (7b), his loneliness in (7d) and



Non-human agents as subjects in English and Dutch

the rolling tumbleweed in (7e). Similarly, M. Hundt (2004: 49) remarks that features like volition, control, deliberateness or responsibility all “entail animacy (or even humanness)”. The way in which these non-human agents are traditionally dealt with, will be reported on in the next section. 1.2

Non-human agents: Instruments and Causers

In the literature, semantic roles other than ‘agent’ have been proposed for agents whose referents are non-human such as the agents in (7b, d, e). C. Fillmore (1968) already separated the agentive from an instrumental case, i.e. “the case of the inanimate force or object causally involved in the action or state identified by the verb” (1968: 46). Further, C. Fillmore (1968: 42) states that an instrument such as “a hammer” in (8c) remains an instrument when it becomes subject, as in (8b), rather than assuming the agentive case. If the agent and the instrument are both present, as in (8c), the agent (John) will be the subject. (8) a. John broke the window b. A hammer broke the window c. John broke the window with a hammer

Building on C. Fillmore’s (1968) example, B. Levin (1993) discusses the instrument/subject alternation and explains how an instrument argument as a hammer in (8c) can become the subject of a derived sentence (8b) in which the human agent argument John is not represented. D. Biber et al. (1999) follow C. Fillmore in writing that the instrument or means used by an agent to perform the action is expressed as the subject in such cases. According to I. Schlesinger (1989: 190), the (unexpressed) agent is then “lurking behind the action”. In accordance with C. Fillmore’s (1968) theory, B. Levin (1993), D. Biber et al. (1999), as well as for instance, R. Quirk, et al. (1972), consider non-human instruments which become subject of a verb denoting an action as instruments. I. Schlesinger (1989), however, analyzes them as agents tout court. Like D. Dowty (1991), I. Schlesinger imagines the agent category as a system with degrees of membership, so that an instrument in subject position like a hammer is actually nothing less than an agent – “though not a prototypical Agent” (1989: 193) – assuming some of the characteristics of the agent category. I. Schlesinger (1989) does not abide by the categorization of instruments proposed by C. Fillmore which includes both ‘forces and objects’. Forces like C.  Fillmore’s (1968: 51) ‘the wind’ in (9) are not taken into consideration as instruments by I. Schlesinger, who does not further elaborate on these non-human agents. B. Levin (1993) and D. Biber, et al. (1999) also distinguish what C. Fillmore

91

92

Steven Doms, Bernard De Clerck and Sonia Vandepitte

describes as ‘forces’ from instruments. D. Biber, et al. (1999: 124) argue that apart from the agent and instrument role, the subject may also express the “inanimate external causer of an event”. The subject in their example in (10) bears strong resemblance to the subject from C. Fillmore’s The Case for Case in (9). (9) The wind opened the door (10) A biting wind gusting to 30 knots threatened to blow the fragile, 15-ft fiberglass hydroplane off course

T. Givón’s (2001: 53) distinction between what he calls “mindless force” (e.g. the wind in (9)) and “mindless tool” (e.g.: a hammer in (8b)) seems to be a clear-cut division of C. Filmore’s (1968) instrumental case category. Mindless force then corresponds to D. Biber, et al.’s (1999) causer and to what L. Talmy (2000: 482) refers to as “an instance of event causation”. While B. Levin (1993), D. Biber, et al. (1999), L. Talmy (2000) and T. Givón (2001) all consider abstract causes like the wind as a distinct semantic role vis-à-vis agents and instruments, R. Huddleston & G. Pullum (2002: 230) even consider the agent role as a subcategory of causer, which according to them “involves direct or immediate causation of an action or event”. From their point of view, an agent subject is attested only when the causer subject is “animate and acts consciously, volitionally” (2002: ibid.). In sum, non-human subjects of action verbs have often been framed as instances of instruments or causers which are considered part of the agent role in only some approaches (e.g. I. Schlesinger 1989; D. Dowty 1991). In the next section, we will return to the agent role itself and argue for a concept of agent which is based on only one agentive feature and which considers human and non-human subjects of action verbs as inherent instances of the agent role. 1.3

Human and non-human agents: Instances of independent existence

The literature discussed so far has provided several features which are assumed to be typical of agents: C. Fillmore’s (1968) instigation and animacy, B. Comrie’s (1981) consciousness and control, D. Dowty’s (1991) volition, sentience/perception, causation, movement and independent existence, B. Primus’ (1999) responsibility and T. Givón’s (2000) deliberateness. All these features are taken into account in Table 1, which contains seven English examples with the action verbs give, offer and show. Four of the subjects accompanying these action verbs (his father in A, she in B, they in D and, she in F) have human referents, while three non-human subjects (maritime expansion in C, a tram card in E, and the scale on the cartridge in G) are presented as well. Every example has been analyzed

Non-human agents as subjects in English and Dutch

93

Table 1. Agentive features and (non-)human subjects of action verbs

Instigation Animacy Consciousness Control Volition Sentience/Perception Causation Movement Independent existence Responsibility Deliberateness

A. His father gave his tape to Van Dyke Parks.

B. She gives trainings in this area.

+ + + + + + + + + + +

+ + + + + + – – + + +

C. D. E. Maritime exThey offered A tram card pansion gave this her chocolate. also offers many period its golden advantages. edge. + – – – – – + – + + –

+ + + + + + – + + + +

– – – – – – – – + – –

F. G. She showed me The scale on the a door hanging cartridge shows about open. how many units you have left. + + + + + + + + + + +

+ – – – – – + – + – –

94 Steven Doms, Bernard De Clerck and Sonia Vandepitte

for these eleven agentive features in order to find out whether human and non-­ human subjects of (these) action verbs have different features in common. From Table 1 it becomes clear that all agentive features are attested in only two sentences (A, F), both of which have a human agent, as subject of give and show, respectively. The examples with non-human subjects (C, E, G), on the other hand, only show some of the eleven agentive features listed here, so that on a par with I. Schlesinger (1989) and D. Dowty (1991) we can say that non-­human agents are not prototypical instances of agents. This, however, does not mean that all human agents are a priori prototypical. In B, for instance, the human subject she heads a divalent sentence which does not include a recipient, i.e. the participant who receives a gift from the agent. The absence of this recipient makes the human subject in B a less prototypical agent vis-à-vis the agents in A and F, because no causation and movement are detected, which according to D. Dowty (1991) are both relative to another participant, in this case the recipient which is absent. In D, the recipient is represented, but causation is not attested, since the meaning of the verb offer does not imply that the recipient actually accepts or even takes what is offered in possession. In fact, in turns out that there is only one agentive feature that is present in all seven examples with (non-)human subjects of verb denoting an action: that of independent existence, which has been described by D. Dowty as not being “brought into being or destroyed by the event named by the verb but […] presumed to exist before and after the event” (1991: 573). Further, features such as consciousness, control, volition and deliberateness, all of which are linked to animacy according to M. Hundt (2004) are indeed only represented in examples with human agents. Therefore, in the present study, only one feature, i.e. independent existence, will be adopted as agentive property. It is a feature which is shared by all instances of the agent role (see also examples (7a–e)). This view, in which traditional features entailing animacy are no longer considered inherent characteristics of the agent category, enables us to further subdivide agents into human and non-human agents, without starting a debate on whether non-human agents have, for instance, control over the action they perform. The semantic role of agent which is proposed in this article, covers an equally wide diversity of agent arguments as I. Schlesinger’s (1989) agent category and D. Dowty’s (1991) agent prototype. It does not, however, take humanness or typically human characteristics as its core. Nor does the agent role which is presented here assume that non-human agents are necessarily instruments or causers. Instead, a detailed lexico-semantic analysis will be made to reveal the nature of non-­human agents which occur as subjects of the action verb give and to disclose which of these non-human agents are accepted in translated Dutch and which are not.



Non-human agents as subjects in English and Dutch

2. Non-human agents as subjects of give and geven Give is a trivalent verb which expresses an act in which an “agent successfully causes [a] recipient to receive [a] patient” (A. Goldberg 1995: 38). A. Goldberg’s (1995) definition of give most definitely calls for human agents, like his father in A above. Instances of non-human agents, however, can be attested quite frequently as well (see for instance example C above). Its Dutch cognate geven, on the other hand, only allows for non-human agents in subject position in a small number of restricted senses. This discrepancy between give and geven can be seen as a case in point for the differences between English and Dutch with regard to non-­ human agents in subject position and seems to proceed from different meanings of both verbs. To compare give and geven, first a concise enquiry into their verb meanings was conducted on the basis of one English and one Dutch online dictionary. Both lemmas were searched: give in the Oxford English Dictionary Online and geven in Van Dale Groot Woordenboek der Nederlandse Taal Online. The Oxford English Dictionary Online records eight verb meanings for give, which express an act in which an agent successfully causes a recipient to receive a gift: I. to bestow gratuitously; II. to deliver, hand over (without reference to change of ownership); III. to make over as a matter of exchange or debt; IV. to sacrifice, devote, dedicate; V. to present, expose, offer; VI. to make partaker of; VII. to allot, apportion, assign; VIII. to yield as a product or result (OED Online, 21 October 2013). Verb meanings I, II, III, IV and VII seem to call for a human agent to perform the action denoted by give. Verb meanings V, VI and VIII, on the other hand, share the emphasis on the effect or result of the action and are illustrated with non-human agents in subject position. Example (11) illustrates verb meaning V, while (12) exemplifies verb meaning VI. Furthermore, OED Online gives (13) as an example of verb meaning VIII. (11) A daily newspaper gave a head-lined account of the speech (12) The Loire, its banks still clad with the broom which gives their title to the Plantagenets (13) The lamps gave an uncertain light

The Dutch cognate of give is geven, which was searched in Van Dale Groot Woordenboek der Nederlandse Taal Online in which ten verb meanings were attested: I. iets aangeven, aanreiken of in iemands nabijheid plaatsen (‘to hand, to pass or to place close to someone’); II. (kaartspel) de kaarten ronddelen (‘(playing cards) to deal the cards’); III. toedienen, toebrengen (‘to administer, to inflict’); IV. verstrekken, verschaffen, bezorgen (‘to supply with, to provide with, to procure’); V. betalen,

95

96 Steven Doms, Bernard De Clerck and Sonia Vandepitte

of aanbieden te betalen, bieden (‘to pay, or to offer to pay, to offer’); VI. toekennen, toeschrijven (‘to ascribe to, to attribute to’); VII. verschaffen, opleveren (‘to provide with, to deliver’); VIII. als gunst of als een bewijs van goedheid schenken (‘to donate as a favour or as a token of goodwill’); IX. overgeven, niet als geschenk, maar als iets dat verlangd of verwacht of waartoe men genoodzaakt wordt (‘to hand-over, not as a gift, but as something that is desired or expected or because someone is obliged to’); X. van zich doen uitgaan (‘to emanate’) (Dikke Van Dale Online, 2 December 2013). All Dutch verb meanings with the exception of the final one are – more or less – hyponyms of the general meaning expressed by geven, each of which typically calls for a human agent as its subject. Van Dale Groot Woordenboek der Nederlandse Taal Online provides less examples for geven than OED Online for give. No examples with non-human agents are attested for verb meanings I until IX. The last verb meaning, however, van zich doen uitgaan (‘to emanate’) also refers to instances such as ‘the sun gives light’ or ‘the stove gives warmth’ and it seems to be limited to these cases which are quite similar to example (13) which expresses English verb meaning VIII, to yield as a product or result. Thus, no clear verb meanings which are typical of non-human agents in subject position are found in the Van Dale description of Dutch geven. As mentioned above, the number of attested examples for geven in Van Dale Groot Woordenboek der Nederlandse Taal is limited. In addition, instances of verb meaning VII are found with non-human agents in non-translated Dutch, as shown in (14). (14) De werken geven een beeld van de evolutie van de Vlaamse grafische kunst tussen 1980 en 2005. (The works give an overview of the evolution of Flemish graphic art from 1980 to 2005)

To gain more insights into how give and geven relate to each other, the Dutch translations of give in Van Dale Groot Woordenboek Engels-Nederlands were also investigated. In total, this English-Dutch translation dictionary registers eight verb meanings for transitive give, for all except one geven is suggested as Dutch translation: 0.1. geven => schenken, overhandigen, aanreiken, toevertrouwen, betalen (‘to donate, to hand-over, to pass, to confide, to pay’); 0.2. geven => verlenen, schenken, verschaffen, bezorgen, gunnen, toegeven, toekennen, toeschrijven (‘to grant, to donate, to provide, to procure, to permit, to concede, to ascribe, to attribute to’); 0.3. geven => opofferen, wijden (‘to sacrifice, to consecrate’); 0.4. geven => uiten, toebrengen, houden, maken, uitbrengen, slaken (‘to utter, to inflict, to hold, to make, to issue, to shout’); 0.5. geven => aanbieden, ten beste geven (‘to present, to donate’); 0.6. (op)geven => meedelen, verstrekken, (ver)noemen, aangeven, tonen, bekendmaken (‘to inform, to supply with, to call, to indicate, to show,



Non-human agents as subjects in English and Dutch

to announce’); 0.7. geven => produceren, voortbrengen, opleveren, afgeven (‘to produce, to bring forth, to yield, to give off ’); 0.8. toasten op (‘to make a toast to’). Although give and geven were shown to have different restrictions on non-­ human agents in two online dictionaries, Van Dale Groot Woordenboek Engels-­ Nederlands prescribes that give can in most cases be translated with geven, except if it has the verb meaning of making a toast. Therefore, in this translation study, we expect that translators will indeed translate give with geven in sentences with non-human agents. However, and due to the restrictions imposed on geven, not all of the Dutch translations will probably contain non-human agents as subjects of geven. We will investigate which non-human agents are combined with geven and which are considered atypical Dutch formulations of geven and are therefore avoided by translators like in (15b), in which dit handboek is a possessor instead of an agent due to the use of the Dutch possession verb bevatten. (15) a. This manual gives instructions on the correct assembly b. Dit handboek bevat voorschriften over de juiste montage (This manual contains instructions on the correct assembly)

If translators consider instances of non-human agents as subject of geven atypical, they can, for instance, introduce a human agent as the subject of Dutch geven (instead of a non-human subject) or they can also opt not to use Dutch geven, but another verb which (in at least one of its verb meanings) typically takes a non-­ human subject, as in (15b). Before analyzing the actual data and the solutions applied by translators (see §4), the selection of the data underlying this study is discussed (see §3). 3. Corpus data and methodology The English source-text and the Dutch target-text sentences which are compared in this enquiry, were gleaned from the 10 million word Dutch Parallel Corpus (DPC). In order to investigate the linguistic data in the DPC, an interface1 was created on the basis of which the entire corpus could be browsed for lemma, partof-speech, context, etc. The lemmas of give were extracted automatically which yielded a total of 788 English source-text sentences. Passive sentences and instances that did not have a subject were filtered out (e.g. participial, prepositional and infinitival clauses; nominalizations and adjectives). Phrasal verbs, i.e. give combined with a particle (e.g. ‘give in’), expressions (e.g. ‘give someone the eye’) 1. I am very much indebted to Isabelle Delaere (Ghent University) and Alexander Malfait for supplying me with their easy-to-use search tool for the DPC.

97

98 Steven Doms, Bernard De Clerck and Sonia Vandepitte

Table 2.  Instances of give with(out) agents in the Dutch Parallel Corpus Noise Human agent Non-human agent Total

386 248 154 788

and idioms (e.g. ‘give birth’) were not taken into consideration either, since these structures do not longer express the typical verb meanings of give and constitute specific translation issues by themselves, regardless of the human/non-human feature of their agents. All these instances were labeled as noise. As shown in Table 2, a total of 402 relevant utterances have been distilled, more than a third of which (38.3%) occur with non-human agents. These utterances consist of trivalent instances, i.e. including three participants (Agent, Patient and Recipient) as in (12), and divalent instances, i.e. including only two participants (Agent, Patient, but no Recipient) as in (15a). This selection method leaves us with in total 154 English source-text sentences and their Dutch translations, a detailed analysis of which is provided in the following sections. The majority of the source-text sentences do not involve modality, whereas in 33 English source-text sentences modality is attested, either by means of modal verbs (e.g. will, can, must) or other modal phrases (e.g. ‘be able to’). A. Siewierska (1991: 48–49) observes that modality can “cancel” typical agentive features such as control. Since this study does not see control as a feature that will distinguish agents from other semantic roles and – in a more specific way – human from non-human agents, source-text sentences containing a non-human agent both with and without modality are investigated. In §4 we will present the solutions translators opted for to translate the non-human agents into Dutch, while the nature of these non-human agents is systematically described in §4.4. 4. Data-analysis: English instances and their Dutch translations In this section, the 154 English sentences and their Dutch translations are analyzed as to investigate which solutions (see §4.1), referred to as translation tactics here (see §4.2), were chosen by translators to deal with the restrictions on non-­ human agents in subject position imposed by Dutch geven.



4.1

Non-human agents as subjects in English and Dutch

Translation solutions in the Dutch translations

As explained in §2, give can in most contexts be translated with geven. The Dutch language, however, seems to impose restrictions on non-human agents as subjects of geven. Hence, we expect that translators will – at least to some extent – avoid non-­human agents as subjects of geven. In this section, we will first analyze the Dutch translations with regard to two parameters: the target-text subject and the target-­text verb. As is illustrated in Table 3, the target-text subject can be a literal translation of the source-text non-human agent (‘= non-human agent’) or a different non-human agent vis-à-vis the source-text non-human agent (‘≠ non-­ human agent’), but the target-text subject may as well represent a human agent or another semantic role such as theme, possessor or recipient. All these options with regard to the target-text subjects are combined in Table 3 with information on the target-text verb, which may be a literal translation of give, but also another donative verb (e.g.: bieden-offer, verlenen-grant), a non-donative verb which denotes an action (e.g.: aantonen-show, beschrijven-describe), or a verb which does not denote an action like a state (e.g. zijn-be), a possession (e.g. hebben-have) or a reception verb (e.g. krijgen-get). Further, Dutch translations may involve the use of the passive voice or may not make reference to the source-text verb (‘no verb’). Table 3 presents the choices made by translators in this study. More than a third of all target-text sentences have a literal translation of the source-text non-human agent and give, as in (16b), indicating that translators did Table 3.  Dutch target-text subjects and verbs Dutch translations

Total

= non-human agent & geven ≠ non-human agent & geven = non-human agent & donative verb ≠ non-human agent & donative verb = non-human agent & non-donative verb ≠ non-human agent & non-donative verb human agent & geven human agent & donative verb human agent & non-donative verb theme subject & stative verb theme subject & verb in the passive voice possessor subject & possession verb recipient subject & reception verb no verb Total

 54   0  26   0  12   0   1   1   8  16   9   7  15   5 154

99

100 Steven Doms, Bernard De Clerck and Sonia Vandepitte

sometimes but not always follow the translation dictionary. In two thirds of the instances, they searched for other translation solutions. One of these solutions which is found in almost a fifth of all translations results in the use of another Dutch donative action verb like leveren in (17b) in combination with the sourcetext non-human agent translated literally. Literally translated source-text non-­ human agents are found less frequently with Dutch non-donative action verbs such as beschrijven in (18b). Not a single non-human agent different from the source-­text subject is introduced in Dutch. In the instances in which the sourcetext non-human agent is translated literally, all the source-text participants (agent, patient, and in case of a trivalent sentence, recipient) are usually also represented in the target text (see arrows). (16) a. The Convention’s draft gives us an opportunity to break out of this cycle b. Het ontwerp van de Conventie geeft ons de kans deze cyclus te doorbreken (The draft of the Convention gives us the chance to break out of this cycle) (17) a. Cheese gave 72 per cent of the 200 volunteers a very good night’s sleep b. Kaas leverde 72 procent van de tweehonderd vrijwilligers een verkwikkende slaap op  (Cheese provided 72 percent of the two hundred volunteers a good night’s sleep) (18) a. Table 1 gives the adverse reactions observed from spontaneous reporting b. Tabel 1 beschrijft de bijwerkingen gezien uit spontane meldingen  (Table 1 describes the adverse reactions observed from spontaneous reporting)

As a solution to avoid Dutch non-human agents as subjects of geven, translators introduce a human target-text agent in only less than one in ten Dutch translations. Human target-text agents occur especially with non-donative action verbs like emigreren in (19b), in which the source-text non-human agent is referred to as a kind of instrument and the source-text recipient becomes the target-­text agent, as indicated by means of arrows between (19a) and (19b). Furthermore, the information in give a route into Britain is reflected by the Dutch verb emigreren which takes the Dutch complement naar Engeland, revealing the direction of emigrating. Besides this change of perspective in the target text, it also contains the modal verb kunnen (could).



Non-human agents as subjects in English and Dutch 101

(19) a. That gave them a route into Britain b. Langs die weg konden ze naar Engeland emigreren (Along that way they could emigrate to England)

In about a third of all translations, however, the target-text subjects do not represent the agent role. Instead, theme, recipient and possessor subjects are found. Theme subjects are attested most often, either as subjects of state verbs like zijn in (20b) or as subjects of passive sentences, as in (21b). In (20), the source-text subject his life and patient a powerful message are translated literally. Give, however, is translated with zijn and the source-text recipient the enlarging Europe becomes a Beneficiary in Dutch. In (21), on the other hand, the source-text non-human agent which takes the form of a clause is a pronominal adverb in the target text (21b) which contains the Dutch donative verb bieden in the passive voice. The source-­text patient a cost effective solution becomes the subject in (21b) in which it fulfills the theme role. (20) a. It is a powerful message that his life has given to the enlarging Europe b. Zijn leven was een krachtige boodschap voor de uitbreiding van Europa (His life was a powerful message for the enlargement of Europe) (21) a. The sensor can easily be installed, giving a cost effective solution b. Hierdoor wordt een kostefficiënte oplossing geboden (In this way, a cost-effective solution is offered)

Besides theme subjects, recipient subjects like ik in (22b) and possessor subjects like wij in (23b) are also attested in the Dutch translations instead of (non-­human) agents. The recipient subject in the Dutch translation in (22b) refers to the sourcetext recipient. The source-text patient is translated as a patient, split up into two parts by the Dutch verb krijgen. The source-text subject it, which explicitly refers to the main clause we also performed frequently is left implicit in the Dutch translation. In (23b), the Dutch subject also corresponds to the source-text recipient, while the source-text patient is translated literally. There is, however, no rendering of the source-text subject that. (22) a. We also performed frequently, so it gave me experience of being in front of a live audience b.  We mochten ook vaak optreden, zodat ik ervaring kreeg met een echt publiek (We could also perform frequently, so that I got experience with a real audience)

102 Steven Doms, Bernard De Clerck and Sonia Vandepitte

(23) a. That seems to me to give us a particular responsibility to show leadership b. Het lijkt mij dat wij de verantwoordelijkheid hebben leiderschap te tonen (It seems to me that we have the responsibility to show leadership)

Finally, some instances do not have a verb in the Dutch translation, as in (24b) in which the source-text verb give is replaced by the Dutch preposition in. (24) a. the women pickers whose generalised clothing gave them a timeless quality b. olijvenpluksters in eenvoudige, tijdloze kleding (female olive pickers in simple, timeless clothing)

In the next section, the translation solutions presented here are grouped together in what are called translation tactics. 4.2

Re-agentization, humanization and de-agentization

Translations of English non-human agents as subjects of give with non-human agents as subjects of geven may procure atypical instances of the Dutch language. We expect that translators who are confronted with this translation problem will search for solutions and these searches can be grouped as “translation tactics”, following Y. Gambier’s proposal to use this term to refer to “locally implemented steps used to pursue a particular translation strategy” (2009: 412). Translating the source-­text sentences into a typical instance of the Dutch language may be such a translation strategy. Translators will then follow this translation strategy by implementing translation tactics. Our results seem to indicate that three major translation tactics are applied in order to translate non-human agents as subjects of give with a typical Dutch target-text sentence. In the first tactic, the English subject is translated literally with its Dutch dictionary equivalent and the verb give is translated with a Dutch action verb (geven, another Dutch donative verb or a Dutch non-donative action verb). This tactic is referred to as re-agentization, because it provides a Dutch translation with a non-human agent in subject position. It comprises the translation solutions ‘=non-human agent & geven’, ‘=non-human agent & donative verb’ and ‘=non-human agent & non-donative verb’ presented in Table 3. In total, more than half of the Dutch translations are the result of this translation tactic. In the second translation tactic, which will be called humanization, the English non-human agent is replaced in Dutch with a human agent, in which case give can be translated both with geven or with another (donative or non-­donative) action verb which typically takes a human subject. This translation tactic corresponds to



Non-human agents as subjects in English and Dutch 103

translation solutions ‘human agent & geven’, ‘human agent & donative verb’ and ‘human agent & non-donative verb’, all of which give birth to humanization of the agent, and is attested in twelve instances. The third and last translation tactic which can be discerned from the translation solutions in Table 3 is de-agentization and comprises three different ways of realization. De-agentization refers to Dutch translations which do not have agents as subjects. It can be achieved through a Dutch translation which does not contain an action verb, so that the Dutch subject is not an agent, but plays, for instance, the theme (‘theme subject & stative verb’), recipient (‘recipient subject & reception verb’) or possessor role (‘possessor subject & possession verb’). Similarly, the use of the passive voice gives rise to a Dutch translation without an agent as subject (‘theme subject & verb in the passive voice’). And finally, omission of the verb (‘no verb’) is also an instance of this translation tactic. De-agentization is found in about a third of all translations. Re-agentization is used by translators in almost sixty percent of the instances and in about a third of all Dutch translations give was translated with Dutch geven in combination with non-human agents as subjects. In §4.3, an analysis is made of the verb meanings of English give which are attested in this study. 4.3

Re-agentization: give and geven

As was revealed in §4.2, more than a third of all instances displayed a literal translation of give with a non-human agent in subject position, although it was argued in §2, following Van Dale Groot Woordenboek der Nederlandse Taal, that geven does not have any specific verb meanings which typically allow for non-human agents as its subjects. In part, the solutions suggested by translators correspond to the information in translation dictionaries like Van Dale Groot Woordenboek Engels-­Nederlands which states that give can be translated with geven in all its verb meanings except if it means ‘making a toast’. For the give lemma, it was shown in §2 that the OED Online provided eight verb meanings, three of which contained examples with non-human agents as subjects: verb meanings V (to present, expose, offer), VI (to make partaker of) and VIII (to yield as a product or result). The other verb meanings were only illustrated in the OED Online with examples containing human agents in subject position. Table 4 confirms this implicit assumption of the English dictionary, in that only three English source-text sentences with give have a verb meaning which typically calls for a human agent. From our dataset, we can conclude that English non-­ human agents are almost exclusively the subjects of give denoting verb meanings V, as in (18), and VIII, as in (17).

104 Steven Doms, Bernard De Clerck and Sonia Vandepitte

Table 4.  Attested verb meanings of give Meaning I Meaning II Meaning III Meaning IV Meaning V Meaning VI Meaning VII Meaning VIII Total

  2   1  –  –  66   7  –  78 154

In those cases in which non-human agents are attested as subjects of geven, the respective source-text sentences always contain give expressing one of its three verb meanings typical of non-human agents according to the examples in the OED Online. Verb meaning V of give is found in almost half of the English source-­text sentences which were translated with a non-human agent and geven. Verb meaning VIII of give was nearly as frequently attested in the source-text sentences of such Dutch translations. In these translations, Dutch geven seems to express either verb meanings IV verstrekken, verschaffen, bezorgen (‘to supply with, to provide with, to procure’) and VII verschaffen, opleveren (‘to provide with, to deliver’), as attested in Van Dale Groot Woordenboek der Nederlandse Taal (see §2). This, however, does not explain why a third of the Dutch translations had non-­human agents as subjects of geven. Perhaps particular non-human agents can occur more easily as subjects of geven, while others cannot? We will deal with this question in §4.4 on the lexico-semantics of the non-human agents. 4.4 Lexico-semantic features of the subjects In order to get a better idea about possible links between kinds of non-human subjects and attested translations, we will first provide a more detailed description of the lexico-semantic features of the English and Dutch subjects. First, an analysis is made of the English source-text subjects which are all non-human agents. In §4.4.2, the English non-human agents are compared with the Dutch subjects which do not all represent the agent role in order to establish whether particular English non-human agents were more easily translated with Dutch non-human agents as subjects of geven.

Non-human agents as subjects in English and Dutch 105



English non-human agents 60

51

50

38

40 27

30 20

16

22

10 0

non-instrumental concrete cause object

abstract cause

event

embedded event

Figure 1.  Lexico-semantic features of the English non-human agents as subjects of give

4.4.1 Lexico-semantic features of the English non-human agents In §1.2, it was argued that non-human agents in subject position are often subcategorized in terms of their semantic roles as instruments and causers. If we compare all non-human agents attested in this study as subjects of give, however, no instances of instruments or causers are found. Although some non-human agents refer to objects, they are not instrumental. And although several subjects denote some form of causation, they are not causers. The subjects in the English source-­text sentences reveal more complex characteristics and have been divided into five subgroups according to the lexico-semantic features they possess: non-­ instrumental objects (i.e. tangible objects), concrete causes (i.e. non-tangible, but perceivable), abstract causes (i.e. non-tangible and non-perceivable), events (i.e. something taking place) and embedded events (i.e. a situation or action by itself, usually as a clause). How many of the English non-human agents belong to each group is shown in Figure 1. The first group of non-human agents consists of non-instrumental objects such as body-parts (e.g. which part of the face), food (e.g. cheese) or other objects such as table in (25a). Further, English non-human agents are attested which denote concrete items like the honourable Member’s report in (26a) or abstract concepts like altruism in (27a). The latter group of concrete items accounts for more than a third of all English non-human agents and most instances refer to ways of (written) communication. The fourth group of non-human agents comprises events or situations which take or have taken place, as is illustrated by the World Expo in (28a). Finally, the second largest group of non-human agents are embedded events, an example of which is given in (29a). These embedded events differ from events, in that they not only refer to an event, but are expressed as a separate state of affairs, which becomes the subject of a larger sentence.

106 Steven Doms, Bernard De Clerck and Sonia Vandepitte

(25) a. one of the tambourine-shaped tables that gave the establishment its name b. een van de tamboerijnvormige tafels die het etablissement zijn naam gaven  (one of the tambourine-shaped tables that gave the establishment its name) (26) a. The honourable Member’s report gives a very accurate overview of what has been achieved so far. b. Het verslag van de geachte afgevaardigde geeft een heel nauwkeurig foverzicht van wat er toe nu toe is bereikt.  (The report of the honourable delegate gives a very accurate overview of what has been achieved so far) (27) a. A  ltruism gives longer-lasting pleasure than a bar of chocolate or buying a new outfit b. Altruïsme geeft langer plezier dan een reep chocola of een nieuw stuk in uw garderobe  (Altruism gives longer pleasure gives than a bar of chocolate or a new piece in your wardrobe) (28) a. The World Expo gives us the chance to meet customers. b.  De Wereldexpo zal ons ook de mogelijkheid bieden bestaande klanten te ontmoeten. (The World Expo will also offer us the possibility to meet our customers) (29) a. Our new product line was among the very few in the world to meet the DICOMcriteria for medical imaging, which gives them a reliable indication of their intrinsic quality. b.  Onze nieuwe productlijn voldeed als één van de weinige in de wereld aan de DICOM-criteria voor medische beeldvorming, wat voor hen een betrouwbare aanwijzing was van de intrinsieke kwaliteit van onze beeldschermen.  (Our new product line met as one of the few in the world the DICOMcriteria for medical imaging, which for them was a reliable indication of the intrinsic quality of our displays)

Three of the Dutch translations in (25)–(29) contain geven with a non-human agent as its subject. The instances with an English non-human agent denoting an event (28a) and an embedded event (29a), on the other hand, are translated with a different verb in Dutch. Whether this is based exclusively on coincidence or not, is revealed in the next section, which deals with the ways in which these English non-­human agents of each lexico-semantic subgroup are translated into Dutch. 4.4.2 Translations of English non-human agents In this section, the Dutch translations of each lexico-semantic subgroup discussed in §4.4.1 are analyzed, so as to investigate which subgroups are translated how

Non-human agents as subjects in English and Dutch 107



Dutch translations of English non-human agents

40

36

35 30 25 20

10 5 0

17

14

15

15 11

10 1 2 1 1 1

3

6

6 0

2

0

non-instrumental concrete cause object

1

3

2

abstract cause

non-human agent human agent

5

theme possessor

3

1 0 1 0 event

1

5 4

2

embedded event

recipient no verb

Figure 2.  Dutch translations of the English non-human agents in terms of lexico-­ semantic subgroups

often with non-human agents in Dutch. As shown in Figure 2, English non-­human agents referring to non-instrumental objects (in ten out of sixteen instances), concrete causes (in thirty-six out of fifty-one instances), abstract causes (in fourteen out of twenty-seven instances) and events (in seventeen out of twenty-­two instances) are very often translated into Dutch with a non-human agent. English non-­human agents represented as an embedded event, on the other hand, were translated in only fifteen out of thirty-eight instances with a Dutch non-­human agent. Instead, translators especially opted for Dutch theme subjects, as well as possessors, recipients and human agents as subjects. The Dutch non-human agents in Figure 2 are subjects of either geven or another Dutch (donative or non-donative) action verb. If only the Dutch translations with geven are taken into account, especially concrete causes, as well as, abstract causes and events are very often attested as non-human agents in subject position, as shown in Table 5. Given the low number non-instrumental objects (in four out of sixteen instances) and, especially, embedded events (in six out of thirty-eight instances) which resulted in non-human subjects of geven, and taking into account the possibility of interference from the source texts, we conclude that translators accept Dutch non-human agents which refer to abstract, concrete causes and events more easily in Dutch as subject of geven than non-human agents denoting non-­ instrumental objects or embedded events. Taking into consideration, however, that non-human agents as subjects of geven are attested in only about a third of all

108 Steven Doms, Bernard De Clerck and Sonia Vandepitte

Table 5.  Lexico-semantic subgroups: Dutch non-human agents as subjects of geven

Non-instrumental objects Abstract causes Concrete causes Events Embedded events Total

English non-human agents as subjects of give

Dutch non-human agents as subjects of geven

 16  27  51  22  38 154

 4 11 22 11  6 54

Dutch translations, we assume it is fair to say that non-human agents as subject of geven are considered atypical instances of the Dutch language by translators. The findings presented in §4.5 might support this claim.

Valency changes as an effect of translation

4.5

In the previous sections, we have established that only one third of the 154 Dutch translations have a non-human agent as subject of geven. In the other instances, translators opt for specific solutions, referred to as translation tactics in §4.2, to introduce (more) typical Dutch translations. These tactics, however, also bring about changes in terms of valency of the Dutch translations. If we compare the English source-text sentences with their Dutch translations in Figure 3, we find that two thirds of the English sentences are trivalent, i.e. consisting of three participants (agent, patient and recipient), whereas the Dutch translations are mainly divalent, containing only two participants. The trivalent structure is typical of donative action verbs like give and geven. In English, this structure is often found in instances with non-human agents in subject position. The results shown in Figure 3 suggest that translators have tried to rephrase the English instances in order to avoid ‘atypical’ Dutch translations, Valency in English and Dutch 60

53

NL trivalent 44

40 20 0

NL divalent

32 14

11

0 EN trivalent

NL monovalent

EN divalent

Figure 3.  English trivalent and divalent constructions and their Dutch translations



Non-human agents as subjects in English and Dutch 109

i.e. the combination of geven with a non-human agents as its subject in trivalent constructions. In most of these cases, the translation tactics humanization and de-agentization have led to a loss in valency. This might point out that translators esteemed the trivalent structure, typical of give, inappropriate for non-human agents as subjects of Dutch geven. Although fewer participants are attested in Dutch, the source-text information is still represented in most Dutch translations. Lexical source-text elements were very often given a different place, grammatical function and – in case of participants – a different semantic role, as was shown in §4.1 by means of arrows between source-text and target-text sentences. This seems to suggest that translators did not merely translate individual words or lexical units, but source-­text utterances as a whole. It appears as if translators visualize the total meaning of the source-text sentence or scene (see e.g. C. Fillmore 1977b; P. Kussmaul 1994), before translating that meaning into a target-text sentence or frame. Evidence for this is provided by instances of de-agentization which involve the use of a non-­ action verb. In these cases, source-text scenes of giving are translated with scenes of being, getting and having, resulting in frames which include verbs like zijn (to be), krijgen (to get) and hebben (to have), calling for different participants than the non-human agents in the source-text sentences. Secondly, if the source text includes an embedded event as its subject, this can be perceived as a scene which is subject of another scene. In this study, translators tend to avoid these non-­ human agents as subjects of geven, which may suggest that translators indeed think in terms of scenes and frames, making it hard to translate two scenes in a single frame. 5. Discussion Analysis of the source-text sentences and their Dutch translations revealed three translation tactics: re-agentization, humanization and de-agentization. Re-­ agentization occurs in about sixty percent of the instances. Further, one third of all instances had Dutch non-human agents as subjects of geven. Why do translators opt for a literal translation in some cases and why do they propose other solutions in other situations? This study has tried to answer that question. In the source-­text sentences, give is found with three verb meanings which allow for non-­human agents in subject position which Dutch geven does not have. Nevertheless, a third of the Dutch translations does contain geven with a non-­human agent, indicating that perhaps translators have followed translation dictionaries in these instances. Or perhaps, particular non-human agents can occur more easily as subjects of geven than others? It is shown that non-instrumental objects

110 Steven Doms, Bernard De Clerck and Sonia Vandepitte

and embedded events are attested in only some Dutch translations as subjects of geven, while concrete causes, abstract causes and events are more often translated literally in Dutch with geven. These findings, however, do not constitute a clearcut division between the kinds of non-human agents which are accepted in Dutch and which are not. We have investigated a number of variables (translation of the subjects, translation of give, semantic roles of the subjects, lexico-semantics of the subjects, verb meanings of give and geven, valency), but the enquiry into non-human agents as subjects of Dutch geven appears to be a complex study. The role of interference from the source text, for instance, remains unclear. Are the instances with a non-human agent as subject of geven the result of interference or are they generally accepted and even idiomatic in Dutch? J. Delsoir argues that non-human agents “that are very common in English, [are] apparently (…) not as unidiomatic in Dutch as thought before” (2011: 41). If this is true, translations of non-human agents with geven attested in this study could be considered typical instances of the Dutch language. 6. Conclusion This paper aimed to investigate how a set of 154 English source-text sentences with a non-human agent as the subject of the verb give is translated into Dutch. In Dutch, non-human agents as subjects of geven, the cognate of give, are considered atypical instances of the language, since geven only occurs with non-human agents in subject position in a few restricted senses found in Van Dale Groot Woordenboek der Nederlandse Taal. Our dataset analysis revealed, however, that approximately one third of the Dutch translations do show non-human agents as subjects of geven. In the other instances, translators applied several translation solutions to avoid atypical instantiations, most of which led to valency loss, but not necessarily to loss of information, in the Dutch translations. Further research involving non-human agents in subject position may disclose why non-human agents referring to concrete or abstract causes and to events are more readily accepted as subjects of geven than non-human agents denoting non-instrumental objects and embedded events. Is this an effect of interference from the source texts? Or could it be – as J. House (2008) assumes – that Dutch language norms are implicitly drifting towards Anglo-American norms, thus introducing new verb meanings to geven, similar to the ones which are already available for English give?



Non-human agents as subjects in English and Dutch

References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. Longman grammar of spoken and written English. London: Longman. Chomsky, N. 1957. Syntactic structures. s’-Gravenhage: Mouton & Co., N.V. Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago: University of Chicago Press. Delsoir, Jan. 2011. The acceptability of non-prototypical agents with prototypical agent requiring predicates in Dutch. Gent: Hogeschool Gent. Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–619.  doi:  10.1353/lan.1991.0021

Fauconnier, Stefanie. 2012. Constructional effects of involuntary and inanimate agents: A cross-­ linguistic study. Unpublished PhD, Katholieke Universiteit Leuven. Fillmore, Charles J. 1968. The case for case. In Universals in linguistic theory. Fillmore, Charles J. 1977a. The case for case reopened. Syntax and Semantics, vol. 8. New York/ San Francisco/London: Academia Press. Fillmore, Charles J. 1977b. Scenes-and-frames semantics. In A. Zampolli (ed.), Linguistic structures processing, 55–88. Amsterdam: North Holland. Gambier, Yves. 2009. Stratégies et tactiques en traduction et interpretation. In G. Hansen, A. Chesterman, & H. Gerzymisch-Arbogast (eds.), Efforts and models in interpreting and translation research, 63–82. Amsterdam: John Benjamins. doi: 10.1075/btl.80.07gam Givón, Talmy. 1993. English grammar: A function-based introduction, 2 vols. Amsterdam: John Benjamins. Givón, Talmy. 2001. Syntax: An introduction. vol. 2. Amsterdam: John Benjamins.  doi:  10.1075/z.syn2

Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago: Chicago University Press. Hawkins, John. A. 1986. A comparative typology of English and German: Unifying the contrasts. Austin, London and Sydney: Croom Helm. House, Juliane. 2008. Towards a linguistic theory of translation as re-contextualisation and a third space phenomenon. Linguistica Antverpiensia 7. 149–175. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Hundt, Marianne. 2004. Animacy, agentivity, and the spread of the progressive in modern English. English Language and Linguistics 8(1). 47–69. doi: 10.1017/S1360674304001248 Klaiman, M. H. 1991. Grammatical voice. Cambridge: Cambridge University Press. Kussmaul, Paul. 1994. Semantic models and translating. Target 6(1). 1–13.  doi:  10.1075/target.6.1.02kus

Levin, Beth. 1993. English verb classes and alternations: A preliminary investigation. Chicago: University of Chicago Press. Palmer, Frank R. 1994. Grammatical roles and relations. New York: Cambridge University Press.  doi:  10.1017/CBO9781139166638

Primus, Beatrice. 1999. Cases and thematic roles: Ergative, accusative and active. Tübingen: Niemeyer Verlag. doi: 10.1515/9783110912463 Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1972. A grammar of contemporary English. London: Longman.

111

112 Steven Doms, Bernard De Clerck and Sonia Vandepitte

Rura, Lidia, Willy Vandeweghe & Maribel M. Perez. 2008. Designing a parallel corpus as a multifunctional translator’s aid. Proceedings of the XVIII FIT World Congress, 4–7 August 2008, Shanghai, China, from http://www.kuleuvenkulak.be/DPC/algemeen/src/FIT2008-rural. pdf. Sanfilippo, Antonio. 1990. Grammatical relations, thematic roles and verb semantics. Unpublished PhD, University of Edinburgh. Schlesinger, Izchak M. 1989. Instruments as agents: On the nature of semantic relations. Journal of Linguistics 25(1). 189–210. doi: 10.1017/S0022226700012147 Siewierska, Anna. 1991. Functional grammar. Routledge: New York. Tamm, Anne. 2012. Scalar verb classes. Scalarity, thematic roles, and arguments in the estonian aspectual lexicon. Firenze: Firenze University Press. Talmy, Leonard. 2000. Toward a cognitive semantics, vol. 1. Cambridge: The MIT Press. Vandepitte, Sonia & Robert J. Hartsuiker. 2011. Metonymic language use as a student translation problem: Towards a controlled psycholinguistic investigation. In C. Alvstad, A. Hild, & E. Tiselius (eds.), Methods and strategies of process research. Integrative approaches in translation studies, 67–92. Amsterdam: John Benjamins. doi: 10.1075/btl.94.08van Velupillai, Viveka. 2012. An introduction to linguistic typology. Amsterdam: John Benjamins.  doi:  10.1075/z.176

Online dictionary items “geven”. Dikke Van Dale. Van Dale Uitgevers. 02 December 2013. “give”. Oxford Dictionaries. Oxford University Press, n.d. Web 21 October 2013. .

Part 2

Valency-changing devices and non-finite verb forms

The argument-structure configuration of English middle and related structures Casilda García de la Maza

University of the Basque Country UPV/EHU

This paper assesses the place of middle structures in relation to transitivity and intransitivity in English. Unlike some accounts that have questioned the grammatical status of middles as a self-standing category, this research justifies the recognition of middles as a structural category that constitutes an independent class of intransitive sentences. It does so through an analysis of the underlying thematic structure of middles themselves, as compared to structures which deviate from the middle prototype and other non-canonical structures. A discussion of these issues highlights the great fluidity shown by English verbs entering diathesis alternations, as well as the intransitivisation process exhibited by the English verbal paradigm. The paper concludes with some theoretical considerations regarding the notion of argument structure.

Introduction English middle structures (Cookery books translate easily) are characterized by having a grammatical subject that is understood as the Patient of the verb, and by an Agent argument which is not overtly expressed, but is semantically present (Keyser and Roeper 1984; Roberts 1987; Fagan 1988; Hoekstra and Roberts 1993; Ackema and Schoorlemmer 1994). Middles bear a certain resemblance to passive sentences (The books were translated), in that an argument bearing a Patient theta-­role, or what seems to be a Patient theta-role, appears in subject position. Yet, unlike passives, no morphological marking is present on a middle verb to indicate this ‘deviation’ from the canonical way of mapping theta-roles to grammatical positions, which usually associates Agent theta-roles with subject positions and Patients with object positions (Baker 1988, 1997). Middles also differ from other types of structures with which they do, however, share the superficial similarity of being intransitive, one-argument clauses: ergative sentences (The vase broke), which lack the implied Agent argument that characterises the middle construction. Furthermore, doi 10.1075/lis.33.05gar © 2016 John Benjamins Publishing Company

116 Casilda García de la Maza

they exhibit a semantic idiosyncrasy that goes beyond what we would expect from the grammatical rearrangement of their arguments and that yields a highly marked constructional meaning (Jackendoff 1992; García de la Maza 2011). All this makes middles a particularly fertile ground for research on mapping and configurational issues. This paper explores the underlying thematic configuration of middles vis-à-­vis other argument structure constructions and constructions which deviate from the middle prototype itself. A discussion of these issues highlights the great fluidity shown by English verbs entering diathesis alternations as well as the intransitivisation process exhibited by the English verbal paradigm. Section 1 contains a detailed description of the grammar and semantics of middles. Section 2 presents the various transitivity alternations which English verbs can enter, with particular attention to the English middle construction and the different possibilities that the system affords. The final section presents a reflection on the very need for argument structure in syntactic theory. 1. What are middles? The English middle construction (Van Oosten 1977; Keyser and Roeper 1984; Fellbaum 1985; Roberts 1987; Fagan 1988, 1992; Hoekstra and Roberts 1993; Ackema and Schoorlemmer 1994; Rosta 1995; Hundt 2007; García de la Maza 2011; see also García de la Maza 2008) can be illustrated with the following sentences: (1) Love stories read easily. (2) Cookery books translate easily. (3) Isabel Allende’s novels sell like hot cakes.

In all of them, the grammatical subject is understood as the patient of the verb. The love stories are the thing being read, cookery books are the thing being translated, and Isabel Allende’s novels are the thing that is being sold. And, even though the Agent is not superficially expressed, it is semantically present. Our knowledge of the world tells us that books do not read or translate by themselves, and neither do novels sell alone. A reader, a translator and a seller are somehow implied. Several other labels have been proposed to refer to the middle construction. Jespersen (1909–1949: III 347), for example, uses the term ‘activo-passive’. Lyons (1968: 366) refers to it as a kind of pseudo-intransitive construction. Van Oosten (1977: 459) and Fellbaum (1985: 21) call it patient-subject construction, thereby capturing the idea that “the NP which is semantically the patient of the verb is expressed as the subject” (Van Oosten 1977: 459). Rosta (1995: 123) and Hundt



The argument-structure configuration of English middle and related structures 117

(2007) prefer to use the term ‘mediopassive’. That was also the term used by Grady (1965: 270), who defined it as ‘an active voice syntactical pattern wherein the subject-­verb relationship is notionally passive’.1 We will refer to them as ‘middles’. It is a crucial feature of middles that they are essentially stative, property-­ predicational constructions. We could draw on Halliday’s (1967: 38ff.) notion of transitivity, which he characterises in cognitive terms, according to the type of process expressed in the clause. Middles can be seen to be attributive clauses, involving primarily a process of ascription of an attribute (“a property of […] one of the participants in the clause” p. 62) to an attribuant (the participant to whom the property is ascribed). Halliday’s (p. 47) example of an attributive clause is She looked happy, where an attribute (happiness) is being ascribed to an attribuant (she). The subject of a middle sentence, like This book reads easily, could equally be seen as having the role of an attribuant to which the property of being easy to read is being ascribed. Erades (1975: 36) refers to this characteristic of middles by noting that “the construction in question is only found when the subject is represented as having certain inherent qualities which promote, hamper or prevent the realisation of the idea expressed by the predicate.” The same insight is captured by Van Oosten (1977: 461) who claims that in middles, “the subject, or a property of it, is understood to be responsible for the action of the verb.” It is, in fact, generally taken to be a characteristic of middles that they “lack specific time reference” (Levin 1993: 26), and that they are generic statements, and, as such, they state propositions that are generally true and do not describe particular events in time (Keyser and Roeper 1984: 384; see also Roberts 1987: 195ff.). Sentence (2), for example, is attributing some property to cookery books that makes them easy to translate, but no actual translating event is being referred to. Whereas this is a characteristic that applies to many middles, examples of ‘eventive middles’ can also be found, as in (4): (4) I thought that this book was going to be really difficult to translate, but it wasn’t. It translated really easily.

Middles are typically associated with a notion of ability or modality, as pointed out by Fagan (1992: 22). Thus, (2) could be paraphrased using a modal: (5) Cookery books can be easily translated.

Middles usually require some kind of modification. The adverbs that appear in middles are usually of the type referred to as ‘facility adverbs’ (Vendler 1984: 305) or ‘middle adverbs’ (Fellbaum 1989: 126), typically easily, with difficulty, smoothly, 1. See also Kemmer (1993) for a typological account of the ‘middle voice’ systems.

118 Casilda García de la Maza

well, or the like. As indicated by their label, their meaning points to the ‘facility’ or ‘ease’ with which the action denoted by the verb is accomplished, which is precisely why they fit so well the semantics of middles. If not an adverb, some other element is typically present, such as negation, as in (6), a modal verb, as in (7), or a combination of the two, as in (8): (6) That book didn’t sell. 

(from Dixon 1991: 326)

(7) The car will steer, after all. 

(from Rosta 1985: 132)

(8) This book won’t translate.

Hoekstra and Roberts (1993: 192) have argued that adverbial modification is a compulsory part of middle structures, and have accounted for it on strictly derivational terms. Their analysis integrates adverbs into the grammatical makeup of the construction and therefore predicts that unmodified middles are “ungrammatical”, as shown in (9).2 The fact is, however, that, adjunctless middles are possible. As noted by Fellbaum (1985: 23) and Rosta (1995: 132), in order for middles to be acceptable, they have to provide ‘newsworthy information’. Consider examples (10) and (11) (from Fagan 1992: 57): (9) ???The book reads. (10) Glass recycles. (11) This dress buttons.

Thus, the difference between examples (9) on one hand and (10) and (11) on the other lies in the different relationship that holds between the properties of the subject and the action denoted by the verb. Under normal circumstances and assuming no context, it is an inherent property of books that they can be translated, and no newsworthy information is being provided by stating it. Similarly, there would not be much relevance in stating, for example, Rubbish recycles, since, after all, it is well-known that recycling consists precisely in turning ‘rubbish’ into reusable material. On the other hand, stating that glass (as opposed to plastic or cardboard, for example) recycles, conveys relevant information, since not all types of materials can be recycled. Finally, saying This dress buttons implies that it does not zip or ‘velcro’, and so conveys newsworthy information.

2. More precisely, they argue that the suppressed subject argument in a middle sentence is a pro which needs to be “theta-licensed for content.” This requirement is satisfied by adverbs like easily, whose Experiencer theta-role is identified with pro. Lack of adverbial modification means that pro is left unlicensed and that the sentence is, therefore, ungrammatical.



The argument-structure configuration of English middle and related structures 119

2. English diathesis alternations This section analyses the structural make-up of middle sentences and shows how these are clearly distinct both from unergative and ergative constructions. Let us begin by introducing a classification of verbs on the basis of their argument structure, paying particular attention to the expression of subjects and objects (see also García de la Maza 2008). Pure transitive verbs are those like cut or sell, which take an object and involve two participants, an Agent, and a Theme or Patient.3 Having both Agents and Patients, pure transitive verbs do meet the structural requirements for middle formation, and they form unambiguous middles, like Bread cuts easily. Intransitive verbs are those that occur without an object and involve one single participant, which is grammatically a subject. Semantically, it can be a Patient or an Agent. This is the basis for Perlmutter’s (1978: 160ff.) Unaccusative Hypothesis, which distinguishes two types of intransitive verbs. Unergative verbs, such as run, laugh or swim, have a subject that performs some action or activity and is associated with an Agent theta-role. In Isabel laughed, Isabel is an Agent, and a subject both underlyingly and on the surface. Since they do not take objects, unergative verbs do not form middles. A sentence like Isabel laughs easily cannot be understood as a middle. The second type of intransitive verb is the unaccusative verb. These are verbs like fall or die, whose subject position refers to the participant that undergoes some change of location or state and is associated with a Patient or Theme theta-­ role. What is more, the Patient theta-role is represented underlyingly in object position, and raised to subject (external) position later on in the derivation of the sentence. In standard Government and Binding analyses, a sentence like Paula died would have an underlying structure __died Paula. Movement of Paula to subject position is forced for reasons of Case. Non-alternating unaccusative verbs fail to assign accusative Case to their objects and they do not assign an external theta-­role either, as captured by Burzio’s generalisation (Burzio 1986: 178). Some unaccusative verbs, like break, open, melt, can also appear in transitive sentences. These are the alternating unaccusative verbs. Along with the intransitive 3. The Agent theta-role is normally associated with the participant that is the ‘doer’, ‘performer’ or ‘instigator’ of an action (Gruber 1967: 943; Fillmore 1968: 24; Lyons 1968: 356; Jackendoff 1972: 32; Cruse 1973), whether volitional, as Mary in Mary ate an apple or non-volitional, as in She coughed. The Patient and Theme theta-roles are often employed indistinctively to refer to the participant which is, in a broad sense, “affected by the action or state identified by the verb” (Fillmore 1968: 25). This includes cases where physical motion is involved, as the cup in I broke the cup, but is can also include non-moving participants or participants which are not physically affected, as the book in Max owns the book (Jackendoff 1972: 30).

120 Casilda García de la Maza

The door opened, the transitive I opened the door exists. The relationship between the transitive and intransitive variants of alternating unaccusative verbs is usually known as the ergative alternation, and the intransitive variant of the verb, and therefore the construction in which it appears, has been called ‘ergative’. Non-alternating unaccusative verbs, like fall, arrive or die, do not form middles, given that they lack Agents. A sentence like The leaves fall easily cannot be interpreted as a middle, since no Agent is implied. Alternating unaccusative verbs (ergatives), on the other hand, do produce middles. A sentence like This door opens easily could, in principle, receive a middle interpretation. The distinction between a middle and an ergative is not always clear. The main difference lies in the presence or absence of an Agent. The middle, derived from the transitive variant of the alternation, involves an Agent argument. The middle interpretation implies that someone, at some stage, could open the door, or that it is easy for someone to open the door. The ergative, in the derivation of which no Agent is involved, implies that the door opens ‘all by itself ’.4 In fact, the incompatibility of middles with the phrase ‘all by itself ’, and the fact that ergatives admit it freely, has been taken as an indication of the presence of an Agent in the former, and of the lack thereof in the latter (Fagan 1992: 52; Keyser and Roeper 1984: 405; Rapoport 1999: 150), as the following examples show: (12) *This wood carves easily all by itself. 

(from Rapoport 1999: 150)

(13) The cup broke all by itself.

So, from a strictly structural point of view, middles can be formed straightforwardly with pure transitive verbs, not so straightforwardly with alternating unaccusative verbs, since ambiguity with ergative sentences exists, and not at all with non-­alternating unaccusative verbs and with unergatives.5 A word of caution is needed at this stage. Saying that middles must be formed from an input structure in which two participants, both Agent and Patient, are present does not mean that any such structure yields a satisfactory middle sentence. There are restrictions at work. Roberts (1987: 207ff.) and Fagan (1992: 64ff.) both argue that the aspectual properties (Vendler 1967: 100) of the verb phrase in question determine its eligibility for middle formation. More precisely, they claim that accomplishments and activities are eligible for middle formation whereas 4. I am assuming that ergatives are the basic variant of the alternation and that they are obtained from an underlying configuration in which the argument bearing the Theme theta-role is in object position and the subject position is empty. But other proposals have been made regarding the derivation of ergatives, as noted below. 5. See Davidse and Hayvaert (2003) and Marín-Arrese (2011).



The argument-structure configuration of English middle and related structures 121

states and achievements are not. This would explain the impossibility of middles like those in (14) and (15) below, given that the verb phrases know the answer and spot the mistake are a state and an achievement respectively. It is also common to find accounts of the middle construction which describe it as being restricted to verbs with affected objects (Doron and Rappaport-Hovav 1991; Levin 1993: 26). Examples like (16), however, show that this is not the case, as Fagan (1992:65) points out. The sentence is a perfect middle, and there is no way in which the book can be seen as affected or undergoing a change of state. There are, furthermore, those who have claimed that middles like that in (14) are “grammatical in principle, although deviant on semantic terms” (Keyser and Roeper 1984: 383). This is an issue, in sum, that is far from settled, and which clearly falls beyond the scope of the present paper, but see García de la Maza (2013, 2014) for some proposals in this direction. (14) *The answer knows easily. (15) *This type of mistake spots easily. (16) This book reads easily.

Having made this clarification, let us not lose the thread. Leaving aside the ability of the different classes of verbs to form middles, and focusing now on the more general picture of transitivity and intransivity, our discussion so far can be summarised as follows: Unergatives: Transitive: Ø Intransitive: I walk Non-alternating unaccusatives: Transitive: Ø Intransitive: She died Alternating unaccusatives: Transitive: I broke the window Intransitive: The window broke Pure transitives: Transitive: I cut the wood Intransitive variant: Ø

How do middles fit in this picture? As intransitives, prototypically obtained from pure transitive verbs, they occupy the intransitive variant slot corresponding to pure transitive verbs. Pure transitive verbs are thus on a par with alternating unaccusative verbs, which also have transitive and intransitive variants. The parallelism with alternating unaccusatives is important. If it were just a matter of

122 Casilda García de la Maza

presence or absence of an object, it is uncontroversial that pure transitive verbs can appear without an object (Huddleston 1984: 192; Payne 1997: 170; Goldberg 2001, 2004: 434). This happens when the object is generic, recoverable from the context, particularly within a certain jargon, and “de-emphasised in the discourse vis à vis the action” (Goldberg 2001: 514), as shown by the following examples, which we can refer to as ‘reduced transitive sentences’: (17) Have you eaten? (Generic object: Food) (18) England won. (Understood object: The match) (19) I would like to submit by May. (Student jargon. Understood object: PhD thesis) (20) When would be a good time to sell? (Understood object: Shares. Stock market jargon)

What we are rather trying to show is that pure transitive verbs are able to participate in a transitivity alternation, in the sense of exhibiting changes in the number and syntactic alignment of their arguments, just as alternating unaccusative verbs do. Middle sentences thus constitute the intransitive counterparts of pure transitive sentences. Middles seem to constitute a step towards the ‘levelling’ of the transitivity paradigm in English, by creating the possibility for pure transitive verbs to have intransitive counterparts. It is noteworthy that middles are not alone in contributing to this move towards symmetry. As further evidence of how structural possibilities made available by the transitivity paradigm can be exploited, consider the following constructions: (21) This suggests he is worried that those who can might rather walk the dog or paint the house than work for minimal benefit. (British National Corpus (BNC), AHB 206) (22) One big mistake I made, being used to handling Dawn, was trying to walk the hawk all the way back to its perch.  (BNC, CHE 1548) (23) Captain Ramirez called them to attention and marched them off to the main camp. (from Davidse and Geyskens 1998: 155)

The existence of this type of construction has been noticed by a number of authors (Levin 1993: 31; Levin and Rappaport Hovav 1994: 70; Davidse and Geyskens 1998). The verbs walk and march are canonically unergative verbs, and should have a single argument, the Agent, occupying the external argument position (I walked, I marched). In these examples, however, the original Agent argument is occupying the object position, and a new external argument, the causal Agent, has



The argument-structure configuration of English middle and related structures 123

been introduced. These sentences have a causative meaning, that of making the dog walk, or making the soldiers march, and are most systematically found with manner-­of-motion verbs (walk, jump, gallop, march, race…), as noted in all the works mentioned at the beginning of this paragraph (especially in Davidse and Geyskens (1998), who offer a description of the sub-types in which this construction can appear with manner-of-motion verbs). Examples (21)–(23) thus instantiate the structural possibility for unergative verbs to have transitive counterparts. It should be noted that the ‘transitive unergative’ structural possibility can be used in ways that do not fully conform to the characteristics of the ‘mother construction’, as shown by the following data: (24) Ian’s working me hard. He made me get up at seven… (Attested) (25) Yoko ran me to Tesco’s. (Attested) (26) Walk the boat out until the water is deep enough for you to put the daggerboard down far enough so that you’ll be able to sail away.  (BNC, J3X 215) (27) Obviously the rig doesn’t always conveniently lie in this position for you, so you must swim the rig into position with a strong leg action.  (BNC, AT6 1041)

In example (24) we find a causative use of the verb work, which is not a manner-­ of-motion verb. Its object corresponds to the original Agent argument in its intransitive counterpart (I/you work hard). Sentence (25) involves a different type of causation from the one found in the previous set of data. It does not mean that Yoko made me run to Tesco’s, but that she gave me a lift to Tesco’s. It would correspond to a transitive sentence Yoko ran the car, itself a causative derived from The car ran. In our example, the original Agent, Yoko, remains in place, but a new participant is introduced in object position, me, yielding a double causative: Yoko caused me to go to Tescos’s by causing the car to run.6 Finally, examples (26) and (27) also diverge from the ‘walk the dog’ pattern. Clearly, they cannot be taken to mean ‘cause the boat to walk’ or ‘cause the rig to swim’. Rather, it is the original Agent argument who does the swimming and the walking. The verbs walk and swim are motion verbs which involve a “manner” semantic component (Talmy 1985). Their meaning becomes in turn “conflated” with what Talmy calls a “path” component, contributed by the directional phrases out, or into position. The introduction of an object transitivises the sentence, introducing the ‘cause’ component, so that the whole thing comes to mean ‘take the boat out by walking’ and ‘put the rig into position by swimming’; or rather, more accurately, ‘cause the boat to be

6. See Davidse and Geyskens (1998: 167) for an account of a similar example with the verb fly.

124 Casilda García de la Maza

out by walking’ and ‘cause the rig to become in position by swimming’.7 From the structural point of view, they all represent genuine instances of the transitivisation of an unergative verb. To complete the parallelism with our discussion of middles, notice the following examples: (28) A curfew was imposed after nightfall, but it was ill-observed, and almost anyone who had a reasonable excuse was permitted to walk the streets without hindrance. (BNC, G17 20883) (29) Among the pioneers teaching before Pavlova’s arrival, the most important one was Helen Webb (whose father was famous as the first man to swim the English Channel). (BNC, ASC 459)

In (28) the streets is a locative object (Greenbaum and Quirk 1990: 212), just like the English Channel in (29). They do not introduce a new participant, and thus differ from our examples (24)–(27), which did instantiate the transitivisation of an unergative verb, or what we could call the ‘unergative alternation’. Unergatives now are on a par with alternating unaccusative verbs and with pure transitive verbs in their ability to have both transitive and intransitive variants. The picture of the transitivity alternations in English is now one of great symmetry, and would look as follows: Unergatives: Transitive: I walk the dog Intransitive: The dog walks Non-alternating unaccusatives: Transitive: Ø Intransitive: She died Alternating unaccusatives: Transitive: I broke the window Intransitive: The window broke Pure transitives: Transitive: I cut the wood Intransitive variant: This wood cuts (easily)

7. More generally, see Levin and Rappaport Hovav (1994: 69ff.) for an account of the interaction of the causative use of verbs of manner of motion with a directional phrase, which seeks to account for the contrast in The rider jumped the horse over the fence vs *The rider jumped the horse (their examples, p. 70). It should be noted, however, that the presence of a directional phrase, though very typical, is not always required, as shown by (21), and that it can sometimes be replaced by a resultative phrase, like hard in (24).



The argument-structure configuration of English middle and related structures 125

Coming back to middles, what matters is that we have identified middles as a distinct and independent grammatical category which comes to occupy an otherwise empty slot in the picture of transitivity alternations in English. We thus depart from existing views (Condoravdi 1989; Lekakou 2002), according to which no such thing as a middle construction exists, ‘middle’ being only “a type of interpretation certain sentences receive and [can therefore be] seen as a notional category independent of its grammatical properties” (Condoravdi 1989: 24). From a structural point of view, according to them, then, middles do not exist, and the sentences receiving a middle interpretation are subsumed under general transitivity alternations already existing in the language. Thus, Condoravdi does not distinguish middles from ergatives, whereas Lekakou asserts that what we call middle in English is nothing but an interpretation that unergative sentences receive. We argue that, in English, middles do constitute a distinct class of intransitive clauses in addition to the ones predicted by the Unaccusative Hypothesis, unergatives and unaccusatives. The differences between the three of them can be found in their underlying thematic structures (García de la Maza 2008). Unergatives are characterised by having a single Agent argument represented both underlyingly and on the surface as an external argument (Isabel laughed). Ergatives and unaccusatives are underlyingly represented as having also one single argument, a Patient/Theme, which occupies the internal argument position, and which then surfaces as an external argument at S-structure (The cup broke, or Paula died). Middles are represented as having originally two arguments, an internal one, a Patient/Theme, and an external Agent argument. On the surface, the internal argument appears in subject position. The Agent argument is left unrealised, but we know that it is a crucial, defining feature of middles that it is semantically present, and that no middle sentence could exist with an input verb which did not have an Agent as part of its thematic structure. So far, so good. Confusion might arise, however, because of the existence of the superficially identical unaccusative, or ergative, alternation (I broke the cup/ The cup broke), and some of the analyses it has received, as noted in García de la Maza (2008). On an analysis in which the intransitive (ergative) variant of the alternation is the basic one whereas the transitive variant is derived from it by inserting an Agent in subject position (see Note 4), middles are still fundamentally distinct from ergatives in having one more argument (the Agent) than them. The confusion might come in on analyses of the ergative alternation in which the transitive variant (I broke the cup) is taken to be the basic one from which ergatives are derived, since then, as explained above, the ergative and the middle alternation would be apparently identical. The key to distinguishing ergatives (The cup broke) from middles (This meat cuts well) would then lie in the manner in which we deal with the Agent argument, which would need to be deleted in the case of ergatives,

126 Casilda García de la Maza

so as to obtain an intransitive sentence in which the action happens ‘all by itself ’. No such deletion of the Agent could take place in the case of middles, as just noted. What is crucial is that it is this Agent argument that is inherent to middles that warrants the recognition of a ‘middle’ intransitive category, and that makes it distinct from both ergatives and unergatives. What is more, an approach to the ergative alternation in which the intransitive variant is more basic seems preferable to me, if only for theoretical reasons, i.e. to set the different classes of intransitive sentences apart from each other in a more simple and transparent way. This is in line with Williams (1981), Brousseau and Ritter (1991: 57–62), and Rappaport Hovav and Levin (2012) who also take the intransitive variant of the alternation to be basic and the transitive variant derived from it by the addition of an argument.8 3. The representation of argument structure In view of our discussion above and the data we have presented, and by way of conclusion, some remarks about the great fluidity and variability that English verbs allow in the expression of their arguments are in order. We have seen, for example, that the verb walk can be used transitively, intransitively, or with a locative object, even though it is classed as an unergative verb. We have also seen that a pure transitive verb like sell can be used without an object, whether this is in a middle construction or in a reduced transitive sentence. And this is just a small sample. In a more comprehensive study, Thompson and Hopper (2001) show that there is a tendency in conversation for clauses to be low in transitivity. This is shown, among other things, by the fact that 73% of the sentences in their database have a single participant, the number of participants being one of the key ingredients of the scalar notion of transitivity (Hopper and Thompson 1980) which they use, and clearly the defining feature of the structural notion of transitivity we are 8. But accounts in the opposite direction have also been proposed. Amongst these we find Levin and Rappaport Hovav’s (1995: 79–133) earlier work, who posit that the causative (transitive) variant is more basic. In favour of this analysis they argue that the selectional restrictions on the set of subjects that the intransitive variant can appear with tend to be greater than the restrictions on the objects in the transitive variant (compare The book will open your mind vs. *Your mind will open from this book (their examples, p. 85)). The variant with looser selectional restrictions is taken to be the basic one. Furthermore, they invoke evidence (from Nedjalkov 1969) that shows that, across languages, the transitive forms of verbs like break are morphologically unmarked, and hence considered to be more basic than the intransitive forms, which, in 22 out of the 60 languages included in the study, show derived forms (in 19 out of 60 languages both forms are identical). See also Mendikoetxea (2000).



The argument-structure configuration of English middle and related structures 127

dealing with at present. The evidence they gathered from conversational contexts also led them to note that the lexical entries for verbs specifying their argument structures have little to do with actual usage, high-frequency verbs being particularly unpredictable in the number of participants they can appear with. They mention the case of verbs like drive or forget, traditionally held to be two-­participant verbs, which most often occur in a single-participant clause in everyday conversation. The most immediate conclusion this gives rise to is the need to discard the idea of transitivity or intransitivity as a fixed and paradigmatic feature of a verb, and to define it instead in terms of the syntagmatic and meaning relations in the clause, a point over which, on the other hand, there is no discrepancy in modern syntactic theory (Jespersen 1909–1949: III 319–320; Halliday 1967: 52). A slightly more controversial issue that stems from the ideas just set forth concerns the very need for argument structure. Thompson and Hopper (2001), on the basis of the conversational evidence alluded to above, reject the traditional notion of argument structure as the fixed and pre-determined entry with information about the number and type of semantic roles a verb takes. They defend instead the view that no such thing as ‘argument structure’ exists, and that what appears to be fixed structure is actually a “set of schemas, some more “entrenched” (Bybee 1985, 1998; Langacker 1987) than others, arising out of many repetitions in daily conversational interactions” (p. 47). Instead of a theory of argument structure, they advocate a “probabilistic theory capturing the entire range of combinations of predicates and participants that people have stored as sorted and organised memories of what they have heard and repeated over a lifetime of language use”. This, they admit, will look very much like “a good unabridged dictionary” (p. 47). In other words, there is nothing fixed or basic about the way clauses are put together; everything arises from usage and communicative needs. Any structure that we might want to attribute to language has to be explained through frequency and actual usage. As appealing as some of these ideas about linguistic theory are, Thompson and Hopper’s approach to clause structure seems too radical, and clearly not very practical. It seems preferable to adhere to Payne’s (1997: 170) traditional view of argument structure. He notes that verbs evoke ‘scenes’ which are associated with a specific number of participants. He talks about the semantic valence of a verb, the number of participants that a verb has conceptually, and its grammatical valence, the number of arguments present in any given clause. He gives the example of the verb eat, which has a semantic valence of two, since “for any given event of eating there must be at least an eater and an eaten thing” (p. 169), but can also have a grammatical valence of one, as we already saw above. Doing syntactic theory without having any such representation of the core or basic meaning of a verb seems a very difficult task. For one thing, the classification of verbs and the patterns and generalizations we have outlined in this chapter

128 Casilda García de la Maza

would not have been possible. It is then the linguist’s task to decide and justify which argument structure should be considered more basic, how different argument structures should be related to one another, or when completely different lexical entries should be posited. A good deal of linguistic insight can be gained from such decisions and their justification.

References Ackema, Peter & Maaike Schoorlemmer. 1994. The middle construction and the syntax-semantics interface. Lingua 93. 59–90. doi: 10.1016/0024-3841(94)90353-0 Baker, Mark C. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Baker, Mark C. 1997. Thematic roles and syntactic structure. In Liliane Haegeman (ed.), Elements of grammar. Handbook of generative syntax, 73–137. Dordrecht: Kluwer. British national corpus, compiled by the British National Corpus Consortium. http://sara. natcorp.ox.ac.uk/lookup.html Brousseau, Anne-Marie & Elizabeth Ritter. 1991. A non-unified analysis of agentive verbs. In Proceedings of the Tenth west coast conference on formal linguistics, 53–65. Stanford, CA: Centre for the Study of Language and Information. Burzio, Luigi. 1986. Italian syntax: A government-binding approach. Dordrecht: Reidel.  doi:  10.1007/978-94-009-4522-7

Bybee, Joan. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins. doi: 10.1075/tsl.9 Bybee, Joan. 1998. The emergent lexicon. Proceedings from the Panels of the Chicago linguistic society’s meeting 34. 421–435. Davidse, Kristin & Liesbet Heyvaert. 2003. On the so-called ‘middle’ construction in English and Dutch. In Sylviane Granger, Jacques Lerot, & Stephanie Petch-Tyson (eds.), Empirical approaches to contrastive linguistics and translation studies, 57–73. Amsterdam: Rodopi. Condoravdi, Cleo. 1989. The middle: Where semantics and morphology meet. MIT Working Papers in Linguistics 11. 18–30. Cruse, D. Alan. 1973. Some thoughts on agentivity. Journal of Linguistics 9. 11–23.  doi:  10.1017/S0022226700003509

Davidse, Kristin & Sara Geyskens. 1998. Have you walked the dog yet? The ergative causativization of intransitives. Word 49(2). 155–180. doi: 10.1080/00437956.1998.11673881 Dixon, Robert M. W. 1991. A new approach to English grammar, on semantic principles. Oxford: Clarendon Press. Doron, Edith & Malka Rappaport Hovav. 1991. Affectedness and externalization. Proceedings of the Annual meeting of the north eastern linguistic society 21. 81–94. Erades, P. A. 1975. Points of modern English syntax. Contributions to English studies by P.A. Erades. Amsterdam: Swets and Zeitlinger. Fagan, Sarah M. 1988. The English middle. Linguistic Inquiry 19. 181–203. Fagan, Sarah M. 1992. The syntax and semantics of middle constructions. Cambridge: Cambridge University Press.



The argument-structure configuration of English middle and related structures 129

Fellbaum, Christiane. 1985. Adverbs in agentless actives and passives. Papers from Parasession on causatives and agentivity. Chicago linguistic society 21(2). 21–31. Fellbaum, Christiane. 1989. On the reflexive middle in English. Papers from the Annual regional meeting of the Chicago linguistic society 25(1). 123–132. Fillmore, Charles J. 1968. The case for case. In Emmon Bach & Robert T. Harms (eds.), Universals in linguistic theory, 1–90. New York: Holt, Rinehart and Winston. García de la Maza, Casilda. 2008. Intransitivity, ergatives and middles. Estudios Ingleses de la Universidad Complutense 16. 31–50. García de la Maza, Casilda. 2011. The semantics of English middles and pseudo-middles. In Pilar Guerrero Medina (ed.) Morphosyntactic alternations in English: Functional and cognitive perspectives, 161–181. London: Equinox. García de la Maza, Casilda. 2013. The conventionalisation of contextual effects in middle structures. International Journal of English Studies 13(1). 111–131. García de la Maza, Casilda. 2014. Affectedness revisited. DELTA: Documentação de Estudos em Lingüística Teórica e Aplicada 30(1). 53–60. doi: 10.1590/S0102-44502014000100004 Goldberg, Adele E. 2001. Patient arguments of causative verbs can be omitted: The role of information structure in argument distribution. Language Sciences 23. 503–524.  doi:  10.1016/S0388-0001(00)00034-6

Goldberg, Adele E. 2004. Pragmatics and argument structure. In Laurence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 427–441. Oxford: Blackwell. Grady, Michael. 1965. The mediopassive voice in modern English. Word 21. 270–272.  doi:  10.1080/00437956.1965.11435431

Greenbaum, Sidney & Randolph Quirk. 1990. A student’s grammar of the English Language. London: Longman. Gruber, Jeffrey S. 1967. Look and see. Language 43(4). 937–947. doi: 10.2307/411974 Halliday, Michael A. K. 1967. Notes on transitivity and Theme in English. Part I. Journal of Linguistics 3(1). 37–81. doi: 10.1017/S0022226700012949 Hoekstra, Teun & Ian Roberts. 1993. Middle constructions in Dutch and English. In Eric Reuland & Werner Abraham (eds.), Knowledge and language, Volume II, lexical and conceptual structure, 183–220. Dordrecht: Kluwer. Hopper, Paul J. & Sandra A. Thompson. 1980. Transitivity in grammar and discourse. Language 56. 251–299. doi: 10.1353/lan.1980.0017 Huddleston, Rodney. 1984. Introduction to the grammar of English. Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139165785 Hundt, Marianne. 2007. English mediopassive constructions. A cognitive, corpus-based study of their origin, spread and current status. Amsterdam: Rodopi. Jackendoff, Ray S. 1972. Semantic interpretation in generative grammar. Cambridge, MA: MIT Press. Jackendoff, Ray S. 1992. Babe Ruth homered his way into the hearts of America. In Timothy Stowell & Eric Wehrli (eds.), Syntax and the lexicon, 155–178. San Diego: Academic Press. Jespersen, Otto. 1909–49. A modern English grammar on historical principles, vol. 7 Heidelberg: Carl Winters (volumes I–IV); Copenhagen: Munksgaard (volumes V–VII). Kemmer, Suzanne. 1993. The middle voice. Amsterdam: John Benjamins. doi: 10.1075/tsl.23 Keyser, Samuel Jay, & Thomas Roeper. 1984. On the middle and ergative constructions in English. Linguistic Inquiry 15. 381–416. Langacker, Ronald W. 1987. Foundations of cognitive grammar, Volume 1: Theoretical prerequisites. Stanford, CA: Stanford University Press.

130 Casilda García de la Maza

Lekakou, Maria. 2002. Middle semantics and its realization in English and Greek. UCL Working Papers in Linguistics 14. 399–416. Levin, Beth. 1993. English verb classes and alternations. Chicago: The University of Chicago Press. Levin, Beth & Malka Rappaport Hovav. 1994. A preliminary analysis of causative verbs in English. Lingua 92. 35–77. doi: 10.1016/0024-3841(94)90337-9 Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the syntax-lexical semantics interface. Cambridge, MA: MIT Press Lyons, John. 1968. Introduction to theoretical linguistics. Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139165570 Marín Arrese, Juana I. 2011. Spontaneous and facilitative events revisited. In Pilar Guerrero Medina (ed.), Morphosyntactic alternations in English: Functional and cognitive perspectives, 137–160. London: Equinox. Mendikoetxea, Amaya. 2000. Relaciones de Interficie: los Verbos de Cambio de Estado. In Ana Bravo, Carlos Luján, & Isabel Pérez (eds.), Cuadernos de Lingüística VII, 125–144. Madrid: Instituto Universitario Ortega y Gasset. Nedjalkov, Vladimir P. 1969. Nekotorye verojatnostnye universalii v glagol’nom slovoobrazovanii. In Ivan F. Vardul (ed.), Jazykovye universalii I lingvisticeskaja tipologija, 106–114. Moscow: Nauka. Payne, Thomas E. 1997. Describing morphosyntax. Cambridge: Cambridge University Press. Perlmutter, David M. 1978. Impersonal passives and the Unaccusative Hypothesis. Proceedings of the Annual meeting of the Berkeley linguistics society 21. 157–189. Rapoport, Tova R. 1999. The middle, agents, and for-phrases. Linguistic Inquiry 30(1). 147–155.  doi:  10.1162/002438999554011

Rappaport Hovav, M. & B. Levin. 2012. Lexicon uniformity and the causative alternation. In Martin Everaert,, Marijana Marelj, & Tal Siloni (eds.), The theta system: Argument structure at the Interface, 150–176. Oxford: Oxford University Press.  doi:  10.1093/acprof:oso/9780199602513.003.0006

Roberts, Ian G. 1987. The representation of implicit and dethematized subjects. Dordrecht: Foris.  doi:  10.1515/9783110877304

Rosta, Andrew. 1995. The semantics of English mediopassives. In Bas Aarts & Charles F. Meyer (eds.), The verb in contemporary English, 123–144. Cambridge: Cambridge University Press. Talmy, Leonard. 1985. Lexicalization patterns: Semantic structure in lexical forms. In Timothy Shopen (ed.), Language typology and syntactic description, 57–149. Cambridge: Cambridge University Press. Thompson, Sandra A. & Paul Hopper. 2001. Transitivity, clause structure, and argument structure: Evidence from conversation. In Joan Bybee & Paul Hopper (eds.), Frequency and the emergence of linguistic structure, 27–60. Amsterdam: John Benjamins.  doi:  10.1075/tsl.45.03tho

Van Oosten, Jeanne. 1977. Subjects and agenthood in English. Papers from the Annual regional meeting of the Chicago linguistic society 13. 459–471. Vendler, Zeno. 1984. Adverbs of action. Papers from the parasession on lexical semantics. Chicago linguistic society 20(2). 297–305. Williams, Edwin. 1981. Argument structure and morphology. The linguistic review 1. 81–114.  doi:  10.1515/tlir.1981.1.1.81

Non-categorical categories Aspect, Voice, Pred and the category of Participles Elena Soare

Université de Paris 8 & CNRS UMR 7023 Structures Formelles du Langage

This paper aims at circumscribing the range of structures that underlie participial constructions, with a particular emphasis on Romanian data. Participial constructions exhibit high flexibility, e.g., the participial stem labeled “supine” in Romanian grammars, which can be used in both nominal and verbal environments: (i) a. cărţi de citit b. cititul cărţilor c. am citit books of read.Sup read.Sup.the books.Gen have read books intended for reading the reading of books I have read The challenge these data raise to linguistic theory is to find the key property that would allow distinguishing a natural class of participles inside the larger class of non-­finite verbal forms. It is assumed, in line with much recent work, that this common property should be their truncated character: participles are deprived of some verbal layers; they are lower verbal domains. Taking this as a working definition, the term ‘participle’ is given a wider use; for instance, it is argued that a part of Romance infinitivals are participial constructions; apparently the same morphology can be used to cover different structures from language to language – defining morphology as the locus of variation. The main idea is that participles are truncated clauses of different heights, and not categories of a special kind. In the view that I propose, having a theory of lexical categories is not an interesting goal in itself (contra Baker 2003–2005). Rather, I show how the structural make-up of the different kinds of participles accounts for their overall behavior, making them behave like nouns, verbs, or adjectives. I make a distinction between non-finite clauses which (i) include a subject layer and (ii) involve defective Tense, like infinitives and English Acc-Gerunds, and participial constructions that (i) do not include a subject layer (ii) do not involve Tense, even defective, but at most personal and agreement marking. I focus on reduced participial domains and refer to non-finite tensed domains and participial nominalizations for comparative matters. I base my study mainly

doi 10.1075/lis.33.06soa © 2016 John Benjamins Publishing Company

132 Elena Soare

on Romance data (with key data from Romanian), and punctually refer to English for comparison.

1. Introduction 1.1

Grammatical indeterminacy

The present paper addresses a well-known case of atypical predicates, which have posed a longstanding problem to all grammatical theories, namely, mixed categories, and more precisely, participles. There is no agreement whatsoever among scholars and grammatical traditions with respect to the category of participles in language. Is it indeed a category to begin with? Does participial morphology have content, or is it just a sound form? If it is contentful, how can this content be characterized? How does it differ from other non-finite forms, e.g. infinitives? Traditional grammarians have long hesitated whether to classify them in the verbal, nominal, or adjectival paradigm. A case in point is the variety of categories taking the phonological shape of an -ing suffix in English, which can be Nouns (building), Verbs (avoiding), Adjectives (unwilling), and even Prepositions (concerning). English grammars further distinguish the gerund, which has both nominal and verbal properties, from the participle, which has only verbal properties. The wide distribution available for participials raises a problem to any classical theory of categorization in terms of N/V features; one would be forced to acknowledge that all the combinations of features are available (Cornilescu 2003). Accordingly, the participial marker (for instance, -ing for the “present participle” and -en/-ed for the “passive-past participle”) has been given different statuses by scholars across languages and frameworks. For English, it has been proposed that -en/ed is a derivational affix that creates adjectives (Borer 1993); the same is argued for Swedish participles by Lundquist (2008), who takes them as being simply adjectives. Doron and Reintges (2011) propose that participial modifiers in Hebrew and Ancient Egyptian are nominal structures, and the participial morphology is a nominalizing head. Other recent contributions treat the gerundial morphology for instance like nominal inflection with defective properties (Baker 2005). Some participial constructions, in Romance for instance, have been considered to be reduced (e.g., relative) clauses (Kayne 1995). Syntactic approaches to word formation (e.g., Distributed Morphology – Marantz 1997; Embick 2004; Alexiadou et al. 2012) address the verbal/adjectival ambiguity of participles as structural ambiguity: eventive participles are realizations of AspP while stative-­ adjectival participles are lower root phrases.

Non-categorical categories 133



Given their diversity and “flexibility”, participials can appear in a wide range of contexts: adnominal position (a shining light, a fallen leaf), nominalizations (the shining of the sun), periphrastic constructions (like the progressive or aspectual periphrases – the sun is shining; I started reading this book). Other participial forms may appear as complements or adjuncts to the verb or to the whole clause (I remember Mary’s performing the sonata; God willing, we will succeed). Consider also, in Romance, past participles, which are employed in perfect (1a), passive (1b), reduced relatives (1c), and the absolute participial construction (1d): (1) a. j’ai lu le livre ‘I have read the book’ b. le livre est lu ‘the book is read’ c. un livre lu ‘a read book’ d.  (une fois) le livre lu, elle est sortie se promener… ‘the book read, she went out’

As well, participles can have different interpretations across languages: they can be active, passive, aspectually marked perfective, or imperfective, progressive, dynamic or stative. In morphosyntactic terms, participles display a wide range of properties: some can take a subject and assign it Accusative or Genitive case; some can assign Accusative to their object, while others cannot; some allow clitics and sentential negation; some can bear agreement markers and some may give rise to modal readings. This diversity is due both to the internal make-up of the participle itself (e.g., if it is inflected for aspect or even tense) and to the contribution of external material (e.g., an auxiliary, a determiner). The variation in Voice, Aspect and Agreement can be morphologically marked; in other cases a single form can have multiple uses (e.g., passive and active uses). Such a situation is encountered with the Latinate participial stem, appearing in Latin Future Active Participle (2a), Passive Perfect Participle (2b, e), Latin and Romanian Supine ((2c–d), (3a)), and Romanian Supine Nominalizations (3b), all realized by the same segment V-T/S, in combination with tense/mood/agreement markers, auxiliaries, or determiners. (2) a. d.

amaturus b. amatus c. amatumLatin ‘about to love’ ‘loved’ ‘for loving’ abiit piscatum e. potus ‘went fishing’ ‘which has been drunk’ / ‘which has drunk’

(3) a. cărţi de citit b. cititul cărţilor c. am citit Romanian books of read.Sup read.Sup.the books.Gen have read ‘books for reading’ ‘the reading of books’ ‘I have read’

134 Elena Soare

This particular range of uses is a challenge both for morpheme-based morphology (cf. Aronoff 1994 who argues on the basis of similar data for the lack of content in participial stems) and for feature-based theories of categorization. The framework adopted here, which is a generative, syntactic approach to categorization, allows us to find an answer to both kinds of challenges. 1.2

Categorization and modern grammatical theory

Some attempts have been made in generative grammar to convey the category of participles. In general, they are taken to be derived categories, a view I share in the present paper. However, the details of the derivation often remain to be clarified. In Baker’s (2003) work on lexical categories, participles only get a footnote containing two alternatives (already mentioned above): adjectives or reduced relative clauses (Note 4: 193). In his terms, verbs are defined by the property of projecting a specifier, while nouns are categories that have a referential index. According to his Reference-Predication Constraint (RPC), a category cannot have both properties. Hence, adjectives are given a negative definition, and participles are not addressed (except in the above-mentioned footnote). To accommodate RPC and gerunds/participles, Baker (2005) traces the defective morphology in participles/gerunds to a special kind of (nominal) inflection. These nominal inflectional layers Ger, Ptp, would have the function of combining a higher functional layer (the D or the auxiliary layer) and the bare VP. Doron and Reintges (2011) similarly consider the participial morphology to involve nominal features. But the precise combination of features that gives rise to a particular type of behavior in non-finite domains still has to be worked out. Another idea is that the role of this morphology is just to (re)categorize and introduce an extra specifier into the structure; such a view is proposed by Alexiadou et al. (2013) where the relevant morphemes are specific voice heads, parallel to nominal affixes like -ation. Still, we need to clarify in which precise way the internal make-up of a projection explains the various combinations of properties (e.g. verbal and nominal) exhibited by specific participial forms. Whatever status they would be assigned, the important part of the job – which still remains unachieved – is to determine what restrictions they are subject to, and how exactly they assemble the high and the low domains in the clausal structure. The solution, I believe, is to continue in the direction defined by constructionist approaches, but carefully define the structural configuration behind each type of non-finite domain. Categorization, in constructionist approaches, is also a matter of structure. Word formation starts with uncategorial roots (Marantz 1997 and subsequent work in Distributed Morphology, DM), or Listemes (Borer

Non-categorical categories 135



2005), and categorization takes place through grammatical layers. In this view, which I adopt here, defining participles amounts to defining domains wherein only a part of the grammatical layers are specific to a clause. They do not “lose” anything; they simply have truncated architecture and are categorized by the syntactic structure and the contribution of relevant functional projections. In the view that I propose, a theory of lexical categories is not an interesting goal in itself (contra Baker 2003, 2005). I will rather try to show how the structural make-up of the different kinds of participles accounts for their overall behavior, and make them behave like nouns or verbs. I will propose that their atypical verbal behavior results from the absence of Tense, correlated with the absence of a subject,1 and the specification of other verbal categories, namely Aspect and v. With the adoption of such an approach – wherein careful mapping of functional projections can account for the variation encountered in participial constructions – there is no need in grammatical theory for constraints like Baker’s RPC followed by various sorts of accommodations. 1.3

The goals of this paper

My aim in what follows is to circumscribe the range of structures that underlie participial constructions. More specifically, what is the key property, in the diversity that comes out from the above sketch, that could help us to distinguish a natural class of participles inside the larger class of non-finite verbal forms that includes infinitives and gerunds? In the spirit of much recent work in syntactic approaches to word formation which treat differences between verbal domains in terms of height of structure, I will assume that this common property should be their truncated character: participles are deprived of some verbal layers, and are therefore lower verbal domains. Thus, they are truncated clauses of different heights, and not categories of a special kind. Taking this as a working definition, I will argue, for instance, that a part of Romance infinitivals are participial constructions, i.e. truncated clauses, others being full clauses with defective Tense. Thus, apparently similar morphology can be used to spell out different structures from language to language – defining morphology as the locus of variation. I will show, based on Romance data, with a special focus on the Romanian supine (a particularly difficult construction for classical theories of categorization due to its peculiar combination of verbal and

1. Tense is a mere shortcut here. I admit a decompositional view of TAM (including Agreement). My study here only involves forms that are deprived of the upper T/Agr layers, including personal agreement only, like Portuguese infinitivals.

136 Elena Soare

nominal properties), that the perceived “absence of supines” in Western Romance is a matter of terminology only: Western Romance and Romanian simply do not use the same stem in the participial constructions studied. A more specific goal within this general frame is to find further criteria and clear properties that would contribute to distinguishing a natural class of participles across languages. The core distinction that I assume is between non-­finite (full) clauses involving defective Tense, like infinitives and English Acc-­Gerunds2 – which include a subject layer (either overtly realized or silent) – and participial constructions that do not include a subject layer (even when inflected). I will start by an overview and a classification of non-finite forms that may appear in English and Romance languages. I define participles as truncated clauses, in contrast to infinitives and gerunds, which are full clauses with defective tense (§2). Then I will show in which particular way participial clauses may be truncated, through a structural comparison of structures corresponding to the distribution of the traditional Romanian supine: participial nominalizations, reduced participial relatives, and some participial periphrases. 2. A typology of non-finite forms 2.1

In search for diagnostic properties

Non-finite forms often have nominal origins, that is, they are verbal nouns (i.e. verbal roots with nominal declension) which later acquired verbal properties. Diachronically, we can see that forms can evolve from nominal to verbal and the other way around: the English infinitive was originally a verbal noun with nominal declension, which developed its Dative declension mark into a T-like element (to).3 Putting aside diachronic matters, non-finite forms in modern English and Romance languages are a link between the verbal and the nominal domain. They may function in combination with a determiner (English nominal gerunds), or with auxiliaries (the English progressive); it is common in grammars to talk about their “double” (or triple – nominal/verbal/adjectival) nature. However, simply admitting the existence of defective or ‘more nominal’ forms in the verbal paradigm is not enough. Rather, one could consider the whole range of non-finite forms as being ‘participial’, inasmuch they are somewhat verbal, although not fully so. To

2. Gerunds that assign Accusative case to their complements, like e.g. Pat watching television. See below §2.2. 3. Similar tendencies can be seen in the evolution from Latin to early Romance languages.



Non-categorical categories 137

determine if terms like participles can define a natural class in grammar, we must look for some common properties and criteria. One candidate for a common, diagnostic property in participles (present or past-passive, etc.), is the absence of morphological tense and person agreement that would license overt, case-marked subjects. Another property is their open structure – in order to form a complete domain, they need to combine with a functional layer (e.g., a determiner, a light verb); otherwise they attach as modifiers to a maximal projection, through a relativization mechanism which includes some operator layers. In view of this, one may also define participles as inherently modifier constructions – without any added material (no Complementizer layer), they can be adjoined to another domain. This defines a special kind of predication which does not make use of overt, typically clausal, functional material. One might suspect that these three properties are related in some specific way. The present study aims at finding elements for a theory of participles that makes this connection more precise. 2.2

Un-tensed clauses with subjects vs. reduced participial clauses

In this section, I will sketch a classification of non-finite clauses. I will distinguish complete clauses with internal subjects from reduced clauses that lack an internal subject (the subject is added in an outer domain). I propose that this reduced character is the core property defining participials across languages. English has two types of non-finite clauses: gerund clauses (“Acc-ing” gerunds, cf. Abney 1987) and infinitival clauses. Traditional English grammars (cf. Huddleston & Pullum 2002), reserve the term gerund to verbal/clausal uses of the -ing form, and the term (present) participle to its adjectival uses (and use “gerund-­participle” in reference to the form itself). Of course, what is meant by verbal or adjectival here is typical distribution associated with verbs or adjectives; the -ing form retains verbal properties in adjectival contexts (for instance, it can have complements, as we will see in a moment). They are both non-finite clausal domains. The -ing forms are non-finite TPs [DP/t V-ing] which take a position inside another DP or clausal domain. In the first case, they are reduced relatives with subject gaps (4a); in the second case, they contain a non-nominative subject (4b) and can be either adjoined to a sentence, or in a complement position (4c–d). Acc-­ing Gerunds can include auxiliary layers (4c) and license Accusative objects ((4c–d), (5a–b)).

138 Elena Soare

(4) a. a jumping cow / a cow jumping in the fields b. God/Him willing, we will succeed c. I am aware of him having made such a good point. d. She argued against buying a television set. (5) a. a man singing the Marseillaise b. (With) him singing the Marseillaise, we can’t hear each other.

However, this first (essentially distributional) distinction that can be drawn inside verbal gerunds in English correlates with an important difference in terms of structure. In (4a), the gerund is predicated of a DP to which it is adjoined, and the internal mechanism of this predication is supposedly comparable to that of a subject-­gap relative clause. In (4b), the gerundial clause licenses a subject and is a complete clause. The first type of construction is similar to another participial relative based on the past/passive participle: a wounded man. The -ing construction is a subject-gap reduced relative, while the -en/ed construction is an object-­gap reduced relative (present vs. past participle in the traditional grammar). This distribution is of course found in Romance, for instance, cf. French (6)– (7). Examples (6a–b) illustrate a present-participial, subject-gap reduced relative, while in (6c) we have a ‘gerund’ clause expressing Cause. In (7), we have reduced past-­participial clauses.4 (6) a. un homme vivant en France ‘a man living in France’ b. une femme vivant en France5 ‘a women living in France’ c. Vivant en France, elle connaît toute la fine gastronomie ‘Living in France, she knows all the fine gastronomy’ (7) un homme blessé, une femme mariée… ‘a wounded man, a married woman’

The same distinctions hold in Romanian (8)–(10). (8) o femeie trăind singură  ‘a woman living alone’

(Ro.)

4. In fact, such constructions are discussed in the literature as ambiguous between a ‘verbal’ participial clause and an adjectival participle, according to their more reduced or more complex structure (cf. e.g. Embick 2004; Sleeman 2011). 5. Romance Gender marking provides another test for distinguishing reduced participial relatives from Adjectives built on the present participle stem (cf. une femme vivante ‘a living.F.sg women’ / *une femme vivante en France ‘a women living.F.sg in France’).

Non-categorical categories 139



(9) trăind singură, are adesea dificultăți living alone has often difficulties ‘living alone, she often has difficulties’ (10) o femeie căsătorită a women married ‘a married women’

We note, in addition, that Romance past participles can also appear in an absolute construction, adjoined to the main clause: (11) a. une fois mariés, ils ont vécu heureux pour toujours one time married, they have lived happy for ever ‘Once married, they lived happily everafter’ b. o dată căsătorită, s-a mutat one time married se-has moved ‘once married, she moved out’ c. cognosciuta Maria, Gianni è cambiato del tutto known Maria Gianni is changed completely ‘having known Maria, Gianni changed completely’

(Fr.)

The adjunct gerundial clause is a non-finite TP, whose special semantics allows it to express for instance an event cause, goal, etc.; there is a constraint on the non-­ finite tense that has to be correlated with the one of the main event. One may call it a gerund for morphological reasons (see Huddleston & Pullum’s 2002 term of “gerund-participle”), but its essential property is that it is a full clause containing a subject. The participial reduced relative construction, which I will address in §3, is an incomplete sentence, containing a gap, which in this case (unlike in full relative clauses) is situated either in the subject position, or in the object position. A similar distinction can be drawn among infinitives. These are non-finite TPs taking argument positions (including the complement position of prepositions) – subordinate clauses, or adjunct positions (infinitival relatives, prepositional adjuncts, e.g. in order to + infinitive). TP projections may or may not include an overt subject, but always have a specified subject (an empty subject PRO or an overt Accusative-marked full DP). Infinitival relatives can be either subject or object-gap; some of them have been argued to represent “reduced” relative clauses in English (see Bhatt 1999; Iatridou et al. 2000). Romance languages generally have infinitives in argument positions (e.g. complement of modal or aspectual verbs), or in a relative position (livre à lire ‘book to read’). As we will see, the latter are in fact (participial) reduced relatives (cf. Giurgea & Soare 2010) with modal readings and passive structure – they are

140 Elena Soare

always object-gap and contain no subject. In turn, Romanian has lost infinitival complement clauses (with one, but still marked as archaic exception, of the bare infinitive in the complement position of can). Crucially for this study on participles, Romanian has a participial stem that is used not only in contexts traditionally called “past-passive participle”, but also in positions that the infinitive can take in other languages (non-finite relatives, periphrastic constructions). A participial form is thus used in contexts where a reduced non-finite form is needed. This wider use of a participial stem brings interesting evidence in favor of a syntactic approach to categorization. Participial constructions, disregarding their morphology, are reduced structures in the sense that they always lack a part of the full clausal spin; all of them lack full Tense marking and person marking; some lack a subject (and consequently have an outer subject, possibly introduced by a Predication (Pred) head); others lack Aspectual Voice layers. And in the end, the participle can be only a root; in this case one may derive any category by adding either overt, or covert categorizing layers. For instance, on a bare present-­ participle stem like marrant ‘amusing’ in French, one can derive an adjective marrant ‘funny’ which will have agreement markers as any adjective. The picture that emerges from this overview is one in which non-finite morphology covers different areas of the clausal spin across languages; infinitives can have participial (i.e., reduced) uses and the other way around. Each language has a particular range of non-finite morphology, and makes a particular use of it; that is to say, morphology makes the languages look different. 2.3

Making the definition work

Let us see now how this working definition of the participial, as a domain which only has an outer subject, applies in the different participial constructions. First, it is trivially fulfilled by nominalized participial forms, where if a subject is licensed, it has to appear in the DP domain. These are sentential nominalizations (‘big’ nominalizations that include much of the clausal spin, most importantly grammatical layers like outer Asp), but are unable to license an internal subject.6 English Poss-Gerunds like Mary’s watching television are instances of such “highest” nominalizations, which involve nominalized non-finite TPs with a Poss 6. Note that I do not include here derived nominals like -ation, which have a nominalizer, but only take into account nominalized structures based on clausal layers (“converted” from participles in the absence of an n-layer). This property leads, I believe, an important distinction among nominalizations: the absence of an n-layer is responsible for the ability to include a high clausal structure. Once a nominalizer is present, the inclusion of non-nominal grammatical layers is no longer possible. See Soare (2013, to appear).

Non-categorical categories 141



subject. They are mixtures of verbal properties and nominal properties in that they take a nominal distribution having argument positions, but accept auxiliaries and license Acc objects; one may say it is the highest nominalization, i.e. it contains a high range of verbal properties. Nevertheless, they are still DPs.7 Nominalized infinitivals in argument positions involving implicit arguments and/or genitive subjects (e.g. nominalized Italian infinitives like il (suo) mormorare parole dolci ‘the/his whisper words sweet / whispering sweet words’) are similar to Poss-ing by their large structure which includes internal arguments marked by Accusative case. Another structurally “big”, near-clausal nominalization is the Romanian Supine Nominal, cf. §3.2 below. Both English and Romance have so-called “Past” participles, which are tenseless, lower AspP (auxiliary-less) domains, Voice layers and vPs with different flavors. Active voice and outer Aspect only appear as a contribution of higher functional domains. One crucial property that distinguishes participles from full verbs (meaning, a verb in a fully tensed clause) is the relationship to the external argument of the base verb. In some participial constructions, the external argument is not projected in the expected position; hence the absence of a specifier can be a defining property of a participle. In Baker’s (2003) terms, such a participle is not a verb (see his Reference-Predication Constraint, RPC), but rather a derived category (Baker 2005), a view that I partially share in this study.8 Other participles may project the external argument but do not assign it the prototypical subject case. This is a property that has to be related to the combination of specific inflectional or clausal layers and not to some lexical property (or alternatively to a change in argument structure). To illustrate, take ‘past/passive’ participles across languages. In a passive structure, for instance, one standardly assumes that the participle does not realize the external argument in the canonical Specifier position, which is not a thematic position anymore. (12) a. John *(has) eaten an apple b. The apple is eaten by John c. An apple eaten by John

Accordingly, participles in perfect and passive constructions might be considered as reduced as the lower VP layer. The size of the associated structure for the participle is bigger in other languages (and in connection with other properties

7. The -ing of Gerunds is not included in this study, as they are fully nominal structures, and the -ing affix is a nominalizer. See Borer (2005), Alexiadou et al. (2010) for details. 8. Although I do not have a definition of lexical categories, nor do I endorse Baker’s RPC, a constraint that would be redundant in the framework I am building, cf. next page.

142 Elena Soare

like auxiliary selection), e.g. in Bulgarian which also allows subject participial relatives (cf. Iatridou et al. 2009). In contrast, ‘Poss’-Gerunds do include an external argument position, but do not case-mark the subject DP in the canonical way – hence the subject is licensed in the nominal layer by a Possessive Determiner. Again, the subject is not in its canonical clausal position. (13) John’s eating an apple (surprised everyone).

The truncated character of participials may thus be seen as one key property distinguishing participials from other types of (non-finite) clauses. They are more or less reduced structures. I assume that a (phrasal) participial may involve one of the three structures in (14); the presence of the external argument is accounted for by the inclusion of a classical specifier-projecting vP. My analysis differs from Baker’s (2005) in that the projection of a specifier does NOT amount to a lexical category (V as opposed to N); rather, specific functional layers (types of v) are responsible for categorization. (14) a.

Ptp=AspP Asp -en/-ing

vP v¢

DP v

VP V

b.

…

PtpP=vP v -en

VP V

c.

…

PtpP=VP V -en

…

Depending on the content/type of the categories involved (more precisely, Asp and v), these structures allow for active, passive, agentive, causative, perfective or imperfective, and stative participles. It is clear that some forms are ‘bigger’ (e.g. -ing in English, infinitives in Romance) and that some forms are ‘smaller’ (past

Non-categorical categories 143



participles in English and Western Romance), but in Romanian the equivalents of -en forms are underspecified and can be big or small – this is a matter of variation in the functional vocabulary characterizing each language. Yet, this is only a very general frame. First, the Ptp layer is just a shortcut to accommodate morphological marking, but does not have content, nor does it introduce the event, or categorial information. Nothing in my analysis hinges on the identity of this layer, and as such, it can be simply omitted. In contrast, the semantic features attached to the layers above (Asp, v) are important; they are spelled out by specific morphology (-ing, -ant, -en, -a/i/ut, -é, -er) according to the range of affixal forms existing in the language. Participial morphology is simply the expression of different truncated structures. Second, these structures appear in different sentence positions, of course, and are selected by auxiliaries, Determiners, or Pred heads. Agreement may be involved, depending on the structure that embeds the participle, in a passive or an adnominal structure for instance – cf. Fr. le livre / la lettre est lu(e) ‘the book / letter is read’, le livre / la lettre lu(e) ‘the read book/letter’. When the participle has a more clausal structure, the presence of upper clausal layers, as in the Romanian supine (cf. below §3.1), blocks agreement. Another difference among non-finite domains is passivization and the presence of auxiliaries. Both are available in the English verbal gerunds, as well as in the infinitive, but neither is available in past participles. This is also accountable in terms of the structure involved, which, in the spirit of (14), may be more or less complex. (15) a. Having received the book… b. The book being received… (16) *the had received book, *the been received book

The truncated status is not the only characteristic of participles. To distinguish gerund modifiers from infinitival modifiers, for example, one would also have to consider the former as having an inherent modifier status, that could be put in terms of selection by a silent Pred head, allowing these constructions to be plugged into a modifier position of another domain, e.g. the clause, a DP. Infinitivals do not include this layer and hence they require extra overt material (e.g. a non-finite complementizer, a preposition, a particle) to be inserted into such positions. Pred can be considered as internal or external to the participle, but if it is considered external, participles should include a property that allows them to be selected by a Pred head, whereas infinitives should lack this property. Note finally that the structure in (14) is not automatically attributed to all segments that would include -en or -ing endings. I assume an idea standard in various syntactic approaches, according to which affixation (participial or nominal)

144 Elena Soare

can take place above a complex structure (as illustrated in (14)) or with a simple root (see Embick 2004; Alexiadou et al. 2013 among many others). Therefore, the criteria that emerge at the end of this section for defining participial constructions are: (i) their truncated structure, (ii) the absence of an internal subject (but the possibility to add an external subject by a specific outer layer, for instance, in a Predication projection PredP), (iii) the incompleteness of the construction (a participle is not a full clause but must contain a gap, either in an argument or in a variable position). With this general background, we can now approach our case study, the Romanian “supine”. 3. Underspecified Participles in Romanian 3.1

The data

In Romanian, the participial stem can be used in a wide range of syntactic contexts, as illustrated in examples (17)–(24) below: (i) as a perfect-passive participle, in perfect and passive constructions with have and be respectively (17)–(18), (ii)  in nominal contexts with a determiner (19a–b), (iii) in adjectival position (20), also giving rise to nominalization (21), (iv) with a functional particle de in verbal periphrases (22) and adnominally as a reduced relative (23), and finally, (v) with various lexical prepositions marking goal or source in particular (24). (17) am citit ‘(I) have read’, am mâncat ‘(I) have eaten’, am alergat ‘(I) have ran’ (18) poemul este citit ‘the poem is read’, mărul este mancat ‘the apple is eaten’ (19) a. cititul poemului ‘the reading of the poem’ b. mâncatul merelor ‘the eating of the apples’ (20) un copil iubit ‘a loved child’, un măr mâncat ‘an eaten apple’ (21) iubitul meu ‘my (be)loved = my love’ (22) am de citit ‘I have to read = I’ve got reading to do’ (23) carte de citit ‘book to read’ (24) a merge la pescuit ‘go fishing’

Note again that I do not attribute any semantics to the participial stem itself; however, I think we need a theory that accounts for all these uses and for the apparent homonymy as well. Traditional grammar groups together ‘past participle’ uses (verbal in (17), (18) and nominalized in (20)–(21)) and ‘supine’ uses (nominal in (19) and verbal in (22)–(24)). This wide use of a participial stem cuts across both

Non-categorical categories 145



the nominal/verbal distinction and the verbal/adjectival distinction. In the rest of this paper, I will concentrate on the first distinction in an attempt to isolate the source of nominal vs. verbal behavior. I will show that the three structural patterns in (14) enable us to propose an account for both the nominal and the verbal contexts in which these forms appear. 3.2

The Romanian nominal supine and “canonical” verbal properties

I start with a description of nominal supine structures and show that they involve a large part of the verbal/clausal spin, as they exhibit canonical verbal properties such as grammatical aspect and agentive Voice. The supine nominal, illustrated above in (19) and below in (25), cumulates some peculiar properties. It can only have eventive readings (cf. Cornilescu 2001; Iordăchioaia & Soare 2008 among others), unlike other nominals across languages, that are ambiguous between eventive and referential readings (cf. the building of the cathedral during a century vs. an impressive building). From this point of view, it can be considered a more verbal than ‘perfect’ nominalization in Vendler’s sense (hence the difference with -ation nominals or the nominalized infinitive in Romanian). (25) a. cântatul acestor cântece (mi-a trezit sentimente patriotice) sing.Sup.the these.Gen songs.Gen (me-has stirred feelings patriotic) ‘singing these songs (stirred me patriotic feelings)’ b. cântatul lui Ion la micul dejun (mi-a dat dureri de cap) sing.Sup.the Gen Ion at breakfast (me-has given headache) ‘John’s singing at breakfast (gave me headache)

A hallmark of this imperfect nominal status is the realization of the subject, which does not affect the eventive meaning, a property which distinguishes the supine nominal (26a) from the nominalized infinitive (26b): (26) a. cântatul lui Ion timp de ore în șir pentru a-și enerva sing.Sup.the Gen Ion time of hours in row for to-his irritate soacra mother-in-law.the ‘John’s singing for hours in order to irritate his mother in law’ b. cântarea Mariei *pentru a-și enerva soacra/ sing.Inf.the Maria.Gen for to-her irritate mother-in-law.the/ a fost apreciată de toți has been appreciated by all

146 Elena Soare

The supine allows resultant state modification (as shown in (27), cf. Iordăchioaia 2008) and verifies tests for the presence of a Voice level, i.e. agent modifiers in (28): (27) umflatul cauciucurilor de către tata (cu pompa) pump-up-Sup-the tires-Gen by father (with the pump) ‘the pumping up of the tires (with the pump) by the father’ (28) a. demolatul intenţionat al clădirilor vechi demolish.Sup.the intentionally of buildings old ‘intentionally demolishing old buildings’ b. condusul maşinii beat drive.Sup.the car.Gen drunk ‘driving the car drunk’

It does not allow impersonals (29a) or pure inchoatives (29b), and has no reflexive reading (29c): (29) a.  *plouatul raining.Sup.the ‘the raining’ b.  *ruginitul fierului, *putrezitul frunzelor rust.Sup.the iron.Gen,  rot.Sup.the leaves.Gen ‘the rusting of the iron, the rotting of the leaves’ c. anunţatul oaspeţilor (#the guests announced themselves) announce.Sup.the guests.Gen ‘the announcing of guests’

The Romanian supine nominalization conveys imperfective Aspect (cf. Iordăchioaia & Soare 2009; Alexiadou, Iordăchioaia & Soare 2010). The supine maps [+bounded] events into [–bounded] events, and thus introduces an aspect shift, which indicates the presence of an outer Aspect layer. (30) a. Sositul lui Ion cu întârziere timp de 2 ani i-a arrive.Sup.the John.Gen with delay time of 2 years him-has adus concedierea. brought firing ‘John’s arriving late for two years brought about his being fired.’ (31) b. Muncitul lui Ion *(până la miezul nopţii) o îngrijorează work.Sup.the John.Gen until at middle night.Gen her worries pe soţia lui. Acc wife his ‘John’s (habit of) working till midnight worries his wife.’



Non-categorical categories 147

The fact that both inner and outer-aspectual modifiers are allowed attests to the presence of an aspectual shift indicating an outer aspect projection. (32) Traversatul râului de către Ion în cinci minute timp de doi ani cross.Sup.D river.Gen by part John in five minutes time of two years a surprins pe toți has surprised of all ‘John’s crossing the river in five minutes for two years surprised everybody’

In light of these and much more data, Iordăchioaia & Soare (2008, 2009), Alexiadou, Iordăchioaia & Soare (2010) proposed that the semantics of the supine contains an aspectual pluractional operator (PO) which correlates with its [–b] property, the aspect shift it introduces, and the semantic effects it triggers, standardly associated with POs. The Romanian supine is odd with one-time events, but becomes fine when the cardinality of a plural object matches with the event plurality introduced by the supine, as shown in (33). Other indications in favor of pluractional semantics are frequency adjuncts like ‘at once’, impossible with the supine (see Laca 2006 for Spanish andar periphrases), cf. (34); modifiers like ‘when’-­clauses are allowed by the supine (35). For more details and a complete analysis, see Iordăchioaia & Soare (2009), Alexiadou et al. (2010). (33) a.  *ucisul jurnalistului/ unui jurnalist kill-Sup-the journalist-the-Gen/ a-Gen journalist b. ucisul jurnaliştilor kill-Sup-the journalists-the-Gen ‘the killing of journalists’ (34) ??cititul unui roman de 300 de pagini dintr-o dată read-Sup-the a novel-Gen of 300 pages at once (35) fumatul lui Ion ori de câte ori iese în pauză smoke-Sup-the John-Gen whenever goes.out in break ‘John’s smoking whenever he takes a break’

As a conclusion, one can say that the Romanian supine nominal has verbal properties, that it includes an imperfective AspP and an agentive vP projection, embedded under a D head that conveys the nominal status. Its lack of nominal properties (Gender, Number and adjectival modification, cf. Iordăchioaia & Soare 2008, 2009) confirms this analysis. The supine nominal is a nominalization of a (+active, +aspectually marked) participial clause; it includes the whole structure in (14a) but obviously no Tense layer. However, the pluractional properties make

148 Elena Soare

it more verbal and comparable to the mixed Poss-Gerund:9 hence, the Asp layer involved materializes grammatical aspect and not Aktionsart. The nominal supine combines the property of projecting the external argument (that is assigned case in the DP projection) and the property of it being imperfective. In line with recent work, a structure like (14a) can, I assume, be elaborated to account for the verbal properties of the supine nominal, producing a result as in (36). The participial phrase contains a specifier-projecting vP and an (imperfective) outer aspectual node that hosts but does not case-mark the external argument, which is therefore assigned Genitive by D. (36)

DP D -l

AspP DP

Asp′ Asp -t…

vP v¢

DP v

VP V citi

DP unui roman

In this representation, D hosts the definite article -l (to which the structure above has to raise to check nominal features; the article thus incorporates into the participial form by head movement and gives rise to the form cititul unui roman ‘the reading of a novel’). The participial morphology instantiates an Asp head, which is a grammatical layer realizing an imperfective value. The vP layer allows the projection of an agentive external argument, which under certain conditions can realize as a Genitive phrase in the higher structure of the DP projection. As a consequence, the supine nominalized construction is an active and imperfective verbal domain.

9. See Alexiadou et al. (2010) for a more extensive comparison: both the supine and the Possing Gerund are characterized by (imperfective) aspect, the projection of a subject (but licensed by D in Romanian, hence Genitive), the compatibility with adverbial modifiers; they differ, however, in the case marking of arguments and the availability of an auxiliary.



Non-categorical categories 149

This type of structure directly captures the fact that the supine may lexicalize the external argument of the verb base. Note that the supine projection (AspP, for concreteness) does not contain a tense layer, and hence cannot license its subject. The external argument is projected inside the supine but has to move out to the DP layer to be assigned case.10 Let me stress again that theories like Baker’s, in which participles are nominal or adjectival (or have nominal inflection preventing them for projecting a specifier), are in difficulty when faced with such data. First, this type of derived nominal must involve a specifier-projecting category (v, for instance), because it allows for the projection of an external argument. If we would want to say that it is contributed by the D-layer, we would miss its argumental status and would have to say that it is introduced like a by-phrase, which would be impossible in the case of unergatives. Second, this kind of approach also has to deal with the “nominal” inflection and the “adjectival” distribution and the combinations thereof. Baker’s ‘elsewhere’ theory of participles does not provide a way of accounting for the different combinations of mixed features in the participial structure. I assume that constructionist approaches to word formation provide a better account for the variety of uses to which a participial stem can correspond. In particular, they can account for the fact that the Romanian participial stem can spell-­out a near-clausal structure with imperfective aspect and an active reading.11 Hence, Romanian shares with Slavic, but also with Latin, the property of having an imperfective and active participle. However, as we will see in a moment, this is only true for the nominal environment.12 10. With unergative and transitive bases, this is straightforward. With unaccusative bases (and achievement predicates), it imposes the iterative-habitual reading but also the unergativization of the VP. I assume that the two phenomena are somehow related (cf. Guéron & Hoekstra 1989). 11. It is assumed here that the active reading is the result of an outer AspP selecting a specifierprojecting vP and an active Voice (that we might add to the structure on top of vP). A similar intuition is developed in Alexiadou et al. (2013), according to which Voice contains active/ passive morphology and alternates with -ation nominal Voice; there, Voice also participates in categorization. In a certain way, I develop a similar intuition by assuming that -ation and the participial construction with different Voice flavors alternate in the structure of derived nominals. In Doron (2003) the property of turning the external argument into an actor is actually the result of a Voice head, in the intensive template in the Hebrew verbal system. The supine may be reminiscent of this intensive pattern, but since it provides positive evidence for the presence of a pluractional operator, I take it to contribute an aspectual head instead. 12. This has to be related to (i) the well-known regression of infinitive in Romanian and (ii) the specific interaction with the Determiner system, which call for detailed study in future research.

150 Elena Soare

3.3

Truncated participial clauses and less “canonical” verbal properties

Let us now turn to the so-called verbal uses of the participial stem. I will address two of these here, namely, participial modifiers (reduced relatives) and verbal periphrases. These constructions have been categorized as ‘verbal’ by traditional grammarians, in view of the fact that the supine appears as determinerless. But, as shown below, their verbal properties are seriously defective. 3.3.1 Participial modifiers, Pred, and modal properties One of the shared properties of participles is that they can inherently be modifier structures; this was already pointed out in §2 above. Without apparent addition, they directly modify DPs or clauses, forming a natural class by being designated adjuncts. As suggested above, this is realized by the presence of a gap in the structure, which is correlated with the modified nominal by means of a Pred projection. (37) a fallen leaf, a shining star (38) un livre à lire ‘a book to read’ 

(French)

(39) o carte de citit ‘a book to read’

(Romanian)

What is the structure of these participles?13 A standard account would treat them as a special type of (non-finite) relatives. The question that I raise in this section concerns the property that designates participles to enter modification structures. I will assume that this property is again their truncated character: something in the structure is absent, and as one of the arguments remains unlicensed, it creates an open structure which can be plugged, into a DP-modifier position, for instance. Let me briefly examine some facts related to the first case. Reduced participial relatives have been examined in recent work in generative grammar. In Giurgea & Soare (2010), they are analyzed in terms of a Pred head; relativization is thus not achieved through a wh C element like in full tensed relatives, but through a Pred head that merges with a nominal specifier and acquires a nominal status, as in Bhatt’s (1999) analysis of relative clauses.

13. In the French construction un livre à lire and their other Romance counterparts, I assume that we also deal with a participial construction. This is not so surprising, given the similar properties of the supine and the à-Inf construction, both unable to undergo passivization and obligatorily object-gap. It is a mere accident (resulting from morpho-lexical variation) that in some languages the form is infinitive, and in others it is a participle or a gerund. This, however, does not imply that the structure is strictly the same, as suggested by the availability of negation only in some cases and not in others (cf. (44b) below, ungrammatical with the supine in Romanian, but available with the infinitive in French).



.

Non-categorical categories 151

(40) a. Cărţile de citit sunt pe masă (Ro.) books-the of read.sup are on table b. I libri da leggere sono sul tavolo(It.) the books to read.Inf are on-the table c. Les livres à lire sont sur la table(Fr.) the books to read.Inf are on-the table

In Romance, these participial relatives exhibit two properties: (a) they always induce a modal-deontic reading, and (b) they are always object-gap (and never subject-­gap). The modal reading is identified by modifiers like ‘obligatorily, at any costs’. (41) a. livres à lire à tout prix(Fr.) b. cărţi de citit neapărat(Ro.) books to read.sup obligatorily

A related approach is the teleological interpretation, which can be derived in the same way. In their analysis, Giurgea & Soare (2010) also observe the unavailability of an unaccusative construction for this type of relatives, as well as of the passive periphrastic construction (the latter being available in English). (42) a.  *flori de ofilit(Ro.) flowers to wilt.sup b. ploaia spală maşina -/-> *maşină de spălat de către ploaie rain-the washes car-the car to wash.sup by part rain ‘the rain washes the car – car to wash by the rain’ (43) a. a book to be read b. *un livre à être lu c. *o carte de a fi citită

These properties of Romance participial relatives have brought Giurgea & Soare (2010) to treat them as reduced clauses, in which the participle itself does not involve voice layers that would be responsible for the projection of an external argument. Instead, the functional particle de, analyzed as a modal inflection, contributes both the modal reading and the external argument, which is thus not brought by the lower verb, that contains only a passive vP. The same analysis has been assumed by Giurgea & Soare (2010) for Tough constructions, which have similar properties (they are passive clauses with a modal meaning) and also involve, in Romanian, a supine-based construction.

152 Elena Soare

The supine used in these constructions never shows up with a subject and never allows the sentential negation nu (but only the constituent negation ne14), nor any clitic material,15 thus providing further support for a reduced structure analysis of the supine domain. (44) a. o carte pe care să nu o citești a book that which subj not it read.2Sg ‘a book that you don’t (have to) read’ b.  *o carte de nu citit a book of not read.sup c.   *o carte de o citit a book of it read.sup

This brings us to an interesting observation about Romanian supine constructions. So far, we saw that when appearing in a nominal environment, the supine still projects many of the typically verbal properties, i.e. grammatical Aspect and layers responsible for the projection of an external argument, which are the most clausal-­like properties of a verb. Now, in contexts where in traditional grammar supine constructions are considered as ‘verbal’, we see that the verbal properties are more defective than in the nominal contexts, as the aspectual layer and the specifier-projecting vP are absent. This, of course, supports the view that (i) it is possible to capture the gradience of verbal and nominal properties by means of specific functional projections, (ii) categorization is the result of such projections, and (iii) a wide range of mixed combinations of properties is allowed. As postulated in the introduction of this paper, it is possible to derive these participial constructions, if one takes the last of the derivations in (14) as a starting point. We must assume a reduced structure in which the supine contributes only a VP with an externalized internal argument (an object gap coindexed with the modified nominal). AspP and vP are not present, since no grammatical aspectual value is contributed, and there is no external argument projected. In the simplified structure in (45), the VP-layer is embedded under a defective Mood inflection, which is only contributing the deontic reading (cf. Giurgea & Soare 2010), and corresponds to the introductory element de. One can clearly see in

14. The position of negation is a further indication of the absence of Asp, see Iordăchioaia & Soare (2013). 15. The availability of negation in Western Romance distinguishes the Romanian supine from the Western-Romance participial relative realized by ‘infinitival’ forms. However, all the other properties being shared, I maintain the view in which they all represent reduced participial structures.

Non-categorical categories 153



(45b) that the structure includes (14c), with the Ptp being in this case the spellout of VP. (45) a. [DP cartei [MoodP de [VP citit ei]] b.

MoodP Mood de

VP V citi-t

…

3.3.2 Periphrastic uses of the supine: Between aspect and modality I will now look at another set of structures where the supine is determinerless, namely verbal periphrases, and show that they are not structurally homogeneous. The supine that introduces the lower part of the periphrastic constructions exhibits different mixtures of nominal and verbal properties, and correspondingly has to be assigned different structures. Correlatively, the main verb varies from a light verb to a lexical verb. The light verb construction presumably involves the formation of a complex predicate, while in the lexical verb construction, the supine acquires more nominal properties. In fact, we deal here with different kinds and degrees of complex predications. From an interpretive point of view, they express modality, aspect and goal of motion. From a syntactic point of view, they illustrate different degrees of cohesion. In all the constructions studied, the supine is incompatible with D (i.e., according to traditional grammars, it has verbal properties), and appears in the complement position of a light verb such as avea ‘have’, fi ‘be’, lua ‘take’, începe ‘begin’, termina ‘stop’. The supine provides the “lexical” material in the formation of a complex predicate. In my view, this means it provides a part of the argument structure that comes with the functional layers responsible for it. It is obligatorily introduced either by a lexical P or by an invariable functional particle de;16 this is traditionally considered as the diagnostic for the supine, which is otherwise homophonous with the past participle (see (3) and (17)–(24) above). Like in the case of reduced relatives, the supine in these constructions never shows up with a subject, never allows the sentential negation nu (but only the

16. As already shown above (§3.1 and 3.3.1.), the same particle de introduces reduced relatives, and has been analyzed as introducing modal meaning and an external argument (Giurgea & Soare 2010).

154 Elena Soare

constituent negation ne17), nor any clitic material. This is already a clear indication of its reduced structure. However, the various constructions differ in several respects, namely: (i) the choice of the particle, (ii) the alternation with nominal constructions, and (iii) the semantic and syntactic properties of the verbal supine. Class 1° modal periphrases This class includes constructions with a avea ‘to have’, a da ‘give’, a fi ‘to be’, a rămâne ‘to remain’ + de + Supine. (46) Am de cules porumb(ul) have of harvest.sup maize-the ‘I have to harvest the maize’18 (47) Mi-a dat de citit me-has given of read ‘He/she has given me something to read’ (48) (Mai) e de cules porumbul still is of harvest corn-the ‘It remains to harvest the corn’ (49) A rămas de cules porumbul (internal argument promoted as subject) has remained of harvest corn-the ‘It remains to harvest the corn’

The whole construction has a modal deontic flavor. Clitic climbing appears (50), as opposed to one of the futural periphrases in Romanian (51), involving a subjunctive form in the lower part; therefore restructuring applies and we have a monoclausal construction (a complex predicate formation). (50) a. îl am de cules vs. b. *am de îl cules it have of harvest have of it harvest ‘I have to harvest it’ (51) a. *îl am să culeg vs. b. am să-l culeg it have subj harvest have subj-it harvest ‘I will harvest it’ 17. The position of the negation is a further indication of the absence of Asp, see Iordăchioaia & Soare (2013). 18. The supine here is again translated with an infinitive in English, as it is in Western Romance. For these constructions, I do not assume a parallel between Western Romance infinitive and the Romanian supine. I only show here, on the basis of the structural cohesion of the periphrase, that it is based on a reduced participial structure (in line with Avram 1994) – a small clause. I have no commitment to an analysis of the corresponding Romance periphrases.



Non-categorical categories 155

I do not consider (in line with Kayne to appear, and unlike Huddleston & Pullum 2002) that the modal reading is to be put on the main verb. Rather, I consider the modal reading to be contributed by the supine part, more precisely by the de particle, which I take to be a Modal Infl (in line with previous work on reduced relatives, Giurgea & Soare 2010). The supine structure is reduced and does not project the subject (no vP, no external argument, no specifier), which is contributed from outside – an outer specifier, through the Modal Infl. In this sense, the proper supine structure remains faithful to the definition of the participial domain as a truncated structure without an inner subject. The construction lacks aspectual properties: it does not undergo habitual or iterative meaning, and the only aspect present is the Aktionsart of the base verb (whether telic or atelic). In this construction, the supine can combine with all aspectual classes of verbs, and does not impose an aspectual shift: in (52), the supine construction preserves the Aktionsart of the lexical base; the periphrasis adds only the modal-deontic meaning. (52) Am de găsit soluția, de săpat grădina, de cântat, de umblat etc. I have to find the solution, to dig the garden, to sing, to walk…

The outer-aspect modifier for x time (in (53)) does not modify the embedded supine but the whole sentence. In this last case it takes scope over the deontic operator (it is the obligation that holds for three years, and not the crossing of the river). On the other hand, the inner-aspectual modifier in x time does modify the supine, and denotes that the action of crossing the river has to be completed in three minutes. The same is true for the PPs modifying the supine in (54) and (55). (53) Am de traversat râul înot în trei minute/ have to cross river.the swimming in three minutes/ ??timp de trei ani time of three years ‘I have to cross the river by swimming in three minutes/for three years’ (54) Am de găsit soluția în cel puțin două ore have to find solution.the in at least two hours ‘I have to find the solution in at least two hours’ (55) Am de sărit într-un picior timp de trei minute have to jump on one foot time of three minutes ‘I have to jump on one foot for three minutes’

The pluractional operator in the supine nominal is responsible for licensing a dependent indefinite, câte ‘one-each’ (Iordăchioaia & Soare 2011).

156 Elena Soare

(56) Cititul câte unei cărţi îl relaxează pe Ion read.Sup.the one-each a.Gen book him relaxes Acc John ‘Reading a book at times relaxes John’

In verbal periphrases, it is the main verb (when habitual) and not the supine that licenses câte: (57) a. Ion are (mereu/tot) de citit câte o carte John has (always/at times) of read each a book ‘John has always a book to read’ b. Ion are azi de citit *câte o carte John has today of read each a book

I conclude that the supine does not contain an outer Asp layer in this class of constructions, and also lacks a Spec vP, thus not having an external argument layer, much in the same way as the reduced participial relatives illustrated in the previous section. The structure of the two domains is here assumed to be identical; they are only selected by different main categories – a DP in participial relatives and an auxiliary verb in the modal periphrase. (58)

… MoodP Mood de

VP V citi-t

…

Class 2° aspectual periphrases This second semantic class of supine periphrases includes main verbs such as a termina ‘finish’, a se apuca (de) ‘start (of)’, a se sătura (de) ‘have enough of ’, a se opri (din) ‘stop (from)’, a se pune (pe) ‘begin (on)’. In this case, the perfective or inceptive aspect comes from the main verb, and the supine does not have an aspectual contribution in itself. I assume that in most of the cases (except for a termina ‘finish’ which has a more cohesive construction and formally patterns with periphrases from the Class 1°) the introductory P is subcategorized by the main verb and counts as an edge for the supine VP. The hypothesis of a null D in the structure is fully plausible, given the possible presence of a full determiner (which can be weak DPs). (59) m-am săturat de carte / de citit / ??de rămas la Ion peste noapte me-has sick of book / of reading / of staying at John’s over night ‘I’ve had enough of the book / of reading / of staying at John’s over the night’

Non-categorical categories 157



(60) m-am oprit din citit / din fumat me-has stopped from reading / from smoking ‘I have interrupted the reading/ the smoking’ (61) m-am lăsat de fumat me-has leave of smoking ‘I have quit smoking’

In the a termina construction, de is a particle and not a preposition (it is not subcategorized by a termina ‘finish’). This difference between the aspectual periphrases with the supine is accompanied by another, namely that clitic climbing only applies with a termina and not with the other main verbs, (62) a. b. c.

le-am terminat de citit them-have finished to read ‘I finished reading them’  *mi le-am apucat de citit me them-have started to read  *mi le-am oprit din citit, etc. me them-have stopped from reading

In this same class I include another group, which could be taken semantically as a Class 3° goal of motion: a merge, a pleca (la) ‘to go’ + supine; however, from a structural point of view, they behave very much alike. (63) Am plecat la cules porumb(ul) / de porumb have gone at harvest maize-the / of maize

Thus, both so-called aspectual and goal of motion constructions can be grouped under a purpose nominal label. From a structural point of view, then, we finally can distinguish two types of construction, taking clitic placement as a basic criterion. One is tighter, and involves a functional particle introducing modality and an external argument. The second involves a subcategorized preposition and a supine with more nominal properties. The supine does not have the same status in all periphrases: I assume it is more nominal (i.e., has a null D) in the complement position of a lexical preposition, and that it is a bare VP in the ‘tighter’ case, i.e. modal have and aspectual finish periphrases. (64) a. [TP am terminat [MoodP de [VP citit ei]] b. [TP m-am lăsat [PP de [DP ø [NP fumat]]]]

The supine complex predications in Romanian exhibit in their “lexical” part, instantiated by the supine, several degrees between nominal and verbal participles.

158 Elena Soare

Further research is needed to establish the exact division of verbal/nominal properties in these contexts, but it is clear that the supine domain is also reduced, and in it, the grammatical, aspectual layers and the layers responsible for the external arguments are not contributed by the supine itself, but by the external layer of the main verb. 4. Conclusions In this paper, I approached the problem of the canonical verbal and nominal properties of mixed categories, which may be defective: nominal projections can trigger fully verbal behaviors, while verbal projections might not. This situation is hard to account for, whether in traditional grammatical systems, or for modern approaches with categorial systems in which lexical categories are defined through binary feature specifications. I hope to have shown that the comparison between nominal and verbal instantiations of different participial constructions (more concretely, those known in Romanian grammar as supine) provides an argument for a syntactic approach of categorization. Participial domains can correspond to structures of different heights, and thus may contain or lack layers which are responsible for the introduction of event structure, and more precisely, of grammatical aspect and the realization of an external argument. Exploring some specific properties of such domains, I rejected the view that categorization is the result of binary feature specifications. Instead, I proposed that categorization is the result of different layers of structure. It is perfectly possible to consider the participial stem as contributing at least a layer which is undetermined for voice (active or passive) and introduces an event variable. The projection of argument structure and the upper layers of grammatical aspect are specific only to some participial constructions. While there are still grey areas that will have to be explored in future research, the account/analysis proposed in this paper shows that participles are open structures of different heights, which never have an inner subject layer.

References Abney, Steven Paul. 1987. The English noun phrase in its sentential aspect. PhD dissertation, MIT. Alexiadou, Artemis. 2001. Functional structure in nominals: Nominalization and ergativity. Amsterdam: John Benjamins. doi: 10.1075/la.42



Non-categorical categories 159

Alexiadou, Artemis, Mariangeles Cano, Gianina Iordăchioaia, Fabienne Martin & Florian Schäfer. 2013. Direct participation and agent exclusivity effects in derived nominals and beyond. In Gianina Iordăchioaia, Isabelle Roy, & Kaori Takamine (eds.), Categorization and category change in morphology, 153–180. Cambridge: CSP. Alexiadou, Artemis, Berit Gehrke & Florian Schäfer. 2012. Adjectival participles revisited. Handout, Workshop on Aspect and argument structure of adjectives and participles, University of Greenwich. Alexiadou, Artemis, Gianina Iordăchioaia & Elena Soare. 2010. Number/aspect interactions in the syntax of nominalizations: A distributed morphology approach. Journal of Linguistics 46(3). 537–574. doi: 10.1017/S0022226710000058 Avram, Larisa. 1994. Auxiliaries and the structure of language. Editura Universității București. Baker, Mark. 2003. Lexical categories. Cambridge University Press.  doi:  10.1017/CBO9780511615047

Baker, Mark. 2005. On gerunds and the theory of categories. Ms. Rutgers. http://www.rci. rutgers.edu/~mabaker/gerunds&cattheory.pdf. Bhatt, Rajesh. 1999. Covert modality in non-finite contexts. PhD dissertation, University of Pennsylvania. Borer, Hagit. 1993. Parallel morphology. USC Manuscript. Borer, Hagit. 1999. The form, the forming and the formation of nominals. Presentation 2nd Mediterranean Morphology Meeting. Borer, Hagit. 2003. Exo-skeletal vs. Endo-skeletal explanations: Syntactic projections and the lexicon. In John Moore & Masha Polinsky (eds.), The nature of explanation in linguistic theory, 31–67. CSLI Publications. Borer, Hagit. 2005. Structuring sense. Vol. I: In name only. Oxford University Press. Borer, Hagit. 2005. Structuring sense. Vol. II: The normal course of events. Oxford University Press. doi: 10.1093/acprof:oso/9780199263929.001.0001 Borer, Hagit. 2013. Structuring sense. Vol. III: Taking form. Oxford University Press.  doi:  10.1093/acprof:oso/9780199263936.001.0001

Chomsky, Noam. 1970. Remarks on nominalizations. In Roderick A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English transformational grammar, Ginn and Company, 184–221. Waltman, Mass. Cornilescu, Alexandra. 2001. Romanian nominalizations: Case and aspectual structure. Journal of Linguistics 37(3). 467–501. doi: 10.1017/S0022226701001074 Cornilescu, Alexandra. 2003. Complementation in English. Editura Universității din București. Bucharest. Doron, Edit. 2003. Agency and voice: The semantics of the semitic templates. Natural Language Semantics 11(1). 1–67. doi: 10.1023/A:1023021423453 Doron, Edit & Chris Reintges. 2011. On the syntax of participial modifiers. Ms, Jerusalem. http://pluto.huji.ac.il/~edit/edit/31.pdf. Embick, David. 2004. On the structure of resultative participles in English. Linguistic Inquiry 15(3). 355–392. http://www.ling.upenn.edu/~embick/result.pdf (28 May, 2013).  doi:  10.1162/0024389041402634

Giurgea, Ion & Elena Soare. 2010. Predication and the nature of non-finite relatives in Romance. In Anne-Marie DiSciullo & Virginia Hill (eds.), Edges, heads, and projections. Interface properties, 191–214. Amsterdam: John Benjamins. doi: 10.1075/la.156.12giu Huddleston D. Rodney & Gefforey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press.

160 Elena Soare

Guéron, Jacqueline & Teun Hoekstra. 1988. T-Chains and the constituent structure of auxiliaries. In A. Cardinaletti, G. Cinque, & G. Giusti (eds.), Proceedings of GLOW conference in Venice, 1987, 35–99. Dordrecht: Foris. Halle, Morris & Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In K. Hale & S. Jay Keyser (eds.), The view from building 20, 111–176. Cambridge: MIT Press. Iatridou, Sabine, Elena Anagnastopoulou & Roumyana Izvorski. 2001. Observations about the form and meaning of the perfect. In Michale Kenstowicz (ed.), Ken Hale: A life in language, 189–238. Cambridge, Mass.: MIT Press. Iordăchioaia, Gianina. 2008. External argument PPs in Romanian nominalizations. In Florian Schäfer (ed.), SinSpeC – working papers of the SFB 732 Incremental Specification in Context, 01, 71–84. Publikationsverbund der Universität Stuttgart (OPUS) Iordăchioaia, Gianina & Elena Soare. 2008. Two kinds of event plurals: Evidence from Romanian nominalizations. In Oliver Bonami & Patricia Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics 7, 193–216. Paris: CNRS. Iordăchioaia, Gianina & Elena Soare. 2009. Structural patterns for plural blocking in Romance nominalizations. In Enoch Aboh, Elisabeth Van der Linden, Josep Quer, & Petra Sleeman (eds.), Romance languages and linguistic theory. Selected papers from “Going Romance” 2007, 145–160. Amsterdam: John Benjamins. doi: 10.1075/rllt.1.08ior Iordăchioaia, Gianina & Elena Soare. 2011. A further insight into the syntax-semantics of pluractionality. Proceedings of SALT 21, 95–114. http://elanguage.net/journals/index.php/ salt/article/view/21.95. Iordăchioaia, Gianina & Elena Soare. 2013. Pluractionality with lexically cumulative verbs: The supine nominalization in Romanian. Submitted. Kayne, Richard. 1995. The antisymmetry of syntax. Cambridge; MA: MIT Press. Kayne, Richard. 2008. Antisymmetry and the lexicon. Linguistic Variation Yearbook 8. 1–31.  doi:  10.1075/livy.8.01kay

Kayne, Richard. To appear. Comparative syntax and English Is-to. To appear in Linguistic Analysis. Laca, Brenda. 2006. Indefinites, quantifiers and pluractionals: What scope effects tell us about event pluralities. In Liliane Tasmowski & Svetlana Vogeleer (eds.), Non-definiteness and plurality, 191–217. Amsterdam: John Benjamins. doi: 10.1075/la.95.10lac Lundquist, Björn. 2008. Nominalizations and participles in Swedish. University of Tromsoe. Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. In Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, & Alexander Williams (eds.), Proceedings of the 21st Annual penn linguistics colloquium. Penn working papers in linguistics 4(2). 201–225. Negoiță Soare, Elena. 2002. Le supin roumain et la théorie des catégories mixtes. Thèse de doctorat: Université de Paris 8. Roy, Isabelle & Elena Soare. 2012. L’enquêteur, le surveillant et le détenu: les noms déverbaux de participants aux événements, lectures événementielles et structure argumentale. Lexique 20. 207–231. Sleeman, Petra. 2011. Verbal and adjectival participles: Position and internal structure. Lingua 121(10). 1569–1587. doi: 10.1016/j.lingua.2011.05.001 Soare, Elena. 2013. La représentation des événements dans les domaines non temporalisés. Mémoire d’Habilitation à Diriger des Recherches. Université de Paris 8. Soare, Elena. To appear. Nominalizing with or without ‘N’. Submitted.

Part 3

Variations in transitivity

The semantic motivation of non-canonical predicative relations The French transitive construction Meri Larjavaara

Åbo Akademi University

The paper discusses the semantic motivation behind the non-standard transitive usage of verbs in contemporary French. An example would be the following: On m’a démissionné hier (‘they fired me yesterday’). In the normative language variety, démissionner is used only in the intransitive construction as in J’ai démissionné (‘I gave in my notice’). The non-canonicity of the predicative relations discussed is of a lexical nature: the verbs in question do not normally appear in the transitive construction. Nevertheless the transitive construction itself belongs to the constructions in frequent use in French, and the novel usage of the verbs can be easily understood. Speakers use the verbs transitively despite the lack of canonicity of this usage, and the paper proposes a semantic explanation for this: the syntactic choice is semantically motivated. The transitive construction carries in itself – despite its frequency – a semantics that the speaker wants to make use of. The study takes advantage of the Construction Grammar approach. The paper starts by presenting definitions of semantic transitivity in a functional, typological perspective (G. Lazard 1994; S. Kittilä 2002; Å. Næss 2007). Then it goes through different cases where a pair of a transitive construction and a non-transitive (oblique) construction can be attested (penser (à), toucher (à), effected objects and others). Many of the examples are from the web where non-normative usage flourishes. The two constructions are compared and it is claimed that more transitive semantic features can be found in the transitive constructions. Other factors possibly relevant here are also briefly discussed: iconic motivation, semantic generalization in a particular technolect. The paper shows that the French transitive construction indeed has a semantics and that these novel usages can thus be semantically motivated.

doi 10.1075/lis.33.07lar © 2016 John Benjamins Publishing Company

164 Meri Larjavaara

1. Introduction The aim of this paper is to investigate whether novel transitive usage of verbs can be semantically motivated in contemporary French. This is done by defining semantic transitivity from a typological point of view as a prototypical category (Section 3) and by analysing pairs of transitive and intransitive constructions to see if the syntactically transitive one corresponds semantically more to transitive semantics than its intransitive counterpart.1 To be able to discuss non-canonicity of predicative relations one has to start by defining in what sense the relations that are going to be discussed are non-­ canonical. I take as my starting point an utterance presented as containing a typical non-­canonical predicative relation in Modern French. (1) On m’a démissionnée hier. ‘they fired me yesterday’

The verb démissionner, ‘to fire’, is in the normative standard language variety used only in the intransitive construction as in J’ai démissionné (‘I gave in my notice’). In the novel construction of the example (1), the same verb, without any morphological modification, is used in a transitive construction, with the NP referring to the person having lost his/her job as an object, while it is the subject of the corresponding intransitive verb. The verb seems thus to be a labile verb, and there are also other labile verbs in Modern French that do appear in both transitive and intransitive constructions without morphological changes (for labile verbs, see M. Rothemberg 1974; C. Bassac 1995; M. Larjavaara 2000). The transitive construction in (1) – subject, verb, (direct) object – is a stan­ dard one and there is nothing novel in the utterance as such. The predicative relation in (1) is thus non-canonical only in the sense that the verb in question normally is used exclusively in an intransitive construction. The non-canonicity of the predicative relations discussed in this paper is thus of a lexical nature: the verbs in question do not normally appear in the transitive construction. Nevertheless the construction itself belongs to the constructions in frequent use in French, and the novel usage of the verbs can be easily understood by the speakers. It is also worth adding that the verb démissionner in the transitive construction is indeed used in colloquial French sufficiently frequently to appear in the dictionary Petit Robert (2011) (with the remark “ironically”; we will return to this), and the fact that it is sometimes used in a non-standard way is precisely

1. I thank the two anonymous reviewers for their valuable comments and suggestions.



The semantic motivation of non-canonical predicative relations 165

why it is taken as an example. The non-canonicity may thus be of a sociolinguistic nature as well. The paper does not have a quantitative purpose. 2. The transitive construction The non-canonical transitive constructions investigated in this paper do pertain to different categories. As we have noted, the non-canonicity or atypicality of the verb structures discussed is lexical: they do not correspond to the lexical properties of the verbs, which, in turn, are an abstraction of the usages in which the verbs do figure. According to this way of defining the canonical structure of a verb, a verb is not transitive if it is not used in transitive constructions. And the other way round, if a verb is found in transitive constructions, it is transitive. In passing, this seems fairly simple but, surprisingly, is not always the way the grammarians see the question especially in languages with a long grammar writing tradition. For example, in the French tradition, many articles are written about the cognate object category (objet interne) which is basically used to entitle the use in transitive constructions of lexically intransitive verbs (see M. Larjavaara 1997; but also A. Bourmayan 2013: 25–28). As to the non-canonical transitive utterances in question, if they are attested and do correspond to existing structures, they do belong within the limits of the possible constructions in contemporary French and are not slips of the tongue possibly corrected afterwards. It is particularly important to investigate them to be able to see where the limits of the possible are drawn in the language in question. All the non-canonical transitive uses in French discussed in this paper are not instances of labile verbs. The most striking difference between labile and other uses is illustrated by (2) and (3): if in the case of the labile verb démissionner in (2) S (subject of the intransitive construction) has roughly the same semantic role in relation to the process referred to by the verb as O (object of the transitive construction), in (3) the semantic role of S corresponds to that of A (subject of the transitive construction): (2) J’ai démissionné → On m’a démissionnée S V → A O V (3) Elle a triché (pour me tromper) → Elle m’a trichée S V → A O V

166 Meri Larjavaara

The non-canonical transitive use of the verb tricher is thus clearly different from that of a labile verb. The non-canonical transitive constructions we are interested in fall into the following cases when the transitive construction is compared to the canonical one (cf. M. Larjavaara 2000) (the translations given here and elsewhere in the paper are literal): (4) Labile verb: On m’a démissionnée. cf. canonical construction: J’ai démissionné. (5) Novel object: Je le coupe. ‘I interrupt him’ cf. canonical construction: Je lui coupe la parole. ‘I interrupt the speech to him’ (6) Explicit effected object: J’aboie un discours. ‘I bark a speech’ cf. canonical construction: J’aboie. ‘I bark’ (7)

Missing preposition: Je plonge le monde. ‘I dive the world’ cf. canonical construction: Je plonge dans le monde. ‘I dive into the world’

(8) Other cases: Elle m’a trichée. ‘she fooled me’ cf. canonical construction: Elle a triché. ‘she cheated’

I will argue in what follows that these different types of non-canonical transitive uses have something in common: namely, that the motivation behind the transitive use is semantic. The speaker makes use of the transitive construction available in the language for verbs that in the standard variant of the language do not appear in a transitive construction, and this because there is a meaning s/he wants to express. That meaning is carried by the construction. The language can be seen as a grammatical system where lexemes are used in constructions. For those working within the Construction Grammar framework (see Construction Grammar 2013), constructions are stocked and learnt in the



The semantic motivation of non-canonical predicative relations 167

same way as lexemes, they simply are at different levels of abstraction. According to this way of seeing language, it meets the expectations that a verb can appear in a construction even if it is not one of the canonical ones for the verb in question. This paper is situated within a functionalist, typological framework, with a constructionist flavour – this is a cautious way of stating it, because no formalism is used but reference to studies pertaining to Construction Grammar is made. The aim of the paper is, to formulate it in a different way, to find a potential common denominator for the non-canonical transitive uses in French, keeping in mind that lexemes and constructions in which they appear together build the meaning of utterances. We can take, for example, the case of the pairs of constructions where there either is or is not a preposition but the utterance otherwise seems to be the same (ex. (7)). According to functional linguistic theories (e.g. Cognitive Grammar (R. Langacker 1991), Construction Grammar (see A. Goldberg 1995; Grammatical Constructions 2005)), variation is never meaningless. If two or more forms seem to correspond to one meaning, there most probably is a difference to be sought and found. It may be a difference related to different language varieties – regional, sociolinguistic, stylistic – or a semantic or pragmatic difference. The variational differences pertain to another kind of problematics, but in one variety, language tends to reach the economic balance one form–one meaning, and the situation where two forms correspond to one meaning would be transitory when seeking the balance. The two forms would tend to have different meanings. It would thus be natural to think that the non-canonical usages of verbs in transitive constructions would have even a slightly different meaning from that of the canonical construction. Indeed, according to J. François (1998: 185), there is a small but tangible semantic difference between (9) and (10): (9) J’ai pensé à ce problème. ‘I thought about the problem’ – with preposition à (10) J’ai pensé et repensé ce problème. ‘I thought and rethought the problem’ – with direct object

J. François says that in (9) the two participants, ‘me’ and ‘the problem’, are in an agent–locative relation, while in (10) ‘the problem’ would be affected by the process (and thus be more of a patient). The transitive constructions investigated in this paper can be classified in different categories, but I do argue that they share a common motivation: the transitive construction – subject, verb, object – has a meaning in French. The question is then to know what the transitive construction means in French, that is, what its meaning is, if there is one.

168 Meri Larjavaara

3. The meaning of the transitive construction The typologists agree to a certain point about the semantics of the transitive utterance.2 To investigate that meaning, one takes as a starting point the morphosyntactically transitive construction, which is defined as the construction that is used to refer to the most prototypical two participant action where the referent of the subject acts agentively and the referent of the object is affected (G. Lazard 1994). In French, this morphosyntactic construction can be schematically represented as a subject-verb-object structure. A typical predicate of that type would be, as unpleasant as it is, ‘to kill’. It is then studied which types of action are referred to by this same construction in a given language and their semantic properties are investigated. I report here the properties given in three typological studies (G. Lazard 1994; S. Kittilä 2002; Å. Næss 2007). G.  Lazard (1994: 250) states that the characteristics of semantic transitivity are agent and object individuation (whether they are definite, human, animate, etc.) and the completed character of the process (its reality, perfectiveness, etc.) including agentivity and effectiveness. The semantic transitivity is a gradual notion: the more individualised the agent and the object are and the more completed the process is, the more likely it is that the sentence is even morphosyntactically transitive. In the same way, Å. Næss (2007: 15) says: A prototypical transitive clause is taken to be one which describes an event involving: – a volitionally acting ‘agent’ participant – performing a concrete, dynamic action – which has a perceptible and lasting effect on a specific ‘patient’. In addition, the event should be presented as real and concluded; that is, the clause should show perfective rather than imperfective aspect, realis rather than irrealis mood, be positive rather than negated, etc.

As to S. Kittilä (2002), he reports several definitions given by different linguists and states (p. 38) that the asymmetric character of the process is the most important semantic property of transitivity: there are two participants, one of them instigates the event and the other one is directly affected by it. He adds that this kind of event is also typically transitive even when a child learns his/her first language. He also points out, as the other linguists mentioned, that semantic transitivity is not clearcut but resembles more a continuum, that is, it is a gradual, scalar notion.

2. For important methodological discussion, see G. Lazard (2011).



The semantic motivation of non-canonical predicative relations 169

This antagonism between the two participants – this asymmetry – seems indeed to be a basic property of semantic transitivity. It can be referred here to the two “proto-roles” presented by David Dowty in his article in Language in 1991. He presents the agent proto-role, that is the prototypic semantic role which has the most prototypic agent as its core, and the patient proto-role. As a matter of fact, the aim of his article is to study the semantic properties of subjects and objects and particularly their contrastive properties. It is worth noting that he stresses as well that the prototypic transitivity is based on the antagonism of the subject and the object: one is ideally approaching the agent proto-role and the other one the patient proto-role. D. Dowty’s way of defining the proto-roles is by assigning to them characteristics that are neither necessary nor sufficient. The two proto-roles are not discrete categories – as the term “proto-role” already tells us – but rather concepts that bring together roles resembling each other (D. Dowty 1991: 572). D. Dowty makes use of the proto-roles when choosing between the participants the one which is most appropriate as a subject. Even though this is his primary goal, the two proto-roles are salient categories that can be used for other purposes as well. For instance, A. Goldberg (1995: 117) takes advantage of D. Dowty’s proto-roles when defining the transitive construction. The proto-roles can indeed be seen as a cross-linguistic definition of the semantic transitivity: in a semantically transitive relation, the two participants tend to have roles that approach the proto-roles. Semantic transitivity is a scalar notion: an utterance is more or less transitive semantically; it does resemble more or less a prototypical transitive utterance. The subject role tends to resemble the agent proto-role and to be different from the object role and the other way round for the object role. It is not surprising to notice that different languages may then stress different factors involved. 4. The meaning of the transitive construction in contemporary French: Case studies Let us now return to the transitive construction in French. In this section, I will discuss several different cases where a transitive construction can be compared with an intransitive one. According to A. Goldberg (1995: 118), there is a transitive meaning cross-linguistically, but this meaning is either very general and abstract or there is a family of related meanings corresponding to the transitive construction, that is, there are multiple senses involved. A. Goldberg continues:

170 Meri Larjavaara

Languages differ in how and to what extent the transitive construction is extended to express nonprototypical semantically transitive scenes. (…) Language-­ specific idiosyncrasies would arise, according to this view, in just how languages extend their inventory of grammatical argument structure constructions to cover expressive requirements.

French is an Indo-European language and G.  Lazard (1994: 63) is not the only linguist who has pointed out that many languages but particularly many Indo-­ European languages in Europe extend the transitive morphosyntactic model to utterances that express a wide variety of meanings. A striking example is the utterance in (11) (G. Lazard 1998: 75): (11) Des hangars jouxtaient des ateliers. ‘warehouses were situated alongside workshops’

In this example, it is difficult to see how the semantic role of the warehouses would be more agentive than that of the workshops. The sentences which have the same morphosyntactic (transitive) form as sentences which refer to prototypical actions do not all refer to actions with agent and patient – or be it proto-agent and proto-patient – but to diverse types of events. Thus, it seems clear that in contemporary French the transitive construction cannot categorically assign agent and patient roles to subject and object respectively. Nevertheless, the existence of the transitive construction may present a semantic model, and if the transitive construction is contrasted to a construction which is not transitive when analysing non-­canonical predications – that is, minimal pairs –, tendencies can be detected. We will now go through several such cases. The corpus has been collected by hand during several years, partly for two other studies (M. Larjavaara 2000, 2009). Many of the examples are from the web where non-normative usage flourishes (the original orthography has not been corrected in the examples). The transitive semantics related to the transitive construction can naturally be seen when two constructions with the same verb are compared. In examples (12) and (13), the verb jargonner (‘to speak in an unintelligible way’) is used in a non-­ canonical transitive construction (the translations are simplified and correspond only to the essential part of the example). (12) Il est encore aujourd’hui quasiment impossible de se faire comprendre dans ce domaine [informatique] autrement qu’en jargonnant anglais, même entre congénères.  (ex. M. Larjavaara 2000: 207) ‘it is almost impossible to be understood in the IT branch if you do not speak English (even if it is bad English)’



The semantic motivation of non-canonical predicative relations 171

(13) Ainsi, le cosmopolite d’aujourd’hui est essentiellement un colonisé qui jargonne le yanqui (comme dit Etiemble) et ne se plaît qu’à l’ombre des gratte-ciel poussés comme champignons sous toutes les latitudes, ébloui par la technique, la verroterie et le catéchisme de l’american way of life. (http://archaion.hautetfort.  com/archive/2007/01/25/michel-mourlet.html 6.11.2012) ‘today’s cosmopolite is a colonized who speaks bad yankee’

When someone speaks bad English and that is expressed using jargonner in the transitive construction, the action is determined and the subject referent is volitional, as in examples (12) and (13). On the other hand, if jargonner is used in a canonical non-transitive construction, the subject referent is not intentionally using bad language, s/he would like to speak better but is not able to – the representation of the badly speaking is different. This is the case in ex. (14). (14) pour me payer assez de temps a l hotel pour pouvoir rancontrer ma fille retrouver on as besoin plus de temps car elle parle pas francais et moi je jargonne en anglais donc on as toujours besoin de un interprette,etc. etc.(http://www.tvqc. com/2009/08/le-banquier-inscriptions-hiver-2010/6.11.2012) ‘we need more time because she doesn’t speak French and I speak bad English so we need an interpreter’

The subject role thus is different in the two constructions: more agentive in the transitive one, less agentive in the non-transitive one. Another verb presenting a clear-cut difference is penser (‘to think’). I already mentioned J. François’s (1998: 185) remark on the verb (ex. (9) and (10) in this paper). In another study (M. Larjavaara 2009), the same verb is investigated in different text genres: on one hand, non-literary, normative standard French, on the other hand, literary or colloquial French. In all these genres there was a slight semantic difference between the transitive and the non-transitive construction, but interestingly, it was not the same in the genres in question. In normative but non-­literary French the transitive construction was frequent and above all motivated by the agentivity of the subject referent: (15) Quel est votre sentiment? – Avant tout, je veux exprimer un sentiment très profond: nous devons penser les problèmes du Proche-Orient en termes de paix, nous devons avant toute chose vouloir un règlement qui assure la paix. (Frantext: Pierre Mendès-France 1990; ex. in M. Larjavaara 2009) ‘we have to think about the problems in Near-East with peace in mind’

In this example, the subject role is decisive: we have to think about, our reflection will do something good.

172 Meri Larjavaara

On the other hand, in literary or colloquial genres, the transitive construction with the verb penser is less frequent and it is primarily motivated by the fact that the object is effected, that is, it refers to the product of the event; thus, the construction is highly transitive semantically: (16) Microsoft a bien pensé le système : (…)  (WebCorp: http://mhmag.free.fr/  zine/efs/EFS.TXT; ex. in M. Larjavaara 2009) ‘Microsoft has conceived well the system’ (17) Ce n’est que gorgé de compagnie et de vin que j’ai pu penser une telle ânerie. (Frantext: Hervé Guibert 2001; ex. in M. Larjavaara 2009) ‘because of the wine and the good company I did think such a foolish thing’

Interestingly, the different genres seem to take advantage in different ways of the semantic difference between the two constructions. In the first case, the typically transitive subject role was central, in the second case, the object role is typically transitive (namely that of an effected object). The verb toucher (‘to touch’) sheds light on a different side of semantic transitivity. According to the recent French verb dictionary (L. Florea & C. Fuchs 2010), toucher would appear in transitive constructions in several different meanings, both concrete and figurative, but the complement would be oblique with the preposition à in figurative meanings exclusively. The authors add, however, that if the sentence is negated, ne touchez pas à cet appareil (‘do not touch this device’) is used, that is, the non-transitive construction, even though the meaning is concrete. As a matter of fact, according to the typological studies, a negated sentence is not semantically typically transitive. In the case of toucher, the morphosyntax thus adds a preposition. The non-transitive, negated construction corresponding to this prevision can be seen in (18); example (19) presents, however, a transitive construction even though the predication is negated: (18) La seule fois où vous avez essayé de repasser une chemise de l’Homme, vous l’avez brûlée, et il vous a fait jurer sur la tête de Petit Chat de ne plus jamais toucher à ses vêtements.  (Frantext: Nicole de Buron 1998) ‘you burnt his shirt and will never more touch his clothes’ (19) Je lui avoue que je n’ai jamais assisté à un match, ni même touché un ballon de ma vie.  (Frantext: Françoise Dorin 1997) ‘I have never assisted in a match and not even touched a ball’

The variation seems thus more complicated than that.3 3. An anonymous reviewer points out that it is possible to say: “Touchez à cet appareil et il démarre tout seul” (´touch this device and it will start working automatically’), where the



The semantic motivation of non-canonical predicative relations 173

There are also verbs whose different constructions are strongly specialised and even polysemic. For example, the verb vivre (‘to live’) in transitive constructions pertains to advertising or journalistic genre (apart from the “cognate object” construction vivre sa vie ‘to live one’s own life’). In this transitive usage its meaning is nearly ‘to act in an agentive way to be able to enjoy the good thing a product/an experience can bring’. Even if this paraphrase may be somewhat strange, the context is anyhow active and dynamic: (20) Pour vous aider à vous projeter et vivre sereinement la découverte des différents campus où vous passerez vos oraux d’ici quelques mois, nous avons imaginé un supplément au magazine Espace Prépas: Business Schools, les plus beaux campus de France.(http://www.studyramagrandesecoles.com/home.  php?idRubrique=815 6.11.2012) ‘to help you to calmly discover the different campuses (…) we have created a special issue of the magazine (…)’ (21) Pour vivre pleinement votre capital beauté, NINA RICCI crée l’Extrait de Jeunesse … (advertisement / Air France Madame 52/1996: 17;  ex. M. Larjavaara 2000: 208) ‘to profit entirely from you beauty’

This usage does not exclusively belong to advertising: (22) (title:) Comment vivre la mort? (article:) «Le plan Kouchner pour adoucir la mort» est à la Une du Monde, qui publie un entretien avec le secrétaire d’État à la santé. Ce dernier annonce «le lancement de d’un (sic) plan triennal de lutte contre la douleur et de développement des unités de soins palliatifs.» (…) Ces mesures sont jugées «courageuses et indispensables» par la rédaction du Monde. Et elles démontrent «ce que ne cessent de proclamer depuis des années les spécialistes des soins palliatifs: il reste beaucoup à faire quand on croit qu’il n’y a plus rien à faire.»  (Le Petit Bouquet 347 24/9/1998; ex. M. Larjavaara 2000: 208–209) ‘how to live one’s death’ (about palliative care)

The cotext shows clearly that it is again an active way of experiencing death: a lot can be done when waiting for death. Comment (‘how’) stresses the choices there are to be made. Il reste beaucoup à faire (‘there is a lot to be done’): again an agentive use of the verb. oblique construction is used in the concrete meaning; this is not what is predicted by L. Florea & C. Fuchs (2010). I would suggest that this possibility could be due to the irrealis which decreases semantic transitivity (see Section 3): if you touched the device (irrealis), it would start working.

174 Meri Larjavaara

We have now seen several cases where the transitive construction can be coupled with more transitive semantics. It is important to note though that it is a question of tendencies and very subtle semantic differences. Furthermore, it is not always possible to detect any difference at all between the two constructions: (23) A part le couscous, avez-vous déjà goûté à un plat africain?(http://fr.answers.  yahoo.com/question/index?qid=20121005092649AAn1EXq 6.11.2012) ‘apart from couscous, have you tasted an African meal?’ (24) Qui a déjà goûté un plat éthiopien? (http://www.jeuxvideo.com/forums/ 1-50-74825509-1-0-1-0-qui-a-deja-goute-un-plat-ethiopien.htm 6.11.2012) ‘who has tasted an Ethiopian meal?’

The two examples seem to be almost exactly alike and still there is the preposition à in (23) and a transitive construction in (24). A semantic difference is difficult to see in these examples. However, according to L. Florea & C. Fuchs (2010), there is a difference between the two constructions: the transitive one would correspond to ‘to taste, to test’ (‘savourer, tester’), the oblique one to ‘to take a little of ’ (‘prendre un peu de’). Possible regional differences cannot be taken into account here. The cases analysed so far have presented a clear distinction between a transitive usage and an oblique usage with a preposition. We can add to these contrasted constructions other non-canonical constructions where the problem is less straightforward. Effected objects that are easily given a linguistic representation with an intransitive verb (we already mentioned them with the verb penser) form a clear-cut group of non-canonical transitive constructions. Semantic transitivity is evident in their case: (25) (…) et l’Anglais lui bramait des mots d’amour dans le cou (…)  (Gabily 1992: 142; ex. of the corpus of M. Larjavaara 2000) ‘the Englishman shouted words of love down her throat’

The object is effected and the subject referent agentive. The effect of the transitive construction in French can be seen when a verb which canonically is intransitive and has an agentive subject (so called inergative verb) can sometimes be used in a non-canonical transitive construction as a labile verb. This is not frequent but does happen when the context is strong enough to reorient the interpretation: it must be clear that the object referent cannot possibly be agentive in the extralinguistic situation in question. The referent may for example be sleeping (ex. (26)) or it can be a little child (ex. (27)). The lack of volitionality is important here.



The semantic motivation of non-canonical predicative relations 175

(26) Ils ont enlevé Julie, tout simplement. Ils l’ont fait sortir de l’hosto, sous prétexte de l’emmener à la radio. Ils (…) l’ont chargée toute dormante dans la bagnole de Louna et, fouette cocher, l’ont grimpée dans ma chambre. Voilà.  (Pennac 1994: 222; ex. M. Larjavaara 2000: 224) ‘they transferred Julie into my room’ (with the verb grimper ‘to climb’) (27) Par cette température, on ne voyage pas un enfant de cet âge.  (ex. in J. Damourette & É. Pichon 1911–1940: III §863,  p. 162 (Mme EJ 1/2/1931)) ‘at this temperature, you don’t travel a child of this age’

In the canonical usage, the subject referents of grimper and voyager are agentive (Julie grimpe; L’enfant voyage). The context makes it clear, however, that these referents (‘Julie’ and ‘the child’) are all but agentive in the context, and a transitive construction is thus possible. In these cases the utterance is easily ironic or there is a hint of wordplay in the choice of construction. The speaker explores the limits of the possible in the language. Let me also recall our very first non-canonical construction, on m’a démissionnée (ex. (1)). As the verb is canonically used only in an intransitive construction and the subject referent is agentive in that usage (J’ai démissionné), the non-­canonical transitive use contains something one could call a wordplay: the transitive construction entails (cf. ex. (26) and (27)) that the object referent is powerless, and yet the verb démissionner itself contains the idea of doing it voluntarily. This is why this use is “ironical” according to the Petit Robert (2011). It is always a question of how the speaker conceives or wants to present the extralinguistic situation. The following examples are interesting in that sense: (28) Ça manquait pas, les filles qui auraient bien aimé qu’il les rentre du bal.  (Picouly 1995: 178; ex. M. Larjavaara 2000: 234) ‘there were lots of girls who would have wanted him to see them home’ (29) Attendez monsieur, je peux vous traverser si vous voulez.  (attested spoken utterance 10/3/1993; ex. in C. Bassac 1995: vol. II, p. 110) ‘wait sir, I can walk you cross the street’

These two examples resemble the previous ones, but the object referent – ‘the girls’ and ‘the man’ – are not totally powerless (unagentive). They are thus in conflict with what I just said. One could maybe suppose that this is made possible by the context where the object referent happily gives his/her permission to be under the power of the subject referent. The question remains open.

176 Meri Larjavaara

We have gone through several different cases where the transitive construction can be seen as being more transitive semantically as its intransitive counterpart. These indicate that the transitive construction does have a meaning in contemporary French. 5. The meaning of the transitive construction in contemporary French: Deductions The semantic differences analysed in Section 4 have, of course, been noted and discussed by several linguists. I already mentioned J. François (1998) about the verb penser (see J. François 2003 as well); in the same book, J.-P. Desclés (1998: 177– 178) discusses pairs of utterances such as Nous avons approché l’ennemi (‘we got closer to the enemy’) and Nous nous sommes approchés de l’ennemi (same meaning, but preposition de with the complement and reflexive verb morphology). He says that the meaning of the transitive construction resembles more the semantics of the prototypical action: there are differences along the parametres of intentionality, control, willingness, aspect, telicity, success of the action, etc. The representation of the situation thus is changing when the construction changes. In the same way G. Alvarez (1968: 19) writes that when the following constructions are analysed, there is always a difference to be found: discuter un problème / d’un problème / sur un problème / à propos d’un problème (‘to discuss a problem’) and frapper le mur / à la porte / sur le sol / contre le mur (‘to hit a wall/ to knock on the door/to hit the floor/against the wall’). According to him, the construction chosen is a proof of a different way of conceptualising the relation between the verb and its complement. In the transitive discuter un problème, the relation is intimate, but it becomes more and more loose, and finally, in the case of the construction discuter à propos d’un problème, the syntactic and semantic distance is at its biggest. H. Geisler (1988: 28–29) says as well that the constructions with a preposition prove to be weak in transitivity.4 After having analysed the non-canonical transitive constructions in French, it can be stated that the speakers do use the transitive construction in order to express the transitive meaning: the transitive utterances tend to approach semantically the two-participant prototypical action with the subject referent in the agent 4. However, G. Gougenheim (1959: 11–12) presents a divergent interpretation: according to him, if we compare commander une armée and commander à une armée (‘to command an army’) or toucher à un objet and toucher un objet, the difference is that the oblique construction is more dynamic and the subject referent has an active role.



The semantic motivation of non-canonical predicative relations 177

proto-­role and the object referent in the patient proto-role. The speaker utilizes analogy – one of the strongest forces in language (cf. E. Itkonen 2005) – and takes up the model of an already existing construction to express the extralinguistic situation. However, I would still like to sketch two other possible motivations to the speaker’s choice of construction. G. Serbat (1994) in a small two-page article treats the verb jouer (‘to play’) and the variety of objects it can take more or less non-canonically: jouer une pelouse durcie / un champion basque / une balle plus grosse (objects: hardened lawn, basque champion, bigger ball), etc. It can be sometimes considered that the relation between verb and object, if it is the shortest possible, compact, that is direct with no preposition, can be treated as representing iconically the fastness of the action (see also L. Kolehmainen & M. Larjavaara 2004). That would be an iconic motivation. On the other hand, the “technical uses” are well known. In a definite context where speakers know which kind of a relation there is between the action and the referent there is no need to specify it. It can be left open and expressed just by the most simple relation there is, that is, the direct one. This can also be due to wearing out for example in some professional jargon. These cases could be more thoroughly explored. I will finish by adding a last example, which is, this time, contrary to the non-­canonical transitive constructions we have seen so far but which proves the existence of the semantic motivation coupled with the transitive construction. In example (30) the construction is not canonical: the verb réussir (‘to manage’) is easily used in a transitive construction, but here, there is a very complex preposition – which, according to G. Alvarez (1968), would be considered as representing a very distant relation – par rapport à (‘in relation to’): (30) Donc oui, c’était un objectif pour moi que les étudiants soient capables de continuer à faire des party. En gros de défendre les activités étudiantes, c’était un des principaux objectifs et je pense qu’on a réussi par rapport à ça.  (http://www.polyscope.qc.ca/spip.php?article1152 6.11.2012) ‘defending the student activities was one of the main goals and I think we have managed to do that’

It seems that in this utterance the complex preposition has been added to avoid self-­praise. The morphosyntactic as well as the semantic transitivity have been hidden or diminished, and the agentivity of the subject referent (‘we’) is thus decreased. This construction could perhaps be called a non-transitive construction of modesty.

178 Meri Larjavaara

After having discussed several different cases of transitive/non-transitive variation in contemporary French, it can be concluded that novel transitive usage of verbs can be semantically motivated in contemporary French, but that there would still be more to investigated.

References Alvarez, Gerardo H. 1968. Transitivité et économie de l’objet dans l’énoncé français. Doctoral thesis. Faculté des lettres et sciences humaines Paris-Nanterre. Bassac, Christian. 1995. Le statut de verbe dit ergatif: étude contrastive anglais–français, vol. 2. Doctoral thesis, Université de Nancy II. Bourmayan, Anouch. 2013. Les objets implicites en français: une approche pragmatique. Doctoral thesis, École des hautes études en sciences sociales. Construction grammar. 2013. http://www.constructiongrammar.org/. Damourette, Jacques & Édouard Pichon. 1911–40. Des mots à la pensée: essai de grammaire de la langue française. Volumes I–VII. Paris: Éditions d’Artrey. Desclés, Jean-Pierre. 1998. Transitivité sémantique, transitivité syntaxique. In André Rousseau (ed.), La transitivité, 161–180. Travaux et recherches. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67. 547–619.  doi:  10.1353/lan.1991.0021

Florea, Ligia Stela & Catherine Fuchs. 2010. Dictionnaire des verbes du français actuel. Paris: Ophrys. François, Jacques. 1998. Théorie multifactorielle de la transitivité, “différentiel de participation” et classes aspectuelles et actancielles de prédication. In André Rousseau (ed.), La transitivité. Travaux et recherches, 181–201. Villeneuve d’Ascq: Presses Universitaires du Septentrion. François, Jacques. 2003. La prédication verbale et les cadres prédicatifs (Bibliothèque de l’Information Grammaticale 54). Louvain – Paris: Peeters. Frantext. http://www.frantext.fr. Fried, Mirjam & Hans C. Boas (ed.). 2005. Grammatical constructions: Back to the roots (Constructional approaches to language 5). Amsterdam: John Benjamins. doi: 10.1075/cal.4 Geisler, Hans. 1988. Das Verhältnis von semantischer und syntaktischer Transitivität im Französischen. Romanistisches Jahrbuch 39. 22–35. doi: 10.1515/9783110244953.22 Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago – London: The University of Chicago Press. Gougenheim, Georges. 1959. Y a-t-il des prépositions vides en français?. Le français moderne XXVII. 1–25. Itkonen, Esa. 2005. Analogy as structure and process: Approaches in linguistics, cognitive psychology, and philosophy of science. Amsterdam: John Benjamins. doi: 10.1075/hcp.14 Kittilä, Seppo. 2002. Transitivity: Towards a comprehensive typology (Yleisen kielitieteen julkaisuja 5). Turku: Turun yliopisto.



The semantic motivation of non-canonical predicative relations 179

Kolehmainen, Leena & Meri Larjavaara. 2004. The “bizarre” valency behaviour of Finnish verbs: how a specific context gives rise to valency alternation patterns. Constructions 1. http://elanguage.net/journals/constructions. Langacker, Ronald W. 1991. Foundations of cognitive grammar, vol. II. Stanford: Stanford University Press. Larjavaara, Meri. 1997. À quoi sert l’objet interne?. In: Les objets: relations grammaticales et rôles sémantiques. In Dominique Willems & Ludo Melis (eds.), Travaux de linguistique, 35. 79–88. Larjavaara, Meri. 2000. Présence ou absence de l’objet : limites du possible en français contemporain (Humaniora 312. Annales Academiæ Scientiarum Fennicæ). Helsinki: Academia Scientiarum Fennica. Larjavaara, Meri. 2009. Penser (à) l’emploi: schémas actanciels dans deux genres de textes. In Eva Havu, Juhani Härmä, Mervi Helkkula, Meri Larjavaara & Ulla Tuomarla (eds.), La langue en contexte: Actes du colloque “Représentations du sens linguistique IV”, Helsinki 28–30 mai 2008 (Mémoires de la Société Néophilologique de Helsinki LXXVIII), 83–91. Helsinki: Société Néophilologique. Lazard, Gilbert. 1994. L’actance. Paris: Presses Universitaires de France. Lazard, Gilbert. 1998. De la transitivité restreinte à la transitivité généralisée. In André Rousseau (ed.), La transitivité. Travaux et recherches, 55–84. Villeneuve d’Ascq: Presses Universitaires du Septentrion. Lazard, Gilbert. 2011. Vers une science des langues – un exemple: la transitivité. Faits de langues 38. 17–27. Næss, Åshild. 2007. Prototypical transitivity. Amsterdam: John Benjamins. doi: 10.1075/tsl.72 Le Petit Robert. 2011. Éd. Josette Rey-Debove & Alain Rey. Paris: Le Robert. Rothemberg, Mira. 1974. Les verbes à la fois transitifs et intransitifs en français contemporain (Janua linguarum. Series practica 215). The Hague – Paris: Mouton. Serbat, Guy. 1994. Le golf: une démocratisation et une transitivation galopantes. L’information grammaticale 62. 57–58.

Atypical argument structures in French From metaphorical uses to atypical ones* Geneviève Girard-Gillet

Université Sorbonne Nouvelle Paris 3 / EA 4398 PRISMES

The article analyses some atypical constructions of French intransitive verbs, such as ils ont suicidé le ministre (they suicided the minister), la crème a explosé les poux (the cream exploded the lice), and tries to account for their existence, as the construction is not always possible, and differs semantically from ils l’ont fait se suicider (they made him commit suicide), il a fait exploser la bombe (he triggered the bomb), where the FAIRE faire construction is compulsory. The hypothesis is that the atypical constructions can be explained within the domain of causative constructions, and more particularly with regard to the distinction between internal and external causation (Guéron, Talmy, Levin, McKoon and Macfarland). The atypical constructions are only possible when no internal causation is available, namely when the causee does not possess the internal characteristics enabling the change of state to occur: the minister was not the suicidal type and the interpretation is then that he was murdered. A change of state did occur, but not the one denoted by the lexical verb. The motivation for the structure is the wish to highlight an unexpected event.

Introduction The issue we propose to address in this article is a threefold one: – how can we account for atypic uses of some French verbs and in particular the transitive uses of some intransitive verbs for change-of-state events, such as se suicider, démissionner, exploser? – what are the semantic constraints limiting this use, since the phenomenon does not apply anywhere? – what is the pragmatic function of such uses? * I would like to thank two anonymous reviewers for their comments, criticisms and insights in a previous version of this article. Any remaining errors and shortcomings are of course my own. doi 10.1075/lis.33.08gir © 2016 John Benjamins Publishing Company

182 Geneviève Girard-Gillet

The following occurrences, which seem more and more frequent in oral French and news bulletins, exhibit the transitive construction of three intransitive verbs, se suicider (to commit suicide), exploser (to explode), démissionner (to resign), are felt to be atypic: (1) Et oui ils l’ont suicidé…comme bien d’autres… C’était l’hécatombe à l’époque… (www.age-des-celebrites.com/age-de-pierre-beregovoy.php) they-him-have-suicided (they suicided him1) (2) Les poux, les lentes, ça les explose Lice-nits-it-them-explodes (It explodes lice and nits) (advertisement at a chemist’s for a cream against lice and nits) (3) Ghilardenghi n’est enfin plus le représentant des Taxis, ils l’ont démissionné comme on dit (www.laprovence.com/…/les-taxis-mobilises-demain-contrela-concurrence) They-him-have-resigned (they resigned him)

But the same verbs do not accept the atypic construction in other contexts, because the interpretation construed is an entirely different one: (4) Nombre de documents démontrent qu’on a tenté de faire se suicider Paulette Cooper (scientologie.fraude.free.fr/02/morts3.htm) (A number of documents show that they tried to make Paulette Cooper commit suicide)/ *ils l’ont suicidée / *they suicided her (5) il a fait exploser la bombe dans le bus / * ? Il a explosé la bombe dans le bus (he triggered the bomb / he caused the bomb to explode)

1. The same phenomenon exists in English, but does not raise the same questions, due to the flexibility of the language. This point will be dealt with in the course of the article. The following examples appear with inverted commas, which seem to indicate that it is a borderline construction: The Nazis would have executed Plettenberg, so it is highly unlikely that they ‘suicided’ him. (books.google.fr/books?isbn=0199532567)/ Contrary to the TIME report, Deborah Jeane Palfrey, the DC Madam who reportedly committed suicide today, apparently warned that she would be “suicided” – murdered but made to look like a suicide. (http://www.huffingtonpost. com/2008/05/01/dc-madam-dead-in-apparent_n_99653.html) / How much money was Troy Tulowitzki making before they resigned him? (http://answers.yahoo.com/question/index?qid= 20080220193550AAIpw01).



Atypical argument structures in French 183

(6) Au moment où elle est mutée en Région parisienne, il l’a fait démissionner de la police / *il l’a démissionnée (he forced/compelled her to resign)

The French causative construction resorts here to the FAIRE faire construction, which encodes another type of cause-consequence relation. The existence of this analytical construction requires a closer analysis of the arguments at play, in order to understand to what extent the atypic uses differ, since the two constructions are not semantically interchangeable. Our hypothesis is that the atypic occurrences can be explained within the domain of causative constructions, and in particular with regards to the opposition between external and internal causation.The study will be divided into three parts: the first part will deal with the distinction between external and internal causation, the second part will analyse the two descriptive modes of causative events that exist in French under certain circonstances, along with possible metaphorical interpretations, and the third part will show how the atypical constructions are only possible when no adequate internal causation is available. This article presents the first tentative conclusions of an ongoing research. There are about 3142 French verbs which have a causative/unaccusative functioning3 – articles about this causative alternation in French sometimes use the term “symetrical” – such as casser (to break) in: le vent casse la branche/la branche casse (the wind breaks the branch/the branch breaks). They do not present an “abnormal” alteration of their argument structure, but some of them are semantically interesting for us, as their two kinds of functioning might be the link between typical and atypical uses. These verbs have been analysed by various linguists, among whom we can mention J. P. Boons, A. Guillet & C. Leclère (1976),4 A. Zribi-­Hertz (1987), R. Forest (1988), J. Dubois and F. Dubois-Charlier (1997), I. Ben Salah Tlili (2007), R. Pasero, P. Sabatier and M.-H. Stefanini (2010). They do not all agree on what their syntactic and semantic status is, and G. Lazard (1994: 155), for instance, does not think that they make up a semantic class.5 We agree with him, 2. The list is given by I. Ben Salah Tlili (2007), drawing on J. Dubois & F. Dubois-Charlier’s dictionary, Les Verbes français (1997). 3. In the causative alternation we are interested in, “the semantic role of the subject of the intransitive use of the verb is the same as the semantic role of the object of the transitive use of the verb” (B. Levin 1993: 25), hence the term “unaccusative”. There is only one protorole, Patient or Theme in the unaccusative construction. 4. We wish to thank D. Le Pezant for giving us this reference. 5. “Il est probablement vain de chercher à délimiter une sphère sémantique commune aux verbes réversibles”. It is problably useless to try and define similar semantic properties for these

184 Geneviève Girard-Gillet

but we suggest that a possible semantic delimitation might exist for some of them. This will be tackled in Section 2. 1. Distinction between external and internal causation 1.1

Theories on causation

1.1.1 Leonard Talmy’s theory In his force-dynamic theory the roles played by the causer (the agonist) and the causee (the antagonist) are fundamental to understand how the relation is conceptualized.6 For him, the “basic causative situation […] consists of three main components: a simple event (that is one that would otherwise be considered autonomous), something that immediately causes the event, and the causal relation between the two” (2000: 480), and in this relation “one force exerting entity is singled out for focal attention – the salient issue in the interaction is whether this entity is able to manifest its force tendency or, on the contrary is overcome. The second force entity, correlatively, is considered for the effect it has on the first, effectively overcoming it or not” (2000: 409). This implies that no causative process can be understood without considering how each participant in the event acts, or reacts to the force exerted. 1.1.2 External and internal causation The distinction between external and internal causation has been discussed by various linguists, even if they do not all use the same terms to refer to the two kinds of phenomema. The necessity to distinguish between the two participants in the causal process is present in J. Guéron’s analysis of causal construction (2008), and it leads her to consider that “the grammar distinguishes two types of causality on the basis of the lexical content of vP and syntactic structure, intentional causality and inertial causality. Under intentional causality, the direct cause of the event is the intention or goal of a human factor. […] Under inertial causality, the change of state which Sn denotes is attributed not to human intention but rather to the laws of the discourse universe revealing themselves over time” (J. Guéron 2008: 161–162).

verbs. Lazard uses the term “reversible” instead of “symetrical”. Dubois prefers “symetrical” for such verbs as ressembler (resemble). 6. We wish to thank S. Chatti for discussing L. Talmy’s theory with us.

Atypical argument structures in French 185



B. Levin posits a conceptual difference between two ways of bringing about a change of state: it can be internally caused or externally caused: a. an internally caused state: [x ] takes place without any outside help or influence. Such verbs as bloom, blossom, decay, flower, rot, rust, sprout …, express this kind of inherent modification. By internal causation B. Levin means that “the changes will occur in the normal course of event. However they may be externally triggered, facilitated, or regulated by certain natural forces or other environmental phenomena. […] The events will happen inexorably in the natural order of things” (2009). b. an externally caused state implies the action of a Causer that acts, and which is responsible for the change of state: [[x ACT ] CAUSE [BECOME[y ]]] and the verbs break, dry, melt, open, split, … suppose, in their causative interpretation, this chain of events.7 B. Levin’s distinction correlates with a syntactic one: the (a) type is not supposed to accept transitive constructions. The verbs are intransitive verbs. In the (b) type the verbs are transitive and intransitive. But the claim about the (a) type was questioned by G. McKoon and T. Macfarland (2000, 2002), who showed that verbs belonging to this class accepted transitive constructions, which for B. Levin is the proof of a causative construction. (7) the intense heat wilted the flowers (8) the severe frost withered the tulips (9) They cover their strawberry patch most commonly with straw … but not too soon! … or you will rot the plants (http://www.idigmygarden.com/forums/archive/ index.php/t-38074.html) (10) If you leave this process too long, you will sprout the seeds or they will darken and rot. Add water to your container slowly (http://www.eagleridgeseeds.com/ blog/growing-heirloom-tomatoes-secrets-to-success)

Examples (9) and (10), which are our examples, show that the causer can also be a human being, even if it is an indirect one. This will be important for our discussion. Drawing from data from Greek, A. Alexiadou and E. Anagnostopoulou (2003) posit that there are four types of change-of-state verbs: agentive (e.g., murder, assassinate), externally caused (e.g., destroy, kill), internally caused (e.g., blossom, wilt), cause unspecified (e.g., break, open), and propose that all change of state 7. We shall not discuss whether the causative use is derived from the inchoative (B. Levin 1993), or whether the inchoative verb is derived from the causative (J. Holmes 1999).

186 Geneviève Girard-Gillet

verbs should receive a causative analysis, both in transitive and intransitive uses. And A. Alexiadou, E. Anagnostopoulou and F. Schäfer (2006) suggest that the occurrence of a causer PP, such as a from phrase in the flowers wilted from the cold tends to prove that it has a causative analysis. In her comments on their hypothesis, B. Levin (2009) examines the validity of the causative diagnostic. She considers that the causative uses of English internally caused change-of-state verbs have subjects that, most of the time, are natural forces, or are in some sense “nature-related” phenomena. They are very rarely human, let alone animate, and, hence, would not qualify as truly agentive. We have found some occurrences with human subjects, and they raise the question of what is meant by “direct causation”. We will not have time to address this issue here, which is widely discussed in J. Collins, N. Hall, and L. A. Paul (2004). What is interesting to note is that negative consequences are more often expressed via a transitive verb than positive ones. Although the sunshine bloomed the flowers is not felicitous, the heat wilted the flowers is. McKoon and Macfarland’s studies show that there is a high probability of transitive uses for corrode, erode, a lower one for rot, rust, wilt and wither, and no occurrence for bloom or blossom, except for a shrub blooms white flowers in summer. The reason might be that the flowers here are part and parcel of the shrub. The question of whether the two types of causative change-of-state events, external and internal, are poles apart or not can interestingly be addressed in French. While the causative alternation is a widespread linguistic phenomenon (F. Schäfer 2009), languages differ in the way the alternation is realised. It is indeed the existence in French of the FAIRE faire construction that suggests that the linguistic expression of the domain of causality takes into account various parameters, and involves a variety of causers and intruments (S. DeLancey 1985). Most French verbs do not accept the English alternation examplified by John bounced the ball/the ball bounced, and resort instead to FAIRE: John a fait rebondir la ball (*John a rebondi la balle)/la balle a rebondi, and the examples given in (7) and (8) raise the question of their translation into French. The issue of whether a verb can or cannot appear in a transitive and intransitive construction is, to a certain extent, the core of the problem, and we think that a better understanding of the semantics of the construction will help us to explain why the atypical constructions we are interested in are linguistically possible.

Atypical argument structures in French 187



1.2

The expression of causation in French

As regards the translations of (7) and (8) we can indeed hesitate between: (7′) la forte chaleur a fané les fleurs (same structure as the English one), (the intence heat wilted the flowers), or (7′′) (la forte chaleur a FAIT faner les fleurs) the intense heat-has made-wilt-the flowers (8′) la sévérité du gel a flétri les tulipes (same structure as the English one), (the severe frost withered the tulips), or (8′′) la sévérité du gel a FAIT flétrir les tulipes the severe frost-has made-wither-the tulips, (the severe frost caused the tulips to wither)

In (7′) and (8′) the verbs faner and flétrir function as transitive/causative verbs. In (7′′) and (8′′’) they function as intransitive/unaccusative, with the following paraphrasis: la forte chaleur a fait que les fleurs ont fâné, et la sévérité du gel a fait que les tulipes ont flétri.8 In the FAIRE faire construction the linear order is deceptive, as the Noun Phrases (the flowers/the tulips here) following the lexical verbs (fâner/ flétrir) are their semantic subjects, and not their objects. What is the semantic role of FAIRE in a causative construction? We suggest that the subject of FAIRE is conceptualized as the Agent/Actor of the change of state, but the change of state cannot take place if the element undergoing the change does not possess the inherent characteristics enabling this change of state. If I say Jean a fait bouillir de l’eau, I mean that Jean put some water in the pan, turned on the gas under the pan, and waited for the water to boil. He could even leave the kitchen, since the process he initiated by turning the gas on will continue of its own accord, without any help from Jean. We have first the expression of an “external” causation, but if the water was not prone to heat and to come to the boil after a while, under these particular circumstances, nothing would have happened. The boiling occurs because in a hot environment the water in the pan can but only gradually reach 100 °C. The successful change of state depends on Jean (External causation) and on the water itself (Internal causation). If Jean had put a stone in the pan instead of water, the stone would not have boiled, since stones require a much higher temperature to become liquid, let alone gaseous. What matters for the change of state to occur is the existence of what we call the “adequacy parameter”: an Agent/Actor may impart some kind of physical or 8. We suggest the following rendering of the French sentence: the intense heat acted in such a way that the flowers wilted.

188 Geneviève Girard-Gillet

mental energy upon an entity, but the realization of the change will entirely depend on the existence of the adequate characteristic in the entity concerned.This point will be taken up when we discuss metaphorical interpretations. 2. Two descriptive modes of causative events We posit that there are two linguistic manners of describing causative events, depending on whether it is only the Causer who is foregounded/profiled (External Causation, ExtC), or whether the Causer and the Causee are both considered as possessing some kind of active/instrumental influence9 (External Causation, ExtC, plus Internal Causation IntC). By “active/instrumental influence” we do not mean that the Causee is the willful agent of the change of state; we only mean that its inherent characteristics work toward the end desired by the Causer. Lemmens (1998) considers that Mary is the Agent of her own death in John starved Mary, but we prefer to say that Mary’s human characteristics are the internal causation of it.10 If Mary is placed in an environment where she receives no food, she will gradually lose all her strength, her organism will weaken, and death will occur after a while. Human beings require food to survive, and this inherent characteristic is the internal cause of the starvation. But no agentivity can be ascribed to Mary. A distinction must be made between “Mary” as a being endowned with decision powers and “Mary” as a physiological entity obeying the laws of nature.11 With inanimate entities, French speakers can resort to two constructions, depending on the existence or not of the adequate Internal Causation:

9. The term is not really adequate, because if the role of the Causer is an agentive role, that of the Causee is not actually agentive, even if the physiological properties of the latter will play a role in the causative process. The term “Actor” used by R. Van Valin would be misleading for the Causee, but the term “Undergoer” which sounds better does not capture the Causee’s contribution to the process, although it is an entirely reluctant contribution. 10. The French translation is: John a fait mourir de faim Mary, with the FAIRE faire construction. 11. One reviewer suggested that “the adequacy parameter cannot be hastily equated with internal cause”, and argued that “Mary will die if nothing extra happens, e.g. if she does not receive food or simply as time goes by, but the water will not boil if nothing extra happens, in particular if there is no source of heating under the pan. It is not in the nature of water to become hot as time goes by”. We quite agree with him/her, but the question we address here is the linguistic expression of the circumstances that might alter the natural course of events. Since Levin (2009) argued that the changes “may be externally triggered, facilitated or regulated by certain natural forces or other environmental phenomena” we think that it is worthwhile to understand how these forces/phenomena are linguistically encoded in French.

Atypical argument structures in French 189



(9) Faire tremper les raisins dans du gin pendant une heure (ExtC + IntC) make-soak-the raisins in gin (soak the raisins in gin for one hour)12 (10) tremper le croissant dans le café (ExtC only, or weak IntC) (Dip the croissant in the coffee) (11) faire ouvrir les moules sur le feu (ExtC + IntC) make-open-the mussels (steam the mussels open) (12) ouvrir les huitres (ExtC only) (open the oysters)

In (9) and (11) the raisins and the mussels will undergo a change of state, due to their inherent properties, and to the environment, the liquid for the raisins, or the heat for the mussels. In (10) and (12)13 the focus is on the Causer of the process, and the entities do not contribute to their change of state. We now need to examine these data in detail. 2.1

Transitive or intransitive uses

In J. Dubois and F. Dubois-Charlier’s dictionary, Les Verbes français (2001) transitive/causative structures are given for: (13) le cuisiner mijote le râgout (the cook simmers the stew) (14) le cuisinier marine les harengs dans l’huile (the cook marinates the herrings in oil)

and The Grand Robert, taking up examples from Furetière, mentions: (15) le boulanger cuit deux ou trois fois par jour (the baker cooks twice or three times a day)

whereas in recipies, in cooking and gardening blogs, the preferred construction in French is the FAIRE faire construction:

12. There also exists an opposition between a long process, during which the raisins will absorb the liquid, and a short process. We shall not have time to discuss this parameter here. 13. The translations into English with a change of verbs suggest a conceptual difference.

190 Geneviève Girard-Gillet

(16) FAIRE cuire les pâtes n’est pas difficile (www.barilla.fr/le_plaisir_d…/ Comment_faire_cuire_les_pates.htm) make-cook-the pasta isn’t difficult14 (cook pasta is not difficult) (17) En utilisant le four principal au ralenti, vous pourrez FAIRE mijoter les blanquettes (http://www.rayburnfrance.fr/92.htm) you-can-make-simmer-the blanquettes (you can simmer the blanquettes) (18) tu ne dois pas donner d’engrais trop souvent, ça FAIT faner les boutons (http://forum.aufeminin.com/forum/loisirs10/__f10576_loisirs10-Orchideeboutons-qui-fanent.html) do not give too much fertilizer too often, it-makes-wilt-the buds (it wilts the buds)

How can we account for the two syntactic constructions? Those are not atypic ones. 2.1.1 The domain of specialists The examples given by J. Dubois and F. Dubois-Charlier have all specialists as the causative subjects of the verbs: (19) Le cultivateur vieillit son vin en cave (the farmer ages his wine in a cellar) (20) le cuisinier gratine la purée (the cook fries the mashed potatoes)

to which can be added: (21) le menuisier vieillit les commodes pour les vendre comme des antiquités (the joiner distresses wardrobes to sell them as vintage furniture) (22) Par définition, patiner un meuble signifie le vieillir de façon artificielle (http://www.teva.fr/actualite/488256-comment-patiner-meuble.html#ixzz28 DkcscPF) (patinating a piece of furniture means ageing it artificially)

but a quick search on Google – even if Google is not entirely reliable – for “vieillir le vin”, “gratiner la purée” with an animate subject, yields faire vieillir (make-age) and faire gratiner (make-fry) etc.

14. The interpretation of the French construction Faire cuire is: to act in such a way that the pasta will cook.

Atypical argument structures in French 191



(23) Voici les conditions de bonne conservation pour tous ceux qui FONT vieillir nos vins d’après les conseils des vignerons d’Anjou (books.google.fr/books?isbn= 2909757676) here are good preserving conditions for those who-make-age our wines (here are good perserving conditions for those who age our wines) (24) Ces vignerons récoltent la vendange, assemblent les cépages, FONT vieillir le vin et enfin le mettent en bouteille (www.vins-saint-emilion.com/plaisir-des-sens) these winegrowers-make-age-wine (25) Pour FAIRE vieillir votre vin dans les meilleures conditions, entreposez-le dans une cave avec une température ambiante stable entre 11° et 14° (www.dico-duvin.com/maturation-du-vin-ou-le-vieillissement/) to-make-age-your wine in the best conditions

The transitive construction seems to be used more felicitously when it is the instrument of the ageing that is taken into account: (26) Aucune lumière naturelle n’est admise dans une cave à vin car elle vieillit le vin trop rapidement. (http://www.emballage-bouteille-vin.fr/7-regles-a-respecterpour-une-cave-a-vin-de-qualite) (No natural light should be allowed in a wine cellar because it ages the wine too rapidly)

This distinction requires a closer analysis. 2.1.2 The laws of nature For Levin, inherent modifications are motivated by the internal properties of the entities concerned, even if the environment can play a role. This interaction between internal characteristics and favourable environmental data seems to be linguistically relevant in French as two different constructions exist. The transitive construction is preferred when the instrument/agent15 is a natural force, even if (7′′) and (8′′) are possible, though a bit awkward, as the following examples suggest: (27) la neige a brûlé l’herbe / *?la neige a FAIT brûler l’herbe the snow-has-burnt-the grass / *the snow-has made-burn-the grass (28) les feuilles sont abîmées. Le soleil a flétri vos salades. Ne les jetez pas (www.udir. org/…1/L/La-pa-moyen-déride-in-sousou-flétri.-2394/) (the leaves are wilted. The sun withered your lettuces; don’t throw them away) 15. The distinction between “agent” and “instrument” does not seem relevant for our present discussion, but should be considered in a more complete analysis. Koenig et al. (2007) propose that sentences with instruments should have a “double causative” structure.

192 Geneviève Girard-Gillet

(29) le soleil a fané les fleurs, flétri les mousses (www.biblisem.net/meditat/ tiercaum.htm) (the sun wilted the flowers, withered the mosses) (30) il ne faut surtout pas arroser pendant la journée lorsque les plantes sont en plein soleil: la réverbération risque de les brûler (www.quintonic.fr/loisirs/ magazine/…/astuces-pour-un-arrosage-efficace) (you mustn’t water the plants when they’re in the sunlight: the reverberation might burn them) (31) Le sel cuira la chair du saumon (http://www.mangerbouger.fr/bien-manger/ la-cuisine/le-dictionnaire-de-la-cuisine) (the salt will cook the flesh of the salmon)

The examples seen so far show that when verbs such as bouillir, mijoter, mariner, cuire, vieillir, faner, flétrir are used with FAIRE, the Agent initiates/triggers the change of state and the change of state follows inexorably, due to the inherent property of the entity acted upon:16 Jean a FAIT bouillir de l’eau. When they are used as transitive verbs, what is highlighted is the initiator of the change of state and the inherent properties are backgrounded, which explains why the construction is preferred with specialists of a domain – they are powerful enough to act upon the entities (examples (13)–(15) and (19)–(22)) – and as such are considered as the “direct cause” of the change of state, without any other influence; the construction is also preferred with the laws of nature (examples (26)–(31)), which are actually the direct causers. The same syntactic distinction is resorted to in order to oppose voluntary and unvoluntary processes: (32) il a brûlé le livre où leur histoire avait été écrite (he burnt the book in which their story had been written) (33) il a fait brûler le fond de la casserole he-has-made-burn-the bottom of the pan (he burnt the bottom of the pan)

In (32) the subject is responsible for the change of state: he planned to burn the book. The External Causation is foregrounded, and meant to be the sole source of the event, even if books are prone to burn in the adequate environment. 16. R. Forest (1988) claimed that these verbs expressed the realisation of a potentiality inscribed in the entity undergoing the change. “Ces verbes expriment la réalisation d’une virtualité inscrite dans l’entité participant au changement”. We agree with his hypothesis, and we shall develop it here in order to understand the atypic examples.

Atypical argument structures in French 193



In (33) the burning of the pan took place because the person busy cooking left the kitchen for a couple of minutes. Here the sentence expresses the irresponsibility of the cook, and the bottom of the pan burnt because it remained too long in a hot environment. Another opposition can be found: (34) Une adolescence brûle une lettre; 15 hectares partent en fumée (www. laprovence.com/…/une-adolescente-brule-une-lettre-15-hectare.) (a teenager burns a letter; 35 acres go up in flames) (35) Ce fait remet en cause la théorie selon laquelle Néron aurait fait brûler des quartiers pour y reconstruire un immense palais (realite-histoire.over-blog. com/article-22561209.html) the theory-that Nero had-made-burn-vast areas in Rome (This fact questions the theory that Nero burnt vast areas in Rome in order to build a huge palace)

In (34) it is the External Causation that is focused on, because of the voluntary action, and because of the short duration of the process: a letter does not take long to burn. In (35) Rome took longer to burn than a letter; hence both types of causation are expressed in French: Nero is the External Causer, and the agentive subject of FAIT, whereas Rome is the Causee, and the lexical subject of burn. Before arguing why this distinction might pave the way to the use of atypical uses, we shall consider how the two constructions differ as regards to literal or metaphorical interpretations. 2.2

Metaphorical interpretations

We can contrast a literal meaning and a metaphorical one, because, interestingly enough, they seem to correlate with the two syntactic constructions: (36) FAIRE mijoter le râgout pendant une heure → Le râgout mijote (ExtC + IntC) make-simmer-the stew (simmer the stew for one hour) → the stew simmers (37) Il mijote un mauvais coup → le mauvais coup ne mijote pas (ExtC only) he-simmers-some mischief (he’s cooking up some mischief) → the mischief does not cook

Something will happen but nothing simmers, * ?le mauvais coup mijote, because the “adequacy parameter” which is necessary for the simmering change of

194 Geneviève Girard-Gillet

state to occur cannot apply here, as mischief is not endowned with a simmering characteristic. (38) Ca FAIT ramollir les biscuits si on ne les met pas dans une boite17 → Les biscuits ramollissent/ se ramollissent (ExtC + IntC) it-makes-soften-the biscuits (the biscuits get soft if they’re not kept in a box) (39) Trop de publicité ramollit le cerveau → *?le cerveau ramollit/se ramollit (ExtC)18 (too many advertisements soften the brain) (40) Si tu arroses trop tes salades, tu vas les FAIRE pourrir → Les salades pourrissent. (ExtC + IntC) you will-them-make-rot (you will cause them to rot) → the lettuces will rot (41) l’alcoolisme a pourri la vie de ma sœur → *la vie a pourri (ExtC) (alcoholism spoilt/ruined my sister’s life)

With FAIRE, the sentence denotes external Causation through the subject of FAIRE + Internal Causation via the inherent characteristics of the subject of the lexical verb. With the transitive construction, the sentence refers to an External Causation, but the change of state cannot generally be expressed by the same verb, since the External Causation could not literally trigger the realisation of a non-existing potential property in the Causee. And not all occurrences express a change of state: (42) Ce chapeau vieillit Marie this hat-ages-Marie (this hat makes Marie look older)

Marie does not get older. Only her appearance is modified by the hat. Marie does possess the ageing faculty – inherent characteristic – but the hat does not impart any energy on the ageing characteristic for it to develop. Whereas with: (43) le stress chronique peut FAIRE vieillir prématurément (chronic stress can make one age prematurally)

the speaker implies that the person will actually age (ExtC + IntC), because the ageing process can be activated by worries. A possible paraphrasis is one can age 17. We wish to thank M. Delmas for this example. 18. It can of course be understood metaphorically, but the lack of the “adequacy parameter” still operates.

Atypical argument structures in French 195



prematurely from too much stress, with the PP from too much stress indicating the cause. A similar paraphrasis is impossible with (42) *Marie aged from the hat. (44) C’est sa barbe qui le vieillit sinon pour son âge ça va (www.youtube.com/all_ comments?v=aLadS82rUfc&page=1) it is his beard-that-him-ages (It’s his beard that makes him look older)

Various parameters are responsible for each use. We have only mentioned a few, but they already seem relevant to help us to tackle the phenomenon of atypic uses. 3. Atypic uses Our hypothesis is that the atypic examples given in (1) to (3) can be explained by the distinction between the two kinds of causation (ExtC and IntC). We shall begin our discussion with example (2), repeated here: (45) les poux, les lentes, ça les explose

The advertisement for this specific cream/shampoo says that by using it, all lice and nits will die. It does not say that lice and nits will explode under its effect, because lice and nits do not possess an “exploding” characteristic, and consequently no internal exploding virtuality will be activated by the cream. The slogan only means that they will be destroyed. Similarly to the metaphorical examples, the interpretation must not be literal. But what differs is that a change of state will actually take place, and this change of state will be final. The construction is meant to highlight the destructive power of the shampoo. This use must be contrasted with: (46) il a fait exploser la bombe dans le bus (example (5))

where the External Causer pulls the trigger, and the bomb, which possesses the exploding quality (IntC) necessary for the change-of-state process to start, does explode: (47) IL A FAIT [la bombe a explosé]

Pinker (2008: 81) notes that some languages have various ways of describing a causative event, and that “whenever a language contains more than one of these services, it reserves the more concise one for more direct causation and the more prolix one for less direct causation. It’s as if the morphemes were laid out like a

196 Geneviève Girard-Gillet

little diagram depicting the links in the causal chain, and the chains with fewer links were conveyed by fewer morphemes”.19 This quotation supports our claim to a certain extent, but does not explain the syntactic construction of metaphorical and atypic sentences, since (47) cannot be felicitously replaced by: (47′) ?*il a explosé la bombe

even if the speaker wished to highlight the responsibility of the terrorist. Our hypothesis can be expressed as follows: atypicity is only possible when there is no Internal Causation at work in the change-of-state process that is triggered. The necessary adequate parameter does not react. There may be a change of state but it cannot be literally described by the lexical verb which is used: ?les poux ont explosé.20

The following examples are all atypic uses of the exploser construction: (48) L’incroyable Usain Bolt a encore frappé. Il a explosé le record du monde du 200 m (http://www.purepeople.com/article/l-incroyable-usain-bolt-a-encorefrappé) (he exploded the world record) (49) Tony Martin n’est pas champion du Monde pour rien: il a explosé le contre la montre de Villard (www.voixdelain.fr/…/cyclisme-tony-martin-a-lheure-auparc-des-oiseaux (he exploded the time trial against the clock) (50) On a explosé nos chiffres d’inscription, près de 1100 étudiants (we exploded the number of registered students)

The occurrences with other verbs can be explained similarly: (51) Ils l’ont suicidé, comme bien d’autres (example (1))

The man who “was suicided” did not have any wish to kill himself. He was not suicidal, which means that no psychological characteristics (qualia for J. Pustejovsky) pushed him to commit suicide. In other words, no Internal Causation existed in

19. We wish to thank C. Delmas for this reference. 20. The sentence can be understood metaphorically with an exaggeration undertone.



Atypical argument structures in French 197

him to activate the pressure of a Causer, and the interpretation is not that he killed himself, but that he was murdered (ExtC, but no IntC).21 (52) ces russes si tolérants l’ont suicidé en maquillant sa mort en cholera (http://www.forumfr.com/sujet505162-post60-russie-une-loi-contrela-propagande-homosexuelle.html) (they suicided him, faking up his death as a cholera case)

But when the pressure is exterted on a psychologically weak person, he/she may commit suicide as a consequence (ExtC + IntC), because the adequate suicidal parameter reacts to the suicide-suggesting Causer: (53) Il avait la force de pouvoir faire se suicider, s’il le voulait, toutes les personnes qui lui étaient inféodées (books.google.fr/books?isbn=2748122399) the power-to-make-commit suicide-all the people (he had the power to push anyone hooked on him to commit suicide)

In ils l’ont démissionné, it is understood that the person had no intention of resigning, was not suffering from burnout, was not planning to change jobs, but that he was given the sack (ExtC, but no IntC). The pragmatic use of such atypic constructions is to highlight a dramatic event, by paradoxically hiding the real situation. In the case of suicider the construction avoids an accusation of murder in the description of what actually happened, while at the same time giving the addressee a linguistic tool to make him/ her understand that the man did not actually commit suicide. Its use is not ambiguous at all, as regards the situation of the “undergoer”22 and it does not require any shared knowledge between the speaker and the addressee as to the relation between the Causer and the victim of the change of state.23 It is the construction itself that yields the right interpretation. The change of state that takes place is very often a change for the worst, even a final one; with suicider, the undergoer is dead, destroyed. With démissioner, the change may seem less detrimental, but in a world undergoing a severe work crisis losing one’s job may lead to a situation tantamount to social death. The con-

21. An anonymous reviewer suggested that no shared knowledge is necessary to understand (51). We agree. It is the construction itself that leads any adressee to understand that the man did not commit suicide. Otherwise the speaker would have said that he committed suicide. 22. This term is only adopted to refer to the entity acted upon, and must not be mistaken for the term used by R. Van Valin. 23. We wish to thank an anonymous reviewer for pointing out that the notion of shared knowledge, adopted in a previous version, was not relevant.

198 Geneviève Girard-Gillet

struction can also encode a limit which is felt to be the absolute limit, positive or negative: the world record cannot be expected to be broken again, and through the construction the journalist expresses a high degree of enthusiasm. Conversely, the situation can be felt as a disaster, if the number of students continues to grow.

Conclusion The two syntactic constructions existing in French to express change-of-state events seem to correlate with the semantic data the speaker wishes to focus on. Either the Causer is thought to be the only active entity, and the entity changing states is backgrounded and appears as the direct object of the causative verb: le cuisinier marine le veau (the cook marinates the veal), le soleil brûle les fleurs (the sun burns the flowers) or the Causer and the Causee have a role to play in the process, and a Faire faire construction is adopted to differentiate the two types of causation: Jean FAIT bouillir le lait (Jean-makes-boil-the milk). As can be understood from our examples, this construction cannot be adopted when no Internal Causation exists that can be triggered – the adequacy parameter – and, quite logically, metaphorical interpretations cannot be expressed via a FAIRE faire construction, all the more so when no actual change of state is perceived: *ce chapeau te FAIT vieillir. The only felicitous construction is: ce chapeau te vieillit (this hat-you-ages) which can be translated by this hat makes you look older. The same kind of conceptualisation seems to be at work in atypic uses, since it is the lack of the Internal Causation properties necessary for the change of state to occur that makes them linguistically possible. A semantic difference exists though: there IS a change of state and usually a dramatic, unexpected one, but it is not the change of state that can be deduced from the semantic features of the verb, whether it be suicider, exploser or démissioner. We have only dealt with three verbs, because these three verbs are often used in this atypic construction. But more data will have to be considered with a view to supporting our hypothesis more clearly in future. It will also be necessary to understand the reasons why it is often unexpected, negative consequences that are expressed with this “atypic” construction, just as detrimental consequences, according to McKoon and Macfarland, are more prone than positive ones to accept a transitive construction.

Atypical argument structures in French 199



Table 1 summarizes our conclusions: External Causation + Internal Causation ExtC triggers the onset of IntC

Result/Consequences

(a) Il a fait bouillir l’eau (he-made-boil-the water); he boiled the water. (b) il a fait pourrir les salades (he-made-rot-the lettuces); he caused the lettuces to rot. (c) Il a fait travailler Marie (he-made-work-Marie);24 he made Marie work. (d) les soucis ont fait vieillir Marie (worries-made-age-Marie); worries caused her to age. (e) lls ont fait mourir de faim Marie (they made-starve Marie); they starved Marie to death. (f) ils ont fait se suicider Marie (they made-commit-suicide-Marie); they made Marie commit suicide.

(a′) l’eau a bouilli (the water boiled)

External Causation / NO “adequate” Internal Causation ExtC cannot trigger the onset of IntC

Results: NO result, or a result different from the suggested result → metaphorical or atypic interpretations

(g) ce chapeau vieillit Marie (this hat-ages Marie)

(g′) Marie n’a pas vieilli (Marie didn’t get any older) → appearance. (h′) la commode n’a pas vieilli (the wardrobe is not older) → appearance. (i′) les poux n’explosent pas (the lice do not explode → ils meurent (they die). (j′) le ministre ne s’est pas suicidé (he did not commit suicide) → il a été tué (he was killed). (k′) il n’a pas démissionné (he did not resign) → il a été licencié (he was fired).

(h) l’ébéniste a vieilli la commode (the joiner distressed the wardrobe) (i) La pommade explose les poux (the cream explodes the lice) (j) Ils ont suicidé le ministre (they suicided the Minister) (k) ils l’ont démissionné (they resigned him)

(b′) les salades ont pourri (the lettuces rotted) (c′) Marie a travaillé (Marie worked) (d′) Marie a vieilli (Marie aged) (e′) Marie est morte de faim (Marie starved to death) (f′) Marie s’est suicidée (Marie committed suicide)

24. We did not have time to take into account the construction when two Agents are concerned: the Causer is an Agent, and the Causee is also an Agent. This interpretation is one of coercion, and is semantically very close to he forced Mary to work.

200 Geneviève Girard-Gillet

References Abeillé, A., D. Godard & P. Miller. 1997. Les causatives en français, un cas de compétition syntaxique. In F. Gadet (éd.), Langue français, La Variation syntaxique, 62–74. Paris: Larousse. Albrespit, J. 2003. Sujet et agentivité en anglais. In J.-M. Merle (éd.), Le Sujet, 125–136. Paris: Ophrys. Alexiadou, A. & E. Anagnostopoulou. 2003. Voice morphology in the causative-inchoative alternation: Evidence for a non-unified structural analysis of unaccusatives. In A. Alexiadou, E. Anagnostopoulou, & M. Everaert (eds.), The unaccusative puzzle: Explorations of the syntax-­lexicon interface, 114–136. Oxford: Oxford University Press. Alexiadou, A., E. Anagnostopoulou, & F. Schäfer. 2006. The properties of anti-causatives crosslinguistically. In M. Frascarelli (ed.), Phases of interpretation, 187–211. Berlin: Mouton de Gruyter. doi: 10.1515/9783110197723.4.187 Ben Salah Tlili, I. 2007. Contribution à l’étude des verbes symétriques en français. contemporain. In J. François & A. Brahim (dir.), Morphosyntaxe et sémantique du verbe, Cahiers du CRISCO 23. 16–37. Boons J.P., A. Guillet & C. Leclère.1976. La structure des phrases simples du français: constructions intransitives. Genève: Droz. Chatti, S. 2009. Sémantique de la causation analytique. PhD Dissertation, Université Sorbonne Nouvelle Paris 3. Collins, J., N. Hall & L. A. Paul (eds.). 2004. Causation and Counterfactuals Cambridge, Mass.: The MIT Press. Cotte, P. 1993. Ces mouvements qui font signe. In Motivation et Iconicité Faits de langues, 129– 136. Paris: PUF. doi: 10.3406/flang.1993.1044 Croft, W. 1994. The semantics of subjecthood. In M. Yaguello (éd.), Subjecthood and subjectivity, the status of the subject in linguistic theory, 29–75. Paris: Ophrys. DeLancey, S. 1984. Notes on agentivity and causation. Studies in Language 8. 181–213.  doi:  10.1075/sl.8.2.05del

Delechelle, G. 1989. L’expression de la cause en anglais contemporain. PhD Dissertation, Université Sorbonne Nouvelle Paris 3. Delmas, C. 2001. La sémantique propose, la syntaxe dispose. QWERTY, 11, Pau, P.U.Pau, 247–255. Delmas, C. 2004. Fragments d’un discours culinaire. In L. Gournay & J.-M. Merle (dir.), Contrast, Mélanges offerts à J. Guillemin-Flescher, 25–34. Paris: Ophrys. Delmas, C. 2006. Variation autour de “feel”. In F. Fredet & A.-M. Laurian (dir.), Linguistique contrastive, linguistique appliquée, sociolinguistique, 75–81. Peter Lang. Dubois, J. & F. Dubois-Charlier. 1997. Les Verbes français. Paris: Larousse. Dubois J. & F. Dubois-Charlier. 2011. Les Verbes français. rali.iro.umontreal.ca/Dubois. Forest R. 1988. Sémantisme entéléchique et affinité descriptive: pour une ré-analyse des verbes symétriques ou neutres en français. Bulletin de la société linguistique de Paris 83(1). 137–162. doi: 10.2143/BSL.83.1.2013659 François, J., G. Rauh & J. Broschart. 1994. Les relations actancielles, sémantique, syntaxe, morphologie. Langages 28. 113–127. doi: 10.3406/lgge.1994.1667 Girard-Gillet, G. 1999. Make/Lassen/Faire, Semantic delimitation and syntactic construction. In Contrast, Comparison and Communication (CIEREC, Travaux XCVI), 49–60. St Etienne.



Atypical argument structures in French 201

Girard-Gillet, G. 2006. Structuration lexicale et structuration syntaxique dans l’expression de la causation. In C. Delmas (éd.), Complétude, Cognition, construction linguistique, 47–61. Paris: Presses de la Sorbonne Nouvelle. Girardin, C. 2003. Polysémie des verbes et traitement argumental dans les dictionnaires du XVIIe siècle: le verbe cuire. Polysémie et Polylexicalité, Syntaxe et Sémantique 5. 131–142.  doi:  10.3917/ss.005.0131

Guéron, J. 2008. Visible effects, hidden causes. In G. Girard-Gillet (éd.), L’envers du décor, études de linguistique anglaise, 158–179. Université d’Avignon. Haspelmath, M. 1993. More on typology of inchoative/causative verb alternations. In B. Comry & M. Polinsky (eds.), Causatives and transitivity, 87–120. Amsterdam: John Benjamins.  doi:  10.1075/slcs.23.05has

Holmes, J. 1999. The syntax and Semantics of causative verbs. UCPWPL 11. 323–348. Koenig, J.-P., G. Mauner, B. Bienvenue, & K. Conklin. 2007. What with?: The anatomy of a (proto)-role. Ms. University of Buffalo. Langacker, R. W. 1987. Foundations of cognitive grammar, vol. 1. Stanford: Stanford U.P. Lazard, G. 1994. L’Actance. Paris: Presses Universitaires de France. Lemmens, M. 1998. Lexical perspectives on transitiviy and ergativity: Causative constructions in English. Amsterdam: John Benjamins. doi: 10.1075/cilt.166 Lemmens, M. 2005. Des patients qui ne veulent pas disparaître, Un complément à l’analyse de Goldberg des constructions causatives sans objet. In Six études de linguistique, Occasional Papers, 1, 117–138. Rouen: CERCLES. Levin, B. 1993. English verb classes and alternations. A preliminary investigation. Chicago and London: The University of Chicago Press. Levin, B. & M. Rappaport-Hovav. 1995. Unaccusativity: At the syntax-lexical semantic interface. Cambridge, Mass.: MIT Press. Levin, B. 2009. Further explorations of the landscape of causation: Comments on Alexiadou and Anagnostopoulou. MIT Working Papers, 239–266. Cambridge: MIT. McKoon, G. & T. Macfarland. 2000. External and internally caused changed of state verbs. Language 76. 833–856. doi: 10.2307/417201 McKoon, G. & T. Macfarland. 2002. Event templates in the lexical representations of verbs. Cognitive Psychology 45. 1–44. doi: 10.1016/S0010-0285(02)00004-X Pasero, R., P. Sabatier & M.-H. Stefanini. 2010. Syntaxe et sémantique formelle des constructions verbales exprimant des relations symétriques. In D. Leeman & P. Sabatier (dir.), Empirie, Théorie, Exploitation: le travail de Jean Dubois sur les verbes français, Langages, 179–180, Paris: Larousse, pp. 115–141. Pinker, S. 2008. The stuff of thought. London: Penguin. Pustejovsky, J. 1995. The generative lexicon. Cambridge, Mass.: MIT Press. Samardzic, T. & P. Merlo. 2012. The meaning of lexical causatives in cross-linguistic variation. LILT 7(12). CSLI Publications. Schäfer, F. 2009. The causative alternation. In Language and linguistics compass, vol. 3, 641–681. Blackwell Publishing. doi: 10.1111/j.1749-818X.2009.00127.x Talmy, L. 2000. Toward a cognitive semantics, vol. 2. Cambridge, Mass.: MIT Press. Tesnière, L. (éd.). 1958. Eléments de syntaxe structural. Paris: Klincksieck. 5th edition, 1988. Van Valin R. D. & R. J. LaPolla. 1997. Structure, meaning and function. Cambridge: CUP.  doi:  10.1017/CBO9781139166799

Zribi-Hertz, A. 1987. La réflexivité ergative en français moderne. Le Français moderne 55. 23–54.

Split intransitivity in Lamaholot (East Flores, Indonesia)* Philippe Grangé

Université de la Rochelle

Split intransitivity has been identified in many languages, including some Eastern Indonesia languages, but the East Adonara Lamaholot (Eastern Indonesia) may be considered as uncommon by the fact it displays all kinds of Split Intransitivity features that have been separately described for a series of languages. This language displays a complex split intransitivity, involving a Split-S (lexically fixed alignment) and a Fluid-S (fluid alignment) triggered by a series of contexts, which can accounted for by Proto-Roles properties as defined by Dowty (1991). The subject (S) is either seen as “agent-like” (SA), unmarked, or as “patient-­like” (SP), thus marked by verbal agreement. The Split-S involves verbs expressing feeling, sentience, biological functions or motion verbs, which must agree with their subject. The properties ‘± Control’ and ‘± Affected’ seem more crucial than ‘± Volition’ for Fluid-S. Aspect features can overlap this Split-S; for instance, motion verbs (excluding displacement) become telic when complemented by a locative prepositional phrase (PP). Perfect aspect is marked by the same verbal agreement as SP, on stative verbs, nouns used as a predicate, or displacement verbs. It appears that in Lamaholot dialects westward from Adonara (Eastern tip of Flores) and eastward (on Lembata Island), the split intransitivity systems are highly eroded, and their remains became lexicalized.

* A preliminary version of this paper was presented at the 11·ICAL (Eleventh International Conference on Austronesian Linguistics) Aussois, 22–26 June 2009. Many thanks to Hein Steinhauer and Marian Klamer for their remarks on a previous version of this paper. I am very grateful to the reviewers, Michael Ewing and Peter Arkadiev, for their appropriate and constructive suggestions and for their proposed corrections to improve the paper. doi 10.1075/lis.33.09gra © 2016 John Benjamins Publishing Company

204 Philippe Grangé

Introduction A special type of semantic alignment, split intransitivity, is a common feature of many East Nusantara languages spoken in the Lesser Sunda Islands (East Indonesia). Split intransitivity triggers a morphological distinction between “agent-like” and “patient-like” subjects in monovalent sentences (this working definition will be refined below). Split intransitivity is not, however, an essential feature of the East Nusantara languages, but only a tendency within them. Moreover, split intransitivity can be found elsewhere in Indonesia, for instance in Acehnese (northern tip of Sumatra), see Van Valin (1990). Klamer (2008: 249) surveyed nine East Nusantara languages, both Autronesian and non-Austronesian, that display a semantic alignment system. Lamaholot, not included in Klamer’s survey, is an Austronesian language spoken at the eastern tip of Flores island, in three nearby islands (Solor, Adonara, Lembata) and in a few coastal pockets further east in the Alor archipelago. There are approximately 200,000 Lamaholot speakers in a chain of highly diversified dialects.1 The island of Adonara occupies a central position, geographically speaking. Its East Adonara Dialect also appears to be the most conservative. Split intransitivity has been identified in other languages throughout the world, but East Adonara Lamaholot may be considered as non-canonical by the fact it displays all the Split Intransitivity features that have been separately described for various languages. It involves a lexically fixed alignment (Split-S) beside a fluid alignment (Fluid-S) triggered by a series of contexts, including aspect. In this paper, I argue that the intransitive verbal agreement in Lamaholot rests on a broad split intransitivity system. After a short definition of split intransitivity, second section presents the transitive and intransitive basic morphology and the characteristics of agreement morphology. Split-S and Fluid-S are dealt with in third and fourth sections. Section five examines aspectual features. Finally, in section six, typological issues within the Lamaholot dialects chain are discussed, and some diachronic hypotheses are proposed.

1. Three Lamaholot dialects have been described respectively by Keraf (1978), Fernandez (1996), Pampus (2001) and Nishiyama & Kelen (2007). Nagaya (2009a, b) is surveying a Lamaholot dialect in the eastern tip of Flores. Our research field site is situated in the eastern part of the Adonara island. This densely inhabited region has two dialects: ‘Dulhi’ and ‘Kiwangona’ (Keraf 1978), that I regard as variants of the same dialect, and tentatively label ‘East Adonara dialect’. Within this dialect area, the few variations are mostly of a phonological nature. The East Adonara dialect is geographically central in the Lamaholot dialect chain.

Split intransitivity in Lamaholot 205



1. Split intransitivity: A definition In some languages, semantic alignment appears when “verbal lexical semantics affects word order, case marking or agreement” (Wichmann 2008: 7). According to their meanings, the verbs can be classed into opposed semantic categories. This semantic opposition may trigger syntactic effects, see Donohue (2008). Split intransitivity is a special form of semantic alignment, which affects intransitive verbs, for instance in Lamaholot. Following Van Valin (1990: 222) and Dixon (1994: 70–110), if we name A the agent of a transitive structure, S the subject of an intransitive structure, and O the patient in a transitive structure,2 we may represent the marking of S in accusative, ergative and split intransitive structures in various languages as in the following figure. The identical hachured squares mean “same mark” (or same lack of mark): A

A

A

S

S

S

O

O

accusative structure

ergative structure

O split intransitivity structure

In a split intransitive system S will be marked as either O or A (or will remain unmarked if this is normally the case for A), which is represented above by an “overlap” zone. Marking S as O leads to the interpretation that the subject of the intransitive sentence is treated as similar to the P rather than to the A of the transitive sentence. Conversely, when S is marked as A (i.e., in Lamaholot, unmarked), playing an agent-like semantic role, it is labelled SA. For instance, the verb tudak ‘to stop’ may appear unmarked (SA) or marked (SP), example (1), this split being triggered by semantic features, as is argued below. (1) Go tudak. Go tudak -ek 1sg stop 1sg stop -1sg ‘I stop.’ ‘I am stopped’ (I’m stuck or hampered)

2. Glossing: agr: agreement; perf: perfective; pfct: perfect; impf: imperfective; det: determiner; gen: genitive; red: reduplication; sg: singular; pl: plural; prn: proper noun; A: subject of a transitive structure; O: object; P: patient; S: subject of an intransitive structure; SA: agentlike S; SP: patient-like S.

206 Philippe Grangé

Thus, in East Adonara Lamaholot, split intransitivity can be schematized more precisely by the following figure. The solid frame means “same mark” and the dotted frame means “unmarked”. A SA SP O

It is impractical to analyze semantically aligned languages as making reference to an S category, since S can appear as either SA or SP . As noted by Wichmann (2008: 4): “It makes little sense to posit such a category [of S], only to have it split up into the subcategories such as SA and SP .” A comprehensive discussion about “Semantic alignment systems: what’s what and what’s not” can be found in Donohue (2008). However, for clarity, I will use the SA and SP labels in this paper. Agent and Patient, although intuitive operating concepts, are not theoritical primitives. We will refer to Proto-roles as defined by Dowty (1991: 572): Contributing properties for the Agent Proto-Role: a. volitional involvement in the event or state b. sentience (and/or perception) c. causing an event or change of state in another participant d. movement (relative to the position of another participant) (e. exists independently of the event named by the verb) Contributing properties for the Patient Proto-Role: a. undergoes change of state b. incremental theme c. causally affected by another participant d. stationary relative to movement of another participant (e. does not exist independently of the event, or not at all)

Dowty’s Proto-Roles are highly relevant for a comprehensive account of split-intransitivity in East Adonara Lamaholot. All these properties, either separately or jointly, are linked to split intransitivity in this language. Proto‑Agent properties mentioned thereafter are Control, Sentience, Affectedness, Motion (corresponding respectively to Agent Proto-Role properties a, b, c, d above). This system is even more complex than in Acehnese, another Austronesian language, where volition is the main criteria accounting for split intransivity, as exposed by Van Valin (1990: 248–251), or in Lakhota, where Volition or Control over the process are crucial, see Mithun (1991: 514–518).



Split intransitivity in Lamaholot 207

Beside the Proto-Roles, split intransitivity in Lamaholot is also related to aspect, as was pointed out by Van Valin (1990) in Italian and Georgian, as well as Mithun (1991: 512–514) did for Guaraní. In Lamaholot, aspect marking (especially perfect aspect) can also trigger intransitive verbs agreement. In some languages split intransitivity is lexically defined, a feature that Dixon (1994) calls Split-S. In these languages, certain intransitive verbs will always align with O (subject is interpreted as SP), while other intransitive verbs necessarily align with A (subject is SA). Beside these semantic features, certain grammatical contexts, involving for instance mood or aspect, can entail in the same language morphological alternations similar to Split-S. In some other languages, marking (SA versus SP) is by no means predetermined, and rests on the choice made by the speaker. Dixon (1994: 79) calls this type of dual intransitivity “Fluid-S”: “In a Fluid-S language the A-type and O-type markings are allocated to intransitive clauses semantically, with each intransitive verb having the possibility of either choice, depending on the semantics of each particular context of use.” Creissels (2007: 3, note 3) writes that “the distinction between split intransitive and fluid intransitive systems is not always easy to establish, since even the most typical split intransitive languages commonly have a handful of intransitive verbs whose behavior shows fluctuations, and languages in which a significant proportion of intransitive verbs allows for fluidity in S alignment may also have verbs rigidly classified as SA or SP verbs.” This remark applies to the East Adonara Lamaholot dialect, where lexically fixed intransitive split (Split-S: some intransitive verbs are always marked, others never) and fluid intransitive split (Fluid-S: marking, or unmarking, is an expression device left to the choice of the speaker) coexist. 2. Agreement morphology in Lamaholot In Lamaholot, agreement is marked on the head of the verb phrase.3 This is consistent with Wichmann’s remark: “semantic alignment is normally associated with head-marking morphology” (Wichmann 2008: 15). In Lamaholot, the intransitive split is displayed by the unmarked status of SA, while SP is marked. This marker is an agreement suffix that cross-references the subject’s person. Prior to analyzing this type of semantic alignment itself, the morphology of verbs and agreement in Lamaholot must be presented.

3. Lamaholot verbs are mostly underived. Therefore, no morpholexical information is available within the verb stem to indicate its valency, agentivity or Aktionsart.

208 Philippe Grangé

2.1

Basic morphosyntax of transitive and intransitive clauses

Lamaholot is a SVO language, which unlike other Malayo-Polynesian languages owns a paradigm of object pronoun suffixes. The 3rd person object pronouns -na / -rona (3sg) and -ra / -wéra (3pl) refer to humans, while -ro (3sg/pl) and ‑wé (3pl) refer to animals or inanimate.4 The suffix -ro is generally reduced to ‑o when suffixing a consonant-final verb. (2) Go guti gelas. Go guti -ro. 1sg take glass 1sg take -3sg ‘I take a glass.’ ‘I take it.’

Lamaholot has no patient voice.5 If a phrase playing the role of a Patient is fronted, it must be cross-referenced with an object pronoun. (3) Tapo ni go gasik -wé. coconut this 1sg count -3pl ‘I count these coconuts’, lit. ‘These coconuts I count it.’

The ‘Agent-position’ must be filled; if the speaker ignores or conceals who acted, then by defaut it is raé (3pl). (4) Kayo nakuné raé poro -ro. wood that 3pl cut.down -3sg ‘That tree was cut’, lit. ‘The said wood, they cut it.’

In Lamaholot, some stative verbs can be used attributively (modifiying a head noun, i.e. as an adjective) or predicatively (as non-verbal predicates). When used attributively, these adjectives display an agreement with the NP head, like adjectives do in German or in French, but the one only attributive agreement morpheme is -n which seemingly originates from the 3sg possessive morpheme.6 If the stative verb ends with a vowel, this agreement triggers the nasalisation of its

4. Following orthographic standards in several languages of Indonesia, [e] is written é, [ə] is written e while the rare [ɛ] is written è. The glottal stop [ʔ] is written ’. 5. Nagaya (2010) argues that voice alternation in Lamaholot, although deprived of any morphological device, can be expressed by the rearrangement of the word order. I interpret these rearrangements as focalisation patterns (extraction and fronting), for instance fronting of the NP which corresponds to the role of a Patient. 6. See Note 18.

Split intransitivity in Lamaholot 209



final vowel, e.g. bura’ “be white” > bur’ã-n [burʔɑ̃ŋ] or [burʔɛ̃n] “white”,7 sometimes along with the epenthesis of a glottal stop /ʔ/ or a /h/ before the final syllable. The attributive agreement will also entail the epenthesis of a /ɛ̃/ if the root verb ends with a nasal consonant, for instance milan “be dirty” > milan-en [milanɛ̃n] “dirty”. The phonetic realisation of this agreement suffix is highly variable among the Lamaholot speakers. (5) Aho bur’ã -n ni hobo-hobo. dog white -agr det bark ‘This white dog barks.’

This attributive agreement internal to a NP will be barred if the adjective itself is modified by an adverb or reduplicated: (6) labu bura’-bura’ / bura’ tua / bura’ wa’ ni cloth white-red / white very / white not yet det ‘this really white / very white / not yet white cloth’

Agreement is also compulsory for stative verbs used as predicates,8 and will be barred under the same conditions, when it is modified by and adverb. (7) Wai ni pelat(é)i -n. / pelaté tua. water det be.hot -agr be.hot very ‘This water is hot / is very hot’

In sum, the position right to the stative verb must be filled, either by the agreement suffix -n or by a modifier (adverb or aspect marker). Surprisingly, the morphology of this suffix does not cross-reference the person: it is -n even for 1st and 2nd persons, for instance go bur’ã-n “I am white, fair-skinned” mo bur’ã-n “you are white”. The suffix ‑n may also be regarded as a demarcative marker, signalling the SN boundary. I will label it ‘default agreement suffix’. Only the indication of the perfect aspect must cross-reference the subject’s person, using the agreement suffix paradigm (1sg /k/; 2sg /o/; 3sg /a/ or /ʔ/, etc.) for instance mo bura-ko “you have become fair-skinned”, as will be shown in Section 5.2.

7. Keraf (1978: 109–114) labels the adjectives (kata sifat) deprived of agreement as “accentuated” (bertekanan), while the adjectives plus agreement suffix -n are labelled “unaccentuated” (tanpa tekanan). 8. They could be also considered as adjectives used as non-verbal predicates.

210 Philippe Grangé

Examples of frequent stative verbs used either attributively or as predicates and undergoing default agreement with -n are as follows:9 laé “be clean”, béle “be big, tall”, pelaté “be hot”, gelete “be cold”, weli “be expensive, precious”, busé “be small”, ba’at “be heavy”, bur’a “be white”, doã “be far”. It is worthy to note that stative verbs expressing feelings or psychological states are excluded from this list, because they are relevant to split intransitivity, as will be argued in Section 3.1. 2.2

Intransitive verb morphology

In the morphology of each agreement affix, we can observe a phoneme recalling the corresponding personal pronoun (see Table 1 below). Table 1.  East Adonara Lamaholot dialect: pronouns and agreement affixes Person

Subject pronoun

Agreement prefix for vowel-initial roots

Agreement suffixes core phoneme

Agreement suffixes10

1sg 2sg 3sg 1pl.incl 1pl.excl 2pl 3pl

go mo na tité kamé mio ra

kmntmmr-

/k/ /o/ /a/ or /ʔ/ /t/ /m/ /e/ /a/

-k; -nek; ‑kek -no; -ko -na; -’ /ʔ/ -net; -ket -nem; -kem -né; -ké -na; -ra

The phonological form of the verb determines the morphology of the agreement affix. We can differentiate verbs ending with a consonant (C-final verbs), or a vowel (V-final verbs), and verbs beginning with a vowel (V-initial verbs). C-final verbs The consonant-vowel alternation (CVCV…) must be preserved by an epenthesis between the final consonant of the root and the core consonant of the agreement suffix. For instance, the 1sg agreement suffix, whose core phoneme is /k/, will be realised /ək/ if the last phoneme of the verb root is a consonant, e.g., soot ‘to fear’ > go soot-ek ‘I fear’.

9. A few stative verbs, when employed predicatively, never occur with the default agreement ‑n, for instance kelemu “be pretty”, mela “be beautiful”, kuma “be yellow”. Nevertheless, they can inflect for perfect aspect with agreement sufffixes. 10. In some other East Adonara sub-dialects, different epenthetic consonants are preferred, /j/ and /r/.



Split intransitivity in Lamaholot 211

V-final verbs The verb roots ending in a nasal vowel, /ɑ̃/ /ɔ̃/ or /ɛ̃/, undergo denazalisation along with the predictible choice of the epenthetic nasal consonant /n/. For instance hopã ‘be out of breath, exhausted’ is suffixed as follows: hopã ‘be out of breath, exhausted’ 1sg go hopanek 2sg mo hopano 3sg na hopana 1pl inclusive tité hopanet 1pl exclusive kamé hopanem 2pl mio hopané 3pl ra hopana

The 1sg agreement morpheme (core phoneme /k/) has two morphophonological behaviors. With vowel-ending root verbs, one would predict that no epenthesis is needed, like in turu ‘sleep’ > go turu-k ‘I sleep’. Nevertheless, other vowel-­ ending root verbs require the shaping of an additional final syllable, through the epenthesis of a consonant plus a schwa /ə/. In this case, the 1sg agreement suffix will be realised /nək/ (roots ending with a nasal phoneme), or /kək/ (other roots, although there is no clear-cut rule).11 For some verbs, two different shapes are perfectly acceptable, for instance sogha ‘make ready for return’: go sogha-k or go sogha-nek ‘I make ready for my return’; both seem adequate in East Adonara.12 Another example is tité pepu-t or tité pepu-net ‘we discuss’. V-initial verbs Regarding the analysis of split intransitivity, I will omit the very rare intransitive verbal stems beginning with a vowel,13 for which the agreement prefix is compulsory in any context. For instance, the verb ‑olo ‘to take leave, to leave 11. The morphology of the agreement morpheme may differ, depending on the variant of East Adonara Lamaholot dialect considered, but the core phoneme is common to all of these variants. In some areas within East Adonara, the suffix may be /jək/ or /rək/. Moreover, in some Lamaho­lot dialects outside Adonara, the epenthesis is further modified by a metathesis, hence the 1sg suffix /fkɑ̃/ (Lamaléra dialect, in Lembata island), or /kən/ (Lewoingu dialect, East Flores). 12. For the root verbs ending with a vowel, a ‘preferred’ epenthetic consonant is inserted, for instance the 2sg suffix /o/ will be realised as /no/ or /ko/ (in Witihama), as /jo/ (in Boleng) or as /ro/ (in Kiwang Ona), all these variants being part of the East Adonara dialect. In this paper, I focus on the Witihama valley sub-dialect. 13. Nishiyama & Kelen (2007: 98), describing the Lewoingu dialect, signal that the verbs receiving this kind of suffix “are not characterized semantically but phonologically: their root

212 Philippe Grangé

before someone’ never occurs under its root stem *olo. This is the reason why I spell it ‑olo. ‑olo ‘to take leave, to leave before someone’ 1sg go k-olo 2sg mo m-olo 3sg na n-olo 1pl inclusive tité t-olo 1pl exclusive kamé m-olo 2pl mio m-olo 3pl ra r-olo

These vowel-initial verbs must agree with the subject and are never marked by an additional agreement morpheme. Intransitivity split rules are therefore not relevant for this particular category of verbs.14 2.3

Agreement suffixes and object pronouns

No morphological differences are found between direct and indirect object suffix pronouns. (8) Na péhé -nek. Na soro -nek buku. 3sg touch -1sg 3sg give -1sg book ‘He/she touches me.’ ‘He/she gives me a book.’

A possible reason why the transitive verbs are never suffixed for agreement is that the position right of the verb is reserved, as in examples (2) and (8), for the suffixed object pronoun. Moreover, almost all the agreement markers on intransitive verbs are homophonous with the object pronouns.15 Malchukov (2008: 77) notices that for is vowel-­initial.” This is also the case in the neighboring language Kédang (Klamer 2008: 230 Note 15). 14. At first view, some vowel-initial root verbs escape this rule of compulsory agreement prefix, but they can be analysed as ambitransitive verbs (object may be erased) or reflexive verbs. This is the case with ag’o ‘to adorn, to dress up elegantly’ or odo ‘to make someone sleep’ (trans.) or ‘to feel asleep’ (reflexive). A verb could be labelled as ‘reflexive’ if the object suffix pronoun cross-references the subject. 15. Nouns also undergo genitive agreement in person with the head noun referring to the possessor. In Lamaholot, there is also a comitative conjunction that must agree with the subject: -o’on ‘and, be with’ (an alternative analysis would describe it as a serial verb), see example (10) below.

Split intransitivity in Lamaholot 213



semantic alignment languages in general “experiencers are usually cross-­ referenced through the same inflection used to cross-reference objects.” In some East Nusantara languages, for which Klamer identifies a semantic alignment, “S is encoded with a dependent pronoun (affix or clitic) attached to the predicate” (Klamer 2008: 248). This is the case in Lamaholot: most of the morphological markers of SP are homophonous, or very similar, to the object pronoun paradigm. Consequently, it is sometimes difficult to ascertain whether Lamaholot ambitransitive verbs are used transitively or intransitively.16 It is certain, however, that in Lamaholot the object pronoun paradigm has been “re-grammaticalized” to form the agreement markers. The cross-reference morpheme is suffixed to the verb, just like the object pronouns. The table below shows the resemblance between the agreement morphemes and the object pronouns. Table 2.  East Adonara Lamaholot dialect: agreement suffixes and object pronouns Person

Subject pronoun

Agreement suffix17

Object pronoun

1sg 2sg 3sg 1pl.incl 1pl.excl 2pl 3pl

go mo na tité kamé mio ra

-k; -nek; -kek -no; -ko -na; -’ /ʔ/ -net; -ket -nem; -kem -né; -ké -na; -ra

-nek; -kek -no; -ko -na; -ro; -rona -net; -ket -nem; -kem -né; -ké -ra; -wé; -wéra

It is only in the 3rd person that a morphological difference appears: the glottal stop -’ /ʔ/ occurs as agreement marker only, while -rona (3sg human) and -ro (3sg non-human) occur as object pronouns only. The same remark applies to the object pronouns -wé / -wéra (3pl). The possessive or genitive paradigm is also very similar,18 for instance langu-k (house-1sg.gen) ‘my house’.

16. Nishiyama & Kelen (2007: 98) assert that all verbs (in the Lewoingu dialect) that receive an agreement suffix are intransitive, or used intransitively. Nagaya (2010) writes about the Lewotobi dialect: “since [the S-agreement clitics] are only used for intransitive verbs, agreement suffixes function practically as markers of syntactic transitivity […].” However, at least in the East Adonara dialect, the presence or absence of a suffix cross-referencing the subject is not sufficient per se to distinguish an intransitive verb from a transitive verb. 17. In some other East Adonara sub-dialects, different epenthetic consonants are preferred, /j/ and /r/, see 2.1. 18. These genitive suffixes are 1sg.gen -k; 2sg.gen -m; 3sg.gen -n; 1pl.incl.gen -t / -et; 1pl.excl.gen -em / -kem; 2pl.gen -ké; 3pl.gen ‑ka / -na. See Grangé (2010).

214 Philippe Grangé

The morphological similarities presented in Table 2 may explain why intransitive split in Lamaholot has remained largely unnoticed. Apart from Arka (2000), surveys of various Lamoholot dialects describe the morphology of the intransitive verb agreement sketchily, without addressing the reasons why certain intransitive verbs always undergo agreement, while others may, and others never do. Keraf (1978: 69–73) defines a class of verbs that can either agree or disagree with the subject in the Lamaléra dialect, but he does not discuss the semantic grounds for this possible agreement. Pampus (2001: xviii) gives fifteen examples of intransitive verbs “which can optionally undergo conjugation”, for example the verb morit ‘to live’, see infra Table 4.2 p. 19. Nishiyama & Kelen (2007: 31) find that the agreement morphemes are “often optional, and they become preferred or obligatory depending on the verb, the pronoun in the subject, or contexts.” The first author who analyzed the intransitive verb agreement feature as a manifestation of split intransitivity in Lamaholot was Arka (2000). His article deals with intransitivity in some East Nusantara languages with a few examples from Lamaholot, especially the East Adonara dialect. Klamer (2008) has published a seminal paper on split intransitivity in Indonesian languages, especially the East Nusantara languages, but she had to omit Lamaholot due to scarcity of data. Interestingly, the split intransitivity system is eroded in most of the Lamaholot dialects, with the remarkable exception of the East Adonara dialect. I will now examine which semantic features do or do not entail agreement of the intransitive verb with its subject. 3. Lexically fixed intransitive split (Split-S) Lexically fixed intransitive split, widely called Split-S after Dixon (1994), means that in some languages, certain intransitive verbs are always marked, and others are never marked. This is relevant for the East Adonara Lamaholot dialect, as well as the following remark by Dixon: “Split-S marking relates to the nature of the verb. It is scarcely surprising that for most languages of this type morphological marking is achieved by cross-referencing on the verb” (1994: 76). 3.1

Feeling verbs

Stative verbs expressing feeling (psychological state) must agree with the subject person when used predicatively.19 19. Agreement is also compulsory when the stative verb is used as adjective in an attributive structure, see Section 2.1 above.

Split intransitivity in Lamaholot 215



(9) Keréak senan -a. child happy -3pl ‘The children are happy.’ (10) Go so’ot -ek k- o’on aho. 1sg afraid -1sg 1sg- with dog ‘I am afraid of dogs.’

This cross reference relies on a paradigm of verbal agreement suffixes, comprising the core phonemes mentioned above: 1sg /k/; 2sg /o/; 3sg /a/ or /ʔ/; 1pl.incl /t/; 1pl.excl /m/; 2pl /e/; 3pl /a/. Feelings are normally out of the control of the subject’s referent,20 a semantic feature consistent with the Fluid-S rules, as will be argued in the next section. Examples of stative verbs that always require agreement with their subject include: kedok’o- ‘be surprised’ mia- ‘be ashamed’ so’ot- ‘be afraid’ suke- [sukə] ‘be sad’ beken- ‘be angry’ senan- ‘be happy’ too- ‘be of the same mind, be in sympathy’

3.2

Sentience verbs

Sentience verbs also must agree. In Lamaholot, they contrast with perception verbs (e.g. see, hear) and ‘propositional attitude verbs’ (e.g. know, believe), generally ambitransitive, for which an object complement is optional. For instance lilé ‘to watch’ in the following example shall not agree with the subject and could be completed by a NP or a suffix object pronoun, while the sentience verb mirè[mirɛ] ‘be sicken’ agrees with its subject. (11) Turun mo mirè -ko, mo aké lilé. if 2sg sicken -2sg 2sg don’t watch ‘If you are disgusted, don’t look.’

20. Klamer (2008: 229) notes that in Kambera (Sumba island), the S of non-verbal predicates “as a typical non-volitional participant of a non-dynamic state of affairs, is marked with an accusative enclitic”. A striking difference, however, is that the agreement is marked on the NP in Kambera, while it affects the VP in Lamaholot.

216 Philippe Grangé

The sentience verbs must agree because the subject is regarded as being deprived of control over the process (hence SP). The trait ‘± control’ is involved in this semantic split, rather than ‘± volition’.21 Sentience intransitive verbs include for instance: mirè- ‘be disgusted, sicken’ hopã- ‘be tired, out of breath, exhausted’ berara- ‘be sick’ gate- ‘to itch’ gele- ‘to have cramps’

Note that some sentience verbs cannot have a pronominal subject. Their subject must refer to some part of the (human) body; this is the case of the three last verbs displayed above and in examples (12) to (14) below. Nevertheless, the verb must agree with the ‘possessor’ of the body, not with the ‘part of the body’. Therefore, under this ‘possessor ascension’ pattern, the SP marking concerns the subject’s referent (typically, a human) as a whole, not a particular part of his/her body. (12) Weki -k gate -k. body -1sg itch -1sg ‘I am itching.’ (lit. ‘my body itches’) (13) Tai -m bohu -m. stomach -2sg be.full -2sg ‘You are fully fed.’ (lit. ‘your stomach is full’) (14) Lei -k gele -k. leg -1sg cramp -1sg ‘I have cramps in my leg.’ (lit. ‘my leg cramps’)

In summary, the intransitive verbs implying a subject’s lack of control over the event, for instance feelings or sentience, require an agreement confirming that the subject is patient-like (SP). This intransitive split is lexically fixed (Split-S). 3.3

Biological function verbs

Verbs of biological functions presuppose that the subject is patient-like, hence SP. Thus, these intransitive verbs bear a compulsory agreement suffix.

21. We will return to the nuance between ‘± control’ and ‘± volition’ in Section 4.1.

Split intransitivity in Lamaholot 217



béle- ‘to grow up’ ewun- [ewũ] ‘to sweat’ suku- ‘to suffocate, to smother’ keduhu- ‘to cough’ te’ure- ‘to dream’

For instance, the verb te’ure- ‘to dream’ must be affixed for agreement: (15) Go te’ure -k. Mo te’ure -no Na te’ure -na 1sg dream -1sg 2sg dream -2sg 3sg dream -3sg ‘I dream.’ ‘You dream.’ ‘He/She dreams.’

Although some of these actions could be performed intentionally, for instance ‘to cough’, native speakers are reluctant to deprive these verbs from their agreement suffix; they would prefer to add adverbials, like ‘on purpose’. Agreement on biological functions verbs appears ‘lexically fixed’. The Split-S patterns described above regarding verbs of feeling, sentience, and biological functions, indicate that the low agentivity of the subject (SP), which corresponds to the Patient Proto-Role (Dowty 1991), properties, especially ‘undergoes change of state’, and in some contexts ‘causally affected by another participant’. 3.4

Motion verbs

The lexically fixed split also involves a subclass of intransitive dynamic verbs which imply some kind of mobility.22 When examining SV sentences, a clear-cut opposition appears between motion verbs (manner of motion, posture) and displacement verbs (indicating a movement, a path of motion): motion verbs must agree with the subject, unlike displacement verbs. Motion verbs (16) Go pana -nek. *Go pana. 1sg walk -1sg 1sg walk ‘I walk.’ (17) Mo nange -no. *Mo nange. 2sg swim -2sg 2sg swim ‘You swim.’

22. Besides, there is a wealth of spatial prepositions and postpositions in Lamaholot. Nagaya (2009a) has described the sophisticated system of directional morphemes, referring to local orientation, global orientation or quasi-cardinal orientation.

218 Philippe Grangé

(18) Téna ba’o -ra kaé. *Téna ba’o kaé. pirogue sail -3pl already ‘Pirogues are sailing already.’ (19) Go dé’i -nek. *Go dé’i. 1sg stand -1sg 1sg stand ‘I am standing up / in upright position.’ Displacement verbs (20) Kamé géré. Go balik. 1pl.excl ascend 1sg return ‘We ascend / go upward / get in (house).’ ‘I go back.’ (21) Go béto nakuné. Mo lodo ki? 1sg arrive while.ago 2sg descend interr ‘I arrived just a short while ago.’ ‘You go down?’

Nevertheless, displacement verbs may agree with their subject, only to mark perfect aspect, as will be shown in Section 5.1. But if no perfect aspect is meant, or if perfect aspect is indicated elsewhere than on the verb (by an adverb for instance), displacement verbs shall not undergo agreement. This Split-S within the verbs of mobility indicates that the subject of a ‘motion verb’ is seen as more patientive (SP) than the subject of a ‘displacement verb’ (SA). This lexically fixed split can be explained by Dowty’s proto-properties: Table 3.  Proto-Patient and Proto-Agent property (Dowty 1991) for motion verbs and displacement verbs in SV sentences Motion verbs → SP

Displacement verbs → SA

Proto-Patient property ‘d’: stationary relative to movement of another participant

Proto-Agent property ‘d’: movement (relative to the position of another participant)

The examples examined above were limited to motion verbs in SV sentences. But when motion verbs (e.g., walk, sail, stand up) are complemented by a locative prepositional phrase (PP), the verbal agreement is barred. Compare examples (22) and (23). (22) Go pana -nek. *Go pana. 1sg walk -1sg 1sg walk ‘I walk.’ (23) Go pana (*-nek) héti Loga. 1sg walk (*-1sg) upward Prn ‘I walk upward to Loga.’

Split intransitivity in Lamaholot 219



There is no self-evident reason why the event’s localisation in space23 should assign an ‘agentive’ role to S. There may be two explanations, not mutually exclusive. On the one hand, locative PP complementation overcomes the Proto-Patient property ‘d’, to the benefit of Proto-Agent property ‘d’ (see table above). For instance, if someone walks on the same place without heading anywhere, he remains ‘stationary’, corresponding to the Proto-Patient property ‘d’. With a locative PP, the subject is considered as more agentive, hence SA (agreement is barred). On the other hand, the locative PP assigns a kind of boundary to the process. The event is supposed to reach its completion; for instance, if someone walks heading somewhere, he shall stop walking when arrived. In other words, the process becomes telic in the presence of a locative PP; otherwise it remains atelic, and entails agreement. The sentence go pana‑nek ‘I walk’ could be the answer to the question “What are you doing right now?”. Alternatively, a locative PP leads to a telic interpretation, for example, go pana héti Loga ‘I walk upward (to) Loga’ could be the answer to “Where are you going?”. Nevertheless, this second explanation is not obvious in some contexts, like ‘swim in the sea’ or ‘stand up on a stone’. Examples of motion verbs24 complemented or not by locative PP are displayed in the table below. Table 4.  Verbs of motion complemented or not by a locative PP Locative PP: S → SA

No locative PP: S → SP

ba’o Téna ba’o (*-ra) lau dai. ‘to float, pirogue sail -3sg/pl sea from to sail’ ‘A / some pirogues sail from the open sea.’

Téna ba’o -ra kaé. pirogue sail -3sg/pl already ‘A / some pirogues sail already.’

nangé Mo nangé (*-no) si tahik onen. ‘to swim’ 2sg swim -2sg at water into ‘You swim into the sea.’

Mo nange -no. 2sg swim -2sg ‘You swim.’

tobo ‘to sit’

Go tobo -nek. 1sg sit -1sg ‘I am seated (I am in a seated position).’

Go tobo (*-nek) si keresi. 1sg sit -1sg at chair ‘I sit / I am sitting on a chair.’

dé’i Go dé’i (*-nek) si wato lolon. ‘to stand 1sg stand -1sg at stone top up’ ‘I stand up on the stone.’

Go dé’i -nek. 1sg stand -1sg ‘I stand up.’

East Adonara Lamaholot cannot solely and simply be classified as a Split-S language: this language also owns a wealth of Fluid-S features, as will be argued in the next section. 23. Complemented with a time PP, or causal PP, a motion verb must agree with its subject. 24. All the examples in this table may appear in a narration, and therefore can recount past events.

220 Philippe Grangé

4. Fluid intransitive split (Fluid-S) For some intransitive verbs, verbal agreement is triggered by contextual semantic factors which are not restricted to the lexicon. When the subject’s referent owns one or several Proto-Agent properties, it is seen as agent-like, the verb appears unmarked (SA), while for patient-like subjects the same verb will be marked (SP). This fluid intransitive split, or Fluid-S in Dixon’s terminology, is complex in the East Adonara dialect, because agreement is not solely determined by the agentivity of S; the indication of aspect is also relevant, and can overlap semantic factors. I will firstly examine Fluid-S related to semantic factors. 4.1

SA controls, SP uncontrols

The fixed intransitive split (Split-S) in East Adonara Lamaholot relies mainly on the trait “± control”. Unsurprisingly, this trait also plays an important role in fluid intransitive split (Fluid-S). Control is slightly different from volition, a major property for the Agent Proto-Role: “volitional involvement in the event or state” (Dowty 1991: 572). Volition means that the subject makes a conscious decision to act, and Dowty adds that “deliberately refraining from action is volitional also.” (1991: 573). Control means that S (referring usually to an human) may stop or suspend the process, even if this process is not volitional: it may have started independently from the subject will. In other words, volition implies control over a process, but the reverse is not true. In Lamaholot, volition is not an essential feature, but the trait ‘± control’ is crucial, either for Split-S or Fluid-S. For an intransitive verb, under Fluid-S the S will be marked (SP) through a verbal agreement if the referent of S does not control the process. For instance, géka ‘to laugh’ will remain unmarked (SA) when it means simply that someone makes sounds to express his amusement, and is able to stop laughing. If the same verb is marked (SP), it implies that the subject does not control his reaction, i.e., he bursts out laughing, dies of laughing, is giggling. géka ‘to laugh’ géka-agr ‘to burst out laughing, to giggle’ 1sg go géka go géka-nek 2sg mo géka mo géka-no 3sg na géka na géka-na 1pl inclusive tité géka tité géka-net 1pl exclusive kamé géka kamé géka-nem 2pl mio géka mio géka-né 3pl ra géka ra géka-na

Split intransitivity in Lamaholot 221



Similarly, Lamaholot speakers can oppose tani ‘to cry, to shed tears’ and the same verb with an agreement suffix tanhi- ‘to weep, to cry one’s heart out’,25 when the subject is supposedly unable to refrain from weeping. (24) Go tani. Go tanhi -nek. 1sg weep 1sg weep -1sg ‘I cry.’ ‘I cry.’ (I can’t stop weeping)

The verb tani can also be used transitively, meaning ‘to weep for someone, to mourn for someone’, for instance: kamé tani-rona ‘we mourn for him/her’.26 The 3rd person agreement suffix -na in na tanhi-na ‘he/she can’t refrain from crying’ is obviously cross-referring the 3sg subject, and cannot be confused with a suffixed 3sg object pronoun, for instance na tani-rona ‘he/she mourns for him/ her”: only a deceased person can be mourned. Therefore, example (24) exhibits a monovalent verb. In the examples above (to laugh, to cry), there is usually no volition by the subject’s referent, but there can be more or less control, therefore, the event of ‘laughing’ may be expressed as under control (SA) or out of control (SP). This was proved by speakers’ judgements on example (25), which was rejected on pragmatic grounds: one can’t use the aspectual marker kaé ‘already, yet, done’, which presupposes an agent intention, because a burst of laughter is in principle not intentional. ( 25) *Go géka -nek kaé. 1sg laugh -1sg perf ? ‘I have burst out laughing (it is done).’

Other examples of Fluid-S in the East Adonara dialect, illustrating the semantic alignment “SA controls the process / SP doesn’t control the process”, are displayed in the table below.

25. An epenthetic /h/ is inserted as the new onset of the stem’s final syllable in tanhina when this verb is followed by an agreement suffix, being intransitive. For instance, the verb tani ‘cry’ becomes tanhi-; koda ‘to say’ > kodha-; médo ‘be bad’ > médho-. 26. This translation would be more literal in French: “nous le pleurons”.

222 Philippe Grangé

Table 5.  Control versus lack of control (S controls) S → SA

(S does not control) S → SP

énénen ‘to dance’

ra énénen 3pl dance ‘they dance’

ra énénen -na 3pl dance -3pl ‘they dance (for joy, spontaneously)’

méké ‘urinate’

mo méké 2sg urinate ‘you urinate’

mo méké -ko 2sg urinate -2sg ‘you make a pee (involuntarily), you pee (in your pants)’

goka ‘to fall’

na goka 3sg fall ‘he/she falls/fell intentionally’

na goka -na 3sg fall -3sg ‘he/she fell by accident’

heto ‘to jump’

Na heto wai. 3sg leap water ‘He leaps (over) the water.’27

Na heto -na. 3sg leap -3sg ‘He jumps (to his feet), (something makes) him jump, he springs.’

A similar example in the Lewotobi dialect was mentioned by Nagaya (2010: 6), quoted as (26) and (27) below, with the verb tedu ‘to collide’: (26) Go tedu knebi. 1sg collide wall ‘I collided with the wall (intentionally)’ (27) Go tedu -ek ia knebi. 1sg collide -1sg dem wall ‘I collided with the wall (accidentally)’

However, in Lewotobi (western edge of the Lamaholot dialect chain), the Split-S is not ‘fluid’ anymore, it seems indeed totally lexicalized.28 The Fluid-S in Estas Adonara dielect also concerns a few ambitransitive verbs, like pelaé ‘to run’:29

27. A location PP is compulsory if heto “jump, leap” has no agreement suffix. 28. From Nagaya’s data on Lewotobi dialect (spoken in the southeastern tip of Flores island) this one and only example possibly displays a unique remnant of an extensive Fluid-S system, still widely used in East Adonara. Nagaya (2010: 6) gives IPA transcriptions of this example, respectively: Go tədu knəbi and Go təduəʔ ia knəbi. 29. The verb plaé ‘to run’ is given as an example of Fluid-S by Arka (2000: 19).

Split intransitivity in Lamaholot 223



(28) Go pelaé. Go pelaé -k. Go pelaé -ro. 1sg run 1sg run -1sg 1sg run -3sg ‘I run.’ ‘I run away, I flee.’ ‘I run away from him/her.’

While at first glance go pelaé-k ‘I run away’ may appear to be a reflexive structure, this is in fact not the case. We have seen that the agreement suffixes are generally homophonous with the object pronouns. It is only with the 3rd person that a morphological difference between the agreement suffix -na and the 3sg object pronoun -ro or -rona conclusively shows that pelaé ‘run’, used intransitively, may or may not agree with its subject: (29) Na pelaé. Na pelaé -na. Na pelaé -rona. 3sg run 3sg run -3sg 3sg run -3sg ‘He/she runs.’ ‘He/she runs away, flees.’ ‘He/she runs away from him/her.’

In this example, the Fluid-S system indicates a semantic opposition between ‘running with control of the subject’ (marked SA) and ‘being forced to run’ (SP). Nevertheless, some example of Fluid-S in which S appears to have control over the process (hence SA) are rejected by speakers on pragmatic grounds, for instance goka ‘to fall’ with the 1st person, because ‘nobody would fall on purpose’. (30) Go goka -nek. *Go goka. 1sg fall -1sg 1sg fall ‘I fell.’ ‘I fell on purpose.’

Arka (2000: 4, ex. (6)) noted that the unmarked *go goka ‘I fall on purpose’ (my translation) is not allowed. In other semantically aligned languages, like Acehnese,30 the fluid intransitive split is linked (mainly or partially) to the subject’s volition. In Lamaholot volition does not seem to be an essential criterion. 4.2

SA unaffected, SP affected

The Fluid-S system also distinguishes between whether or not the subject’s referent is affected by the process. The label ‘affected’ means that the subject’s referent undergoes a change of state, for example receives a new property, or is deprived of a property. The more a subject’s referent is affected, the more it is considered patient-like.31 30. See Donohue (2008: 36) and Klamer (2008: 226). 31. In other East Nusantara languages such as Klon (Klamer 2008: 235–237) and Tanglapui (Klamer 2008: 239–243), the semantic alignment relies principally on whether or not S undergoes a change of state.

224 Philippe Grangé

Generally, the distinction in this Fluid-S system corresponds to the animate-­ inanimate opposition: an animate referent is more likely to be affected by the process. For instance, if the subject of the verb doro ‘to slip’ refers to a road or a pathway, this referent is obviously not affected by its ‘being slippery’. But if S refers to an animate, usually this referent is affected and the verb agrees, showing the patient-like role of the subject, marked SP . Table 6.  Unaffected versus affected (S unaffected) S → SA

(S affected) S → SP

doro ‘to slip, to slide’

Ukut ni doro. pathway det slip ‘This pathway is slippery’

Ra doro -ka. 3pl slip -3pl ‘they slip, they slide’

wédon [wedɔ̃n] ‘to shake, to quake’

Bréro wédon béle -béle. earthq. shake big -red ‘The earthquake shook very hard.’

Go wédo -nek. 1sg totter -1sg ‘I totter, I stagger’

tudak ‘to hinder, to stop’

Go tudak. 1sg stop ‘I stop.’

Go tudak -ek. 1sg stop -1sg ‘I am stopped’ (stuck, hampered)’33

mor’it [mɔrʔit] ‘to live’

Na mor’it uling. 3sg live still ‘He still lives.’

Na morit -a lau tahik. 3sg live -3sg sea water ‘He lives from the sea’ (from sea resources)

Spéda wédo -na héti lali. bicycle shake -3sg up down ‘The bicycle staggers to and fro.’32

Go morit -ek epa ikã. 1sg live -1sg catch fish ‘I live from fishing’ (lit. ‘I live catching fish’)

The second example in Table 6 involves the ambitransitive verb wédon ‘to shake, to quake’. If the subject is inanimate like bréro ‘seism, earthquake’, no agreement shall appear (SA). If the subject is affected, we get SP: go wédo-nek ‘I am tottering down, I shiver’. However, an inanimate subject’s reference can also be considered as affected by the process, for instance spéda ‘bicycle’, thus, it also triggers the SP verbal agreement.

32. This data is quoted from an account by an East Adonara informant watching the film Pear Story. At this stage of the film, a young boy is loosing control of the bicycle he is riding. 33. For instance by a lack of money, a ceremonial obligation.

Split intransitivity in Lamaholot 225



5. Fluid intransitive split and aspect marking Fluid intransitive split (Fluid-S) rests mainly on the semantics of S, but may also be triggered by aspect marking. This feature generally applies to stative verbs, but also some dynamic intransitive verbs. 5.1

Displacement verbs and aspect

The Split-S within verbs expressing mobility has been presented in Section 3.4. In the absence of locative prepositional phrase (PP), motion verbs (which do not imply movement from a location to another) agree with the subject, thus interpreted as SP and corresponding to the proto-patient property ‘d. stationary relative to movement of another participant’ (Dowty 1991). On the contrary, when complemented by a locative PP, motion verbs must not agree, because the subject is seen as more agentive (Proto-Agent property ‘d’), therefore agreement is barred. A locative PP may also convey telicity to the process, a semantic factor which is not unrelated to the aspect features that will be described below. In contrast, displacement verbs (indicating a movement, a path of motion) in SV sentences do not agree with their subject, leading to a SA interpretation of the subject, linked to the proto-agent property ‘d. movement (relative to the position of another participant)’. At first view, unlike motion verbs, for displacement verbs complementation by a locative PP does not interfere with the verb morphology. (31) Kamé géré (héti lango  mo’en). 1pl.excl ascend (upward house  2sg.poss ‘We ascend (upward to your house).’

However, an additional semantic factor may overlap this Split-S, and induce a shift to a Fluid-S: perfect aspect34 marking. As opposed to the Split-S described above, agreement with the subject must go along with a locative PP to indicate perfect aspect. (32) Kamé géré (*-kem) 1pl.excl ascend ‘We ascend.’ 34. According to Comrie (1976: 62) there is a “tendency to confuse perfect and perfective. The perfect links a present state to a past situation, whether this past situation was an individual event, or a state, or a process not yet completed […]” while (p. 21) “perfectivity involves lack of explicit reference to the internal temporal constituency of a situation […] subsumed as a single whole.”

226 Philippe Grangé

(33) Kamé géré -kem héti lango mo’en. 1pl.excl ascend -1pl.excl upward house 2sg.poss ‘We have gone upward to your house.’

In other terms, marking the perfect aspect by an agreement suffix on a verb of displacement forces its complementation by a locative PP. Or, viewed from another angle: a displacement verb must be made telic (by means of a locative PP) before it can be marked for perfect aspect. Other examples are displayed in the table below: Table 7.  Displacement verbs and aspect marking Verbs of displacement35

No aspect marking: S → SA optional locative PP

Perfect aspect marking: S → SP compulsory locative PP

balik ‘return, go back’

Go wulin balik 1sg market return ‘I go back from the market.’

Go wulin balik -ek. 1sg market return -1sg ‘I went back from the market. I am back from the market’

lodo ‘go down, Go héti ilé lodo Kamé sé Balawéling lodo descend, get 1sg up mountain descend 1pl.excl there prn descend out’36 ‘I go down from the mountain.’ -kem. -1pl.excl ‘We, from (distant) Balawéling, went down’37 béto ‘to arrive’ Go béto nakuné. 1sg arrive while ‘I arrived a short while ago’

Go béto -nek si oring. 1sg arrive -1sg at hut ‘I have/had arrived at the hut’

Go béto si oring. 1sg arrive at hut ‘I arrive(d) at the hut’

35. All the examples in this Table 7 may appear in a narration, and therefore can describe past events. 36. It should be noted that the verbs balik ‘return, go back’ and lodo ‘go down, get out’ require an uncommon syntactic order: the locative PP must precede the verb. It is as if the syntactic order mimics the chronological order: the place of origin precedes the displacement verb. Inversely, with géré ‘go upward, get in’, the locative PP always appears at the right of the verb, mimicking that the place of destination comes after the displacement process. 37. The preposition sé ‘there (distant)’ is not mentioned in the surveys of other Lamaholot dialects, see Nagaya (2009a) and Pampus (2001). This preposition may be specific to the East Adonara dialect (and possibly others), and could originate from sédék ‘horizon’.

Split intransitivity in Lamaholot 227



The verbs in the right column will always be interpreted as bearing a perfect aspect (the perfect marker kaé may be added, unless judged redundant). In summary, there is a lexically fixed intransitive split (Split-S) between dynamic intransitive verbs of “motion” and “displacement”. Each of these verb categories splits further, in contradictory manners regarding the cooccurrence of agreement and locative PP. Semantic and syntactic factors are entangled. Aspect marking on displacement verbs can be regarded as a Fluid-S which overcomes the said Split‑S. The table below recapitulates these rules. Table 8.  Verbs of motion versus verbs of displacement, complemented or not by a locative PP without locative PP: atelic

with locative PP: telic

5.2

Verbs of motion

Verbs of displacement

S → SP Go pana -nek. 1sg walk -1sg ‘I walk’

S → SA Kamé géré. 1pl.excl ascend ‘We go upward.’

*Go pana. 1sg walk

*Kamé géré -kem. 1pl.excl ascend -1pl.excl

S → SA Go pana héti Loga. 1sg walk up prn ‘I walk upward (to) Loga’

S → SA Kamé géré héti Loga. 1pl.excl ascend up prn ‘We go upward to Loga.’

*Go pana -nek héti Loga. 1sg walk -1sg up prn

+ perfect aspect Kamé géré -kem héti Loga. 1pl.excl ascend -1pl.excl upward prn ‘We have gone upward to Loga.’

Fluid-S and aspect marking on stative verbs

When an stative verb is used as predicate, it would normally take the default agreement suffix -n (see 2.1 above), but an aspect can be indicated through the paradigm of agreement suffixes,38 as shown in example (35). (34) Karo ni naé lolon bete -n. tree det 3sg.gen leaf green -agr ‘This tree’s leaves are green.’ (lit. ‘This tree, its leaves are green’)

38. The paradigm of agreement suffixes comprises the following core phonemes: 1sg /k/; 2sg /o/; 3sg /a/ or /ʔ/; 1pl.incl /t/; 1pl.excl /m/; 2pl /e/; 3pl /a/.

228 Philippe Grangé

(35) Karo ni naé lolon bete -ka. tree det 3sg.gen leaf green -3sg.pfct ‘This tree’s leaves have become green.’ (lit. ‘This tree, its leaves have become green’)

In example (35), bete-ka ‘have become green’ expresses a perfect aspect, more precisely a perfect of result,39 because the state (‘being green’) results from a prior unmentioned process. Unlike the examples mentioned in the previous sections, where the fluid intransitive split relied on S semantic traits (± control, ± affected, ± in motion), here the split relies on aspect marking. However, Lamaholot split intransitivity does not provide a comprehensive aspect marking system, because the verbal agreement suffixes indicating the perfect of result only alternate with the default agreement suffix40 -n (or -Ø if modified by an adverb). Thus, the sole aspectual opposition expressed by fluid intransitive split is between the perfect of result and any other aspect (in other terms, aspect neutral). Perfect of result marking through verbal agreement is identical for inanimate and animate/human subject’s referents: (36) Wai ni geleti -n. Wai ni gelete -na. water det be.cold -agr water det be.cold -3sg.pfct ‘This water is cold.’ ‘This water has become cold.’

39. Comrie (1976: 56–58) proposes this definition of the “perfect of result”: “a present state is referred to as being the result of some past situation.” The last example of Comrie’s section about the “perfect of result” deals with the Mandarin Chinese particle -le, and could also fit in with our description of the Lamaholot 3rd person suffixes -ka, -na or -ra: “the verbal particle ‑le indicates perfective aspect and relative past time reference […] With stative predicates, the force of this particle ‑le is often to indicate a state resulting from some previous situation, as in dõngxi guì-le ‘things are expensive’ (but with the implication that once they were not, i.e., they have become expensive).” 40. In Lamaholot, adjectives can be used attributively (modifying a head noun) or predicatively (as non-verbal predicates). When used attributively, adjectives exhibit an agreement with the NP head, like adjectives do in German or French, but the attributive agreement morpheme is the suffix -n which obviously originates from the 3sg possessive morpheme. If the stative verb ends with a vowel, this agreement triggers the nasalization of its final vowel, e.g., bura’ ‘be white’ > labu bur’ã-n [burʔã) ] or [burʔɛ̃n] ‘white cloth’, sometimes along with the epenthesis of a glottal stop /ʔ/ or a /h/ before the final syllable. This attributive agreement internal to a NP will be barred if the adjective itself is modified by an adverb or reduplicated, for instance labu bura’-bura’ ‘really white cloth’, labu bura’ tua ‘very white cloth’, labu bura’ wa ‘not yet white cloth’. In summary, when an adjective is used attributively (modifying a head noun), the position right of the adjective must be filled, either by the agreement suffix -n or by a modifier (adverb or aspect marker).

Split intransitivity in Lamaholot 229



(37) Go bur’a -n. Go bur’a -kek. 1sg be.white -agr 1sg be.white -1sg.pfct ‘I am white (fair skinned).’ ‘I have become white.’ Or: ‘Now I look whiter (than other people).’ (38) Na medho -n. Na medho -na (kaé) 3sg be.bad -agr 3sg be.bad -3sg.pfct (pfct) ‘He is bad.’ ‘He has become bad (already).’

The default agreement -n, or the lack of agreement, implies that no aspect is signalled within the stative verb, which remains compatible with any aspect marker. Thus, the perfect aspect has to be indicated by a floating post-verbal aspect marker kaé ‘done, completed’, and the imperfective by a pre-verbal aspect marker mété ‘be …-ing’. On the other hand, if this stative verb predicate agrees with its subject, this agreement indicates a “perfect of result” aspect, thus, only the perfect aspect marker kaé may be added. The aspect marking through an agreement suffix is also compatible with stem reduplication (expressing intensity): (39) Labu ni laé -n. / laé -ka. / laé- laé -ka. cloth det clean -agr / clean -3pl.pfct / clean red -3pl.pfct ‘This/these cloth(es) is/are clean.’ / have become clean.’ / have become very clean.’

A few nouns can also be employed as non-verbal predicates; they obey the same rules. (40) Kayo ni wato -na. wood det stone -3sg.pfct ‘This piece of wood has become stone.’ Kayo ni wato -ka (kaé). wood det stone -3pl.pfct (pfct) ‘These pieces of wood have become stone (already).’

Under this Fluid-S system, the verb agreement that indicates perfect aspect (or perfect of result) cannot be simply regarded as the consequence of a syntactic factor. This agreement, which obviously evokes SP , reflects a the Proto-Patient property ‘incremental theme’. Subjects of these stative verbs refer to inanimate or animate beings which are ‘effected’, undergo an explicit change of state, even destruction. For instance, in example (40), the (former) piece of wood can be considered as an ‘incremental theme’, because it has acquired a new chemical composition.

230 Philippe Grangé

5.3

Fluid-S and aspect marking on intransitive dynamic verbs

The agreement suffixes can mark a perfect aspect on verbs of displacement, as was shown in Section 5.1; compare examples (41) and (42). (41) Kamé géré héti Balawéling. 1pl.excl ascend up npr ‘We go upward to Balawéling.’ (42) Kamé géré -kem héti Balawéling.41 1pl.excl ascend -1pl.excl up npr ‘We have gone upward to Balawéling.’

For other intransitive dynamic verbs the fluid intransitive split is also linked to aspect. The Fluid-S alignment distinguishes “states” from “events”. It may seem odd that a dynamic verb leads to a stative interpretation. It is, however, the case when the event is stabilised (presented as continuous and unbounded) or completed (presented as a resulting state), and generally indicating a background circumstance. The intransitive verb remains unmarked to express a stative meaning, see example (43), while the agreement confers a perfect aspect to the verb, obviously interpreted as an event in example (44). (43) a. Manuk telu -n to’u data. hen egg -3sg.gen one break ‘One hen’s egg is broken’ b. Manuk telu -n rua data. hen egg -3sg.gen two break ‘Two hen’s eggs are broken’ (44) a. Manuk telu -n to’u data -’ hen egg -3sg.gen one break -3sg ‘One hen’s egg has got broken / has been broken’ b. Manuk telu -n rua data -ka. hen egg -3sg.gen two break -3pl ‘Two hen’s eggs have got broken / have been broken’

The semantic alignment affecting the verb which means ‘to break apart’ in the neighbouring language Kédang is seemingly identical (Klamer 2008: 234). In the East Adonara dialect, a slight morphological difference appears between the agreement suffix paradigms for stative verbs and dynamic intransitive verbs. For

41. When a verb of displacement receives a perfect aspect (marked by the agreement suffix), this verb must be completed by a locative PP.

Split intransitivity in Lamaholot 231



stems ending with a vowel, the 3sg person agreement morphology rests on the /a/ core phoneme (-na, -ka, -ya or -ra) for stative verbs (or noun used as predicates). On the other hand, for the intransitive dynamic verbs, the 3sg agreement morpheme is -’ (the glottal stop /ʔ/), and this 3sg suffix clearly contrasts with the 3pl -ka. In addition, there is also a tendency to distinguish the 3sg and 3pl agreement morphologies stative verbs, but not all speakers agree on this issue. When an intransitive dynamic verb expresses a state rather than an event, it is not compatible with a pas time reference. The sentences (43) cannot begin with the adverbs wia ‘yesterday’ or nakuné ‘a short while ago’, which on the contrary would be appropriate for the sentences above (44). Unsurprisingly, an imperfect aspect marker like mété ‘be …-ing’ in not compatible with the agreement which indicates the perfect aspect. It is the opposite for the perfect aspect marker kaé, which optionally follows an agreeing verb. (45) a. Lusi rua (mété) beka. (*kaé) eagle two (impf) fly pfct ‘Two eagles fly (are flying).’ b. Lusi rua (*mété) beka -ka (kaé). eagle two (*impf) fly -3pl (pfct) ‘Two eagles have flown away (already).’

As can be expected, with an inanimate subject referent (for instance a sheet of paper), S is considered patient-like (SP), hence the verbal agreement, which subsumes “SP” and “perfect aspect”. One can still add the floating marker kaé if the perfect aspect has to be explicit. (46) Kertas goen beka -’ / *beka. paper 1sg.gen fly -3sg / fly-Ø ‘My paper has been blown away.’ (not ‘*my paper flies’)

The same rules apply to transitive verbs used intransitively, such as bi’at ‘to tear, rip to pieces’. Still, the verbal agreement signals the “perfect of result”, and excludes the imperfective aspect. (47) Go labhu -k (mété) bi’at (*kaé) 1sg cloth -1sg.gen (impf) tear (*pfct) ‘My clothes are being torn up. Someone is tearing up my clothes.’ (48) Go labhu -k (*mété) bi’at -a (kaé) 1sg cloth -1sg.gen (*impf) tear -3pl.pfct (pfct) ‘My clothes have been torn up (already).’

232 Philippe Grangé

5.4

Fluid-S in serial verbs and subordinate clauses

Within Lamaholot serial verb structure, one but not both of the verbs may agree. I will examine for an example the intransitive verb tékeng [tekɛ̃ŋ] or [tekɛ̃n] ‘to work abroad legally’.42 In example (49), the agreement marks a perfect aspect. (49) Kamé tékén -em lau Malaysia. 1pl.excl work.abroad -1pl.excl.pfct oversea Malaysia ‘We have worked in Malaysia.’

The verb tékeng must agree if preceded by the verb -ai ‘to go’. Here, the verb tékeng agrees, probably to indicate that S is affected by the event, and is therefore patient-­ like (SP). If the perfect aspect has to be indicated, the aspect marker kaé must be added. (50) Na n- ai téke(n) -na. 3sg 3sg- go work.abroad -3sg ‘He goes working abroad.’

In example (51), the lack of agreement on tékeng leads to a stative interpretation. In examples (52) and (53), the agreement on balik ‘return’ or on tékeng can be interpreted as a perfect aspect marking. (51) Kamé tékéng balik. 1pl.excl work.abroad return ‘We are back from working abroad (possibly since a long time).’ (52) Kamé tékéng balik -em 1pl.excl work.abroad return -1pl.excl.pfct ‘We have returned from working abroad.’ (53) Kamé tékén -em balik. 1pl.excl work.abroad -1pl.excl.pfct return ‘We have already worked abroad (and we are back).’

Agreement on intransitive verbs in subordinate clauses also involves aspect marking. In a causal or factitive relationship between two verbal clauses, an optional agreement suffix to the second verb indicates its perfect aspect. In addition, in this case, it signals completion for the verb of the first clause. In other terms, when the 42. The verb tékeng originates from Dutch tekenen ‘to sign’, probably via Kupang Malay (H.  Steinhauer, pers. comm.). In East Flores there is a tradition of working abroad, almost exclusively in the Malaysian states of North Borneo. A significant proportion of the Flores migrants work in Malaysia without permits, while the verb tékeng means explicitly that the worker holds a legal status.



Split intransitivity in Lamaholot 233

verb of a subordinate clause inflects for the perfect aspect, this aspect also applies to the verb of the main clause. This inflection (shown by a suffix) subsumes the agreement feature (the person) and the aspect (perfect). In the presence of causal or factitive subordinate clauses, one can infer the chronology of the events. For instance, in example (54), the agreement on soga ‘raise, lift’ entails that the action of ‘making’ is completed. (54) Ra r- aan jembatan soga -na. 3pl 3pl- make bridge raise -3sg.pfct ‘They have lifted the bridge.’ (lit. ‘They make the bridge being raised’)

The cause or the stimulus logically precedes the effect or result. The examples (55) are quoted from Arka (2000: 16), who first noted this feature in Lamaholot. (55) a. Na huda ana’ kréak tobo. 3sg order infant child sit ‘He orders the children to sit down (but the children did not, or not yet.)’43 b. Na huda ana’ kréak tobo -ka. 3sg order infant small sit -3pl.pfct ‘He has ordered the children to sit down (and the children did)’44

The verb’s aspect in (55b) tobo-ka (or tobo-na in other East Adonara variants) can be understood as a perfect of result: the children are seated at the moment of reference. Therefore, the verb of the main clause, huda ‘to command’ indicates the same aspect. Moreover, the perfect aspect marker kaé could be added to the end of sentence (55b), but not to (55a), where the event in the subordinate clause is not completed. 6. Split intransitivity and the Lamaholot dialect chain Holton (2008: 252) notes that “among the non-Austronesian languages of Eastern Indonesia one commonly finds pronominal systems which exhibit more than one morphological pattern for indexing single arguments of intransitive verbs. In such languages core arguments are semantically aligned. That is, the two patterns for marking the single argument of an intransitive verb are distinguished not only 43. Our translation from “Dia menyuruh anak-anak itu duduk (tetapi anak-anak itu belum/ tidak melakukannya)” (Arka 2000: 16). 44. Our translation from “Dia menyuruh anak-anak itu duduk (dan anak-anak itu sudah mela­ kukannya)” (Arka 2000: 16).

234 Philippe Grangé

formally, but also semantically. Active or agentive arguments follow one pattern, while stative or patientive arguments follow another.” Several split intransitivity examples in East Nusantara languages have been reported by Donohue (2001) and Klamer (2006, 2008). Among the nine East Nusantara languages showing split intransitivity surveyed by Klamer (2008), Split-S and Fluid-S only coexist in three of them: the Austronesian languages Taba and Dobel, and the non-­Austronesian language, Klon. Lamaholot can be added to this group. The Lamaholot language comprises many dialects and variants. Keraf (1978), on the basis of a lexicological survey, delineated 33 languages or dialects among which sharing the Swadesh list vocabulary can be as low as 44%, and not exceeding 89%. The Lamaholot linguistic area consists of a dialect chain where, although speakers of neighbouring dialects understand each other, speakers at the two extremes of the chain do not comprehend each other. Therefore, the most distant dialects, if compared to each other, may be considered different languages. Unfortunately, only a few Lamaholot dialects have been extensively documented. From the available data, it appears that the East Adonara dialect has the most complex inflection system, including attributive agreement, genitive agreement, paradigm of object pronouns, and a relatively complicated morpho-­ phonology (consonant epenthesis, vowel alternation). Moreover, the split intransitivity system in East Adonara can be regarded as a striking particularity, because it combines Split-S and Fluid-S. Under lexically fixed intransitive split, or Split-S, some verbs must agree with their subject, because S is considered as patientive (SP). These verbs belong to the following categories: Feeling verbs, Sentience verbs, Biological function verbs and Motion verbs (excluding movement/displacement). The traits that trigger agreement, or not, on an intransitive verb are displayed in the table below. Table 9.  Semantic factors relevant to Fluid-S Semantic traits

Intransitive verb morphology under Split-S

Agent Control Unaffected (no change of state) Displacement

S → SA (unmarked)

Motion Sentience Verbs of biological function Feeling and psychological states Affected (undergoing a change of state) No agent control

S → SP (verbal agreement)



Split intransitivity in Lamaholot 235

The East Adonara Lamoholot split intransitivity system matches with Dowty’s (1991) distinction between “Proto-Agent” and “Proto-Patient” properties.45 These general, cross-linguistic properties are displayed in the left column of the table below, and compared against Lamaholot in the right column. Table 10.  Proto-Agent and Proto-Patient properties underlying intransitive split in Lamaholot Proto-Agent properties

Proto-Agent properties in Lamaholot (S → SA)

a. volitional involvement in an event or state b. sentience (and/or perception) c. causing an event or change of state in another participant d. movement (relative to the position of another participant)

no; but rather ‘+ control’ perception only (see, hear…) yes yes: verbs of displacement

Proto-Patient properties

Proto-Patient properties in Lamaholot (S → SP)

a. undergoes a change of state b. incremental theme c. causally affected by another participant d. stationary relative to movement of another participant

yes yes, esp. with stative verbs yes yes: verbs of motion

In East Adonara Lamaholot, the main trait on which the split intransitivity system rests is ‘± control’. It is manifest when we consider the Split‑S as well as the Fluid‑S rules. This could testify to a former Lamaholot état de langue where the ‘± control’ trait accounted for an all-embracing fluid alignment system. It appears that in the Lamaholot dialects westward from Adonara (Eastern tip of Flores) and eastward (on Lembata Island), the split intransitivity systems are highly eroded. For instance, the verb for ‘eat’ is the only verb that inflects for person and number, and agrees with S and A arguments in Lewoingu dialect, see Nagaya (2010: Note 3), and the only documented occurrence of SP agreement (indicating that S is deprived of control) seems linked to the presence of a locative PP: tedu ‘to collide’, quoted in Section 4.1 from Nagaya (2010: 6). In the description of the Lamaléra dialect, the remains of a former split intransitivity system appear to be totally lexicalized, see for instance Keraf (1978: 111), and no occurrences of Fluid-S can be found. The one and only example of fai ‘water’ > faika ‘become water’ (Keraf 1978: 214), presented as a lexical derivation, recalls the East Adonara perfect aspect marking.

45. See also Arkadiev (2008: 10).

236 Philippe Grangé

Holton (2008: 275) highlights the cross-linguistic tendency of split intransitivity system “erosion”. In the East Adonara dialect, the split intransitivity system seems especially alive and comprehensive, typically retaining a Fluid-S, while it has faded away and/or been largely lexicalized in the ‘peripheric’ areas of the Lamaholot dialect chain, as represented in the Map 1. East Adonara Lamaholot: alive semantic alignment (Split-S and Fluid-S)

FLORES

SOLOR

Other Lamaholot dialects: eroded semantic alignment

LEMBATA

Lamaholot dialects chain

Map 1. The East Adonara dialect at the center of the Lamaholot dialect chain

It is unlikely that only a single East Adonara dialect evolved into having a more complex split intransitivity system compared to the neighbouring dialects. Besides, the remains of split intransitivity in the peripheral Lamaholot dialects appear to be very different from one dialect to another,46 therefore, they must originate from a morphologically more complex proto-language. If Lamaholot obeys the general tendency of an eroded split intransitivity system, we observe that the East Adonara dialect remains more conservative, typically retaining the Fluid-S and “freezing” the Split-S for certain intransitive verbs. Further investigations are needed to assess the diachrony of the Lamaholot dialect chain and the neighboring related languages, including Alorese (Bahasa Alor) in the north coast of the island of Pantar and in pockets along the bay of Kalabahi of Alor. Klamer (2012: 103) states that “the ancestors of the Alorese were Lamaholot speakers from Flores and/or Solor, who arrived at the coasts of Pantar around 1,300–1,400 AD.” One could suppose that the extension of the Lamaholot linguistic area (maybe from Adonara Island, still the most fertile and densely populated area today), which divided in many dialects and variants forming today’s dialect chain, is anterior to the eastward migration(s) to Alor. 46. However, to date we can only compare the Lamaléra dialect in Lembata (Keraf 1978) with the Lewoingu dialect at the southeastern tip of Flores (Nagaya 2009, 2010).



Split intransitivity in Lamaholot 237

Conclusion The East Adonara Lamaholot dialect displays a complex split intransitivity system, involving a Split-S (lexically fixed intransitive split) and a Fluid-S (fluid intransitive split) triggered by a series of contexts, and sometimes overlapped by aspect marking. Under split intransitivity, the S is either seen as ‘agent-like’ (SA), unmarked, or as “patient-like” (SP), which is marked by verbal agreement. The Split-S involves verbs expressing feeling, sentience, biological functions or motion verbs, which must agree with their subject. Verbs expressing mobility are divided between motion verbs (e.g. walk, swim, sit down) and displacement verbs, involving movement along a path (e.g. go, ascend, enter). Motion verbs agree with the subject, hence SP , and become telic when complemented by a locative PP; while verbs of displacement are optionally complemented by a locative PP. But this locative PP is compulsory when a verb of displacement agrees, to express perfect aspect. The Fluid-S rests principally on the “± control” trait: to express a lack of control over the event, the verb is marked as SP . Another important trait is “± affected”. Unlike many other semantic aligned languages, volition is not relevant, although it is not easy to distinguish it from the “± control” trait: volition implies control over the process, but the reverse is not true. Control only implies that someone can halt an ongoing process. In addition, the verbal agreement also expresses perfect aspect, and in some cases this aspect marking subsumes other Fluid-S inputs. The Lamaholot stative verbs can be employed attributively (as modifiers of a head noun or pronoun), or as predicates, and they are agree with the subject only to mark aspect. Serial intransitive verbs obey a rule of precedence: agreement on the second verb, when expressing perfect aspect, implies that the event referred to by the first verb is completed. Within the Lamaholot dialect chain, the island of Adonara occupies a central position, geographically speaking. This dialect appears to be the most conservative, with a complex inflectional morphology, whose most noticeable feature is split intransitivity. The peripheral dialects (eastern tip of Flores island, western and central part of Lembata Island) show various stages of eroded split intransitivity, with mostly lexicalized remains. Finally, East Adonara dialect may be a ‘textbook case’, which offers a wealth of split intransitivity features. It would even seem that this language displays all sorts of split intransitivity features that have been previously identified cross-­ linguistically.

238 Philippe Grangé

References Arka, I Wayan. 2000. Beberapa aspek intransitif terpilah pada bahasa-bahasa Nusantara: sebuah analisis leksikal-fungsional. In B. K. Purwo & A. Moeliono (eds.), Kajian serba linguistik, 423–454. Jakarta: Universitas Katolik Atmajaya. Arkadiev, Peter M. 2008. Thematic roles, event structure, and argument encoding in semantically aligned languages. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment, 101–117. Oxford: Oxford University Press.  doi:  10.1093/acprof:oso/9780199238385.003.0004

Arndt, Paul. 1937. Grammatik der Solor-Sprache. Ende: Arnoldus. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Creissels, Denis. 2007. Remarks on split intransitivity, septième Colloque de Syntaxe et Sémantique, Paris, http://www.ddl.ish-lyon.cnrs.fr/fulltext/Creissels/Creissels_(2007)c.pdf. Dixon, Robert Malcolm Ward. 1994. Ergativity. Melbourne: Cambridge University Press.  doi:  10.1017/CBO9780511611896

Donohue, Mark. 2001. Split intransitivity and saweru. Oceanic Linguistics 40(2). 322–336.  doi:  10.1353/ol.2001.0017

Donohue, Mark. 2008. Semantic alignment systems: What’s what, and what’s not. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment, 24–75. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199238385.003.0002 Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–619.  doi:  10.1353/lan.1991.0021

Fernandez, Inyo Yos. 1996. Persesuaian Subjek-Verba dalama bahasa Mai Brat dialek Ayamaru dan Lamaholot dialek Ile Mandiri: Studi Perbandingan Pengaruh Bahasa Non-Austronesia terhadap Bahasa Austronesia. In Jurnal Humaniora: Journal of Culture, Literature, and Linguistics III, 41–46. Yogyakarta: UGM. Grangé, Philippe. 2010. The possession in Lamaholot and neighbouring languages. The 6th International East Nusantara Conference on language & culture (ENUS (2010)), Kupang, 5–7 August, (2010). Holton, Gary. 2008. The rise and fall of semantic alignment in North Halmahera, Indonesia. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment, 252–276. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199238385.003.0010 Keraf, Gorys. 1978. Morfologi dialek Lamalera. Ende: Arnoldus. Klamer, Marian. 2006. Split S in the Indonesian Area: Forms, semantics, geography. The Tenth International Conference on Austronesian Linguistics (10-ICAL), Puerto Princesa City, Palawan, the Philippines, http://www.sil.org/asia/Philippines/ical/papers/Klamer-Split% 20S_Indonesian%20Area.PDF. Klamer, Marian. 2008. The semantics of semantic alignment in eastern Indonesia. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment, 221–251. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199238385.003.0009 Klamer, Marian. 2012. Papuan-Austronesian language contact. In N. Evans & M. Klamer (eds.), Documentation & Conservation Special Publication No. 5 (December 2012) Melanesian Languages on the Edge of Asia: Challenges for the 21st Century, 72–108. http://hdl.handle. net/10125/4561. Klamer, Marian & Michael C. Ewing. 2010. The languages of East Nusantara: An introduction. In M. Klamer & M. Ewing (eds.), Typological and areal analyses: Contributions from East Nusantara, 1–24. Canberra: Pacific Linguistics.



Split intransitivity in Lamaholot 239

Malchukov, Andrej. 2008. Split intransitives, experiencer objects, and ‘transimpersonal’ constructions: (Re-)establishing the connection. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment, 76–100. Oxford: Oxford University Press.  doi:  10.1093/acprof:oso/9780199238385.003.0003

Mithun, Marianne. 1991. Active/Agentive case marking and its motivation. Language 67(3). 510–546. doi: 10.1353/lan.1991.0015 Nagaya, Naonori. 2009. Subject and topic in Lamaholot, Eastern Flores. (oral presentation) 11th International Conference on Austronesian Linguistics (11-ICAL), Aussois, France, 22–29 June (2009). Nagaya, Naonori. 2010. Voice and grammatical relations in Lamaholot of eastern Indonesia. Proceedings of the workshop on Indonesian-type voice system, Research Institute for Languages and Cultures of Asia and Africa (ILCAA), Tokyo University of Foreign Studies (TUFS), July 17–18, (2010), http://lingdy.aacore.jp /doc/indonesia(2010)/DrNagaya_ proceedings.pdf. Nishiyama, Kunio & Herman Kelen. 2007. A grammar of Lamaholot, Eastern Indonesia. Muenchen: Lincom Europa. Pampus, Karl-Heinz. 2001. Mué Moten Koda Kiwan – Kamus Bahasa Lamaholot. Frankfurt: Frobenius-Institut. Van Valin, Robert D. 1990. Semantic parameters of split intransitivity. Language 66(2). 221–260.  doi:  10.2307/414886

Wichmann, Søren. 2008. The study of semantic alignment: Retrospect and state of the art. In M. Donohue & S. Wichmann (eds.), The typology of semantic alignment. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199238385.001.0001

Part 4

Norm variation in predicate-arguments relations

Geographic variation in a non-canonical infinitive structure with the modal verb brauchen Pascale Van Praet and Gilbert Magnus

Université de Normandie UNICAEN, CRISCO, EA 4255 / Lille 3, CeLiSo Paris-Sorbonne, EA 4084

The objective of the present paper is to demonstrate that it is possible to measure and situate an atypical construction. The focus will be on a “satellite” modal verb in modern German: the verb brauchen followed by an infinitive construction, and therefore marked or not marked as an infinitive complementation. Referring to descriptive grammars, the first part will summarize the typical construction of modal verbs in German and of the verb brauchen + infinitive and diverging opinions on the structure (nicht) brauchen + infinitive verb without any marker. The second part presents a corpus-based analysis of European newspapers in German-speaking regions over the last ten years areas. Observations and quantifications will show that there are important differences between Northern and Southern Germany, Southern Germany and Austria, Austria and Switzerland. The third part is a contextual study of non-marked infinitive constructions following brauchen which leads to syntactic, semantic and pragmatic analyses of the occurrences. In the pragmatic study we will consider the context of the occurrences so that written and oral conception will be discussed. The pragmatic study will show some recurrent speech acts. The last part reflects upon the semantics of brauchen inside the modality as one of the means to express necessity, similarities and differences with müssen, and at least the geographic factors that could explain the phenomenon of variation.

1. Introduction The questions of atypical syntactic structures, especially when they compete with normative structures, are a particularly important issue in syntactic analysis. The aim of the present paper is to study this phenomenon through the prism of the alternation between the presence versus the lack of a syntactical mark in the doi 10.1075/lis.33.10van © 2016 John Benjamins Publishing Company

244 Pascale Van Praet and Gilbert Magnus

construction of a peripheral modal verb in modern German, the verb brauchen (‘to need’). We focus especially on the way the infinitive phrase is introduced after this verb – that is, with or without the marker zu – which appears in very recent written discourse. In Section 1 of our paper we will approach the normative uses of the structure following the verb brauchen. We then make use in Section 2 of a large corpus to answer the following question: in which contexts and in which geographic areas does the verb brauchen + infinitive phrase appear without the marker zu? Quantitative and qualitative accounts will be discussed, and syntactic, semantic and pragmatic points of view will all be addressed. Section 3 will be concerned with putative considerations about the reasons behind the shift from a syntactically marked structure to a un-marked structure. Among these reasons, we will touch upon the issues of modality and variation. Section 4 concludes and looks forward about this question. 1.1

Normative uses of the verb brauchen

The first section presents the different structures of the verb brauchen in scientific grammars of modern German. The verb brauchen presents two types of construction. 1.1.1

A transitive construction with an accusative complement

(1) Allerdings entstehen gerade die gefährlichsten Unwetter sehr schnell, binnen weniger als einer halben Stunde. Sie fallen heute oft noch durchs Be­obachtungsraster von 15 Minuten. Wir brauchen daher neue Wetter­ satelliten, die ein Unwetter schon erkennen, bevor sich die dunklen Wolken zusam­menbrauen. Mit neuen Technologien und Beobachtungen könnte auch schon Stunden vorher vor dem Ereignis gewarnt werden. (Translation: We need therefore new weather satellites) (SPK/J08.00829 spektrum­direkt, 12.09.2008; Ein Gespür für Unwetter)

The subject of the finite verb can be +human or –human, but an impersonal subject is also possible. (2) Die bisherige systematische Bevorzugung von großen Parteien wurde abgeschafft, und jede Stimme hat bei der Sitzzuteilung gleich viel Gewicht. Sogar in entwickelten Demokratien braucht es also gelegentlich Anpassungen. (Transl. Even in developed democracies you sometimes need adjustments) (Die Zeit, 31.05.2012)

Geographic variation in a non-canonical infinitive structure 245



Potential objects are accusative noun phrases (animate or inanimate). This includes the majority of the examples with the verb brauchen (95% of a corpus of the verb brauchen, inanimate noun phrases 82% / animate noun phrases 13%): (3) „Wir brauchen noch Zeit, (Transl. We still need time to…) um uns einzuspielen“, so Trainer Bertl Hausmann (Burgenländische Volkszeitung, 21.07.2011) (4) Die burgenländischen Kfz-Betriebe locken mit einem neuen Karrieremodell mit dem Motto „Wir brauchen Dich!“ (Transl. “We need you!”)  (Burgenländische Volkszeitung, 28.02.2007)

1.1.2

A construction with an infinitive phrase (5 per cent of the aforementioned corpus)

(5) Die Europäer brauchen dennoch nicht mit dem Finger nach Fernost zu zeigen (Transl. However, Europeans don’t need to point a finger at the Far East): Viele Probleme könnten sie schon innerhalb der Europäischen Union angehen. Warum erlässt die EU kein Fangverbot für den Roten Tun in ihren Gewässern und für ihre Flotten? Oder unterbindet den Handel auf dem Staatsgebiet der Mitgliedsländer? (SPK/J10.00221 spektrumdirekt, 24.03.2010; Aus der Fisch)

On the one hand, the infinitive complement with the verb brauchen is restricted to negative particles nicht / kein or to restrictive particles as nur / erst. The same is found with the English modal need, which, as a modal verb directly followed by the infinitive without the mediation of a “preposition”, occurs in negative contexts such as: You need not worry (Duffley P. and Larrivée P., 1998): (6) People who have coverage and like their plan don’t need to do anything.  (Erin Hannigan, Health Care Campaign Manager,  on twitter @BarackObama, Oct. 10, 2013)

However, the English verb to need has also a modal version used with an infinitive introduced by to, which can be found in non-negative contexts such as (7). In French as well, the infinitive phrase is possible without a negative or a restrictive particle (8): (7) 7 things you need to know about food safety during the government shutdown.  (Rachel Sanders, BuzzFeed Staff, on twitter @BarackObama, Oct. 7, 2013) (8) J’ai besoin de me reposer aujourd’hui. (Transl. I need to rest today)

On the other hand, the prescriptive norm in German requires any infinitive complementation to be signaled by the integration of the marker zu, positioned before the infinitive phrase, except after the modal verbs (wollen ‘to want’, sollen ‘shall’,

246 Pascale Van Praet and Gilbert Magnus

können ‘can’, mögen ‘may’, dürfen ‘to be allowed to’, müssen ‘must’), the verbs lassen (‘to let’), lehren (‘to teach’), lernen (‘to learn’) and a few verbs of perception such as sehen (‘to see’) and hören (‘to hear’). The structure brauchen + infinitive phrase without zu, a deviation from the prescriptive rule, can be found in spoken German; in written German, however, it is perceived as an incorrect construction. Nonetheless, our corpus of examples from German newspapers points to a shift from a structure with zu complementation to a modal-like structure without the marker zu. Indeed, some German grammars state that, in a familiar register, brauchen + infinitive phrase without zu can be used. The construction mirrors that with a modal verb. To quote H. Brinkmann’s grammar (21971: 395–396): “Als negative Entsprechung zu müssen ist brauchen in das Modalfeld eingetreten.”1 Brinkmann cites Herbert Kolb (1964: 64–78)2 who had observed brauchen + infinitive phrase without zu starting in the early 19th century. However, in the corpus of Grimms’ Fairy Tales (first published in 1819) or in the corpus of Goethe’s works (1749–1832) found in the DEREKO corpus (IDS Mannheim3), no variation can be seen. More recently, H. Weinrich (1993: 300–301) goes so far as to speak about the modal verbs müssen and (nicht) brauchen: (9) Als einziges Modalverb der deutschen Sprache steht brauche (nicht) nach der Norm des guten Sprachgebrauchs immer vor einem Infinitiv mit vorangestellter Präposition zu, doch weicht die Umgangssprache nicht selten von dieser Regel ab. Beispiele: Keiner braucht mehr (zu) geben, als er hat. Du brauchst mir das nicht unbedingt (zu) glauben. (Transl. In correct use, the modal verb of the German language brauchen (nicht) is the only one which is placed before an infinitive with the preposition4 zu. But in the familiar use the rule is not always respected. Examples: Nobody needs to give more than he has. You don’t have to believe anyone.)

In the G. Zifonun / L. Hoffmann / B. Streckers’ grammar (1997: 1276), which is the reference grammar of the IDS Mannheim, the authors accord little space to

1. “As a negative counterpart of müssen, brauchen has fallen within the scope of modal verbs.” 2. Herbert Kolb. 1964. „Über brauchen als Modalverb: Beiträge zu einer Wortgeschichte“. In Zeitschrift für deutsche Sprache 20, 64–78. 3. IDS: Institut für deutsche Sprache. 4. We would not speak about a preposition in this case.



Geographic variation in a non-canonical infinitive structure 247

irregularities concerning the use of brauchen + infinitive phrase and point out two features that allow brauchen to be distinguished from modal verbs: (10) Brauchen unterscheidet sich von den Modalverben im engeren Sinne im Wesentlichen nur durch zwei Merkmale: – Es enthält in der Regel den Infinitivbestandteil zu.5 – Es ist auf den Gebrauch in kommunikativen Einheiten mit negativem oder einschränkendem Charakter beschränkt oder auf kommunikative Minimaleinheiten, aus denen auf die Negation eines thematisierten Sachverhalts geschlossen werden kann. (Transl. Brauchen can be distinguished from modal verbs stricto sensu essentially with respect to two main aspects: – brauchen will be used with the infinitive marker zu – its use is limited to communication units with negative or restrictive connotation or to minimal communication units that allow us to identify a denied state of things.)

In Richtiges und gutes Deutsch (Duden 9, 62007: 185), it is stated that (11) Besonders in der gesprochenen Sprache wird das zu vor dem Infinitiv oft weggelassen, d.h., verneintes oder eingeschränktes brauchen wird wie verneintes oder eingeschränktes müssen verwendet: Du brauchst nicht kommen = Du musst nicht kommen. Du brauchst erst morgen anfangen = Du musst erst morgen anfangen. […] In der geschriebenen Sprache wird das zu vor dem Infinitiv meistens noch gesetzt: Du brauchst nicht zu kommen. Du brauchst erst morgen anzufangen. (Transl. Especially in the spoken language one avoids zu before the infinitive, the denied or restricted verb brauchen is used as well as müssen: You don’t need to come = You must not come. You need to begin only tomorrow = You must begin only tomorrow. […] In the written language, zu is placed before the infinitive in the most cases.)

Two points are highlighted here: the equivalence between nicht brauchen and nicht müssen, which we verified by means of a mini-survey of lecturers at our universities, and the marking with or without zu, a differentiation that could be linked to a differentiation between written and spoken language which we also have to verify in our corpus. A. Rousseau (2003: 804), finally, approaches the semantic value of brauchen in his article on modal verbs in German and considers earlier states of the language:

5. One counter-example excerpted from the FAZ (Frankfurter Allgemeine Zeitung).

248 Pascale Van Praet and Gilbert Magnus

(12) Transl. A constant evolution leads from “need” to “necessity”, as we have already demonstrated for Gothic, where þaurban “to need” generates in some way skulan “must”. Modern German resumed this relationship and introduced formal renewals but without interrupting the semantic continuity: bedürfen “to need” is renewed by brauchen which became the equivalent of müssen in a negative utterance as in the examples: He doesn’t need to come. He must not come.6

2. Corpora analysis 2.1

Corpus 1a (Germany)

Corpus 1a is based on the electronic corpus DEREKO developed by IDS in Mannheim, widely known by linguists interested in German. It focuses on the German press in Northern, Central and Southern Germany: Newspapers

Northern Die Zeit Germany Hamburger Morgenpost

Years

Regional/ % of oc- Number of national press currences occurrences without zu

2012 national press 2005–2007 regional press 2011

6%

100/corpus total: 3121

Central Berliner Morgenpost 1998 regional press Germany Braunschweiger Zeitung 2009–2011 regional press Hannoversche Allgemeine 2009–2010 regional press

11%

100/corpus total: 8981

Southern Nürnberger Nachrichten Germany Nürnberger Zeitung

21%

100/corpus total: 3503

2010–2011 regional press 2011 regional press

Nicht brauchen without zu is used in the field of sport (in spoken German or in comments, 15.8% of the identified occurrences without zu7) or in the field of 6. Une évolution constante conduit du “besoin” à la “nécessité”, comme nous l’avons déjà démontré pour le gotique où þaurban “avoir besoin” génère en quelque sorte skulan “devoir”. L’allemand moderne a repris cette relation en procédant à des renouvellements formels sans toutefois interrompre la continuité sémantique: bedürfen “avoir besoin” est renouvelé par brauchen, qui est devenu l’équivalent de müssen dans un énoncé négatif, comme dans les exemples: Er braucht nicht kommen. “Il n’a pas besoin de venir”. Er muss nicht kommen. “Il n’est pas obligé de venir”. 7. By each indication of a percentage in this section, we always mean a percentage of the identified occurrences without the marker zu.



Geographic variation in a non-canonical infinitive structure 249

musical entertainment (quoting the statements of others or interviews in spoken German in the weekly quality newspaper Die Zeit under the heading “Philosophy”, 13%): (13) Als Ausrede wollte die schwache Schiri-Leistung (MOPO-Note 5) beim HSV keiner zählen lassen. Kompany: „Nach so einer ersten Halbzeit brauchen wir uns über gar nichts beschweren.“ (Transl. After such a first half we don’t need to complain about anything) In guter Erinnerung wird er in Hamburg aber sicherlich nicht bleiben. (HMP08/FEB.00295 Hamburger Morgenpost, 04.02.2008, S. 5;  Spot an Günter perlt überhaupt nicht) (14) Um den Nachwuchs für die U-19-Bundesliga braucht man sich in naher Zukunft nicht sorgen (Transl. In the near future you won’t need to worry about the succession in the U-19 Bundesliga). In der neuen U-16-Bundesliga haben es die Basketballer des Teams Braunschweig / Wolfenbüttel nämlich überzeugend in die Play-offs geschafft. Nun wollen sie mehr.  (BRZ10/FEB.13586 Braunschweiger Zeitung, 27.02.2010;  Basketball-Nachwuchs begeistert von neuer Liga) (15) In der Tat. Beim „Eurovision Song Contest“ im finnischen Lordi-Land sind auch diesmal einige Exoten am Start. „Ich habe mir alle 41 Beiträge vorab angesehen. Ein Erlebnis. Allerdings gibt’s keine Monster“, erklärt Roger Cicero augenzwinkernd. Sein Fazit: „Musikalisch brauchen wir uns nicht verstecken.“ (Transl. From a musical point of view we don’t need to hide). Die stärkste Konkurrenz hat er aber auch schon ausgemacht: Irland, die Schweiz mit DJ Bobo, Belgien, die Niederlande und Israel, wobei die letzten vier noch das Halbfinale am 10. Mai überstehen müssen. Im Finale treten dann 23 Konkurrenten gegen unseren Mann aus Ottensen an.  (HMP07/MAI.00215 Hamburger Morgenpost, 03.05.2007, S. 20;  Rogers „Grand Prix“-Konkurrenz) (16) Safranski: […] Unsere Gäste brauchten nicht als Inhaber und Bewirtschafter ihrer Meinungen auftreten (Transl. Our guests didn’t need to pose as owners and operators of their opinions), sie durften von sich selbst überrascht werden. Sloterdijk: Sie kennen sicher Sebastians Haffners Buch Erinnerungen eines Deutschen. Darin wird berichtet, wie gebannt die 14- bis 16-Jährigen während des Ersten Weltkriegs die tägliche Frontberichterstattung verfolgten. (Die Zeit 10.05.2012)

Readers’ letters taken from newspapers, insofar as they are a forum allowing people to let off steam, so to speak, are another genre wherein a certain looseness in the use of nicht brauchen without zu appears (5.25%). The same can be said for

250 Pascale Van Praet and Gilbert Magnus

the column ‘The Question of the Day’. It is an opportunity for the readers to speak about issues which affect them. Such cases are very close to the spoken medium. (17) Ein Ehrenkodex für ausscheidende Politiker? Das heißt im Umkehrschluss, dass aktive Politiker sich an keinen Ehrenkodex halten brauchen (Transl. … That is, on the other hand, active politicians don’t need to hold on to a code of honour). Wer so eine Diskussion ins Rollen bringt, merkt offensichtlich schon nicht, wie lächerlich er sich macht. (HMP05/DEZ.01578 Hamburger Morgenpost, 15.12.2005, S. 7; Leserbriefe) (18) „Im Allerpark braucht nicht noch mehr gebaut werden (Transl. In Allerpark you don’t need to build any more). Wenn kein Geld da ist, sollte man hier anfangen zu sparen. In Wolfsburg wird zuviel gebaut. Man sollte keine neuen Wohnhäuser errichten, sondern die leer stehenden Wohnungen nutzen.“ (BRZ10/JAN.12630 Braunschweiger Zeitung, 30.01.2010;  Umfrage des Tages Wo soll die Stadt sparen?)

Some uses of nicht brauchen can be found in contexts where advice concerning new technologies is given (10.5%). We consider that the use of nicht brauchen instead of müssen will be perceived by the receiver of the message as a less binding suggestion: (19) Apple hat einen Patch bereitgestellt, der das bei MacBooks aufgetretene Problem des plötzlichen Herunterfahrens beheben soll. Nun brauchen die betroffenen Rechner nicht mehr zur Reparatur gegeben werden (Transl. Now you no longer need to bring in the computers concerned for repair). Info: www. apple.com/suport/downloads. (HMP06/NOV.00636 Hamburger Morgenpost, 07.11.2006,  Beilage S. 4; IN KÜRZE) (20) Mit technologisch wie optisch immer anspruchsvolleren Modellen treten die großen internationalen Automobilhersteller den Kampf um Marktanteile in allen Leistungsklassen vom Kleinwagen bis zur Luxuslimousine an. Das wurde im Vorfeld des heute beginnenden 100. Pariser Automobil-Salons bei der Präsentierung der Weltneuheiten deutlich. Autofans aus Berlin und Brandenburg brauchen jedoch nicht unbedingt an die Seine reisen, um Neuheiten und interessante Studien in Augenschein zu nehmen. (Transl. Motoring fans from Berlin and Brandenburg don’t necessarily need to travel to the Seine to see new car models and interesting studies). Zahlreiche Exponate werden auch demnächst auf der „AAA“ unterm Funkturm gezeigt. (L98/SEP.09807 Berliner Morgenpost, 30.09.1998, S. 32, Ressort: QUER;  Kampf um Marktanteile in allen Klassen)



Geographic variation in a non-canonical infinitive structure 251

Other contexts without the marker zu in the infinitive phrase include the following: economy (21%), everyday life (15.8%), politics (5.25%) and, finally, popularized medical contexts (only 2.6%). We also observe a shift towards the structure without zu in some purely syntactic patterns, as in the following particularly noteworthy example: (21) Ich habe ein bisschen darauf spekuliert, dass Geromel den Ball rausköpft, bin nachgegangen – und plötzlich flog der Ball mir wieder vor die Füße. Da habe ich einfach abgezogen. Es gab noch einige gute Gelegenheiten, um das beruhigende zweite Tor zu erzielen … Wir hätten alle gerne das 2:0 gehabt, dann hätten wir nicht am Ende so zittern zu brauchen (Transl. We all would have loved to achieve a 2:0 score, then we wouldn’t have had to tremble so much at the end of the match). Aber es hat gereicht – und allein das zählt. Sie hatten sehr viele gute Szenen und nur ganz wenige Fehler in Ihrem Spiel. Interview: Christian Purbs. (HAZ09/NOV.00091 Hannoversche Allgemeine, 02.11.2009)

Standard German would require dann hätten wir nicht am Ende so zu zittern brauchen, if one considers that the verb brauchen behaves as a modal verb and that the perfect is built with the auxiliary haben (here in the Konjunktiv II) + the past participle in the form of an infinitive in the presence of the infinitive zittern (as in example (22)). A certain casualness results from the speaker omitting to place the marker zu before the real infinitive, however; it is put instead before the past participle in the form of an infinitive. Unless we have an amalgam of the structure we have described above and the periphrastic structure of the modal verb müssen which is haben zu (to have to), it would show a lack of mastery of the written norm. (22) „Um zu bewahren, müssen wir erneuern“, sagte Gröhe. Unter Führung des hessischen Fraktionschefs Christian Wagner hatten vier Landespolitiker Merkel vorgeworfen, sie habe das mäßige Abschneiden der CDU mit ihrem „präsidialen Stil“ zu verantworten, der die CDU um ihr Profil gebracht habe. In Berlin richtete sich der Unmut vor allem gegen die Kritiker. Über „ewig wiederkehrende und fruchtlose Konservatismusdebatten“ wurde geschimpft. Pikiert wurde vermerkt, dass einer der Kritiker, der Thüringer Fraktionschef Mike Mohring, auch in der Rechts-Postille „Junge Freiheit“ für mehr Nationalstolz plädiert. Niedersachsens Ministerpräsident Christian Wulff verwahrte sich gegen allzu schnellen Schulterschluss. „Wenn jeder glaubt, er habe alles richtig gemacht, hätten wir gar nicht zusammenkommen brauchen.“ (Transl. If everyone believes he has done everything properly, we wouldn’t have had to meet). Man müsse jeden ernst nehmen, der etwas zu sagen habe. (HAZ10/JAN.02089 Hannoversche Allgemeine, 15.01.2010;  Wulff: Wir sollten Merkel entlasten)

252 Pascale Van Praet and Gilbert Magnus

2.2

Corpus 1b (Austria and Switzerland)

Here we focused on finding occurrences of the verb nicht brauchen  + infinitive phrase in newspapers published in Switzerland and Austria between 1999 and 2011. Our aim was to closely examine the contexts in which the structures nicht brauchen + infinitive phrase appear with or without the marker zu and to analyse them. The two tables below show contrasting information about the selected occurrences. Table 1 Country

Newspapers

Years

Regional and % of ocGlobal supra-regional currences number of press without the occurrences marker zu

2000– 2011

Supra-regional and regional press: 2%

2%

100/total corpus: 2200

Die Presse 2000– Niederöster. Nachrichten 2011 Burgenländ. Volkszeitung Vorlarberger Nachrichten Tiroler Zeitung

Supra-regional and regional press: 96%

47%

100/total corpus: 3600

Switzerland St Galler Tagblatt die Südostschweiz Zürcher Tagesanzeiger Austria

Table 2 Switzerland

Years

% without the Austria marker zu

Zürcher Tagesanzeiger

2000

0% (0/12)

St Galler Tagblatt 2000 2001 2003 2007 2008 2009 2010 2011

3.5% (2/82)

Die Südostschweiz 2010 2011

0% (0/6)

Years

% without the marker zu

Die Presse

2000

4% (4/10)

Tiroler Zeitung

2000

4.5% (1/22)

Niederöster. Nachrichten

2007–2011 60% (6/10)

Burgenländische Volkszeitung

2007–2011 50% (24/48)

Vorlarberger Nachrichten

2000

2% (2/10)



Geographic variation in a non-canonical infinitive structure 253

2.2.1 Quantitative observation of the Swiss and the Austrian corpora It must first be pointed out that the occurrences of nicht brauchen without the marker zu show an obvious quantitative difference between the Swiss corpus (2%) and the Austrian corpus (47%). This will be further discussed in Section 3. Differences appear within each corpus, however – made up of mini-corpora. It was not possible to find equal occurrences in the different Swiss newspapers, for example, because the electronic corpus DEREKO does not provide the same selection of years for each newspaper. For the newspaper Zürcher Tagesanzeiger only the year 2000 is available: only 10 occurrences of the structure nicht brauchen + infinitive phrase were found. Conversely, the newspaper St Galler Tagblatt spans eight years: 82 occurrences of the structure nicht brauchen + infinitive phrase were found. We can also see comparable discrepancies in the data of the Austrian corpora: 10 occurrences of the structure nicht brauchen + infinitive phrase in the newspaper Niederösterreichische Nachrichten and 45 occurrences of the structure nicht brauchen + infinitive phrase in the newspaper Burgenländische Volkszeitung. The quantitative analysis brings to the fore a notable mismatch between Switzerland and Austria, although both countries are located south of Germany. Collateral questions about the nature of the corpora and of the utterances will now be discussed. 2.2.2 Qualitative observations concerning the Swiss and Austrian corpora Taking as a point of departure the empirical observation of both corpora, we present below our analysis of the occurrences of the structure nicht brauchen + infinitive phrase using four fairly traditional analytical approaches: syntactic, semantic, contextual and pragmatic. A syntactic analysis of both the Austrian (A) and the Swiss (S) corpora reveals that the first position in a sentence may be a facilitating context for the lack of the marker zu in the infinitive phrase: (23) Wir haben das beste Team. Wir müssen nur weiterhin diszipliniert auftreten. Unsere Torhüter sind die besten der Liga. Wir haben vor allen Gegnern Respekt, mehr nicht. Verstecken brauchen wir uns nicht. (Transl. We don’t need to hide.)  (St. Galler Tagblatt, 21.01.2010, S. 41;  SCH: Lust auf mehr.) (S) (24) „Wünschen brauchen wir uns bei dem Wahlergebnis nichts mehr“, (Transl. We no longer need to wish for anything else in the results of the election), sagt Karg. (Vorarlberger Nachrichten, 04.04.2000, S. A5, Ressort: Lokal;  Für Bregenzer FPÖ wird Werner Karg in den Stadtrat einziehen) (A)

254 Pascale Van Praet and Gilbert Magnus

Given only three occurrences in the whole corpus, we cannot draw a generalizable conclusion as to reoccurrences of the phenomenon, however. A semantic analysis could be a more successful line of research insofar as it raises core questions about the significance of the verb (nicht) brauchen as a modal verb. The basic meaning of the verb brauchen with a complement (etwas brauchen / to need something) leads directly to the meaning of necessity. When used modally, brauchen is used only with negative or restrictive elements; it thus conveys non-­ necessity. But is there a semantic difference between the use with or without the marker zu? There would appear to be none. One consequence is that occurrences with the marked and non-marked form of the structure nicht brauchen + infinitive phrase can be found in the same sentence: (25) „Unsere Mannschaft ist derzeit sehr gut drauf. Außerdem, wir brauchen uns zu Hause vor keinem Gegner verstecken, noch zu fürchten.“ (Transl. At home we don’t need to hide behind any opponent or to fear.) (BVZ08/APR.01777 Burgenländische Volkszeitung, 16.04.2008, S. 68;  STIMMEN & SPRÜCHE)

A contextual analysis aims to characterize the text types from which our occurrences are taken. Based on Adamzik (2004), who provides a list of all parameters which define a text, we recognize that most of the attempts to relate communication domains with styles of texts are unsuccessful. Some parameters are strong enough to have a constant influence on the typology of texts, however; one such parameter is the situative context (der situative Kontext). We follow Adamzik, and claim that this includes (1) the domain of communication or the context of interaction in which the text is produced (here, the written press), (2) the local and temporal situation (here, the local origin of the newspapers and the year of their publication), (3) the respective position of the participants (die Partnerkonstellation) (here, the relationship between the political direction of the newspapers and the local or national population) and (4) the conversational situation (die Redekonstellation) (here, the relationship between the journalists and the potential readers of the newspapers). Seen from this perspective the data from the corpora are significant. The Swiss corpus gives only two occurrences of the structure nicht brauchen + infinitive phrase without the marker zu; the occurrences are found neither in the supra-­national newspaper Zürcher Tagesanzeiger, nor in the other regional newspaper of German-speaking East-Switzerland (Südostschweiz) but in another regional newspaper, St Galler Tagblatt. Relatively speaking, the Austrian corpus gives as many occurrences in the supra-­regional newspaper die Presse (40% of the sub-corpus) as in the regional newspapers, such as die Niederösterreichischen Nachrichten (60% of the



Geographic variation in a non-canonical infinitive structure 255

sub-­corpus), die Burgenländische Volkszeitung (50% of the sub-corpus), die Vorarlberger Nachrichten (20% of the sub-corpus), and die Tiroler Zeitung (4.5% of the sub-­corpus). The interpretation of this data leads us to believe that the differences in language use between the supra-regional and regional newspapers in Austria are insignificant. We then looked for more relevant parameters: one parameter concerns the fact that most occurrences appear in a context of “oral text” – which has to be analyzed more closely – and the other parameter concerns the newspaper sections that provide the larger context from which the short texts are extracted. The Swiss occurrence ((25), above) is a typical example of quoted words. In the Austrian corpus, we found only 6 examples which are definitely not citations. All the other occurrences are citations, but in a “journalistic” style: directly quoted with quotation marks or indirectly quoted without quotation marks, the name of the speaker is followed by a colon (36% of the whole corpus). But such a “journalistic” style also implies that the reader is never sure whether the words are reproduced exactly or paraphrased. This result is a hybrid style which borrows from both oral and written styles. These remarks are directly related to the newspaper sections in which they were found. The main newspaper section in which we found our occurrences is sport, as in example (26): (26) Die Zielvorgabe beim Leader ist jedenfalls der Titel. Pressesprecher Ernst Laczko: „Wenn wir vom Verletzungspech verschont bleiben, brauchen wir uns vor keinem Gegner fürchten. (Transl. If we aren’t jinxed with an injury, we don’t need to fear any opponent) Wir sind auf den Meistertitel eingestellt.“  (Burgenländische Volkszeitung, 25.03.2010;  Winterkönig ist der große Meisterfavorit) (A)

The Austrian corpus, which contains most of the occurrences of our structure, is more representative of the variety of the newspapers sections. Among the different sections, the sports section provides the major examples of the structure without the marker zu: this structure is four times more likely to exist here than in the whole Austrian corpus. The other sections are society (33%), politics (6%) and economy (4%). We analyzed occurrences where no oral quotes appear as well. The analysis of the “written” occurrences shows up in sections such as the horoscope, readers’ letters, the day’s (or week’s) pearl of wisdom, advertising, sports or tips as in examples (27) and (28):

256 Pascale Van Praet and Gilbert Magnus

(27) Wenn Sie beim kommenden Ball Schwung ins Gespräch an der Bar bringen wollen, dann brauchen Sie nur den Satz einwerfen: „Schon gelesen, in Eisenstadt gibt‘s heuer kein Stadtfest?“ (Transl. At the next ball, when you want to give impetus to your conversations at the bar, you just need to say the following sentence:…) (Burgenländische Volkszeitung, 16.01.2008, S. 7;  Eigentor oder Tor des Jahres?) (28) Ballsaison 2007 / Die Veranstalter brauchen sich keine Sorgen machen, Bälle bleiben „in“ (Transl. The organizers don’t need to worry, the balls are still “in”.) (Niederösterreichische Nachrichten, 01.01.2007, S. 3;  Interesse ist nach wie vor ungebrochen) (29) Se brauchen ned glauben, dass wir Burgenländer ned vorbereitet auf des Haydn-­Johr san. Sogor die Taxler san geschult worden, wos es mit dem Haydn-­Johr auf sich hot. Wobei, so a Gorschtiger hot dann im Internet gepostet, dass nur zwa Prozent der Taxler wissen, wer der Herr Haydn war. De tatn eahm eher für den Chef vom Haydn-Stüberl halten. Ich glaub, des stimmt ned, und außerdem wissen‘s a, wo die Haydn-Gossn is. (Transl. You don’t need to believe that we people from Burgenland are not prepared to the Haydn-year.)  (Burgenländische Volkszeitung, 01.04.2009, S. 16; GMOA-TRUMMLER)

The above examples, quickly regarded as “oral” or “written” examples, merit further comment. We share the views developed by Koch and Österreicher (1985), who substitute the “frozen” concepts of written and spoken texts with the concept of medial and conceptual written and oral expression (mediale und konzeptuelle Schriftlichkeit und Mündlichkeit). They imagine a graduated scale between a language of proximity and a language of distance (Sprache der Nähe und Sprache der Distanz). According to this concept, the oral conception of some texts may have influenced their form and the choice of some structures such as the structure nicht brauchen without any marker. The conclusion relativizes the remarks made in the first section of this paper about the use of the structure as typical of a colloquial register (Umgangssprache). This is not the case when a press officer comments on a match, because he represents his institution for the public, but he may use oral structures extracted from an oral use rather than those used in written texts, as suggested by Koch and Österreicher’s concepts of language of proximity and a language of distance. We take our analysis one step further and observe the utterances with nicht brauchen in their immediate context. For our pragmatic analysis, we turn to the utterances and their immediate context. It results from the different occurrences that, before expressing the non-­necessity of the process, the speaker uses to establish an overview of it. The illocutionary act here consists for the quoted speaker – the coach of a team, a

Geographic variation in a non-canonical infinitive structure 257



commercial manager of a company or a reporter – in establishing an overview of certain elements of information. These elements may be found in the preceding or following context and help the speaker to achieve an outcome about the necessity or non-­necessity of the overall (nicht) or partial (nur / kaum / fast) achievement of the process. These elements form the external conditions of the achievement of the process as in the following example: (30) „Das ist eine Gewinnsituation für beide Orte, die Ebergassinger brauchen nicht mehr ständig hinfahren und die Verklausungen ausräumen, wir müssen nicht ständig nach dem Rechten sehen“, (Transl. It’s a winning situation for both villages, the Ebergassinger don’t need to drive and evoke jams, we mustn’t look all the time to the right) meint Gramatneusiedls Amtsleiter Andreas Tremml.  (Niederösterreichische Nachrichten, 03.01.2007, S. 30;  Dank Kooperation gelingt die Sanierung)

Furthermore, the intentions of the speaker show different perlocutionary acts such as: – well-intended advice: (31) Mütter, die all das erleben wollen, brauchen einfach nur nebenstehenden Kupon ausschneiden, zur NÖ Landesausstellung mitbringen und beim NÖN/ BVZ-Corner vorlegen. (Transl. Mothers who want to experiment all these things, simply have to cut out the attached coupon and bring it to …) Gültig am 8. Mai von 13 bis 16 Uhr. www.noe-landesausstellung.at  (Burgenländische Volkszeitung, 05.05.2011; Gratis Kaffee und Kuchen)

– advice, reinforced by a threat: (32) „Nachtnebel und Ulram brauchen mir nicht sagen, was ich zu tun habe und was nicht. (Transl. Nachtnebel and Ulram don’t need to tell me what I have to do or not.) Wir nehmen unsere Verantwortung wahr. Die Halbturner ÖVP soll doch endlich selbst aktiv werden.“ (Burgenländische Volkszeitung, 14.02.2007, S. 26;  Umfahrungsstraße sorgt für heftigen Polit-Streit)

It can be observed in a context of advice a great number of collocations. We find the structure nicht brauchen without the marker zu with predicates which express fear, such as Angst haben (‘to be afraid‘), sich scheuen (‘to shy’), sich Sorgen machen (‘to worry about’), sich fürchten (‘to fear’). These collocations represent one third of all occurrences in the Austrian corpus.

258 Pascale Van Praet and Gilbert Magnus

3. Theoretical accounts The discussion of what represents an “incorrect” use for the prescriptivist is in direct relation with questions surrounding typicality and atypicality. In this part we will discuss the probable reasons accounting for why an atypical structure exists, and why it is maintained and extended. A first reason is related to the semantics of the verb brauchen. The transition from the status of a typical transitive verb – with an object complement or an infinitive complement with the marker zu – to the status of a modal verb, with its typical morphological and syntactic characteristics (lack of the marker zu) attests to the semantic integration of nicht brauchen into the group of modal verbs in the domain of non-necessity. It should be considered a consequence of this semantic integration, given that this phenomenon is unstable and locally very variable. This integration operates in the larger frame of “enlarged modality”. Approximately fifty verbs could be integrated into this frame including lassen, heißen, drohen etc… as mentioned by French scholars as F. Raynaud (1975), Schanen / Confais (21989, 1986), Rousseau (2003). Whereas a great number of verbs in German are tending towards the status of a modal verb on a semantic level, none of them attest the same degree of integration through a morpho-syntactic sign (here the marker Ø) as nicht brauchen. In the same order of meaning (necessity), the verb haben becomes a modal verb if and only if it is employed with the marker zu. Other scholars such as Palmer (1986, 2001) and van der Auwera and Plungian (1998) integrate in their studies speech acts and modal particles. There are terminological differences between Van der Auwera (1998) and Nuyts (2011) for dynamic and deontic modality. But for the modal category, they agree to distinguish “needs / necessities which are fully inherent to the first argument participant (henceforth called participant-inherent dynamic)” […] from “needs  / necessities which are determined by the local circumstances […] of that participant (we call this participant-imposed dynamic)” (Nuyts 2006: 3). These external factors can be either explicit, such as in (25), (26), (28) and (29), or implicit such as in (27). Second is the question of whether there are in modern German two modal forms for one use and the specification of their uses. Nicht müssen exists beside nicht brauchen. Do they have complementary uses or is it possible to substitute one for the other one without any semantic or pragmatic difference? Brinkmann speaks of nicht brauchen as an equivalent of nicht müssen. Duden 9 considers nicht brauchen and nicht müssen as synonyms as well. Zifonun et al. (1997: 1903) adopt the same point of view:



Geographic variation in a non-canonical infinitive structure 259

(33) Außerdem lautet die reguläre äußere Negation von müssen nicht nicht müssen, sondern nicht brauchen. Nicht brauchen ersetzt nicht müssen (äußere Negation) in allen Verwendungsweisen (epistemisch, circumstanciell und normativ). (Transl. Moreover, the regular external negation of müssen is not nicht müssen but nicht brauchen. Nicht brauchen replaces nicht müssen in all its uses (epistemic, circumstantial and normative).)

Contrary to these claims, however, we observed in our corpus that the two verbs do in fact have their own uses and specificities. First, we found the two modal verbs in the same extract as in (30). Different conclusions can be made about the two occurrences of the non-necessity through two different forms: for stylistic reasons – to avoid repeating the same verb, although this is not as problematic in German as it is said to be in French, for example; for semantic reasons – but here the speaker expresses the non-necessity of doing something, with an equivalent conversational background; for pragmatic reasons – the speaker aims to nuance his perception of non-necessity in both cases. The speaker aims here to distinguish with different verbs a more normative conversational background for the population of his village (nicht müssen) than for the population of the neighbor village (nicht brauchen). In order to analyze the meaning of the two modal verbs of non-necessity, we tested six native German native speakers (three from North and East Germany; three from Austria (two from Salzburg and one from Innsbruck)) with a set of 10 utterances from our corpus, containing either müssen or brauchen in the original corpus. The native speakers were asked to complete the gaps of the sentences either with müssen or brauchen. We conclude that there is an unstable use of the two modals by the native speakers: we never found for the same utterance a clearcut preference for either nicht müssen or nicht brauchen. This leads us to conclude that the perception of the context can lead to contradictory results. When some speakers comment upon their choice, they recognize a pragmatic aspect to nicht brauchen, namely that the tone of the speaker is less direct and that their choice of nicht müssen depends on the clear perception of an external normative context which points, on the one hand, to the widely used notion of deontic modality (cf. Palmer 1986: 98–100 and 22001: 91): (34) John mustn’t come tomorrow (= There is necessity not to come) John needn’t come tomorrow (= There is no necessity to come)

and, on the other hand, to the notion of ‘participant-external modality’ used by Van der Auwera and Plungian in their seminal article “Modality’s semantic map” (Linguistic Typology 2, 1998: 79–124).

260 Pascale Van Praet and Gilbert Magnus

In fact, each verb seems to keep its core semantic features. The basic semantic feature of müssen is the expression of the norm, first defined by Aristotle, or re-defined by Le Querler (2001) or Chapusot (2002) as aletic modality: “The subject and the behavior of a subject are only conditioned by natural rules and do not depend at all of any norm set by any individual or group of individuals.” Furthermore we have the intuition that there is a graduation between a strong and a weak necessity, graduation that cannot be defined within a modal logic as evoked by Portner (2010)8 but within a pragmatic study as shown in the analysis of our corpora and of the native speaker-tests. The modal verb nicht brauchen seems to maintain the agentive characteristic of the transitive verb, the result being that the feature of agentivity is still discernable in its use and may generate the pragmatic value of advice. We thus conclude that there is a differentiable use of the two modal verbs of non-necessity. Finally, it is necessary to take into account the geographical context of our data. Concerning the presence or the absence of the marker zu in infinitive phrases after nicht brauchen, the native-speaker test we employed reinforces the empiric observations of the corpora: no native speakers from Germany used nicht brauchen without the marker zu, but some Austrian native speakers used the structure without the marker. In the corpora, we noticed an increase in the number of occurrences of the structure nicht brauchen  + infinitive phrase without the marker zu in texts published in the South, inside of the Federal Republic of Germany, and above the borders between South Germany and Austria. We are not surprised to notice a phenomenon of variety in the German spoken space. German is indeed such as English, French, Spanish or Portuguese a “multicentralized” language as evoked by Ammon (2004). The concerned countries, Austria, the German speaking Swiss and Germany have their particularities so that the concept of “variety” is consequently well adapted, but the most varieties are focused “almost not on grammar but partly only on the vocabulary and the pronunciation”.9 Within the phenomenon of variety in German – where grammatical variety is infrequent – emerges clearly an axis: a North-South-axis. Switzerland, 8. “Sometimes people have the intuition that there should be more options between than (necessity modals) and (possibility modals). For example, deontic should seems to be weaker than deontic must, but certainly closer to must than may. […] However, it’s not clear how to define a concept of weak necessity, as contrasting with regular necessity, in a way that makes sense within a basic framework of modal logic” (Portner, p. 32). 9. „Im Gegensatz zu verschiedenen Sprachen unterschieden sie(Varietäten) sich kaum in der Grammatik und nur teilweise im Wortschatz und in der Aussprache. Vor allem an diesen Besonderheiten (Varianten) im Wortschatz und Aussprache können die Angehörigen einer Varietät erkannt werden“ (Ammon, 2004, p. XXXI–II).



Geographic variation in a non-canonical infinitive structure 261

however, is a special case. Switzerland is a federal country with several official languages: French, German, Italian and Romanch. German in Switzerland is in a diglossic relationship with the local dialect: the standard language (Hochdeutsch) spoken and written at school, in public administration and in the media, and dialects, spoken in all other everyday situations. The standard language (Hochdeutsch) is therefore conservative and normative as explained by Fernande Krier (2010: 65–66), who compares the use of the Alemannic dialect with the use of standard German: (35) En Suisse alémanique précisément, le dialecte haut-alémanique ou «Schwyzertütsch», «couvert» par l’allemand codifié comme langue écrite, est le symbole de l’identité de ses locuteurs, au point qu’il assume de multiples fonctions, et qu’il est parlé dans toutes les couches sociales, aussi bien dans les centres urbains que dans les milieux ruraux, l’allemand étant réservé à l’usage écrit, à l’enseignement, au discours à caractère officiel et à l’échange de paroles avec les étrangers germanophones. (Transl. In Alemannic Switzerland, the Alemannic dialect “Schwyzertütsch”, “covered” by codified German as a written language, symbolizes the identity of its speakers and assumes several functions; it is spoken in all social classes, in the urban centers and rural strata, German being reserved for written use, at school, in an official discourse, or in exchanges with foreigner who speak German.)

This may explain the few occurrences in the Swiss corpus. In Austria, there are several spoken dialects; for historical reasons the Austrians have another relationship to their language. The awareness of the specificities of the Austrian dialects and their valorization show a distance with the “great neighboring country” and its language. It is a less complex relationship with respect to the German language and its norm. Through the study of our structure, we see that the social role of the dialects has very different impacts in Austria and in Switzerland. Here, the geographical, historical and political parameters have to be taken into account. Most of the occurrences are direct or indirect quotes, or oral-­conceived written texts. Oral language, and thus dialects, can be considered as having an influence on the use of atypical structures. A logical collateral question is the extent to which this phenomenon is subject to variation. The answer would require great nuance and suggest further questions about the nature of the variation itself. We noticed a phenomenon of variation along a north-south guideline. This variation is clearly of a diatopic kind, and the data show the significance of differentiated geographical factors. It is based on an analogical principle. Nicht brauchen is copying the functional model of nicht müssen, leading one to conclude that this evolution is to be compared

262 Pascale Van Praet and Gilbert Magnus

to the phenomenon of grammaticalization, without the step of desemantization (that is, nicht brauchen keeps the same semantic features of non-necessity as nicht müssen in the domain of modality but never the features of modalization). This variation can be classified as a recategorization from an operator verb followed by an infinitive phrase (normally followed with the marker zu) to a modal verb (never followed by the marker zu). Given that this phenomenon is variable in its proportion and its extension, it is difficult to surmise if the next step will be a change in standard German. 4. Conclusion The linguistic question explored in this paper is that of the distributional behavior of the modal use of the German verb brauchen. The issue that it raises is that of emerging atypical constructions. An un-marked structure first considered as a spoken phenomenon and hence deemed incorrect is in fact more than a spoken phenomenon when it is attested in a large range of extracts from the press. The results of this study show discrepancies from 2% to 45%. They also show that the atypical use of brauchen with an un-marked infinitive structure is an attested phenomenon of variation increasing from Northern to Southern Germany. This process is more than a syntactic mimesis. As can be observed in the corpus, semantic differences appear related to the pragmatic uses. The results illustrate the mutual influence of several factors. One of these factors is register. A growing influence on formally established written texts can be found of what is commonly called “spoken language” and is redefined by Koch as “language of proximity”. The written press obviously shows a heterogeneous “communicative style”. To determine if the phenomenon of variation could be evidence of a change, we would have to test it by analyzing other corpora, both written and spoken. Choosing highly representative corpora such as a conversational corpus and a literary corpus – both extremes of the scale – would help to measure the phenomenon and contribute to defining its extent further. The degree of influence between register has been shown to vary with the regional context. However, only further research will determine whether this variation is influenced by the different dialects or by a relationship of greater or lesser conformity to standard German with respect to geographical parameters. The final factor of variation documented in this study is situational: the thematic aspect of a text influences the choice of structure; we observe that rather few different themes emerge. Research that mainly focuses on what unites these themes could yield interesting results.



Geographic variation in a non-canonical infinitive structure 263

All factors are involved; while the relative weight of the main influences is difficult to measure, recognizing their existence does help to relativize what exactly a typical and an atypical structure is.

References Adamzik, Kirsten. 2004. Textlinguistik. Eine einführende Darstellung. Tübingen: Max Niemeyer Verlag. doi: 10.1515/9783110946642 Ammon, Ulrich, Hans Bickel & Jakob Ebner, et al. 2004. Variantenwörterbuch des Deutschen. Berlin / New York: Walter de Gruyter. doi: 10.1515/9783110905816 Brinkmann, Hennig. 21971, 1962. Die deutsche Sprache. Gestalt und Leistung. Düsseldorf: Pädagogischer Verlag Schwann. Chapusot, Sonia. 2002. Les verbes de modalité et leur élargissement. Mémoire de maîtrise (prepared under the direction of Prof. André Rousseau). Université de Lille 3 – Charles-deGaulle, 86 pages. Duffley, P. & P. Larrivée. 1998. Need, Dare, and Negative Polarity. Linguistic Analysis 28(1–2). 89–107. Duden 9. 62007. Richtiges und gutes Deutsch. Mannheim / Leipzig / Wien / Zürich: Dudenverlag. Fried, Mirjam, Jan-Ola Östman & Jef Verschueren. 2010. Variation and Change. Pragmatic perspectives. Amsterdam: John Benjamins. doi: 10.1075/hoph.6 Frawley, William. 2006. The expression of modality. Berlin, New York: Mouton de Gruyter.  doi: 10.1515/9783110197570 Le Querler, Nicole. 1996. Typologie des modalités. Caen: Presses Universitaires de Caen. Le Querler, Nicole. 2001. La place du verbe modal pouvoir dans le système des modalités en français. In P. Dendale & J. van der Auwera (eds.), Les Verbes modaux, Cahiers Chronos 8, 17–32. Amsterdam: Rodopi. Koch, Peter & Wulf Österreicher. 1985. Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. In Romanistisches Jahrbuch 36. 15–43. Krier, Fernande. 2010. Les dialectes et leurs valeurs. In M. Kauffer & G. Magnus (eds.), Langues et dialectes dans tous leurs états. Hommage à Marthe Philipp, 61–74. Nancy: Presses Universitaires de Nancy. Lampert, Günther & Martina Lampert. 2000. The conceptual structures(s) of modality: Essences and ideologies. Frankfurt/Main: Peter Lang. Magnus, Gilbert. 2003. Aperçu sur les modalités en allemand. In Revue belge de philologie et d’histoire 81/3. Bruxelles, 791–796. doi: 10.3406/rbph.2003.4756 Palmer, F. R. 1986, 22001. Mood and modality. Cambridge: University Press. Portner, Paul. 2010. Modality. Oxford: University Press. Raynaud, Franziska. 1975. Les verbes de modalité en allemand contemporain. Thèse d’État, Université de Paris-Sorbonne (Paris IV). Lille 3: Atelier national de reproduction des thèses. Rousseau, André. 2003. La question des verbes de modalité en allemand moderne. In Revue belge de philologie et d’histoire 81/3. Bruxelles, 611–616 and 797–823. Schanen, François, & Jean-Paul Confais. 21989, 1986. Grammaire de l’allemand: formes et fonctions. Paris: Nathan.

264 Pascale Van Praet and Gilbert Magnus

Schiewe, Jürgen. 1998. Die Macht der Sprache. Eine Geschichte der Sprachkritik von der Antike bis zur Gegenwart. München: C. H. Beck. Van der Auwera, Johan & Vladimir A. Plungian. 1998. Modality’s semantic map. In Linguistic Typologie 2. 79–124. Weinrich, Harald. 1993. Textgrammatik der deutschen Sprache. Mannheim / Leipzig / Wien / Zürich: Dudenverlag. Zifonun, Gisela, Hoffmann, Ludger & Strecker, Bruno. 1997. Grammatik der deutschen Sprache. Berlin / New York: Walter de Gruyter [= Schriften des Instituts für deutsche Sprache].  doi:  10.1515/9783110872163

Verbal constructions in spoken language deviating from the norm Reflections on the concept of atypicality Günter Schmale

Université Jean Moulin – Lyon 3

Starting out from the consideration that conclusions as to what is atypical regarding a linguistic feature have to be drawn on the basis of language norms, the paper discusses three types of norms: those defined in authoritative grammars or dictionaries, those apparent through judgements by native speakers, and those determined by corpus-based analyses of language use in large corpora. As grammars and dictionaries are generally based on – formal – written language productions (“folk belief ” being strongly dependant on school grammar), the norms defined do not cover the phenomena of verbal constructions occurring in oral language production, especially in dialogical contexts. The study of naturally occurring conversations consequently reveals a great number of syntactic, semantic and conversational features a priori deviating from the “official” norm. The paper demonstrates that rather than being atypical these forms are specific for oral language production. What is more, atypicality cannot be defined on the basis of one specific type of speech exchange system, each different genre having its own typical properties. Atypicality is consequently what does not coincide with the forms of oral or written conceptuality applied by the majority of a community of language users in a given situation.

1. Introduction: What is atypical in a language?1 The Online Free Dictionary2 defines the adjective typical as “Exhibiting the qualities, traits, or characteristics that identify a kind, class, group or category: a typical suburban community.” Atypical would thus be what is “not conforming to type, 1. Thank you to Ms Lisa Spicker for her thorough proof reading and useful stylistic advice. All remaining flaws and imperfections are obviously imputable to the author of this text himself. 2. http://www.thefreedictionary.com (17/09/2013). doi 10.1075/lis.33.11sch © 2016 John Benjamins Publishing Company

266 Günter Schmale

unusual or irregular”. However, whereas it seems unproblematic to determine basic traits of a social feature such as a suburban community, as the example used in the quoted definition, to delimit what is atypical regarding a linguistic phenomenon appears to be a thornier endeavour. This is due to the fact that conclusions as to the atypicality of a linguistic feature have necessarily to be drawn on the basis of a language norm which is extremely difficult to circumscribe considering highly complex and varying parameters of language manifestations. Language can be spoken or written, formal or informal, used by different age, social or professional groups. Is there one norm only for all these different types of language productions? For instance the norm described in generally accepted grammars such as Duden Grammatik for German, Grevisse for French or the Oxford English Grammar for English? Or rather the language norm determined by native speaker usage via a corpus-based approach? One might also consider interrogating native speakers who often have strong opinions of what is “good” or acceptable language. The present paper will examine if the three types of aforementioned norms are liable to serve the purpose of describing atypical phenomena in the domain of verbal constructions: those outlined in grammars, dictionaries or guides for good language use (Section 2), those discernible through judgments on speech productions emitted by language users (Section 3), and those determined by corpus-­ based empirical analyses of authentic oral language use augmented by personal observations (Section 4). The aim is to delineate different types of typicality in order to achieve a definition of what is atypical (Section 5). 2. Norms described within “official” grammars In general, normative judgments as to what might be atypical are based on those norms defined in grammar text books considered as authoritative. However, as will be demonstrated (cf. 2.1), the language system in itself is characteristically “suffering” from numerous deviations from the norm, especially where idiomatic expressions are concerned. What is more, normative grammar rules are, apart from rare exceptions, based on those prevalent in written language productions (cf. 2.2), without thoroughly examining if these really apply to oral language use in general and verbal interaction in particular. 2.1

Deviation from norms immanent in the language system

The language system in itself involves, as an integral part, structures which deviate more or less from syntax and/or semantics as defined for canonical use of the



Verbal constructions in spoken language deviating from the norm 267

language in question.3 As a matter of fact, a non-negligible number of idiomatic expressions (= idioms) present deviations from the semantic or morphosyntactic norm of English, French and German. Four different types can be distinguished: – Idioms may contain archaic elements surviving within one expression only, i.e. they are normally not being used without direct reference to their phraseological form or meaning. – English: go scot free.4 – French: ne pas se mettre mar tel en tête (not worry too much about something); – German: Maulaffen feilhalten (gawk, sit on one’s hands); The use of these elements – martel, Maulaffen and scot free – is “unique” (unikale Komponenten), i.e. first and foremost reserved to their phraseological environments, or at least to allusions to the latter. – Idioms can display morphosyntactic features which, like archaic lexemes, are no longer productive. – English: stay put, I am all ears, but not: *stay sit or *I am all nose.5 – French: plier bagages or coûter bonbon, but not: *plier chemises, *coûter argent; – German: auf des Messers Schneide stehen, but not: *auf des Stuhles Rand sitzen; – Idioms may have a regular morphosyntactic structure without being semantically regular, i.e. their phraseological meaning is semantically non compositional. Examples are taken from the paradigm be dead. – English: push the daisies, following the model V + NP as in push the wheelchair; – French: manger les pissenlits par la racine, following the model V + ACC/ NP + PP as in attraper un serpent par la queue; – German: die Radieschen von unten begucken/beschauen, following die Landschaft von oben betrachten (ACC/NP + PP + V); All idioms possess a regular morphosyntactic structure whereas their meaning is idiomatic, thus semantically non-compositional. – According to their context of use, idiomatic expressions may have two lectures, i.e. one which is semantically regular or compositional, having a 3. At least the systems of English, French and German, the languages we shall use as examples in this section. We can of course not make statements about languages we have not studied. 4. Other instances of unique or archaic elements are: avoir maille à partir avec qn, sich nicht ins Bockshorn jagen lassen or play footsie with someone. 5. Same observation for French, je suis tout ouïe, or German, ich bin ganz Ohr.

268 Günter Schmale

non-­phraseological meaning, and another one which is semantically irregular or non-­compositional, having a phraseological meaning. This does not necessarily imply, however, that both meanings are simultaneously activated. Apart from play on words, co- and contexts normally do not leave any doubt about what is meant. – English: have kittens – a cat may actually have kittens (non-phraseological meaning) vs. somebody being very frightened (idiomatic sense). – French: mettre de l’eau dans son vin – non-phraseological literal meaning: pour some water into one’s wine to make it less strong vs. the idiomatic one: tone something down, take something down a notch, be more moderate; – German: die Weichen stellen – non-phraseological literal meaning: work the switches (for a train) vs. idiomatic meaning: prepare the future, set the course.





The discussion of four types of idiomatic expressions reveals that expressions can be lexically, morphosyntactically or semantically deviant from a canonical norm,6 however, they are by no means to be considered as atypical of the language in question. On the contrary, they are particularly typical as they often possess very ancient traits of it; what is more, they serve to conceptualize human experience and knowledge in common metaphorical formats. They so greatly contribute to what is considered as original, specific, idiomatic … typical of a language! 2.2

Grammatical norms for verbal constructions developed on the basis of written language use

Apart from the fact demonstrated in the previous subsection that deviations from the grammatical, lexical, or semantic norm are an integral part of the language system, one has to observe that canonical norms of authoritative grammars have, in the majority of cases, been developed on the basis of written language productions.7 The fact that examples are mainly taken from literary or journalistic texts support this statement. However, one has to ask oneself whether norms developed on the basis of written language productions do apply to oral language. In fact, studies of oral language in general and conversational interaction in particular, applying corpus-driven or corpus-based “analytic mentality”, have revealed and continue to reveal that the structure of spoken language, in particular when 6. That is, from a synchronic, not a diachronic point of view. 7. For German, one may quote as exceptions the so-called IDS-Grammatik (Zifonun et al. 1997), containing rules for oral language use, and Fiehler’s chapter on spoken language in the latest edition of the Duden Grammatik (2009) (cf. also Fiehler 2006).



Verbal constructions in spoken language deviating from the norm 269

used in interactional environments, profoundly differs from the organisation of written language. Section 4 of the present paper will discuss several specific organisational features of spoken – dialogical – language. Furthermore, one has to state that a global opposition of oral and written language no longer corresponds to the state of the art of linguistic research. In fact, following insights into language use, it would be appropriate and necessary to distinguish medial orality and scripturality on the one hand, and conceptual orality and scripturality on the other hand (cf. Koch/Oesterreicher 1985; Bakker/ Kahane 1997). Not all oral language productions correspond in reality to the norms of conceptual orality, and each written language manifestation is by no means guided by the principles of conceptual scripturality as the following table demonstrates: Table 1.  Medial orality/scripturality vs. conceptual orality/scripturality Mediality Oral Conceptuality (phonic code)

Scriptural (graphic code)

Spoken Written

SMS, chat, post-it, email letter, written exam

conversations, discussions play (theatre), speech, lecture

As a consequence, orally produced language such as a play or a lecture obeys the rules of conceptual scripturality as actors memorize a written text and lecturers are likely to read a script. On the other hand, a written text such as a SMS or a contribution to an Internet chat is more likely to display the structure of a spoken than that of a written text. One would therefore be just as much mistaken to apply standards of written language to any type of language production as to consider everything using the phonic code corresponding to the rules of conceptual orality. Instead, phenomena of conceptual orality or scripturality should be attached to certain types of texts or communicative situations as demonstrated in Table 1. What is typical or atypical for a given text or communicative event thus depends not on the norms defined by grammars considered as authoritative, but on their specific organisational structure which can considerably vary from what one might find in a grammar. Linguistic phenomena encountered in naturally occurring communicative contexts may thus not correspond to rules defined in a grammar and be nevertheless highly typical of the register in question. On the contrary, it is rather the grammatical rules which would have to be regarded as atypical for the description of communicative features which do not respond to the criteria of written language use.

270 Günter Schmale

3. Norms according to native speaker judgment of correct language use In order to account for what native speakers of a language conceive as wellformed, and thus typical, Meyer (2011) proposes large surveys. However, even if popular opinions and beliefs on language, studied by folk linguistics, are highly interesting and instructive in many respects, they are most likely to be strongly guided by grammatical norms taught in school or conveyed by the media.8 This attitude results in identifying certain structures as atypical – because they do not conform to an authoritative canonical form – even though a majority of speakers would only use this “atypical” structure which would consequently be typical as far as frequency of use is concerned. This is why an English person might consider “infinitive splitting” (I want you to quickly go to bed.) as deviant from the norm and not well-formed, although this structure is widely used. And an educated speaker of German may well consider the very commonly used structure im Herbst diesen Jahres9 as agrammatical, the canonically correct form being the strong declination in im Herbst dieses Jahres, this latter form nevertheless quickly becoming atypical of common language use. Germany’s most well-known language purist Sebastian Sick (2004) offers a wide collection of non-canonical language use in his various volumes of Der Dativ ist dem Genitiv sein Tod. And the Anglizismen-Index of the Verein für Deutsche Sprache (VdS)10 suggests a German lexeme for every single English word – lexemes which are however totally unknown to the great majority of users. As for a speaker of French, he may reject the utterance l’outil que j’ai besoin, the grammatically canonical form being l’outil dont j’ai besoin, although the absence of the genitive form of the relative pronoun seems more and more common and is likely to become the norm not only in “uneducated speech”.11 One might also mention the slow disappearance of phonetic liaison in French: most speakers 8. Or by conversational norms of what one should or should not do when talking to someone else. Schmale (2008b) demonstrates that participants have very little insight into the conversational principles governing a conversational interaction. Informants believe for instance that one should always wait for the end of partner’s turn before taking it whereas in reality turn-final segments, especially when not contributing to the turn-constructions propositional content, are very frequently overlapped by the beginning of next turn. 9. Cf. Section 4.2, point 1, for a discussion of this phenomenon. 10. URL: http://www.vds-ev.de/anglizismenindex (20/10/2013). 11. The change of meaning of un magasin bien achalandé supports this point of view. Only purists still claim that this is a shop having many clients, a chaland being in fact a client, general use being convinced that this a shop offering a large variety of products, a sense which has even found its entry into Le Petit Robert.



Verbal constructions in spoken language deviating from the norm 271

would say trois # euros or dix # euros12 even though phonetic rules stipulate the production of liaison in these cases. The discussed examples illustrate that native speaker judgements of non-­ standard or even incorrect language forms are in general based upon canonical, lexical or grammatical norms. However, these comments are by no means oriented towards actual language use. Thus, what might be atypical from a canonical point of view is totally typical from a usage perspective. As Teubert (2009) has it: […] we have not authorised linguists to tell us what is right and what is wrong. If people want to pay respect to the normalising intentions of the Académie française, they are welcome to do so. But no one can force them. Corpus linguistics teaches us that there is no ‘true’ or ‘correct’ interpretation of a lexical item, a text segment or a text. The meaning of a lexical item is all that has been said about it in the discourse. We can find it in the negotiations between the members of a discourse community.  (Teubert 2009: 61)

This is why the next section of this paper shall concentrate on linguistic norms established via extensive empirical corpus studies of naturally occurring conversations. 4. Norms discovered via the corpus-based study and observations of language use If typical language use is about structures being used by participants in real-life situations, then these structures have to be studied and described empirically in a corpus-based, ideally: a corpus-driven manner13 using large corpora. Considering that norms deployed in authoritative grammars and dictionaries are based upon – mostly rather formal – written language productions in mainly journalistic or literary writings, priority shall be given to – oral – conversational corpora in the present paper. However, given that medial scripturality is increasingly subjected to the impact of oral conceptuality (cf. supra), it is highly probable that norms for written conceptuality are equally undergoing a profound change. 4.1

Studies of naturally occurring conversations

Four types of specific phenomena of oral language production deviating from canonical norms of written conceptuality as defined in authoritative grammars 12. But then one might come across: vingts=euros. 13. To use Tognini-Bonelli’s (2001) distinction.

272 Günter Schmale

will be discussed in this subsection: syntactic (4.1.1) and lexical features (4.1.2), characteristics of oral encoding or verbalization (4.1.3), and properties of conversational organisation (4.1.4). Examples in this section are taken from German, yet as far as specific features of encoding or conversational organisation are concerned they may well apply to English, French or other languages. 4.1.1 Syntactic features of oral conceptuality Research into German oral syntax widely discusses so-called “main clause structures” following – canonically speaking – subordinating conjunctions such as weil, obwohl, wobei, wenn or wo (cf. Auer/Günthner 2003). In fact, when opening a “main clause structure”, the finite verb occupying not its “normal” final position, weil, obwohl etc. no longer have subordinating functions, instead operating as discourse markers performing altogether different tasks. These different functions of discourse markers will be discussed by using examples from Auer/Günthner (2003) and from Susanne Günthner’s various articles on this subject. Grammars of German almost invariably treat weil as a subordinating conjunction possessing a causal function as in ¤Ich konnte nicht kommen, weil ich krank war,14 i.e. an utterance with the finite verb in final position. However, a causal construction carrying the identical meaning can be produced within a main clause structure as ¤Ich konnte nicht kommen, weil (.) ich war krank, the finite verb being in second position. But as this is an oral utterance one must not forget its multimodal nature and take into account the short pause – 1/10 to 0.25 seconds – which is decisive in order to make the turn construction unit conversationally – and grammatically15 – acceptable. Yet weil can occupy a completely different function from that of a subordinating causal conjunction as in (1) hereafter: (1) den Profs geht das am Arsch vorbei (.) weil (.) denen ist das scheißegal16

weil obviously does not operate as a causal conjunction in this case as one cannot justify the existence of one fact by restating the same fact in a different – though quasi-­synonymic – formulation. In fact, weil functions as a discourse marker with the sole purpose of extending the term, which is possible because weil projects a construction promising an explanation of a previous state of affairs. 14. The sign ¤ indicates that the utterance thus marked has been invented for illustrative purposes. 15. As this is, taking into account oral production’s temporality and linearity, in fact a new start and a new utterance. 16. Meaning something like: the dons don’t give a sh… because (.) they couldn’t care less.



Verbal constructions in spoken language deviating from the norm 273

The same observation holds true for normally concessive obwohl in se­quence (3). (2) der hat vier Brüder (.) obwohl (.) stimmt das eigentlich? drei sinds

obwohl does not designate an exception from what is considered as normal in this case,17 but serves to initiate a self-reflection on the veracity of a previous construction followed by a self-correction, thus serving as a discourse marker: he does not have four brothers, but three. wobei in (3) hereafter has exactly the same function and could in fact be replaced by obwohl in this case. (3) Restmüll ist jeden zweiten Montag (.) wobei (.) das ist seltener nicht alle zwei Wochen

wenn, in sequence (4), is not a subordinating conjunction introducing a conditional construction either, but again a discourse marker. The wenn-construction by no means represents the condition (protasis) of the second part of the utterance which is thus not the conditioned part (the apodosis). (4) wenn du Lust hast und Zeit (.) wir machen morgen einen Kindergottesdienst18

The Sunday school will take place independently of the hearer’s participation; the speaker simply pronounces an invitation. Contrary to the first three discourse markers, however, the wenn-construction implicitly conserves its conditional potential provided one considers it as elliptical, the elided part being in fact the apodosis: ¤you can come and join us. The discourse marker would in reality be not just wenn, but the elliptical conditional structure. On two conditions, nevertheless: to subsequently produce a short pause, and to use a main clause structure. 4.1.2 Lexical structures deviating from the norm If the absence of a lexeme or an expression from authoritative dictionaries is to be considered as a deviation from a lexical norm implying their lexical atypicality, then neologisms have to be classified as atypical. Phraseological neologisms as those found and described by Schreiber et al. (2012; cf. Table 2) have no entries in generally accepted specialized dictionaries of phraseology such as Duden 11 or Pons-Schemann, but they can be found in the online-dictionary redensarten-­ index.de which lemmatizes– after verification by a mediator– expressions sent in by users and which consequently reflect, to at least some extent, norms of usage. Even though these neologisms might never find their entry into a dictionary 17. As in: ¤I have to work even though I’m sick. It would be normal to stay in bed in this case. 18. Meaning: if you feel like it and if you’ve time (.) we’ll organize a Sunday school tomorrow.

274 Günter Schmale

Table 2.  Idiomatic neologisms in German Neologism19

Meaning

einen Clown zum Frühstück gegessen haben rubbel die Katz sich zum Horst machen den Zonk ziehen wie Hulle20

be unbearably funny, good-humoured very quickly, in the blink of an eye make a fool of oneself make a bad choice, have bad luck like mad, like blazes, beat the band

applying high lexicographical standards, owing to a limited range of usage and an ephemeral existence, they can be particularly typical for a certain period and a certain type of user.21 As a consequence, whereas dictionaries based on scientific lexicographic principles do represent what is widely and durably used in a language, they do not necessarily convey what is typical for a limited range of users in a confined period of time. 4.1.3 Characteristics of oral encoding It is almost trivial to state that principles of oral encoding profoundly differ from that of scriptural language use. When expressing themselves orally, speakers dispose of only an extremely limited encoding time, cannot wipe out mistakes once they are made, but, on the other hand, they have direct access to their addressee – all three characteristics entailing specific forms of oral encoding unknown in written language. Such specific and typical phenomena of oral language productions are the following: – Interruptions and new starts. A speaker may abruptly put end to a construction without having completed it in order to start a new one, e.g. ¤I don’t want to’ (.) in fact I’d much prefer a glass of wine.22

19. The authors discovered 15 phraseological neologisms as a total, the others are: jmdm ein Kotelett an die Backe labern etc., Geh scheißen/kacken!, sich freuen wie ein Schnitzel, Dann ist der Drops gelutscht!, Fick die Henne!, jmdm die Kauleiste tiefer legen, Gleich klatscht es, aber kein Beifall! (sic!), jmdm scheint die Sonne aus dem Arsch, Du bist wohl in der U-Bahn geboren!, jmdm geht der Stift (cf. Schreiber et al. 2012: 24). 20. Apparently a phonetic variant of Hülle as in Hülle und Fülle (galore, abundantly). 21. Cf. Schmale‘s (2013) extended definition of prefabricated language including also very short-lived phrasemes as well as those being used by small groups only, both being, however, highly pertinent as a language resource. 22. The apostrophe signifying a phonetic rupture after to’.



Verbal constructions in spoken language deviating from the norm 275

– Repetitions. Speakers frequently repeat parts, words or phrases of their utterance while reflecting on continuation, e.g. ¤If I were you (.) if I were you I’d think twice before accepting this offer. – Encoding errors. At any moment, speakers are liable to encounter articulatory, lexical or syntactic formulation problems which can be remedied by the speaker himself or by the hearer. Errors can also simply be ignored or treated implicitly when they do not endanger the pursuit of commonly engaged activities in conversational contexts. – Pauses. Participants can pause at any instant and subsequently continue their verbalization. Considering the multimodality of conversational interaction prosodic, non-verbal or contextual clues allow such pauses to be treated implicitly whereas they have to be made explicit via comments or orthographic signs when writing. – Unfinished or elliptical turn construction units. Whereas coordinating ellipses such as What’s your name? – (My name/It is) Peter are normal and typical in dialogical language, unfinished turn constructions quite naturally appear in conversations without causing any problem whatsoever for turn-taking or conversational negotiation in general. Even though they might be deviating from the canonical syntactic norm demanding the production of complete sentences, they are perfectly typical for conversational turn-construction and by no means to be considered as elliptical from an interactive point of view (cf. Schmale 2008a and 2016) – and therefore in no case atypical as far as conversational usage is concerned. – Apokoinu constructions. Utterances such as I like tennis is my favourite sport,23 containing a koinon, i.e. an element, in this case: tennis, which can be part of a first construction unit, i.e. I like tennis, but also of a second unit, i.e. tennis is my favourite sport, functioning as an accusative complement in the first case, and as a subject in the second, are fairly common in oral language. Even if canonical – scripturally oriented – syntax would mark them as such, they are not ungrammatical from an interactive point of view taking into account the temporality of oral language production.24 In fact, only the second part of the apokoinu construction would be subject to conversational negotiation. In any case, given the absence of corrections, new starts or metacommunicative comments, this type of construction has to be considered as normal and consequently on no account atypical.

23. Source: http://www.odlt.org/ballast/apo_koinou.html (26/10/2013). 24. What Peter Auer (2005) has chosen to call “online-syntax”.

276 Günter Schmale

– Non-sentential turn-construction units. A look at oral communication quickly convinces the analyst that sentential turn-construction units are the exception rather than the rule (cf. Schmale 2012). Neither Hennig’s (2006) “possible sentence” of spoken language nor Zifonun et al.’s (1997) “interactive units” or “communicative minimal units” can cover the complete range of units used in conversational turn-construction. In fact, any type of turn-­ construction can be employed, be it on sentential or far below this level, as long as it permits participants the pursuit of their on-going activities. Atypical would thus be what does not correspond to participants’ expectations hic et nunc, no matter if it conforms to canonical grammar or not; in certain environments grammatically elliptical constructions might even be more typical than a syntactically complete structure. 4.1.4 Properties of conversational organisation Similarly to typical oral encoding phenomena, conversational organisation recurs to strategies which are totally unheard of in written language production and which remain unmentioned in most “official” grammars.25 Some of these organisational characteristics are: – Interactivity. Whereas written texts are the result of a single author’s verbalizing activities, its reader’s interpretations or reactions being only indirectly and retrospectively available, the organisation of a conversation is interactive in character, consequently by no means the outcome of a single speaker’s communicative intentions. What is more, this interactive character is displayed on the conversational surface and thus visible and reconstructible for the analyst. Participants’ interpretations of a first (non-)verbal activity are observable in subsequent activities, e.g. an activity commonly qualified as a response manifests the hearer’s interpretation of a previous activity as a question. Particularly conspicuous instances of interactivity are collaborative turn-constructions where two speakers conjointly produce one single turn-construction, for instance: ¤A: I think he’s about…. B: eighteen, I believe. A: Yeah, that’s right. – Turn-taking. The stated interactivity of conversational organisation is being manifested in exemplary fashion via the joint arrangement of speaking turns. The most basic task participants have to accomplish is in fact the organisation of who speaks when and for how long. This taking of conversational turns is continuously being negotiated by participants by recurring to different types of – implicit or explicit – techniques. Without the organisation of turn-­taking

25. With the exception of the IDS grammar as far as German is concerned.





–

–

–

–

Verbal constructions in spoken language deviating from the norm 277

a conversation would be chaotic and thematic development would be impossible. Due to this absolutely fundamental status of turn-taking, its detailed study allows us the unveiling of a great number of conversational features such as existing hierarchies between participants, tensions or conflicts between them, misunderstandings, or else cooperativeness or unhelpful behaviour and so forth. Simultaneous turn production. Contrary to common belief that any concomitant speech of two participants means an interruption of speaker A by speaker B, overlap of turn-endings and turn-beginnings – especially of routine constructions such as turn-exit devices – are absolutely normal for conversations and represent in no way failure of organisation, quite the contrary, it is a strong sign for its perfect machinery. Even overlap of elements belonging to the propositional contents of an utterance certainly do not always become subjects of conversational negotiation. Repair. Given the temporal, on-line nature of oral encoding, problems pertaining to articulation, formulation, understanding, interpretation etc. can arise at any point of sequential development. Participants, the speaker as well as the hearer, thus have the right to introduce repair at the first possible completion point. On-going activities will then be momentarily suspended and give rise to side sequences dedicated to treating the trouble source. Non-verbal and prosodic features. Communicative events are, as pointed out, by principle multimodal, integrating segmental, supra-segmental, non-verbal and contextual components. Both prosody and non-verbal communication can consequently play a decisive role in the interpretation of conversational activities, illustrating, attenuating, express the contrary of a verbal “message”, or even, in form of non-verbal emblems, possess a meaning of its own. A systematic integration of all aspects of a communicative activity is therefore indispensable in order to reveal the very nature of conversational negotiation. Laughing or smiling. Both vocal laughing and non-verbal smiling can accompany verbal utterances, but also be executed independently. Whereas laughing or smiling can of course express amusement or happiness, its role is by no means limited to its expression. Varying types have numerous functions in conversation, its most basic one being the (re)constitution of reciprocity between participants. Rather than demonstrating the non-serious nature of an utterance, laughing/smiling most frequently aims at an attenuation of face threat to the partner, for instance when one steps on somebody’s foot, one points out that someone is wrong, one asks somebody a favour … and smiles at the same time.

278 Günter Schmale

Written texts could at best try to reproduce the developed conversational features – indirectly or artificially – through the presentation of invented dialogues, without however being able to take into account phenomena of oral encoding which would render the dialogue illegible for the non-linguist reader (cf. supra). The developed features are nevertheless highly representative and typical of conversational organisation! The fact that authoritative grammars, predominantly reflecting canonical written language production, do not – yet – include these conversational phenomena does by no means imply that they are atypical! 4.2

Observations on specific phenomena of oral language use

The insights of empirical corpus-based studies into specific phenomena of oral and/or conversational language production will be completed by personal observations in the present sub-section. Different phenomena of primarily oral language use, which may however find their entry into scriptural conceptuality, on a lexical and/or syntactic level will be outlined hereafter:26 – Even though Duden 9 Richtiges und gutes Deutsch (Dudenredaktion 2001) claims that its recommendations are based on everyday language use,27 its advice to use the strongly declined dieses Jahres, corresponding to standard language use, rather than weakly declined diesen Jahres, hardly reflects genuine use. In fact, both “on the street” and in the media, native speakers now seem to prefer the weak declination, probably motivated by a felt redundancy of double genitive marking. It will not be long before educated speakers will be considered to be speaking incorrectly as it is majoritarian usage which determines typicality while pretended authoritative use may well be atypical! – Further constructions which might – sooner or later – enter the domain of typicality as far as oral language production is concerned are: – attributive rather than predicative use of the adjective okay as in Es ist ne okaye Ausgangsposition;28

26. Examples are from personal collections of utterances heard on German television. 27. „Die Empfehlungen der Dudenredaktion gründen vor allem auf der Häufigkeit ihres Vorkommens, also ihrem tatsächlichen Gebrauch im Sprachalltag.“ (Duden 9; cf. URL: http://www. duden.de/presse/richtiges-und-gutes-deutsch---der-sprachratgeber-fuer-konkrete-zweifelsfaelle (27/10/2013). 28. Duden Deutsches Universalwörterbuch (DUB) has „es ist alles okay; gestern ging es mir reichlich mies, aber heute bin ich wieder okay; das Mädchen ist wirklich okay (verhält sich kameradschaftlich)“ (quoted following the digital version of DUB) only.



Verbal constructions in spoken language deviating from the norm 279

– verbalization of nouns, e.g. Jetzt wird warngestreikt! whereas DUB lemmatizes the noun only: „ein eintägiger Warnstreik; es kam in vielen Städten zu Warnstreiks“;29 – absence of the “habitual” prepositional complement for certain verbs such as kümmern in da muss sich ja jemand jetzt zu Hause kümmern; – replacement of an infinitive by a noun phrase after modal können as in Kann Ihr Vereinspräsident auch Bundespräsident? or Ich kann Chef; – English loan translations of expressions such as I’m late = Ich bin spät instead of Ich habe mich verspätet/Ich komme zu spät; Ich habe gelernt = I learnt rather than Ich habe erfahren; Du bist sehr hilfreich30 = you are very helpful instead of du hast mir sehr geholfen; ich erinnere den Tintenkiller noch = I remember the ink eraser and not Ich erinnere mich noch an den Tintenkiller.

As we have not yet researched the aforementioned constructions in large corpora we are not in a position to state that they are typical for contemporary language use, on the other hand, given their presence in televised communication one could not conclude to their atypicality either, judging from their absence from a highly official dictionary of German, i.e. the DUB. But what about production errors, produced on purpose or by mistake, which were televised at some stage such as in Saturn’s advert Wir hassen teuer, Verona Feldbusch’s famous publicity for Teleauskunft: Da werden Sie geholfen!, or Giovanni Trappatoni’s equally notorious outbreak during a press conference Ich habe fertig! These utterances have become part of common core, and the last two are used very frequently indeed, although mainly in a humoristic fashion. Are they to be classified as atypical on account of their high degree of deviance from standard grammar in spite of their frequent use? It seems in fact that deviance from a canonical norm is not a relevant criterion for determining typicality and even less atypicality. Expressions of all kinds may be marked by a very high frequency of use, and thus be typical, while at the same time more or less deviate from what authoritative grammar or dictionaries consider as standard language.

29. But: 6000 occurrences found by Google. 30. A Google search results in 58,100 hits for this construction, on the other hand, Wortschatz Leipzig, exclusively presenting examples from written, journalistic or literary sources, only has one single occurrence with an animate subject only. Goethe’s use of hilfreich as in Edel sei der Mensch, hilfreich und gut!, seems to resurface via English after several centuries. A phenomenon which is by no means new: Grimm’s Deutsches Wörterbuch has ‘ablachen = sich müde lachen’ and in Goethes Wilhelm Meister one comes across ‘sich abängstigen = sich im Übermaß ängstigen’, both using the intensifying prefixe ab-, highly popular in modern spoken language (cf. Schmale 2007).

280 Günter Schmale

5. Types of typicality and definition of atypicality As foreseen in the introduction to the present paper the treatment and delimitation of typicality and atypicality concerning verbal constructions has turned out to be a thorny matter indeed as the chosen point of reference, i.e. deviation from existing “official” syntactic or lexical language norms, has not proven to be conclusive for characterizing a specific linguistic construction as typical or atypical. It has in fact been shown that – idiomatic expressions, whose structure deviate both from standard syntax and semantics, constitute an integral part of English, French and German, and are consequently highly typical of these languages; given the high degree of prefabricated elements in all types of language manifestations, texts or conversations without any structurally irregular idioms would be rather untypical; – canonical norms as defined by authoritative grammars or dictionaries can hardly be applied indifferently to any type of – oral or written – language production as most grammars deduce normative rules from written language found in “high quality” journalistic or literary sources; oral language in general and conversational interaction in particular, subject to specific forms of organisation, but also written productions strongly guided by oral conceptuality, underlying totally different norms, are not taken into account; norms of oral language production can consequently not even be called deviating from the – written – norm, and much less atypical because their organisation has not been sufficiently looked into; – surveys of participants’ beliefs cannot yield satisfactory results as speakers more or less strongly orient their opinions and attitudes towards canonical norms acquired in school or through the media; a native speaker may thus classify a certain type of construction as atypical even though he regularly makes use of it himself; – empirical studies of conversational corpora reveal that oral language in general and in dialogical environments in particular is subject to a great number of specific norms both in their syntactic and lexical structure as in their forms of encoding and conversational organisation, the last two profoundly differing from conceptually written productions; obviously these specific forms are by no means atypical but on the contrary typical of conceptually spoken language. As a consequence, one has to conclude that typical and ergo atypical language can evidently not be defined by referring to one particular type of speech exchange



Verbal constructions in spoken language deviating from the norm 281

system, each system disposing of its specific norms on account of which certain types of language forms may be classified as atypical. However, what may be atypical for speech exchange system A could be perfectly typical for system B and vice versa. One has thus to conclude that there are different kinds of typical language and, by deduction, atypical language on the basis of what does not correspond to the hearer’s expectations. One has thus to conclude that language is to be considered as atypical whenever it does not coincide with the forms of oral or scriptural conceptuality applied by the majority of users of a given language community in a specific situational or textual genre. Slightly transforming Wittgenstein’s credo “meaning is use” one might conclude: Atypical is what is being used when not expected by participants in a given situation!

References Auer, Peter. 2005. Syntax als prozess. Interaction and Linguistic Structures/InLiSt 41. URL: http://www.uni-potsdam.de/u/inlist/issues/41/InLiSt41.pdf. Auer, Peter & Susanne Günthner. 2003. Die Entstehung von Diskursmarkern im Deutschen – ein Fall von Grammatikalisierung? InLiSt 38. 1–30. URL: http://www.unipots dam.de/u/ inlist/issues/38/index.htm. Bakker, Egbert J. & Ahuvia Kahane (eds.). 1997. Written voices, spoken signs. tradition, performance and the epic text. Cambridge/Mass., London: Harvard UP.  doi:  10.4159/9780674020467

Dudenreaktion (eds.). 2001. Duden Band 9 – Richtiges und gutes Deutsch. Wörterbuch der sprachlichen Zweifelsfälle. 5. neu bearbeitete Auflage. Mannheim, Leipzig, Wien, Zürich: Dudenverlag. Dudenreaktion (eds.). 2007. Duden Deutsches Universalwörterbuch. 6., überarbeitete und erweiterte Auflage. Mannheim, Leipzig, Wien, Zürich: Dudenverlag. Dudenredaktion (eds.). 2009. Duden Band 4 – Die Grammatik. Unentbehrlich für richtiges Deutsch. 8. überarbeitete Auflage. Mannheim, Wien, Zürich: Dudenverlag. Fiehler, Reinhard. 2006. Was gehört in eine Grammatik gesprochener Sprache? Erfahrungen beim Schreiben eines Kapitels der neuen Duden-Grammatik. In A. Deppermann, R. Fiehler, & T. Spranz-Fogasy (Hrsg.), Grammatik und Interaktion, 21–41. Radolfzell: Verlag für Gesprächsforschung. URL: http://www.verlaggespraechs forschung.de/2006/ pdf/gram matik.pdf. Günthner, Susanne. 2002. Konnektoren im gesprochenen Deutsch – Normverstoß oder funktionale Differenzierung? Deutsch als Fremdsprache 2. 67–74. Hennig, Mathilde. 2006. Grammatik der gesprochenen Sprache in Theorie und Praxis. Kassel: University Press. doi: 10.1515/9783110936063 Koch, Peter & Wulf Oesterreicher. 1985. Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. Romanistisches Jahrbuch 36. 15–43.

282 Günter Schmale

Meyer, Markus. 2011. Wohlgeformtheit als empirisches Konzept. Zeitschrift für angewandte Linguistik 55. 23–58. Schmale, Günter. 2007. ‘abfeiern, ablachen, abtanzen’ – Zur Funktion des Erstglieds abbei verbalen Neuschöpfungen und Neubedeutungen im Gegenwartsdeutschen. In M. Kauffer & R. Métrich (Hrsg.), Verbale Wortbildung im Spannungsfeld zwischen Wortsemantik, Syntax und Rechtschreibung, 133–144. Tübingen: Stauffenburg. Schmale, Günter. 2008a. Constructions inachevées et transfert du tour de parole. In J. Durand, et al. (eds.), Congrès mondial de linguistique française. Paris, 9–12 juillet 2008. Recueil des résumés + contribution sur CDRom, 89. Paris: ILF/EDP Sciences. Schmale, Günter. 2008b. Conceptions populaires de la conversation. In G. Achard-Bayle & M.-A. Paveau (eds.), Linguistique populaire? Pratiques 139/140. 58–80. Schmale, Günter. 2009. Langue de Goethe oder E-Deutsch – Welche Variante für den DaF-­ Unterricht in Frankreich? In R. Grimm & E. Venohr (Hrsg.), Immer ist es Sprache. Mehrsprachigkeit – Intertextualität – Kulturkontrast. Festschrift für Lutz Götze zum 65. Geburtstag, 179–198. Frankfurt/M.: Lang. Schmale, Günter. 2012. Korpusgestützte Beobachtungen zur Besetzung der initialen Position konversationeller Äußerungen. In C. Cortès, et al. (eds.), Satzeröffnung – Formen, Funktionen, Strategien, 15–29. Tübingen: Stauffenburg. Schmale, Günter. 2013. Qu’est-ce qui est préfabriqué dans la langue? – Réflexions au sujet d’une définition élargie de la préformation langagière. Langages 189. 27–45.  doi:  10.3917/lang.189.0027

Schmale, Günter. 2016. Gibt es in Konversationen fragmentarische Äußerungen? – Beobachtungen zur konversationellen Behandlung rechts offener Konstruktionseinheiten aus interaktionaler Perspektive. In J.-F. Marillier & E. Vargas (Hrsg.), Fragmentarische Äußerungen, 247–270. Tübingen: Stauffenburg. Schreiber, David, Cerstin Mahlow & Britta Juska-Bacher. 2012. Phraeologische Neologismen: Identifikation und Validierung. Yearbook of Phraseology 3. 3–30. Sick, Bastian. 2004 (200630). Der Dativ ist dem Genitiv sein Tod. Ein Wegweiser durch den Irrgarten der deutschen Sprache. Köln: Kiepenheuer & Witsch. Schwitalla, Johannes. 20032. Gesprochenes Deutsch. Eine Einführung. Berlin: E. Schmidt. Teubert, Wolfgang. 2009. Corpus linguistics: An alternative. Semen 27. 2–21. Tognini-Bonelli, Elena. 2001. Corpus linguistics at work. Amsterdam: John Benjamins.  doi:  10.1075/scl.6

Zifonun, Gisela, et al. 1997. Grammatik der deutschen Sprache. 3 Bände. Berlin, New York: de Gruyter. doi: 10.1515/9783110872163

Index of authors A Abeillé 200 Abney  137, 158 Ackema  115–116, 128 Ackerman  67, 84 Adamzik  254, 263 Aikhenvald  25, 56, 58–59 Aissen  67, 84 Albrespit 200 Alexiadou  132, 134, 141, 144, 146–149, 158–159, 185–186, 200–201 Alvarez 176–178 Ammon  260, 263 Anagnastopoulou 160 Aranovich  70, 77, 84 Arka  214, 222–223, 233, 238 Arkadiev  235, 238 Arndt 238 Auer  272, 275, 281 Avram  154, 159 B Baker  115, 128, 131–132, 134–135, 141, 159, 189 Bakker  269, 281 Barnes  9, 24 Bassac  164, 175, 178 Belletti  7–8, 24 Belloro  70, 77–78, 84 Bennis  9, 24 Ben Salah  183, 200 Benveniste  30, 44–46, 56 Berman  30, 39, 56 Bhatt  139, 159 Biber  88–89, 91–92, 111 Bickel 263 Bienvenue 201 Birner  28, 51, 56, 60 Blau  43, 56 Bleam  79, 84 Boas 178

Bolinger  29, 31, 56 Bonch-Osmolovskaya  9, 11, 24 Boons  183, 200 Borer  132, 134, 141, 159 Borschev  49, 56 Bossong  19, 24 Bourmayan  165, 178 Brandi  50, 56 Brinkmann  246, 258, 263 Broschart 200 Brousseau  126, 128 Burzio  4, 6, 9, 24–25, 56–57, 119, 128 Bybee  41, 57, 127–128, 130 C Cano 159 Casielles-Suárez  63, 84 Chafe  55, 59 Chapusot  260, 263 Chatti  184, 200 Chomsky  8, 25, 111, 159 Cilianu-Lascu  12, 25 Collins  186, 200 Company Company  82–85 Comrie  90, 111, 225, 228, 238 Condoravdi  125, 128 Confais  258, 263 Conklin 201 Corbett  39, 57 Cordin  50, 56 Cornilescu  132, 145, 159 Cotte 200 Creissels  29, 36, 38, 40, 47, 52, 54, 57, 207, 238 Croft  35, 57, 70, 84, 200 Cruse  119, 128 Cuervo  62, 67, 79, 82, 84 D Damourette  29, 50, 57, 175, 178 Davidse  85, 120, 122–123, 128 DeLancey  186, 200

Delechelle 200 Delmas  194, 196, 200–201 Delsoir  87, 110–111 Demonte  64, 79, 84 Desclés  176, 178 Dixon  25, 56, 58–59, 118, 128, 205, 207, 214, 238 Donohue  205–206, 223, 234, 238–239 Doron  121, 128, 132, 134, 149, 159 Dowty  88, 90–92, 94, 111, 169, 178, 203, 206, 217–218, 220, 225, 238 Dubois  183–184, 189–190, 200–201 Dubois-Charlier  183, 190, 200 Duffley  245, 263 E Ebner 263 Embick  132, 138, 144, 159 Erades  117, 128 Ewing 238 F Faarlund  52, 57 Fagan  115–118, 120–121, 128 Fauconnier  88, 111 Fellbaum  116–118, 129 Fernandez  204, 238 Fiehler  268, 281 Fillmore  35, 43, 57, 88, 91, 109, 111, 119, 129 Finegan 111 Florea  172–174, 178 Forest  183, 192, 200 François  167, 176, 178, 200, 263 Franco  82, 84 Frawley 263 Fried  178, 263 Fuchs  172–174, 178

284 Atypical Predicate-Argument Relations

G Gambier 111 Ganeshan 84 García de la Maza  115–116, 119, 121, 125, 129 García Salido  82, 85 Gast  29, 48–49, 52, 57 Geisler  176, 178 Geyskens  122–123, 128 Girard-Gillet  181, 200–201 Girardin 201 Giurgea  139, 150–153, 155, 159 Givón  31, 37, 39, 46, 57, 70, 82, 84, 88, 90, 92, 111 Glinert  30, 57 Godard 200 Goldberg  35, 57, 95, 111, 122, 129, 167, 169, 178, 201 Goldenberg  27, 40–41, 43–46, 48, 50, 57 Gougenheim  176, 178 Grady  117, 129 Grangé  203, 213, 238 Greenbaum  111, 124, 129 Guéron  3–7, 25, 149, 160, 181, 184, 201 Guillet  183, 200 Günthner  272, 281 Gutiérrez-Bravo 62 H Haas  29, 48–49, 52, 57 Haider  7–9, 25 Halevy  27, 30, 39, 44, 46, 50, 58 Hall  186, 200 Hartsuiker  87–88, 112 Haspelmath  6, 19, 25, 56, 58, 201 Hawkins  88, 111 Hennig  263, 281 Herman 239 Herschensohn  4–5, 7, 25 Heyvaert 128 Hoekstra  115–116, 118, 129, 149, 160 Hoffmann  246, 264 Holmes  185, 201 Holton  233, 236, 238 Hoop 58 Hopper  45, 51, 58, 126–127, 129–130

House 110–111 Huddleston  88, 92, 111, 122, 129, 137, 139, 155, 159 Hundt  91, 94, 111, 116, 129 I Iatridou  139, 142, 160 Iordăchioaia  145–147, 152, 154–155, 159–160 Itkonen 177–178 Izvorski 160 J Jackendoff  116, 119, 129 Jespersen  39, 44, 46, 58, 116, 127, 129 Johansson 111 Junker  5, 25 Juska-Bacher 282 K Kahane  269, 281 Kayne  132, 155, 160 Keenan 58 Kemmer  117, 129 Kempson  53, 58 Keraf  204, 209, 214, 234–236, 238 Keyser  115–117, 120–121, 129, 160 Khan  42, 57–58 Kittilä  163, 168, 178 Klaiman  88, 111 Klamer  204, 212–215, 223, 230, 234, 236, 238 Kleiber  6, 25–26 Koch  40, 52, 58, 256, 262–263, 269, 281 Koenig  191, 201 Kolehmainen  177, 179 König  6, 25 Koontz-Garboden  77, 85 Krier  261, 263 Kuno  29, 51, 58, 63, 85 Kuroda  49, 58 Kussmaul  109, 111 Kuzar  30–31, 33–34, 48, 58 L Laca  85, 147, 160 Lambrecht  28–29, 36, 38–39, 42–43, 47, 49, 51–53, 58, 62–63, 85

Lampert 263 Langacker  127, 129, 167, 179, 201 LaPolla 201 Larjavaara  163–166, 170–175, 177, 179 Larrivée  245, 263 Lauwers  35, 58 Lazard  29, 39, 50, 52, 58, 163, 168, 170, 179, 183–184, 201 Leclère  183, 200 Leech 111 Lekakou  125, 130 Lemmens  188, 201 Le Querler  260, 263 Levi 59 Levin  31, 59, 88, 91–92, 111, 117, 121–122, 124, 126, 130, 181, 183, 185–186, 188, 191, 201 Li  45, 51, 57–59, 84 Lødrup  53, 59 Luján  67, 79, 85, 130 Lundquist  132, 160 Lyons  116, 119, 130 M Macfarland  181, 185, 198, 201 Magnus  243, 263 Mahlow 282 Malchukov  30–32, 47, 52, 57–59, 212, 239 Maldonado  71, 85 Marantz  8, 25, 132, 134, 160 Marín Arrese  130 Martin  25–26, 58, 130, 159, 196 Martineau  5, 25 Masullo  61, 85 Mauner 201 McKoon  181, 185–186, 198, 201 Mendikoetxea  126, 130 Merlo 201 Meyer  130, 270, 282 Miller 200 Milsark  50, 59 Mithun  55, 59, 206–207, 239 Moore  8, 25, 67, 84, 159 N Næss  163, 168, 179 Nagaya  204, 208, 213, 217, 222, 226, 235–236, 239 Nedjalkov  126, 130

Index of authors 285



Negoiţă 160 Nishida  61, 69, 71, 85 Nishiyama  204, 211, 213–214, 239 O Oesterreicher  269, 281 Ogawa  30, 32, 47, 59 Onishi  25, 37, 56, 58–59 Österreicher  256, 263 Östman 263 P Palmer  87, 111, 258–259, 263 Pampus  204, 214, 226, 239 Parodi  67, 83, 85 Partee  49, 56 Pasero  183, 201 Paul  58, 111, 129–130, 158, 186, 200, 238, 263 Perez 112 Perlmutter  4, 7–8, 25, 30–31, 59, 130 Pichon  29, 50, 57, 175, 178 Pinker  195, 201 Plungian  258–259, 264 Portner  260, 263 Primus  88, 90, 111 Prince  51, 59 Pullum  88, 92, 111, 137, 155, 159 Pustejovsky  169, 201 Pylkkänen  82, 85 Q Quirk  91, 111, 124, 129 R Rakhilina 24 Rapoport  120, 130 Rappaport Hovav  59, 122, 124, 126, 128, 130 Rauh 200 Raynaud  258, 263 Reinhart  51, 59 Reintges  132, 134, 159 Reuland  7–9, 25, 57, 129

Reznikova 24 Riegel  6, 26 Rizzi  7–8, 24 Roberts  115–118, 120, 129–130 Roeper  115–117, 120–121, 129 Rosén  29, 59 Rosta  116, 118, 130 Rothemberg  164, 179 Rousseau  178–179, 247, 258, 263 Roy 159–160 Rura  89, 112 S Sabatier  183, 201 Samardzic 201 Sanfilippo  88, 112 Sasse  29, 49, 59 Schäfer  159–160, 186, 200–201 Schanen  258, 263 Schiewe 264 Schlesinger  88, 91–92, 94, 112 Schmale  265, 270, 274–276, 279, 282 Schmerling  29, 59 Schreiber  273–274, 282 Schwitalla 282 Şerbănescu  9, 26 Serbat  177, 179 Seržant  19, 26 Shlonsky  38, 59 Sick  156, 270, 273, 282 Siewierska  31, 47, 52, 57, 59, 98, 112 Sleeman  138, 160 Smith  19, 26 Soare  131, 139–140, 145–147, 150–155, 159–160 Spanoghe  6, 26 Stefanini  183, 201 Strecker 264 Suñer  76, 82, 85 Svartvik 111 T Talmy  57, 84, 92, 111–112, 123, 130, 181, 201

Tamm  88, 112 Tesnière 201 Teubert  271, 282 Thompson  45, 51, 59, 84, 126–127, 129–130 Tognini-Bonelli 282 Traugott  45, 51, 58 U Ulrich  28, 60, 263 V Vandepitte  87–88, 112 Van der Auwera  258–259, 263–264 Vandeweghe 112 Vanhoe  64–65, 85 van Lier  71, 73, 85 Van Oosten  116–117, 130 Van Peteghem  3, 6–7, 26 Van Valin  188, 197, 201, 204– 207, 239 Vázquez Rozas  67–68, 82, 85 Velupillai  90, 112 Vendler  117, 120, 130 Vergnaud  4–7, 26 Verschueren 263 von Heusinger  67, 85 W Ward  51, 60, 129, 238 Weinrich  246, 264 Weissenrieder  70, 77, 85 Wichmann  205–207, 238–239 Willems  35, 58, 179 Williams  126, 130, 160 Woolford  9, 22, 26 Z Zewi  42, 60 Zifonun  246, 258, 264, 268, 276, 282 Ziv  51, 60 Zribi-Hertz  183, 201

Index of subjects A accusative  3, 5, 7–8, 14, 18–19, 24, 55, 56, 62, 67, 111, 119, 133, 136–137, 141, 205, 215 affected  23, 34, 119, 121, 167– 168, 203, 217, 223, 228 affectedness  128, 206 agent  35, 47, 89–90, 94, 101– 102, 122, 169, 206, 220 agent-like  203, 205, 220 agentive  22, 89–91, 93–94, 185–186, 225 agentivity  56, 217 alignment  31, 40, 49–50, 203–207, 213 anaphor 5 anaphora 6 anaphoric 5–6 animacy  71, 91, 94 animacy hierarchy  71, 91, 94 anticausative 20 argument  3–5, 7–12, 14, 37–39, 42–48, 52–54, 71, 90–91, 115–116, 118–120, 122–123, 125–130, 141, 170, 183, 233, 238 external argument  8, 17, 30, 122, 141–142, 149 internal argument  3–4, 7, 14, 125, 141 argument structure  127 arguments  3, 8–10, 12, 14, 17, 22–24, 33, 125–127, 141, 148, 150, 158, 233–235 aspect  148, 155, 207, 226, 229 perfect aspect  229 atypicality  165, 265–266, 280 B Burzio’s generalization  4–5,7– 9, 25–26,30–31,57

C case  8, 22 inherent case  22 morphological case  8 categorical  49, 53 causation  92, 94, 188 external causation  181, 192 internal causation  181, 184–186, 188 causative  183–184, 186, 188 causative alternation  11, 186 cause  15,124, 184–185, 199 causee  181, 184, 188 causer  92, 188, 199 clitic  62, 73, 76, 81–82, 213 clitics 82 cognitive grammar  167 configurational 116 Construction Grammar  35, 54, 166–167 control  37, 39, 90–94, 98, 220 D dative  4, 12, 19–22, 34–35, 75, 78, 82–83 deictic  37, 42, 53 definite 51 definiteness 50 direct object (DO)  78, 81, 164, 198 ditransitive  64, 69, 78 divalent  94, 98, 108 Dutch  52, 87–89, 94–104, 106–111, 128–129, 232 E English  19, 26, 29–31, 38, 40– 41, 44, 46, 50, 52, 56, 58–60, 72, 79, 85, 87–89, 92, 95–98, 102–108, 110–112, 115–116, 119, 122, 124–126, 128–132, 136–143,

151, 154, 158–160, 170–171, 182, 186–187, 189, 201, 245, 260, 266–268, 270, 272, 279–280 ergative  30, 115, 120, 125, 129, 205 existential  28–31, 35, 52, 54 experiencer  3, 7, 18, 24, 61, 82, 213 expletive  27, 48–49 F finite 244 French  4–7, 19, 25–27, 29–31, 36, 38, 40, 42–45, 48–50, 52, 58, 138, 140, 150, 163–165, 167–172, 174, 176, 178, 181–183, 186–191, 193, 198, 208, 221, 228, 245, 258–261, 266–268, 270, 272, 280 G Generative Grammar  3–4, 9, 30, 134, 150 ground  47, 49, 55 grammaticalization  31, 42, 55, 262 H Hebrew  27–33, 35, 37–39, 41–48, 50–52, 54, 56–60, 132, 149 I Idiomatic  88, 110, 266, 268 idioms  98, 267, 280 impersonal  7, 14–15, 28, 31, 34, 40–41, 47–49, 52 inalienable possession  4–5, 7 inanimate  42, 48, 68, 90–91, 188, 208, 224, 228–229, 245 incremental theme  5, 206, 235

288 Atypical Predicate-Argument Relations

indirect object (IO)  5, 61–83, 212 infinitive  68, 135–136, 139–140, 143, 145, 149–150, 154, 243– 247, 251–254, 258, 260, 262, 270, 279 information structure  31, 49, 51, 53–54 instrument  91–92, 100, 191 instruments  89, 91–92, 94, 105, 191 instrumental  91–92, 105, 188 intransitive  10–11, 16, 20, 22, 36, 38, 48, 52–54, 115, 119–126, 163–165, 169, 174–176, 181–183, 185–187, 189, 204–205, 207–208, 210–214, 216–217, 220–221, 223, 225, 227–228, 230–237 intransitives  7, 121, 128 intransitivity  115, 127, 203–207, 210–212, 214, 228, 233–239 split intransitivity  203–207, 210–211, 214, 228, 233–239 L Lamaholot  203–215, 217, 219– 223, 226, 228, 232–239 M mediopassive 129 modality  57, 98, 117, 153, 157, 159, 243–244, 258–260, 262–263 mood  68, 133, 152, 168, 207, 263 monovalent  204, 221 N nominalization  141, 144–147, 158, 160 nominative  3, 5, 7–9, 14, 19, 24, 26 non-finite  52, 131–132, 134–137, 139–140, 142–143, 150, 159 number hierarchy  71 O object  3, 5–7, 9–10, 12, 15–16, 18–19, 22, 29–31, 42–43, 47, 52, 58, 67, 70, 72, 82–85, 91,

115, 119–120, 122–124, 126, 133, 139, 147, 152, 164–170, 172–175, 177, 183, 198, 205, 208, 212–213, 215, 221, 223, 234, 258 cognate object  165, 173 P participle  16–18, 132–133, 137–138, 140–141, 143–144, 149–151, 153, 251 participles  44, 131–137, 139–144, 149–150, 157–160 passive  17–18, 25, 28, 34, 43, 56, 89, 97, 99, 101, 103, 115, 133, 138–139, 141–144, 149, 151, 158 passives  7, 38, 43, 59, 115, 129–130 passivization  143, 150 Patient  19, 27, 31, 38–39, 41, 43, 52–53, 55, 90, 95, 98, 100–101, 108, 115–116, 119–120, 125, 129, 167, 169–170, 177, 183, 205–206, 208, 217 patient-like  203–205, 216, 220, 223–224, 231–232, 237 phraseology  273, 282 phraseological  37–38, 49, 267–268, 273–274 possessor  4–5, 7, 22, 30, 68, 72, 87, 97, 99, 101, 103, 212, 216 possession  4–5, 7, 25, 30, 32, 42, 50, 56, 79, 94, 97, 99, 103, 238 Proto-Agent  88, 90, 170, 206, 218–220, 225, 235 Proto-Patient  170, 218–219, 225, 229, 235 proto-role  169, 177, 206, 217, 220 proto-roles  90, 111, 169, 178, 203, 206–207, 238 psychological verbs  7, 84–85 psych verb  61–62, 64, 66, 68–71, 74, 79–83 psych verbs  24, 47, 61–62, 64–68, 71–73, 83, 85 Q quirky case  8

R recipient  70–71, 77–78, 87, 94–95, 98–101, 103, 108 referential  27, 29, 37, 41, 45–47, 50, 55, 61, 65, 70–75, 77–78, 81, 83, 85, 134, 145 referentiality  71, 75 reflexive  8, 19–20, 22, 64, 129, 146, 176, 212, 223 S semantic role  8, 12, 53, 87, 92, 94, 99, 109, 165, 169–170, 183, 187, 205 semantic roles  88, 90–91, 98, 105, 110, 127 thematic role  55 thematic roles  111–112, 128, 238 theta-role  115, 118–120 θ-role 7–8 S salience  19, 51 sentience  90, 92, 203, 206, 215–217, 234, 237 Spanish  4, 19, 29, 48, 50, 61–63, 65, 67, 69–72, 81–85, 147, 260 stimulus  10, 19, 28, 61, 233 subject  3–10, 12–14, 16, 18–20, 22–24, 27, 29–31, 35, 38–39, 44–48, 52–53, 55–63, 68, 70, 76, 78–79, 84, 87, 89–92, 94–103, 105–110, 115–120, 125, 131, 133–142, 144–145, 148–149, 152–155, 158, 164–165, 167–172, 174– 177, 183, 187, 190, 192–194, 200, 203, 205, 207, 212–221, 223–225, 229, 231, 234, 237, 239, 244, 260–261, 272, 275, 279–280 accusative subject  3, 23 dative subject  9, 19, 22–24 dummy subject  46, 48 nominative subjects  9, 19, 137 supine  131, 133, 135–136, 141, 143–150, 152–158, 160

Index of subjects 289



T telic  155, 203, 219, 226, 237 tense  14, 37, 41, 44, 57, 68, 131, 133, 135–137, 139–140, 147, 149 theme  3–4, 7, 11–12, 19, 23, 31, 38, 43, 45, 51, 61, 87, 99, 101, 103, 107, 119–120, 125, 129, 183, 206 thetic  28–29, 47–49, 53, 58–59 topic  29, 39, 45, 51, 53–54, 57–59, 62–63, 65, 70, 84, 239 transitivity  58, 84–85, 115–117, 121–122, 124–127, 129–130, 163–164, 168–169, 172–174, 176–179, 201, 213 trivalent  95, 98, 100, 108–109 U unaccusative  4, 6–8, 16, 25, 30, 38, 41, 54, 59, 67, 119–122, 124–125, 130, 149, 151, 183, 187, 200 unaccusatives  28, 30–31, 36, 38, 41, 121, 124–125, 200 Unaccusative Hypothesis  7, 25, 30, 59, 119, 125, 130

unaccusativity  4–5, 7, 16, 24, 31, 59, 130, 201 undergoer  28, 188, 197 unergative  8, 24, 36, 38, 41, 54, 119, 122–126, 149 unergatives  28, 30–31, 120–121, 124–126, 149 V valency  87, 108–110, 179, 207 verbs  3–12, 14–17, 19–24, 26, 28, 30, 36, 38, 41–43, 45–48, 52–54, 58, 61–62, 64–69, 71–73, 79, 81, 83–85, 88, 92–95, 97–101, 108–109, 115–116, 119–124, 126–132, 134–135, 137, 139, 141, 155–157, 160, 163–167, 173, 178–179, 181–187, 189–190, 192, 196, 198, 201, 203, 205, 207–220, 222, 225–227, 229–234, 236–237, 243, 245–247, 258–260, 279 stative verbs  203, 208–210, 214–215, 225, 227, 229–231, 237

manner-of-motion verbs  123 modal verbs  98, 243, 245–247, 258–260 motion verbs  42, 123, 203, 217–219, 225, 234, 237 verbs of pain  3–5, 9, 15–16, 19, 26 voice  80, 99, 101, 103, 111, 117, 129, 131, 133–134, 140–141, 145–146, 149, 151, 158–159, 200, 208, 239 middle  115–122, 125–126, 128–130, 146 passive  17–18, 25, 28, 34, 43, 56, 89, 97, 99, 101, 103, 115, 133, 138–139, 141–144, 149, 151, 158 passives  7, 38, 43, 59, 115, 129–130 volition  90–92, 94, 206, 220– 221, 223, 237

This book deals with atypical predicate-argument relations. Although the relations between predicates, especially verbal, and their arguments have been long studied, most studies are concerned with typical telic verbs in the past tense, indicative mood, active voice, with all arguments expressed. Recently, linguists have become interested in other types of predicate-argument relations displaying atypical properties, be they morphological or syntactic, in one language or cross-linguistically. The articles in this book investigate some of these: argument marking with some special groups of verbs, arguments not foreseen in the verb valency and contributed by the construction, verbs in idiomatic constructions, valency-changing operations, arguments in thetic sentences or in participle constructions etc. The authors work within diferent theoretical frameworks and on various languages, from more current languages like English, Spanish, French or German, to Hebrew or Lamaholot, an Austronesian language.

isbn 978 90 272 3143 7

J OHN B ENJAMINS P UB LISHING COMPANY