Subjects in Constructions – Canonical and Non-Canonical : Canonical and Non-Canonical [1 ed.] 9789027269188, 9789027204387

Citation preview

Subjects in Constructions – Canonical and Non-Canonical

Constructional Approaches to Language he series brings together research conducted within diferent constructional models and makes them available to scholars and students working in this and other related ields. he topics range from descriptions of grammatical phenomena in diferent languages to theoretical issues concerning language acquisition, language change, and language use. he foundation of constructional research is provided by the model known as Construction Grammar (including Frame Semantics). he book series publishes studies in which this model is developed in new directions and extended through alternative approaches. Such approaches include cognitive linguistics, conceptual semantics, interaction and discourse, as well as typologically motivated alternatives, with implications both for constructional theories and for their applications in related ields such as communication studies, computational linguistics, AI, neurology, psychology, sociology, and anthropology. his peer reviewed series is committed to innovative research and will include monographs, thematic collections of articles, and introductory textbooks. For an overview of all books published in this series, please see http://benjamins.com/catalog/cal

Editors Jan-Ola Östman

Kyoko Ohara

University of Helsinki, Finland

Keio University, Japan

Advisory Board Peter Auer

Mirjam Fried

Knud Lambrecht

University of Freiburg, Germany

Charles University, Prague, Czech Republic

University of Texas at Austin, USA

Adele E. Goldberg Princeton University, USA

Max Planck Institute for Evolutionary Anthropology, Germany

Seizi Iwata

Arnold M. Zwicky

Kansai University, Japan

Stanford University, USA

Hans C. Boas University of Texas at Austin, USA

William Crot University of New Mexico, USA

Charles J. Fillmore† International Computer Science Institute, Berkeley, USA

Michael Tomasello

Paul Kay University of California, Berkeley, USA

Volume 16 Subjects in Constructions – Canonical and Non-Canonical Edited by Marja-Liisa Helasvuo and Tuomas Huumo

Subjects in Constructions – Canonical and Non-Canonical Edited by

Marja-Liisa Helasvuo University of Turku

Tuomas Huumo University of Turku

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

he 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.

Library of Congress Cataloging-in-Publication Data Subjects in Constructions – Canonical and Non-Canonical / Edited by Marja-Liisa Helasvuo and Tuomas Huumo. p. cm. (Constructional Approaches to Language, issn 1573-594X ; v. 16) Includes bibliographical references and index. 1. Grammar, Comparative and general--Topic and comment. 2. Grammar, Comparative and general--Coordinate constructions. 3. Grammar, Comparative and general--Syntax. 4. Construction grammar. I. Helasvuo, Marja-Liisa, editor. II. Huumo, Tuomas, editor. P163.5.S83 2015 415’.018--dc23 2014035633 isbn 978 90 272 0438 7 (Hb ; alk. paper) isbn 978 90 272 6918 8 (Eb)

© 2015 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · he Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Abbreviations Canonical and non-canonical subjects in constructions: Perspectives from cognition and discourse Marja-Liisa Helasvuo and Tuomas Huumo

vii

1

Section I. Grammatical and discourse perspectives on non-canonical subjects On the subject of subject in Finnish Tuomas Huumo and Marja-Liisa Helasvuo

13

Hidden subjects in conversation: Estonian personless verb forms as referential devices Renate Pajusalu

43

Subjects under generic conditions: Implied subjects in Finnish and Estonian if-clauses Hanna Jokela and Helen Plado

73

Section II. Stretching the limits of subjecthood Abstract locational subjects: Field and settings in French and English Michel Achard

101

Subjecthood of the agent argument in Estonian passive constructions Liina Lindström

141

Categorization and semantics of subject-like obliques: A cross-linguistic perspective Ilja A. Seržant

175

he world is raining: Meteorological predicates and their subjects in a typological perspective Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

205

vi

Subjects in Constructions

Section III. Subjects in networks of constructions he syntactic and semantic history of the Finnish genitive subject: Construction networks and the rise of a grammatical category Jaakko Leino

231

From canon and monolith to clusters: A constructionist model of subjecthood in Russian Aki-Juhani Kyröläinen

253

he role of non-canonical subjects in the overall grammar of a language: A case study of Russian Laura Janda and Dagmar Divjak

293

Construction index

319

Subject index

321

Abbreviations

1 2 3 abl abs acc ade adj adv agptcp all aor app art aux caus com cmpr cond conneg cont conv dat ds deb def dem dmn ela erg exp fem gen ger ill imp impf

First person Second person hird person Ablative Absolutive Accusative Adessive Adjective Adverb Agent participle Allative Aorist Active past participle Article Auxiliary Causative Comitative Comparative Conditional Connegative Continuative Converb Dative Diferent subject Debitive Deinite Demonstrative Deminutive Elative Ergative Expletive Feminine Genitive Gerund Illative Imperative Imperfect

in ind ine inf instr invar iprs jus kvt lat long masc n neg nom obl p par pass pfv pl plain.ind poss ppp pr pre pred prf prs prt prv pst ptcl ptcp q r refl

Location inside a landmark Indicative Inessive Ininitive Instrumental Invariant Impersonal Jussive Quotative (evidential) Lative Long form Masculine Neuter Negation, negative Nominative Oblique Patient Partitive Passive Perfective Plural Plain indicative Possessive Passive past participle Preposition Prepositional case Predicative Perfect Present Particle preverb Past Particle Participle Question marker Realis Relexive

viii Subjects in Constructions

rel rem s sg ss

Relative Remote past Single argument of an intransitive verb Singular Same subject

sta sup tam ter tns trnsl

Stative Supine Tense-aspect-mood Terminative Tense Translative

Canonical and non-canonical subjects in constructions Perspectives from cognition and discourse Marja-Liisa Helasvuo and Tuomas Huumo University of Turku

1.

Introduction

Subjects, and subjecthood in general, have received substantial attention in cognitive and discourse-functional linguistics, and along with the criteria for subjecthood, less canonical instances of subjects have also been discussed (see e.g. Aikhenvald et al. 2001). he purpose of this volume is to discuss constructions with non-canonical subjects in individual languages and cross-linguistically. he volume incorporates insights from both approaches for a deeper understanding of non-canonical subjects, and of the concept of subjecthood in more general terms. Ultimately, the volume scrutinizes the criteria for linguistic functions presented by both traditions. A common denominator for the two approaches is their usagebased conception of language, where the role of the language user or conceptualizer is central in construing linguistic expressions. he individual chapters of the volume assume a constructional perspective rooted on the one hand in cognitive linguistics, on the other in discoursefunctional or interactional linguistics. In cognitive linguistics, one point that is emphasized is that a clause-level linguistic expression always designates a conceptualization of a situation from a particular point of view and in terms of a particular construal, where the speaker has multiple choices (e.g., Langacker 1987, 1991). he function of a grammatical subject is that of a trajector, deined in Cognitive Grammar as a primary focal participant in the designated relationship,1 and it also 1. Langacker (2008: 365) deines the grammatical subject as a clause-level trajector, which is a notion based on the cognitive ability known as the focusing of attention (see also Talmy 2000, Chapter 5). In Cognitive Grammar, the clause-level trajector is the primary focal participant doi 10.1075/cal.16.01hel © 2015 John Benjamins Publishing Company

2

Marja-Liisa Helasvuo and Tuomas Huumo

functions as the semantic starting point for the predication carried by the clausal expression (compare the discourse-functionalist idea of subjects as starting points; Chafe 1994). In a similar vein, coming from a frame-semantic approach, Fried (2005) suggests that the relationship between predicates and the morphosyntactic expression of their arguments is best described with the help of the notion of scene: linking patterns are organized in a network of generalized scene types, each of which has a particular role coniguration where one of the roles represented in the scene is picked out as the vantage point for presentation of the event type. In Construction Grammar, the valence of a predicate (verb) represents a subset of the elements of a frame, and thus a certain conceptualization of the frame. Constructions difer as to which elements of the frame are represented in the valence and are thus projected into the syntactic expression (Östman & Fried 2004: 42–43). For instance, the range of roles available for the grammatical subject difer between languages, English being an example of a language where any lexical construction of the syntactic type ‘verb’ must unify with a linking construction that provides a subject valence slot (Östman & Fried 2004: 43–44). he situation is diferent in many of the languages studied in the present volume (e.g. Russian and Finnic) where there are verbs that do not select a grammatical subject and where the subject status of some subject-like elements is questionable. Alternatively, the construal of linguistic expressions can be approached in terms of observed discourse patternings, for example by looking at the most typical subject-verb combinations (Scheibman 2002) or at the discourse features of subjects (Helasvuo 2001). Interactional linguistics studies the routinization of these recurrent patternings as they are used in interaction in performing various social actions, for example in the expression of stance (see e.g. Kärkkäinen 2003 on the English I think). We might say that while cognitive linguistics takes the viewpoint of the individual conceptualizer, interactional linguistics places more emphasis on the intersubjective nature of linguistic expression.

2. Subjects and non-subjects in constructions Prototypical subjects have oten been characterized in terms of semantic, syntactic and discourse features, such as animacy, agentivity, topicality, referentiality, deiniteness and autonomy of existence of the subject referent (cf. Keenan 1976 on subjects and prototypes). Non-canonical subjects are thus subjects that lack some of these features, and this may be relected in their meaning (as lack of agentivity in the relationship proiled (indicated) by the clause-level construction. his relationship is an elaborated and grounded instance of the process type indicated by the inite verb.

Canonical and non-canonical subjects in constructions

or animacy), their grammatical coding (e.g. deviant case marking or subject-verb agreement pattern), and/or their discourse function (lack of topicality). here are two crucial questions: where the line between subjects and non-subjects is to be drawn in a particular language, and how the grammatical function of a nonsubject is to be deined in case it is not categorized as a subject. One possible answer has been suggested by Radical Construction Grammar (Crot 2001), which considers subjects (and other grammatical functions) construction-speciic, thus allowing the criteria for subjecthood to vary along with the constructions in which the relevant elements are used (for a discussion of the Radical Construction Grammar approach to subjects, see Kyröläinen, this volume). Some of the features identiied with non-canonical subjects – most notably, deviant grammatical coding and lacking semantic features – have been discussed intensively in the literature on impersonal constructions. Impersonalization is associated structurally with the absence of a canonical subject and functionally with agent defocusing. Impersonal constructions may for example exhibit nonnominative (oblique) marking, or may lack referentiality of the primary argument or invariant person marking of the verb. (For further discussion, see Siewierska 2008.) As discussed in several chapters in the present volume, impersonalization oten entails features associated with non-canonical subjects (see chapters by Pajusalu, Plado & Jokela, Achard, Kyröläinen, Janda & Divjak). he criteria for subjecthood discussed in the volume include grammatical, semantic, and discourse properties, and thus lie at the intersection of cognitively oriented and discourse-oriented approaches to language. In discussing noncanonical subjects in individual languages and cross-linguistically, the chapters in the volume address the following more general topics: – – – – – – – –

What kinds of grammatical (morphological and syntactic), semantic and discourse criteria can be used to distinguish subjects from non-subjects? To what extent are subject criteria construction-speciic? What kinds of constructions have non-canonical subjects? What are the semantic and discourse functions of (constructions with) noncanonical subjects? Are subjects which are grammatically non-canonical also atypical in terms of their discourse features? What kinds of meanings tend to be expressed by non-canonical subjects cross-linguistically? What kinds of language-speciic uses do non-canonical subjects have? What counts and what does not count as a subject in a particular language or in a particular construction?

3

4

Marja-Liisa Helasvuo and Tuomas Huumo

he chapters approach these questions in the light of data from a variety of languages. Some writers take the description of an individual language as their starting point (Finnish by Huumo & Helasvuo and by Leino, Estonian by Pajusalu and Lindström, Russian by Kyröläinen, Janda & Divjak); while others assume a contrastive vantage point (Achard comparing English and French, Jokela and Plado comparing Estonian and Finnish) or look at linguistic phenomena from a wider typological perspective (Seržant and Eriksen, Kittilä & Kolehmainen).

3. Grammatical and discourse perspectives on non-canonical subjects Part I discusses subjects from the point of view of their grammatical and discourse properties. he chapters by Tuomas Huumo and Marja-Liisa Helasvuo, Renate Pajusalu and Hanna Jokela and Helen Plado each take certain constructions types as their starting point and discuss whether these construction types have subjects or not. In these chapters several discourse-related concepts such as referentiality, referent tracking, and topicality are discussed in relation to grammatical and semantic properties. In their chapter, “On the subject of subject in Finnish”, Huumo and Helasvuo discuss the existential construction in Finnish and other construction types related to it (constructions describing states and results). he article argues, from a Cognitive Grammar (CG) point of view, that Finnish existential NPs which have traditionally been analyzed as subjects, should not be analyzed this way, as they do not meet most grammatical, semantic and discourse criteria for subjects. More generally, it is argued that the irst argument (trajector in CG terms) of the verb will not necessarily constitute the subject of the clause-level construction. he function of an e-NP (existential NP) is proposed for such elements instead. Huumo and Helasvuo also argue that the meaning of Finnish existential clauses involves features of theme orientation as opposed to the more general agent orientation, where the agent plays a prominent role as initiator of the event (cf. Langacker 2008). hey point out that many expressions with a theme allow both a construal in which the theme is the clause-level trajector (i.e., the subject), and one where another element serves as the clause-level trajector (such as the setting subject construction; see also Achard, this volume). Pajusalu’s chapter, “Hidden subjects in conversation: Estonian personless verb forms as referential devices”, discusses Estonian construction types which lack a grammatical subject and which contain verbs in conditional and impersonal forms. Unlike most other inite verbs in Estonian, these verb forms lack a person marking indexing the subject. hese forms have usually not been viewed as referential because there is no linguistic element capable of referring to something. On

Canonical and non-canonical subjects in constructions

the basis of everyday and institutional conversations, the author shows that these forms can both be used to introduce an unspeciic group referent into the referential frame of the conversation despite the lack of an explicit subject. However, they have diferent contexts of use in conversational data: the personless conditional is preferred when the aim of the conversation is to plan a future activity, while the existential impersonal is mostly used in narratives. he personless conditional is typically inclusive, incorporating one or both of the interlocutors into the group, while the impersonal is typically exclusive and refers to a group that does not include the participants of the conversation. Jokela and Plado also discuss the relationship between referentiality and subjects in construction types creating open or generic reference. heir chapter, “Subjects under generic conditions: Implied subjects in Finnish and Estonian if-clauses”, focuses on certain construction types used in conditional clauses in Estonian and Finnish. he common feature in these construction types, in addition to their lack of subject, is that they all serve to create open reference. By looking at the Finnish and Estonian systems contrastively Jokela and Plado explore the possibilities of expressing generic or open reference in languages.

4. Stretching the limits of subjecthood he second part of the volume focuses on the characteristics of subjects that are in some respect non-prototypical. Michel Achard discusses abstract locational subjects, while Liina Lindström and Ilja Seržant examine oblique subjects. Pål Eriksen, Seppo Kittilä and Leena Kolehmainen explore weather expressions in the world’s languages. In his chapter, “Abstract locational subjects: Field and settings in French and English”, Achard takes as his starting point the view of Cognitive Grammar (Langacker 1987, 1991, 2008) that diferent types of clauses are particularly well suited to express certain types of events. he subject role of many clause types is reserved for an agentive participant that initiates the event, but there are also constructions that deviate from this principle (cf. also Huumo & Helasvuo, this volume, on theme orientation). Achard investigates English and French impersonal constructions that select as their focal igure a location or setting within which the proiled event takes place, and mark it as the subject. A location is a relatively restricted area, whereas a setting is “a global expanse within which events unfold (the diference is one of degree)” (Langacker 2009: 118). Langacker suggests that impersonal constructions select the “ield” as the focal igure in the proiled relation, and that the ield is represented by the impersonal pronoun: “impersonal ‘it’ proiles the relevant ield, i.e. the conceptualizer’s scope of awareness for the issue at hand” (Langacker 2009: 139). In his chapter, Achard argues that in order

5

6

Marja-Liisa Helasvuo and Tuomas Huumo

to describe the English and French impersonals, the notion of “ield” needs to be expanded. He proposes that the conceptualizer’s scope of awareness should be speciied in terms of its viewing organization. Achard thus advances and elaborates existing Cognitive Grammar analyses of constructions with an abstract setting subject. Furthermore, he proposes speciic lines of investigation for abstract locational subjects that could provide a basis for a typology of the diferent types of regions they code in the languages for the world. Lindström’s chapter, “Subjecthood of agent argument in Estonian participial constructions”, takes a historical perspective and discusses changes in expressing the agent in Estonian passive constructions. Estonian has several diverse agentmarking devices, and the use of these devices has changed considerably over the past 200 years. Lindström focuses on the adessive and elative agent adverbial in passive sentences, comparing their agentivity, topicality, and control over the use of the relexive pronoun and over ellipsis in the second clause of a sentence containing coordinated clauses. All of these properties have been considered as subject properties. Lindström shows that the adessive argument has more subject properties than the elative argument. he most important feature appears to be topicality, since it has a strong efect on the interpretation of the whole construction. In his chapter, “Categorization and semantics of subject-like obliques”, Seržant looks at constructions with a non-prototypical trajector (subject) suggesting a categorization of these constructions into syntax-, gram- and lexeme-driven ones depending on the nature of the domain that triggers the oblique case-marking on the trajector argument. Additionally, a uniied semantic account in terms of a radial category is advanced. All three types of constructions conceptualize the event as being a causally consequent event and imply the existence of a causally antecedent event. Seržant shows that in spite of this common semantic core the three types difer in the pragmatic properties of the antecedent event and its main participant: in the syntax-driven type the antecedent event is conceptualized as onstage, explicit and referential, whereas in the gram-driven type, it is ofstage, implicit, and non-referential. he lexeme-driven type merely implies the existence of an antecedent event and does not assess the concept of this event. In their chapter, Eriksen, Kittilä and Kolehmainen study meteorological expressions (such as it is snowing or the wind blows) from a functional-typological perspective. hese expressions linguistically constitute a challenging type of constructions: he situations they describe inherently lack proper participants with distinct and salient semantic roles, such as agent, patient or experiencer. his makes them diferent from canonical intransitive and transitive events that can be distinguished from each other, for example, on the basis of the nature of the participants involved in them. he unorthodox nature of meteorological events has, expectedly, consequences for the meteorological expressions and their subjects.

Canonical and non-canonical subjects in constructions

Eriksen, Kittilä and Kolehmainen show that the argument structure properties of these constructions depend on the peculiar semantic nature of meteorological events, and this is relected both in the semantic and in the formal characteristics of the subjects of weather predicates. he subject may be missing altogether, and in cases in which it is overtly present it shows features atypical of canonical subjects.

5. Subjects in networks of constructions In the third part of the volume the focus shits to the clause-level construction of which the subject element forms a part. he chapters in Part III also discuss the status of such constructions in the grammatical system of the language, which is seen as consisting of a network of constructions. Some theories of Construction Grammar (most notably Crot’s [2001] Radical Construction Grammar) argue for a construction-speciic deinition of grammatical functions, including the subject. his means that the criteria given for canonical subjects in frequent construction types (such as transitive or intransitive clauses) need not concern subject-like elements in other constructions, which can deviate substantially from canonical subjects. Such a point of view may be especially fruitful for an analysis of noncanonical subjects, since they may behave diferently from canonical subjects but nevertheless fulill a subject-like role in their respective constructions. he three chapters in the third part of this volume address the issue of non-canonical subjects, and the criteria for subjecthood in networks of closely related constructions which difer from each other in some respects but involve similar subject elements. In his chapter, “he syntactic and semantic history of the Finnish Genitive subject: Construction networks and the rise of a grammatical category”, Jaakko Leino examines the Finnish genitive subject. It is a genitive-marked element that occurs in many non-inite constructions and has traditionally been analyzed as the subject constituent of these constructions. he verb, being in non-inite form, does not agree with the subject, and the genitive subject may or may not govern possessive suixes and other anaphoric elements. Leino shows that the genitive subject has three diferent origins in diferent constructions: dative-like use of the genitive, syncretism of genitive and accusative, and genitive attributes of deverbal nouns. It is shown that out of these coincidentally similar origins, a fully grammaticalized category has emerged through a complex set of intertwined reanalyses. In his chapter “From Canon and monolith to clusters: A constructionist model of subjecthood”, Aki-Juhani Kyröläinen analyzes the subjecthood of the dative argument in Russian relexive constructions in terms of Radical Construction Grammar. His data consist of the usage patterns of ten constructions (transitive, intransitive, ditransitive, dative-ininitive, four modal constructions and

7

8

Marja-Liisa Helasvuo and Tuomas Huumo

two impersonal constructions) and their co-occurrence with 20 features. he data show that, contrary to postulations in formal approaches, a number of features previously claimed to be ungrammatical are used in contemporary written Russian. he cluster-based approach to subjecthood allows us to simultaneously take into consideration a number of factors which have not commonly been discussed jointly in previous studies on subjecthood: (1) (dis)similarity between diferent constructions, (2) the evaluation of possible subject candidates, (3) the status of proposed tests, (4) the possible hierarchy between diferent features, and lastly (5) a network model of subjecthood in Russian based on inheritance. he last chapter in Part III, and of the volume, is Laura Janda’s and Dagmar Divjak’s “Attenuated agency in Russian”. he chapter takes the perspective of construction grammar (Goldberg 1995 & 2006 as well as Crot 2001) in analyzing the use of non-canonical subjects in Russian. he focus is on the organization of these constructions in larger networks of related personal and impersonal constructions, with impersonal constructions as peripheral members of the system. In addition to addressing the intricacies of the rich system of Russian impersonal constructions, the chapter makes a theoretical contribution to construction grammar. Most construction grammar analyses do not investigate the ways in which constructions can be related to each other. he authors show that construction grammar can be applied to a language in which syntax is largely expressed in terms of grammatical case, and that the entire grammar of a language can be understood as a network of constructions related to each other in terms of the number and case marking of the participants. Overall, this view of grammar is parallel to the radial category, in which members of a category are mutually interrelated. Syntax is thus understood as a network in which neighboring constructions support and inluence each other.

6. Concluding remarks his volume brings together cognitive and discourse perspectives on non-canonical subjects. his approach highlights the importance semantic and discourse criteria in deining syntactic functions. It also puts into focus the charactistics of speciic constructions where the potential subject candidates are used. he constructions discussed in the chapters of this volume include existentials, possessive constructions, setting subject constructions, impersonals and constructions expressing meteorological conditions among others. hus, the focus is on constructions which are usually ignored in discussions of general subject criteria. he chapters combine analyses of individual languages with cross-linguistic and typological perspectives.

Canonical and non-canonical subjects in constructions

References Aikhenvald, Alexandra Y., R. M. W. Dixon, & Masayuki Onishi (Eds.). (2001). Non-canonical marking of subjects and objects. Amsterdam: John Benjamins. DOI: 10.1075/tsl.46 Achard, Michel (1998). Representation of cognitive structures: Syntax and semantics of French sentential complements. Berlin and New-York: Mouton de Gruyter. DOI: 10.1515/9783110805956 Chafe, Wallace L. (1994). Discourse, consciousness, and time: he low and displacement of conscious experience in speaking and writing. Chicago: University of Chicago Press. Crot, William (2001). Radical construction grammar. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780198299554.001.0001 Fried, Mirjam (2005). A frame-based approach to case alternations: he swarm-class verbs in Czech. Cognitive Linguistics, 16(3), 475–512. DOI: 10.1515/cogl.2005.16.3.475 Fried, Mirjam, & Östman, Jan-Ola (2004). Construction grammar: A thumbnail sketch. In Mirjam Fried & Jan-Ola Östman (Eds.), Construction grammar in a cross-language perspective (pp. 11–86). Amsterdam: John Benjamins. DOI: 10.1075/cal.2.02fri Goldberg, Adele (1995). Constructions: A construction grammar approach to argument structure. Chicago: Chicago University Press. Goldberg, Adele (2006). Constructions at work: he nature of generalization in language. Oxford: Oxford University Press. Helasvuo, Marja-Liisa (2001). Syntax in the making: he emergence of syntactic units in Finnish conversation. Studies in Discourse and Grammar 9. Amsterdam: John Benjamins. DOI: 10.1075/sidag.9 Kärkkäinen, Elise (2003). Epistemic stance in English conversation. A description of its interactional functions, with a focus on I think. Amsterdam: John Benjamins. DOI: 10.1075/pbns.115 Keenan, Edward (1976). Towards a universal deinition of “subject”. In Charles N. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press. Langacker, Ronald W. (1987). Foundations of cognitive grammar. Vol. 1: heoretical prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. (1991). Concept, image, and symbol: he cognitive basis of grammar. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110857733 Langacker, Ronald W. (2008). Cognitive grammar: A basic introduction. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780195331967.001.0001 Langacker, Ronald W. (2009). Investigations in cognitive grammar. Berlin and New York: Mouton de Gruyter. DOI: 10.1515/9783110214369 Scheibman, Joanne (2002). Point of view and grammar. Structural patterns of subjectivity in American English conversation. Amsterdam: John Benjamins. DOI: 10.1075/sidag.11 Siewierska, Anna (2008). Impersonalization from a subject-centered vs. agent-centered perspective. Transactions of the Philological Society, 106, 115–137. DOI: 10.1111/j.1467-968X.2008.00211.x Talmy, Leonard (2000). Toward a cognitive semantics. Volume 1: Concept structuring systems. London: he MIT Press.

9

section i

Grammatical and discourse perspectives on non-canonical subjects

On the subject of subject in Finnish* Tuomas Huumo and Marja-Liisa Helasvuo University of Turku

his article examines the category of subject in Finnish. Among the grammatical features examined are case marking, agreement and syntax, semantic role, its relation to other sentence elements and its semantic function as regards the sentence as a whole. One important discourse feature is whether the subject introduces a referent and whether that referent is mentioned again later in the discourse. We discuss diferent construction types and examine whether the alleged subjects fulil the role of the subject as deined in cognitive grammar, i.e. elements which function as a clause-level trajector and as a starting point for the construal of the relationship and its participants. More generally, it is argued that the trajector does not necessarily constitute the irst argument of the verb. he function of an e-NP (existential NP) is proposed for such elements instead.

1.

Introduction

In this article we examine the category of subject in the Finnish language, and explore ways in which it can be described and deined in relation to other syntactic functions. We examine the subject in relation to a variety of diferent clause types and give detailed analyses of certain problematic constructions. Among the grammatical features examined are case marking, agreement and word order. We consider the semantics of subjects, including their semantic roles, their relation to other sentence elements and their semantic functions as regards the sentence as a whole. he latter perspective is linked in particular to the question of whether Finnish subjects fulill the role of the subject as deined in cognitive grammar, i.e. functioning as a clause-level trajector (primary focal participant; cf. Langacker 2008: 325) and as a starting point (or viewpoint, in the terminology of Fried 2005)

* his research has been partly funded by the Estonian Science Foundation (grant ETF7552, Tuomas Huumo) and Academy of Finland (Marja-Liisa Helasvuo).

doi 10.1075/cal.16.02huu © 2015 John Benjamins Publishing Company

14

Tuomas Huumo and Marja-Liisa Helasvuo

for the construal of the relationship and its participants. he starting-point function of the subject creates a bridge between the levels of semantics and discourse, as this function can be viewed as both a semantic and a discursive feature. One important discourse feature is whether the subject introduces a referent and whether that referent is mentioned again later in the discourse. It is apparent that the syntactic, semantic and discourse properties typical of subjects do not always co-occur, and that the category of the subject is thus rather diferent depending on which properties are given more weight. We propose a model for describing syntactic functions which takes into account these grammatical, semantic, and discursive features of the sentence, and which approaches syntactic functions as part of a particular grammatical construction. he model argues for a relatively narrow deinition of subject. he subject is viewed as the semantic and discursive starting point of the sentence; in Finnish, it functions as a nominative NP and the predicate verb agrees with it in number and person. he subject is (almost always) the semantic irst argument of the verb. he subject can appear in transitive, intransitive, and copulative sentences; in terms of cognitive grammar, it expresses the clause-level trajector. We will show that the subject thus deined shares many discourse features as well: it is typically referential, and its referent is usually a given and referentially continuous; in other words, it is not mentioned just once but is tracked throughout the discourse. here has been some debate on the analysis of existential clauses in Finnish linguistics (for a summary of this debate, see Tiainen 1997). In Finnish, similarly to many other languages (cf. Freeze 2001), existential clauses show syntactic, semantic and discursive features which are peculiar to the construction type. In Finnish linguistics, the debate has centered around the status of the irst argument of the verb in the existential clause. Separate from the subject, we postulate a syntactic category which we entitle the e-NP (‘existential NP’, Helasvuo 1996) – an element that appears either in an existential clause or in certain other construction types (see Section 4). It can be a noun phrase, and it is normally the semantic irst argument of the verb, though it does not trigger agreement, and it assumes the role neither of the starting point nor of the clause-level trajector. It appears either in the nominative or in the partitive case, depending on the one hand on its own features relating to countability and quantiication, on the other on whether the sentence is airmative or negative. With regard to discourse features, the e-NP is typically referential and its referent is normally new. Indeed, the referent is not usually tracked further in the discourse (Helasvuo 2001). Non-referential e-NPs in particular can appear in the nominative case in negative clauses. he typical position of the e-NP is ater the verb.

On the subject of subject in Finnish

We irst discuss the criteria for subject with regard to morphosyntactic, semantic and discourse features. We argue for the e-NP as a syntactic category separate from the subject. We discuss e-NPs not only in relation to canonical subjects, but also with regard to objects and predicate nominals. We then proceed to consider several non-agentive construction types regarding the analysis of the irst arguments of these constructions. We suggest that they represent a conceptualization strategy which, following Langacker (2008), can be described as theme orientation as opposed to the more conventional strategy of agent orientation (for these terms, see Langacker 2008: 370–379). Agent orientation is based on the starting point status of the agent and focuses on the dynamic interaction between the participants. heme orientation, in turn, is compatible with non-dynamic situations such as locative relationships, and involves a starting point with the role of a theme – not an energy source but a passive participant situated in a location or undergoing a change. We argue that constructions with an e-NP oten involve a theme orientation irrespective of the verb they select (which may be agentive).

2. Criteria for subjecthood 2.1

Grammatical criteria

Grammatical criteria for the category of subject have been under dispute in Finnish linguistics. he criteria discussed include case marking, agreement and word order, which are all common criteria in discussions of the category of subject in typological research (see e.g. Dixon et al. 2001: ix).1 Many researchers have drawn attention to the fact that in Finnish the subject is not an obligatory constituent; there are in fact many construction types which are subjectless, including passive constructions and clauses describing weather conditions, such as Sataa ‘It is raining’; cf. Eriksen et al., this volume). his has led to the conclusion that the subject may play a less central syntactic role than in some other languages (cf. e.g. Siro 1964: 16; Shore 1992: 98–104). With regard to case marking, in a transitive clause the object is distinguished from the subject by case marking (accusative or partitive); the subject is in the base form, the nominative, and is thus unmarked for case. Hakulinen and Karlsson (1979: 163–166) mention this as the most important criterion for grammatical subjects. his is typical of any language exhibiting a nominative-accusative alignment system: subjects of transitive and intransitive clauses show identical coding 1. In Finnish, alternations in voice (active vs. passive) do not change the syntactic functions of the main arguments, and thus are not relevant criteria for subjects.

15

16

Tuomas Huumo and Marja-Liisa Helasvuo

that distinguishes them from the object. However, subject and object marking is afected by referential features of the phrases functioning as subjects or objects. hese features relate to their lexicosemantic properties and number. As Helasvuo (2001) shows, personal pronouns show nominative-accusative alignment where the subject is always in the nominative and objects are distinguished from it by either accusative or partitive case marking (consider Example 1). However, if the object is some other pronoun (not a personal pronoun) or a full NP, its marking depends on whether or not the clause contains an overt subject (2). Finally, the case marking of plural NPs does not necessarily distinguish between syntactic roles, since both the subject and the object may be in the nominative (3). Helasvuo concludes that in plural NPs the nominative functions similarly to an absolutive. (Helasvuo 2001: 40–57.) (1) Minä tunne-n häne-t. 1sg know-1sg 3sg-acc ‘I know her/him.’ Pekka tunte-e presidenti-n. m.name.nom know-3sg president-acc ‘Pekka knows the president.’ b. Presidentti tunne-taan hyvin täällä. president.nom know-pass well here ‘he president is well known here.’

(2) a.

(3) Lapse-t sö-i-vät tikkari-t. child-pl.nom eat-pst-3pl lollipop-pl.nom ‘he children ate the lollipops.’

he subject is thus typically in the nominative case. Hakulinen and Karlsson (1979: 164) consider the nominative to be the unmarked case for subject, while Hakulinen (1983) mentions the nominative as the case for the “prototypical subject”. Vilkuna (1996) suggests a deinition of subject that includes three layers. Under the narrowest deinition, only nominative NPs which trigger agreement in the predicate verb can function as subjects. A somewhat broader deinition also includes partitive, ininitival and clausal subjects in the category, as well as genitive subjects which occur with certain ininitival constructions. Finally, under the broadest conception, oblique subjects are also included. hus, in a possessive construction the locative constituent expressing the possessor would be a subject, e.g. Liina-lla on kissa [Liina-ade is cat+nom] ‘Liina has a cat’. he idea of oblique or locative subjects has not gained ground in Finnish linguistics, and a recent comprehensive grammar of Finnish does not even mention it. In contrast, genitive subjects are widely considered as subjects in certain modal constructions expressing necessity (cf. e.g. Hakulinen et al. 2004: 868 and Leino, this volume).

On the subject of subject in Finnish

On a par with case marking, agreement functions to code the subject in Finnish. he predicate verb shows agreement with the subject in person (1st, 2nd or 3rd) and number (singular vs. plural). However, agreement functions as a subject coding process only with nominative subject NPs (4a); genitive subjects do not trigger agreement (4b), nor do ininitival or clausal subjects (on so-called “partitive subjects”, see below Section 3). In other words, case marking and agreement are interrelated. Sinä laula-t 2sg.nom sing-2sg ‘You sing beautifully.’ b. Sinu-n täyty-y 2sg-gen must-3sg ‘You must eat lunch.’

(4) a.

kauniisti. beautifully syö-dä lounas-ta. eat-inf lunch-par

In (4a) the predicate verb shows agreement with the subject, while in (4b) the 2nd person singular genitive subject does not trigger agreement in the predicate verb; instead, the predicate is in the 3rd person singular form, which is the unmarked personal form for verbs used with genitive, ininitival and clausal subjects. In colloquial varieties, the number opposition (singular vs. plural) has been neutralized in the 3rd person, and singular verb forms are used with both singular and plural subjects. Moreover, passive forms are oten used instead of 1st person plural verb forms in connection with 1st person plural pronominal subjects. (For discussion, see Helasvuo 2001: 64–75.) In sum, we can say that agreement functions as a clear subject criterion in connection with 1st and 2nd person singular subjects; in the 3rd person and in plural forms, there is more variation. With regard to word order, Finnish favors SV-order (see e.g. Hakulinen & Karlsson 1979: 164). In a comprehensive study of word order in Finnish, Vilkuna (1989) emphasizes the importance of the preverbal “theme” position, which is usually taken up by nominative subjects (in Vilkuna’s terminology, the “default” theme); however, in certain clause types, such as the existential clause, it may be taken up by other elements (see also Hakulinen et al. 2004: 1306–1312). In their discussion of subject criteria, Hakulinen and Karlsson (1979: 164) state that word order helps to distinguish subjects only in cases where other criteria do not work, for example in connection with plural nominative subjects (see also Palander 1991, based on spoken language data). Several corpus studies lend support to the idea that SV is the basic word order in Finnish (see Hakulinen, Karlsson & Vilkuna 1980: 145 and Huumo 1994 on written language data, Helasvuo 2001: 76–77 on conversational data). Helasvuo (2001: 79–81) points out, however, that the tendency toward the SV order is dependent on the structure of the subject NP and its referential properties. In conversational data SV is predominant, and in clauses with 1st and 2nd person

17

18

Tuomas Huumo and Marja-Liisa Helasvuo

subjects over 96% of the clauses exhibit an SV order; in clauses with subjects which are lexical NPs, on the other hand, the subject precedes the verb much less oten, especially if the referent of the subject is non-human (only 63%). In sum: the grammatical features which have been discussed as criteria for subjecthood in Finnish are case marking, agreement and word order. Many researchers agree that the grammatical criteria for subjects are best fulilled by nominative subjects: they trigger agreement in the predicate verb, and they also appear most oten in the preverbal theme slot. Hakulinen (1983) has called them “prototypical subjects”. We ind it problematic, however, that nominative or partitive NPs in existential clauses are oten also considered to be subjects (see e.g. Hakulinen & Karlsson 1979; Vilkuna 1996; Hakulinen et al. 2004: 868), even though they do not fulill any of the grammatical criteria for subjects: they may be in the partitive case, they do not trigger agreement in the predicate, and they most oten appear postverbally. We therefore suggest that they should not be analyzed as subjects, but rather as representing another syntactic function, the e-NP. We will show that a more narrow deinition of the subject role is supported by both semantic criteria (2.2) and discourse features (2.3). 2.2

Semantic criteria

In this section we take a look at semantic criteria for subjects and discuss possible semantic deinitions for the category. We start with a general crosslinguistic overview of the semantics of subjects, highlighting central viewpoints of Cognitive Grammar and Construction Grammar, followed by a discussion of the semantic criteria for subjects proposed in the literature on Finnish. 2.2.1 Crosslinguistic perspectives on the semantics of subjects Since Fillmore’s seminal paper (1968), it has been customary to characterize the semantic functions of clausal elements by assuming a set of semantic roles their referents may have in the designated situation. he number of roles and the basis for their deinition vary remarkably among authors, and no consensus has been reached as to which particular roles are essential for the grammatical analysis of a language (a recent overview is Levin & Rappaport Hovav 2005: 35–50). In most analyses it is customary not to associate a syntactic function (such as subject or object) with one particular role, since no role can apparently characterize all occurrences of a category. hus the possibility that notional deinitions might be provided for grammatical functions is usually rejected. However, some roles clearly correlate with certain syntactic functions: for instance the role of agent with subject and that of patient with object.

On the subject of subject in Finnish

A radically diferent approach is taken in Cognitive Grammar (e.g. Langacker 1991a: 216, 2008: 363–370) which argues for a notional deinition of grammatical functions. In this approach, all occurrences of a certain category, such as the subject, are assumed to have something in common not only in form but also in meaning. his common meaning, which obviously needs to be quite schematic in order to it all instantiations of the function, serves as a foundation for a notionally-based deinition for the category. In particular, the CG approach argues that it is not the cognitive content of the expression as such that serves as the foundation for such deinitions, but rather the way in which it conceptualizes the situation and its participants. More technically, Langacker (2008: 365) deines the grammatical subject as a clause-level trajector, a notion based on the cognitive ability known as the focusing of attention (see also Talmy 2000, Chapter 5). Similar characterizations for subjects have been given in approaches based on Construction Grammar. For instance, Fried (2005) characterizes the subject as a viewpoint, i.e. “a perspective-based semantic correlate of morphosyntactic prominence, realized as the subject”. In Cognitive Grammar, the clause-level trajector is the primary focal participant in the relationship proiled (indicated) by the clause-level construction. his relationship in turn is an elaborated and grounded instance of the process type indicated by the inite verb. hus for instance the expression is grounded in the speech situation and its participants are speciied by nominal elements. In addition to the grammatical subject there are oten other nominal elements, which serve as secondary focal participants and are called landmarks in this framework. What distinguishes the subject from other nominal elements is thus its function of designating the conceptually most prominent participant in the clause-level relationship. his cognitive prominence is then relected in the grammatical behavior of subjects and in their prominent syntactic position. Unlike classic semantic roles, which are based more directly on the conceptual content of the designated relationship (i.e., what kinds of roles the participants play in the event), the functions of trajector and landmark are based on conceptualization. However, conceptualization is not independent of conceptual content but is typically motivated by it – for instance, animate participants and especially agentive ones tend to be more prominent and thus more natural candidates for subjects than inanimate patients, instruments or locations. Agents initiate events by starting a causal chain that leads to some kind of efect, which is a feature that also enhances their prominence. On the other hand, there are constructions such as passives and impersonals, which in some languages demote the agent to a more peripheral function and promote a less prominent participant to the subject function. Another example of this are so-called setting (or locative) subject constructions, discussed by Langacker (2008: 387–390; see also Fried 2005 and Achard’s paper in this volume). hese select a locative element as their subject and assign

19

20

Tuomas Huumo and Marja-Liisa Helasvuo

the participant actually performing the process a more peripheral role, such as a prepositional phrase (e.g., he garden is swarming with bees) or the object (his stadium has seen some thrilling contests). In sum, the criteria for selecting one particular participant as the subject vary between constructions and even within one and the same construction, especially if the construction is a highly schematic one such as the transitive clause, which allows non-agentive subjects (e.g., He received a letter). According to Langacker (1991a: 223), diferent instances of subjects in fact set up a continuum based on the extent to which their subjecthood is motivated by the conceptual content of the clause. At the less prototypical end of this continuum there are subjects whose selection is based purely on conceptualization. For instance, in a pair such as Caleb resembles George vs. George resembles Caleb the choice of one or the other participant as subject depends on our choice of the participant we wish to compare to the other one. hat other participant is then made the standard of comparison and is indicated by the object. In terms of Cognitive Grammar, the clause-level trajector also serves the function of a starting point, a concept used in CG to refer to the initial element in a natural path – a cognitive ordering of participants in a relationship. In a similar vein, using a Construction Grammar approach, Fried (2005: 484) argues that the subject opens up a perspective on the designated relationship – “the speaker’s assessment of the hierarchical relations among participants in an event […] which leads to a particular construal of that event”. he concept of cognitive ordering is a broad one, covering phenomena extending from information structure (the information status of diferent elements) to semantically based qualities such as the position of an element in a causal chain (for this term, see also Crot 1991). Cognitive ordering may but does not have to coincide with the actual word order of the elements in a clause. More technically, Langacker (1991b: 382–383) deines the starting point as “the initial element encountered when a complex structure is accessed via some cognitively natural ordering of the participants on a natural path”. As noted above, one kind of a natural path is the given vs. new distinction, which divides elements on the basis of their information status. It is especially in this respect that the CG concept of the starting point coalesces with the use of this term by Chafe (1994: 82–83). 2.2.2 he semantics of subjects in Finnish In studies on Finnish, the starting point status of diferent kinds of subjects has been discussed by Huumo (1997, 2003), who argues that only canonical subjects function as starting points, whereas e-NPs (“existential subjects” in the terminology of Huumo 1997 and 2003) do not fulill this function. his is demonstrated by a study of minimal pairs in which one member has a canonical subject (nominative case, verb agreement), the other an e-NP (partitive case, no verb agreement; cf.

On the subject of subject in Finnish

Examples 5 and 6 below). he lexical content of the clauses in the pairs is identical. It turns out that only canonical subjects, not e-NPs, provide a starting point for the relationships indicated by other clausal elements. For instance, the spatial extent (of paths) and the temporal duration of events are construed from the point of view of the clausal trajector (indicated by the canonical subject) in such a way that the trajector is the participant traversing the path or experiencing the duration. he e-NP fails to provide a starting point for such construals, and the spatiotemporal extent of the event is therefore construed independently of it, oten by assuming an external point of view. his shows that the e-NP indicates a “downstream” participant on a natural path, a referent that is accessed only ater the mental space constituting the relevant dimension (e.g., a spatial path or a temporal duration) has already been established. his is why an e-NP oten evokes a reading where its referents (especially in the plural) are scattered along the path and not in motion along it. Even when the verb indicates motion, that motion is not the basis for the construal of the path. he reference of the e-NP may also change during the event. he following pair of examples illustrates this diference. In (5), the canonical nominative subject designates a mover that traverses the path indicated by the directional adverbials (‘from the church to the station’), and the path is construed from the point of view of this motion. In (6), the partitive e-NP also indicates a mover, but the path is now construed independently of the motion: the movers are merely scattered along the path, and the direction of the path may not coincide with that of their motion – in (6) the path between the two landmarks (the church and the station) is occupied by runners along its full length. (5) Kirko-lta asema-lle asti maratoonari-t juoks-i-vat church-abl station-all ptcl marathon.runner-pl run-pst-3pl tien-varre-ssa. road-side-ine ‘From the church to the station the marathon runners were running alongside the road.’ (6) Kirko-lta asema-lle asti maratoonare-i-ta juoks-i church-abl station-all ptcl marathon.runner-pl-par run-pst.3sg tien-varre-ssa. road-side-ine ‘From the church to the station there were marathon runners running alongside the road.’

hese and similar pairs studied by Huumo (2003) suggest that the e-NP is not a starting point for the construal of the event. his of course invites the question whether the e-NP is a clause-level trajector (the CG deinition for subjects) at all, or whether it serves a less prominent function.

21

22

Tuomas Huumo and Marja-Liisa Helasvuo

In sum: the semantic factors considered in this section do support our grammatical analysis (Section 2.1), where we argued that there is no clear subject candidate in the Finnish existential construction. While the e-NP indicates the trajector of the verbal process, it fails to indicate the trajector of the whole clause. As our comparison with impersonals, passives, and setting/locative subject constructions shows, this is not at all linguistically exceptional. For comparison, consider the Czech alternations of case discussed by Fried (2005: 491–495) in her study on locative expressions indicating the perceptibility of a stimulus in a location, as in ‘he kitchen smelled of cinnamon’. In Czech it is possible to code alternatively the stimulus (‘cinnamon’) or the setting (‘kitchen’) as the nominative subject, or to use a third strategy that lacks a nominative subject altogether. In this third strategy, the setting expression occurs in the locative case and the stimulus in the instrumental case. Fried argues that in such expressions “there is inherently no commitment to either participant as the more prominent one”. On the basis of what we said above about Finnish existentials, it seems that such a characterization might be appropriate to their meaning as well. In addition to the e-NP, the other potential subject candidate in the Finnish existential construction is the locative element, which typically occupies the initial position (see Vilkuna 1996: 157). It has some starting-point-like features, and it has oten been pointed out in the literature (since Ikola 1954) that the Finnish existential construction is “a predication about a location” indicated by the locative element. However, as argued in Section 2.1, in grammatical terms the initial locative is an even more problematic subject candidate than the e-NP. It is also semantically clear that the initial locative is not the verbal trajector – the Finnish existential construction allows a wide range of verbs including agentive ones, and in the processes they designate, it is always the e-NP that designates the verbal trajector. he semantic function of the locative case in the clause-initial locative is also quite productive and transparent. In CG terms it indicates a relationship where the case-inlected NP is a landmark, making it unnatural to analyze the locative element (together with its local case ending) as the trajector for the whole clause. he potential subject properties are apparently divided between the two NPs: the locative is more topical and starting-point-like than the e-NP, which in turn indicates the verbal trajector. Morphology and semantics, however, do not support the analysis of the locative element as a subject. 2.3

Discourse criteria

In the typological literature, prototypical subjects are oten characterized by agentivity and topicality (see e.g. Comrie 1981: 106–107) and referential continuity (see Givón 1983). Furthermore, subjects have been said to carry a light information

On the subject of subject in Finnish

load: they most typically refer to referents which are either given or accessible in discourse (Chafe 1994: 85–88; see also above on starting points; cf. Langacker 1991b: 382–383). In the literature on subjects in Finnish, discourse features have received much less attention than grammatical and semantic criteria. Topicality has been mentioned as a subject feature by both Hakulinen (1983) and Vilkuna (1996). Hakulinen (1983: 244) mentions topicality among the characteristics of what she calls “prototypical subjects”; existential “subjects”, that is, e-NPs in our terminology, lack this property. Vilkuna (1996: 82) notes that in addition to formal factors (grammatical form, syntactic features), semantic role (more speciically agentivity) and topicality also have an efect on which element becomes the irst argument of the verb, and thus a subject. Vilkuna further mentions that agentivity and topicality are closely interrelated. Hakulinen (1983) and Vilkuna (1996) do not base their views on an analysis of actual discourse, or on a study of topicality in a speciic data set, but rather on their own intuitions. Referential continuity has been studied by Helasvuo (2001) and Tiainen (1998). Helasvuo (2001) has analyzed referent tracking in conversational data, and concludes that referent tracking is related to syntactic function: arguments of the clause core, subjects and objects, have referents which are usually mentioned several times, that is, they are tracked in the discourse, whereas referents of NPs in other syntactic roles are less likely to be tracked (Helasvuo 2001: 92). Referential continuity is thus a characteristic feature of NPs functioning as subjects, but is fairly common in objects as well. Tiainen (1998) has studied referential continuity in third person subjects in terms of referential chains, which she deines as consisting of at least three co-referential NPs, at least two of which appear in the preverbal theme slot and at least one of which occupies the syntactic role of subject (see Tiainen 1998: 507). Tiainen shows that referential chains tracking animate referents consist of NPs functioning as (i) transitive subjects, (ii) intransitive subjects which are agentive and (iii) NPs expressing the possessor in possessive constructions. In contrast, chains which track inanimate referents usually consist of (i) subjects of predicate nominal clauses, (ii) intransitive subjects which are non-agentive and (iii) objects of transitives. (Tiainen 1998: 519.) Referential continuity is thus closely interrelated with the semantics of the referents of the NPs functioning as subjects. he most important semantic feature afecting referential continuity and tracking appears to be animacy. he relationship between information load and subject role in Finnish has been studied by Huumo (1994). On the basis of an analysis of ictional texts, Huumo notes that referents of subjects have either been mentioned earlier in the discourse or are inferrable. Subjects referring to referents which are new form a minority (15%; Huumo 1994: 195). Helasvuo (2001: 90) has studied the distribution of new referents across syntactic roles in conversational discourse. Her

23

24

Tuomas Huumo and Marja-Liisa Helasvuo

indings regarding the subjects of transitives, intransitives and copular clauses are very similar to those of Huumo (1994): only 16% referred to referents which were new to the discourse. Unlike Huumo (1994), however, she looked at e-NPs separately and found that over 75% of e-NPs referred to new referents. Helasvuo (2001: 1001–101) notes that with respect to discourse features e-NPs show a proile which is quite diferent from that of subjects: e-NPs usually refer to referents which are new to the discourse. hese referents are usually not tracked in discourse; in other words, the referents introduced by e-NPs are usually not referred to in subsequent discourse. In the existential construction, the sentence-initial locative element could be considered another subject candidate on a par with the e-NP. We noted in regard to semantic features (cf. 2.2 above) that the locative element shares some features which are typical of semantic starting points.With regard to discourse features, we may add that in the existential construction sentence-initial locative elements typically refer to referents which are given in the discourse and which are further tracked in the discourse (Helasvuo 1996: 350; see also above on referential chains, Tiainen 1998). However, they lack the grammatical features related to subjects (see 2.1 above). When e-NPs are excluded from the category of subject, discourse features yield a fairly consistent characterization of the subject role. Subjects thus deined usually refer to referents which have been previously mentioned or which are inferrable from context, and their referents are mentioned again and tracked in subsequent discourse. In contrast, e-NPs usually refer to referents which are new to the discourse and which do not become topics, but rather are mentioned just once.

3. E-NPs: Between subjects, objects and predicate nominals he intermediate status of e-NPs between subjects, objects and predicate nominals has oten been discussed in the literature: although the e-NP has most oten been classiied as a subject, scholars have pointed out that it shares many features with objects (Wiik 1974; Itkonen 1976) and predicate nominals (Toivainen 1986). In fact, it is the two last-mentioned categories that the e-NP resembles morphologically, in that all three display a nominative vs. partitive alternation in their case marking. In object marking the opposition distinguishes the partitive object from the so-called total object which is a semantically uniform but morphologically heterogeneous category. he total object indicates aspectual and quantitative boundedness, and it may be morphologically marked with the nominative or the accusative, depending on syntactic factors (for details see Heinämäki 1983, 1994; Helasvuo 2001 & Huumo 2005, 2010). Very much like the partitive e-NP, the partitive object may represent its referent as constituting an unbounded quantity (of a substance

On the subject of subject in Finnish

or of entities; Example 7), but it also has the function of indicating unbounded (atelic or progressive) aspect (8). In many examples the partitive object may be understood as indicating both aspect and quantity (9). (7) Löys-i-n raha-a. Find-pst-1sg money-par ‘I found [some] money.’ (8) Katsel-i-n maalaus-ta. Look-pst-1sg painting-par ‘I looked / was looking at a / the painting.’ (9) Sö-i-n puuro-a. Eat-pst-1sg porridge-par ‘I ate [some] porridge’ / ‘I was eating porridge.’

In (7) the partitive must indicate pure quantity (of the money), since the event of ‘inding’ is punctual and therefore aspectually bounded. In (8) the partitive indicates unbounded aspect but not quantity, since the object is a count noun and thus quantitatively bounded. he general function of the object-marking partitive is thus the indication of unboundedness, which can be either nominal (quantity) or verbal (aspect). Sands and Campbell (2001: 266–269) argue that the partitive case relects reduced transitivity both when used on objects and on e-NPs (or in their terminology, subjects). What distinguishes the partitive object from the e-NP is its purely aspectual function (in 8), which is independent of nominal quantity. Like e-NPs, object NPs also take the partitive marking when under negation (10). (10) E-n huoman-nut sinu-a. neg-1sg notice-ptcp you-par ‘I didn’t notice you.’

he system of case marking of the predicate nominal (in most cases a predicate adjective) is, at least apparently, the simplest of the three: it relects only the nature of the subject as count vs. mass noun, but not aspect or negation (Sadeniemi 1950: 46–47; Itkonen 1976; for a detailed discussion in English, see Denison 1957; Huumo 2009). If the subject is (conceived of as) a count noun, i.e. as indicating an indivisible entity, the predicate adjective takes the nominative case marking (11). If the subject is a mass noun or a plural form, indicating a substance or a multiplicity of referents, then the predicate adjective takes the partitive case (12–13). As Example (13) shows, the predicate adjective also agrees with the subject in number. (11) Pöytä on pyöreä. Table.nom be.prs.3sg round.nom ‘he table is round.’

25

26

Tuomas Huumo and Marja-Liisa Helasvuo

(12) Kahvi on musta-a. Cofee.nom be.prs.3sg black-par ‘(he) cofee is black.’ (13) Pöydä-t o-vat pyöre-i-tä. Table-pl.nom be-prs.3pl round-pl-par ‘(he) tables are round.’

he case alternation of the predicate nominal thus resembles the nominal (quantitative) functions of the case alternation in the e-NP and the object, but it must be emphasized that it is not entirely similar. he partitive marking of the e-NP and the object indicate the unbounded quantity of their referents, but the partitive predicate nominal merely relects the fact that the subject is understood as a mass noun, not that its referent is quantitatively unbounded – in fact, this is usually not the case in copulative constructions, since their subject is in the nominative, which in mass nouns and plural forms correlates with a bounded quantiication (for exceptions, however, see Huumo 2009, 2010). To sum up: we can see that there are both factors more directly related to the phrasal meaning of the NP which carries the case ending (quantity) and factors related to the clause-level meaning, such as aspect and negation. Object marking seems to be the most complex system, since it involves three factors (quantity, aspect and negation); the marking of the predicate nominal appears to be the simplest one, since it only involves the conceptualization of the subject as a mass noun or a count noun. he e-NP occupies an intermediate position between the two; it lacks the purely aspectual function of the object (though it can sometimes trigger an iterative or incremental reading and afect aspect more indirectly; see Huumo 2003 for discussion) but nevertheless is sensitive to negation.

4. Non-agentive constructions with theme orientation In Section 2 we presented a detailed analysis of the grammatical, semantic and discourse features of subjects and contrasted them with those of the e-NPs, suggesting that the latter are not subjects. he only subject-like feature in e-NPs is their function as verbal trajectors, or irst arguments of verbs; as Langacker’s (2008) analysis of passive, impersonal and locative subject constructions shows, however, not all verbal trajectors end up as clause-level trajectors. All the other factors considered in Section 2 speak against the analysis of e-NPs as subjects. If we nevertheless wish to analyze them as subjects (which is the received view in Finnish linguistics), it becomes very diicult to assign any formal, semantic or discourse-related criteria to the overall subject category – indeed, most criteria

On the subject of subject in Finnish

given for Finnish subjects apply only to canonical instances but not to e-NPs. A more coherent picture arises if subjects are given a more narrow deinition which excludes e-NPs. In this section we discuss certain borderline instances which have traditionally been problematic for the Finnish subject category. hese include constructions with an NP that resembles, at least in some relevant aspects, the e-NP, but is even less subject-like than the typical e-NP. A characteristic feature of such NPs is that they are oten non-referential and their semantic function seems to be closer to predication than to reference. he comprehensive grammar (Hakulinen et al. 2004) analyzes such NPs as predicate nominals, but we argue that this analysis is problematic as well. hough such NPs deviate from typical e-NPs in some respects, they nevertheless resemble e-NPs more than predicate nominals in their grammatical behavior. In our view, the problem follows from the analysis whereby e-NPs are subjects: the NPs under discussion are even less subject-like than typical e-NPs. In the following discussion we show that it is hard to draw a strict line between e-NPs and such alleged predicate nominals, making it diicult to treat them as separate syntactic categories. heir classiication as predicate nominals is also problematic since they difer from typical predicate nominals even more that they difer from typical e-NPs. Our approach, whereby e-NPs are not subjects, ofers a solution to the problem: when e-NPs are not categorized as subjects at all, there is no need for such problematic NPs to meet subject criteria either, and thus they can be categorized as e-NPs. he categories to be analyzed include certain idiomatic construction types, which are relatively ixed in form and carry a very speciic construction-level meaning; in Section 4.3, we characterize this meaning as theme orientation in the sense of Langacker (2008). Such features distinguish them from more schematic clause types such as transitive and intransitive clauses, and even from existential clauses, which themselves are more idiomatic than the most schematic clause types. It has been noted in recent studies (Hakulinen et al. 2004: 849) that it is more natural to approach such constructions at a holistic level than to analyze them as combinations of the components they consist of. hus a constructionbased approach, such as that of Construction Grammar, seems fruitful for the analysis of such expressions. Here we discuss three constructions that ofer new insights for the general classiication of e-NPs. he irst is the setting subject construction (SSC; Section 4.1). Finnish SSCs have an SVO structure, and their subject indicates a setting or location occupied by the entity that is indicated by the grammatical object. In Section 4.2 we discuss two constructions which feature a nonreferential clause-inal NP. hese constructions have traditionally been known as state clauses and result clauses. State clauses typically indicate natural conditions that prevail in

27

28

Tuomas Huumo and Marja-Liisa Helasvuo

a location, such as weather or temperature, sometimes also the psychophysiological state of an animate undergoer (e.g., hunger, thirst, cold, heat). he category of state clauses, as deined in the literature, is heterogeneous, and includes microconstructions that have a common schematic meaning but only few if any morphosyntactic features binding them together. Result clauses indicate a change in the internal state, role, or occupation of a topical, clause-initial participant which is marked with a locative case. he phrase indicating the result of such a change occupies clause-inal position and resembles the e-NP in many ways. his element turns out to be particularly relevant for the present study, as it shows how a category can blend not only the meanings but also the formal markings of two diferent grammatical functions (an e-NP and a predicate nominal). 4.1

Setting subject constructions

Finnish setting subject constructions resemble the English ones discussed in 2.2.1, in that they select the expression of a setting or a location as their subject (see also Achard’s analysis of abstract setting subjects in this volume). In contrast to English SSCs, the entity occupying the setting is not indicated by an oblique element but by a grammatical object, which is marked with the partitive case. Consider Examples (14), (15) and (16): (14) Puutarha kuhis-i mehiläis-i-ä. garden.nom swarm-pst.3sg bee-pl-par ‘he garden was swarming with bees.’ (15) Pello-t tulv-i-vat vet-tä. ield-pl.nom lood-pst-3pl water-par ‘he ields were looded with water.’ (16) Sinä vuoda-t ver-ta. you leak-prs.2sg blood-par ‘You are bleeding.’

In addition to case marking, the subjecthood of the initial setting element is corroborated by verb agreement in number (in (15) the plural subject triggers plural agreement in the verb) and person (in (16) the subject is a 2nd person pronoun and the verb takes the 2nd person form). To see the intermediate status of setting subject constructions between existentials and other clause types, consider the following set of weather-expression examples (cf. also Eriksen et al., this volume) that share (roughly) the same semantic content but have a quite diferent morphosyntactic structure:

On the subject of subject in Finnish

(17) Sata-a lun-ta. precipitate-prs.3sg snow-par ‘It is snowing.’ (18) Pilve-stä sata-a lun-ta. cloud-ela precipitate-prs.3sg snow-par ‘he cloud is snowing’ (lit. ‘From the cloud precipitates snow’) (19) Pilvi sata-a lun-ta. cloud precipitate-prs.3sg snow-par ‘he cloud is snowing’ (lit. ‘he cloud is precipitating snow’) (20) Ensilumi sata-a lääni-n itäos-i-in irst.snow precipitate-3sg county-gen east.part-pl-ill tavallisesti lokakuun 15–20. päivän hujakoilla. usually October-gen day-gen around ‘he irst snow usually falls around October 15–20 in the eastern parts of the county.’ (Kolehmainen 2010: 18) (21) Sato-i lume-n. precipitate-pst.3sg snow-acc ‘It snowed (a cover of snow).’

Examples (17) and (18) resemble existentials in form. Especially Example (18) is a typical existential clause with an initial locative element, which in this case indicates the source of the motion of the snow (sources and goals, in addition to static locations, are commonly expressed by the initial locatives in Finnish existentials). he inal NP lunta ‘snow’ looks like an e-NP in both examples. Example (19) is a setting subject construction, formally a transitive clause, but it resembles the existential (18) in its information structure and meaning. he presence of the nominative subject, however, makes it clear that grammatically the clause-inal NP must be analyzed as an object. Example (20) shows that the verb sataa ‘precipitate’ even allows for a nominative subject (the example comes from corpus data analyzed by Kolehmainen 2010). Example (21), like (17), lacks the initial locative but has its inal NP in the accusative case, indicating telicity of the event by its case marking (‘it snowed a layer of snow’). Since accusative marking is not possible in e-NPs, the clause-inal NP in (21) needs to be analyzed as an object. his being so, one might ask whether the partitive e-NP in Example (17) might also be an object, the progressive/atelic counterpart of the accusative in (21) (‘it was snowing’; cf. Section 3 on object marking). Indeed, since the partitive marks both objects and e-NPs and since there is such a close resemblance between Examples (21) and (17), this alternative analysis cannot be dismissed outright. he problem, of course, is that the inal NP in (17) looks like a typical e-NP – are all e-NPs objects, then?

29

30

Tuomas Huumo and Marja-Liisa Helasvuo

What this group of examples shows is how construction types intertwine and how diicult it is to draw boundaries between categories. In terms of meaning, the functions of the e-NP in Example (18) and the object in Example (21) are very similar if not identical. It is only the overall construction that distinguishes the two elements; as shown by Examples (17) and (21), however, even in formal terms the overall constructions are oten diicult to keep apart. Setting-subject constructions such as (19) are relevant for the overall analysis of e-NPs at a more general level as well; recall Langacker’s (1991a, 2008) analyses of English setting subject constructions, briely discussed in Section 2.2. In Langacker’s characterization of the meaning of SSCs, the selection of the locative as the subject relects a conceptualization whereby the relationship between the location and its content is given a prominent position in the clause; the meaning is, approximately, ‘be the setting for crawling / swarming’. Langacker’s analysis of setting subject constructions is closely compatible with our analysis of Finnish existentials in Section 3. It is based on the observation that there are constructions where the verbal trajector does not end up as the clause-level trajector or as the grammatical subject of the clause. In such expressions the verbal trajector takes on another function that is available in the construction: in English locative subject constructions it becomes a prepositional phrase, in Finnish locative subject constructions the object. 4.2

Referentiality as a criterion for a grammatical category: Between e-NPs and predicate nominals

In addition to setting subject constructions, there are a few other construction types that resemble existentials formally and/or semantically. hese constructions include NPs which resemble e-NPs grammatically but difer from typical e-NPs in being nonreferential. Unlike typical e-NPs, they do not introduce new referents into locations. Space does not permit a thorough discussion, but to shed light on the problem we briely discuss two relevant constructions. In the literature, these are referred to as state clauses and result clauses. 4.2.1 State clauses he term state clause (SC), as used in the comprehensive grammar of Hakulinen et al. (2004), is actually a cover term for a heterogeneous group of constructions that resemble each other to a lesser or greater extent semantically and grammatically. Some subtypes of the SC resemble existential constructions and include an NP that resembles the e-NP (Example 22 below). Another subtype consists of a mere verb (23), yet another of a verb accompanied by a locative (24):

On the subject of subject in Finnish

(22) Helsingi-ssä on lämmin-tä. Helsinki-ine be.prs.3sg warm-par ‘It is warm in Helsinki.’ (23) Sata-a. rain-prs.3sg ‘It is raining.’ (24) Mere-llä myrskyä-ä. sea-ade storm-prs.3sg ‘It is storming at sea.’

What unites such diferent subtypes of SCs is thus meaning rather than form, which is quite variable – except for the fact that they all lack a grammatical subject. hus an approach that emphasizes the formal unity of putative constructions might actually ind the category of the SC too incoherent to be useful. From the point of view of existentials and e-NPs it makes sense to concentrate on those SCs that resemble the existential construction and have a possible e-NP candidate. In addition to expressions of weather and other natural circumstances, there are SCs that express physical or psychological states. In such expressions the clause-initial locative introduces an animate (typically human) experiencer, while the psychological or physiological state of this experiencer is indicated by a clause-inal NP. his inal NP resembles the e-NP and is usually in the nominative (Example 25), sometimes in the partitive (26), while the initial NP that indicates the experiencer takes the adessive case. he Finnish adessive (basic spatial meaning ‘at/on’) is a productive means for the indication of many kinds of relationships with an animate reference-point (for this term, see Langacker 1993); it also indicates the possessor in canonical possessive constructions (27). (25) Opettaja-lla on kylmä. teacher-ade be.prs.3sg cold ‘he teacher is cold.’ (26) Opettaja-lla on tylsä-ä. teacher-ade be.prs.3sg boring-par ‘he teacher is bored.’ (27) (Canonical possessive) Opettaja-lla on kirja ~ kirjo-j-a. teacher-ade be.prs.3sg book.nom ~ book-pl-par ‘he teacher has a book ~ [some] books.’

Finnish grammars (e.g. Hakulinen and Karlsson 1979) oten classify possessive constructions as a subtype of existentials. Indeed, they conform to the constructional scheme of existentials, the main diference being that the clause-initial locative

31

32

Tuomas Huumo and Marja-Liisa Helasvuo

indicates a possessor. In contrast, the comprehensive grammar of Hakulinen et al. (2004) keeps state clauses and possessive–existential constructions apart as distinct clause types, in spite of SCs that formally resemble possessives and existentials. his is also relected in the solution of Hakulinen et al. (2004), classifying the inal NP in state clauses such as (25)–(26) as a predicate nominal rather than an e-NP (an existential subject in their terminology). However, Hakulinen et al. (2004) also argue that state clauses are a “subtype of existentials”, which seems to contradict their diferent analyses for the inal NPs of these constructions (i.e. existential subject vs. predicate nominal). his solution seems to be at least partly motivated by the analysis of e-NPs as subjects: the clause-inal NPs of state clauses are even less subject-like than typical e-NPs. In fact, they are oten adjectival phrases (as in 25), and mostly non-referential. hey resemble predicate nominals in their semantic function, which is the characterization of the location (or experiencer, in psychophysiological expressions) indicated by the clause-initial locative. In semantic terms, it can be argued that they predicate rather than refer. A morphosyntactic feature that associates these elements with predicate nominals is the fact that they do not take the partitive under negation; consider (25′): (25′) Opettaja-lla ei ole kylmä. teacher-ade neg.3sg be.conneg cold.nom ‘he teacher is not cold.’

If the inal phrase kylmä in (25′) were a typical e-NP, it would take the partitive under negation. However, it maintains its nominative case, which is a feature typical of predicate nominals (see also Section 3). his kind of behavior is especially typical of adjectival phrases, but some NPs, especially non-referential ones, also share it; consider (28): (28) Huomen-na ei ole hyvä hiihto-keli. Tomorrow-ess neg.3sg be.conneg good skiing-conditions.nom ‘he skiing won’t be good tomorrow.’

Like a predicate nominal, the inal NP in (25) and (28) can be argued to predicate a quality of the entity indicated by the clause-initial locative. However, both their morphosyntax and the overall constructions in which they are used make them resemble e-NPs more than predicate nominals. It might also be pointed out that if boundaries between clause types are established on the basis of such relatively vague concepts as referentiality, it becomes extremely diicult to draw boundaries between them. Another problem is that some clause-inal NPs allow both the nominative and the partitive under negation, in which case it would be altogether impossible to decide whether the corresponding airmative version is an existential or a state clause.

On the subject of subject in Finnish

In our view the inal NPs of state clauses should not be analyzed as predicate nominals, for both formal and semantic reasons. As already noted, the inal NPs of state clauses resemble e-NPs more than predicate nominals. Another important factor that distinguishes them from typical predicate nominals is that their case marking does not relect the nature of their presumptive correlate in the same way it does in predicate nominal (discussed in Section 3). Consider the following set of examples, where (29) and (30) are canonical copulative constructions with predicate nominals while (31) and (32) are state clauses. In (29) and (30) the count vs. mass status of the subject determines the case marking of the predicate nominal, while (31) and (32) allow both variants irrespective of the count/mass status of the initial locative: (29) Talo on lämmin ~ *lämmin-tä house be.prs.3sg warm.nom ~ *par ‘he house is warm.’ (30) Muta on lämmin-tä ~ *lämmin mud be.prs.3sg warm-par ~ *nom ‘[he] mud is warm.’ (31) Talo-ssa on lämmin ~ lämmin-tä house-ine be.prs.3sg warm.nom ~ par ‘It is warm in the house.’ (32) Muda-ssa on lämmin ~ lämmin-tä. mud-ine be.prs.3sg warm.nom ~ par ‘It is warm in the mud.’

his relative autonomy of the clause-inal NPs of state clauses is also relected in the looseness of the semantic connection between them and the initial locatives (as opposed to that between the subject and the predicate nominal in copulative clauses). In copulative clauses, the predicate nominal indicates a quality of the subject, which thus functions, in traditional terms, as its correlate. In contrast, the inal NPs of state clauses like (22), (25) and (26) introduce a circumstance that (currently) prevails in a location – they are not purely predicative but perhaps better characterized as semi-referential. For instance, the copulative clause (29) can be understood as predicating a permanent quality of the house (e.g., it is warm because it is well insulated) while (31) can indicate the current temperature inside the house. In a sense, the general meaning of (31) resembles that of an existential: it selects a location as the starting point and introduces the content of the location as new information. What distinguishes it from typical existentials is that the content is not an entity but a circumstance.

33

34

Tuomas Huumo and Marja-Liisa Helasvuo

In sum, our brief discussion of state clauses shows that their inal NPs share semantic and morphosyntactic features with both typical e-NPs and typical predicate nominals. As such this is not surprising, keeping in mind the continuum-like nature of linguistic categories. However, we have argued that the inal NP of state clauses resembles the e-NP more than the predicate nominal, and that it would thus be natural to analyze it as an instance of the former rather than the latter category. First, the overall construction in which it occurs is formally similar to the existential clause but very diferent from the copulative clause. Second, the clausal position and case marking of the problematic NPs mostly follow the rules of the e-NP (with the exception of the partitive of negation). hird, this NP does not have a similar relation to its putative correlate as the predicate nominal does. In our view, there is thus no compelling evidence for excluding such elements from the category of the e-NP. 4.2.2 Result clauses he category of result clauses is another construction type that resembles existential clauses but difers from their prototypical instances both semantically and morphosyntactically. A result clause predicates a quality acquired by an entity that undergoes a change. Its general schematic meaning is ‘X becomes Y’. he undergoer of the change is indicated by the clause-initial locative NP that carries the elative ‘from [inside]’ case, which also relects the dynamic conceptualization of the change – the elative can be understood as indicating the initial stage of the undergoer of the change as a metaphoric source location. By using this kind of coding, these expressions can be argued to blend components of a spatial motion event (the locative coding of the undergoer) with an internal change experienced by the undergoer, very much in the spirit of classic localist analyses (Anderson 1971). he quality or role acquired by the undergoer is indicated by the clause-inal element (an AP or an NP) that grammatically resembles the e-NP. Examples are provided below. (33) Liisa-sta tul-i opettaja. name-ela become-pst.3sg teacher ‘Liisa became a teacher.’ (34) Kahvi-sta tul-i vahva-a. cofee-ela become-pst.3sg strong-par ‘he cofee became / turned out strong.’

Both examples predicate an incipient quality of the undergoer introduced by the elative NP. he construction is idiomatic and is strongly associated with the verb tulla ‘become’ (as well as ‘come’), although it also allows other verbs of change as in (35) and (36):

On the subject of subject in Finnish

(35) Pennu-sta kasvo-i iso koira. puppy-ela grow-pst.3sg big dog ‘he puppy grew into a big dog.’ (36) Touka-sta kehitty-y perhonen. caterpillar-ela develop-prs.3sg butterly ‘he caterpillar develops into a butterly.’

he inal NP of a result clause resembles the e-NP in that its case marking varies between the nominative and the partitive. However, it also resembles the predicate nominal, in the sense that case selection depends on the count vs. mass status of the elative-marked undergoer. For instance, the inal element in (34) is in the partitive because the initial elative refers to a substance (‘cofee’). he partitive is also used with plural undergoers; see (37). (37) Pennu-i-sta kasvo-i iso-j-a koir-i-a. puppy-pl-ela grow-pst.3sg big-pl-par dog-pl-par ‘he puppies grew into big dogs.’

On the other hand, again, under negation the inal element of result clauses takes the partitive, which is a feature shared by e-NPs but not by predicate nominals: (38) Pennu-sta ei kasva-nut iso-a koira-a. puppy-ela neg.3sg grow-ptcp big-par dog-par ‘he puppy did not grow into a big dog.’

With regard to case marking, the inal phrase of result clauses thus seems to be a genuine blend of e-NPs and predicate nominals: it partly follows the case marking rules of both elements. In our view, the analysis of Hakulinen et al. (2004), where such phrases are classiied as predicate nominals, makes more sense in result clauses than in state clauses. In semantic terms as well, the inal phrases of result clauses have a clear correlate, the elative-marked undergoer, of which they predicate an incipient quality. At the same time, however, the overall constructions of result clauses are reminiscent of existentials, and the classiication of their inal NPs as non-referential, predicating e-NPs would therefore be an attractive alternative. What makes it problematic in classic analyses is again the basic assumption that e-NPs are subjects. In our analysis where e-NPs are not subjects, such elements can be naturally classiied as a subtype of this category, albeit an idiosyncratic one with some features typical of predicate nominals.

35

36

Tuomas Huumo and Marja-Liisa Helasvuo

4.3

Existential constructions and theme orientation

In the semantics of Finnish existentials we can see features of what Langacker (2008: 370–379) calls theme orientation.2 heme orientation is a conceptualization strategy that difers from the more canonical strategy called agent orientation, which is based on observation of an agent and a force-dynamic interaction initiated by it. In contrast, theme orientation selects as its starting point a participant with the role of a theme, which Langacker uses as a common term for nonagentive roles such as patients, movers, experiencers and zeroes. Such participants are not initiators of events but either function as undergoers or are not afected by events. he concept of the theme role is based on non-dynamic relationships, such as having a quality or being situated in a location. Despite their non-agentive nature, themes, according to Langacker, are participants that can naturally constitute focal, prominent participants of situations and thus act as starting points for their construal. It is easy to recognize features of theme orientation in Finnish existentials. he construction as a whole can be considered as assigning a theme-like role to the verbal trajector, which then becomes the e-NP. his is relected in the observation, oten repeated in the literature, that the existential clause suppresses the active role of the referent of the e-NP. he e-NP typically indicates an entity that occupies a location, and is introduced as a discourse-new content of the location. It inherits the theme role from the overall construction, and the construction backgrounds its role as a verbal trajector. his happens even if its role as the verbal trajector is agentive – the Finnish existential construction allows highly agentive verbs (e.g., ‘run’, ‘study’, ‘ight’); in actual usage, however, such occurrences are rare, as the construction clusters around typical existential verbs such as ‘be/exist’, ‘appear’, ‘come’ and so on. A deviation from Langacker’s theme orientation is the fact that – if one accepts our analysis above – the element with the theme role, the e-NP, is not a subject. Our proposal can be compared to Langacker’s analysis of the English existential there construction (e.g., here was a lion in the clearing). Langacker (1991: 352 and 2008: 496) analyzes the initial there element as the grammatical subject. He bases this analysis on the behavior of the element in grammatical modiications (e.g., the fact that it follows the irst auxiliary in questions: Was there a lion in the clearance?) and on its semantic function of establishing an abstract starting point, a “locus of existence” (cf. also Achard’s paper in this volume). hus analyzed, the English there construction represents another instance of a construction where the verbal trajector (the lion in the example) does not carry the function of a clause-level trajector (= subject) but that of a landmark (see Langacker 1991: 352). 2. We thank Tiina Onikki-Rantajääskö for this observation.

On the subject of subject in Finnish

According to Langacker (1991: 354), the efect of placing there in subject position is “to background the participant [whose existence is being indicated] without foregrounding anything else of real substance”. As we argued in Section 2.2, there are good reasons not to analyze the initial locative of Finnish existentials as a clause-level trajector. However, what Langacker says about the NP that follows the verb in the English there construction seems to apply to the Finnish e-NP as well: it is an element backgrounded by an overall construction in which no other participant is foregrounded. his results in a construction that introduces a discourse-new participant while observing the usual pattern of beginning a clause with an element that is given, i.e., accessible from the current discourse space (cf. Langacker 1991: 354). Summing up, and considering the similarities between Finnish e-NPs and other rhematic clausal elements (predicate nominals and objects, especially the object of the Locative subject construction) which in Cognitive Grammar are called landmarks, there are good reasons to classify the e-NP as a landmark in the Finnish existential construction as well.

5. Conclusions We have argued for a fairly narrow deinition of the subject role. We have discussed subjects with respect to several construction types, most notably to existential clauses. We have suggested that the irst argument of the existential clause should not be considered a subject, but rather as representing another syntactic role, that of the e-NP. We have discussed e-NPs and their relations to canonical subjects, predicate nominals and objects. Here we summarize the discussion. We deine the subject role as the semantic and discursive starting point of the clause. In Finnish, the subject appears in the nominative case and triggers agreement in the predicate verb. It usually appears in the preverbal theme slot. Semantically, the subject is the irst argument of the verb and expresses the clause level trajector. It is typically referential and its referent is usually given. he referent is most oten human and is typically tracked in discourse (see Helasvuo 2001). Separate from the subject we have proposed the category of the existential NP or e-NP, which usually functions semantically as the irst argument of the verb but does not trigger agreement nor does it represent the semantic starting point or function as the clause level trajector. It is in either the nominative or the partitive case, depending on its own features (such as countability or quantiication) or clausal features (such as polarity). he e-NP is typically referential and its referent is usually new to the discourse. Its referent is usually not tracked further in the discourse (Helasvuo 2001: 99–100). It typically occupies the postverbal position.

37

38

Tuomas Huumo and Marja-Liisa Helasvuo

We can see that the characteristics of e-NPs are quite opposite to those of subjects, with respect to both grammatical and discourse features. We have also argued that the meaning of Finnish existential clauses involves features of theme orientation in the sense of Langacker (2008). heme orientation is a conceptualization strategy common in expressions that proile a relationship in which a participant undergoes a change, moves (non-volitionally) or is located somewhere. he alternative strategy, called agent orientation, is based on the concept of an action chain, in which energy is transmitted from an agent to a patient and the agent plays a prominent role as initiator of the event. In contrast, the theme role is based on non-dynamic relationships, such as having a quality or being situated in a location. Since themes are not initiators of events, their role in the clause-level expression is more variable and oten less prominent than that of the agent. his is why many expressions with a theme allow both a construal where the theme is the clause-level trajector (the subject), and a construal where another element serves as the clause-level trajector (an example is the setting subject construction). In Finnish existentials, as we have argued, both the initial locative element and the e-NP display some subject features but also features that speak against their classiication as grammatical subjects. Our general conclusion is that there is no reason to consider either of them as a grammatical subject; thus both the existential construction and its sister constructions discussed in Section 4 are genuinely subjectless.

References Anderson, John (1971). he grammar of case: Toward a localistic theory. Cambridge: Cambridge University Press. Chafe, Wallace L. (1994). Discourse, consciousness, and time. he low and displacement of conscious experience in speaking and writing. Chicago: University of Chicago Press. Comrie, Bernard (1981). Language universals and linguistics typology (2nd ed.). Chicago: he University of Chicago Press. Crot, William (1991). Syntactic categories and grammatical relations. Chicago: he University of Chicago Press. Crot, William (2001). Radical construction grammar. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780198299554.001.0001 Denison, Norman. (1957). he partitive in Finnish. Annales Academiae Scientiarum Fennicae. Series B. 108. Helsinki: Suomalainen Tiedeakatemia. Dixon, R. M. W., Aikhenvald, Alexandra Y., & Onishi, Masayuki (2001). Preface. In Alexandra Y. Aikhenvald (Ed.), Non-canonical marking of subjects and objects (pp. ix–xi). Amsterdam: John Benjamins.

On the subject of subject in Finnish

Freeze, Ray (2001). Existential constructions. In Martin Haspelmath (Ed.), Language typology and language universals, Volume 2: An international handbook (pp. 941–953). Berlin: Mouton de Gruyter. Fried, Mirjam (2005). A frame-based approach to case alternations: he swarm-class verbs in Czech. Cognitive Linguistics, 16(3), 475–512. DOI: 10.1515/cogl.2005.16.3.475 Givón, Talmy (1983). Topic continuity in discourse: A quantitative cross-language study. Amsterdam: John Benjamins. DOI: 10.1075/tsl.3 Goldberg, Adele (1995). Constructions: A construction grammar approach to argument structure. Chicago: he University of Chicago Press. Goldberg, Adele (2006). Constructions at work. he nature of generalization in language. Oxford: Oxford University Press. Hakanen, Aimo (1972). Normaalilause ja eksistentiaalilause. Sananjalka, 14, 36–76. Hakanen, Aimo (1978). Kontrastiivista lauseanalyysia: Eksistentiaalilauseet. Turun yliopiston suomalaisen ja yleisen kielitieteen laitoksen jukaisuja 8. Turku: Turun yliopisto. Hakanen, Aimo (1980). Existential clauses. Introduction. In Osmo Ikola (Ed.), Congressus Quintus Internationalis Fenno-Ugristarum Pars III (pp. 238–251). Turku: Turun yliopisto. Hakulinen, Auli (1983). Subjektikategoria vai nomiaalijäsenten subjektimaisuus? In Auli Hakulinen & Pentti Leino (Eds.), Nykysuomen rakenne ja kehitys 1. Tietolipas 93. Helsinki: Finnish Literature Society. Hakulinen, Auli, & Karlsson, Fred (1979). Nykysuomen lauseoppia. Helsinki: Finnish Literature Society. Hakulinen, Auli, Karlsson, Fred, & Vilkuna, Maria (1980). Suomen tekstilauseiden piirteitä: Kvantitatiivinen tutkimus. Helsingin yliopiston yleisen kielitieteen laitoksen julkaisuja 6. Helsinki: Helsingin yliopisto. Hakulinen, Auli, Vilkuna, Maria, Korhonen, Riitta, Koivisto, Vesa, Heinonen, Tarja Riitta, & Alho, Irja (2004). Iso suomen kielioppi. Helsinki: Finnish Literature Society. Heinämäki, Orvokki (1983). Aspect in Finnish. In Caspar de Groot & Hannu Tommola (Eds.), Aspect bound. A voyage to the realm of Germanic, Slavonic, and Finno-Ugrian aspectology (pp. 153–177). Dordrecht: Foris. Heinämäki, Orvokki (1994). Aspect as boundedness in Finnish. In Carl Bache, Hans Basböll & Carl-Erik Lindberg (Eds.), Tense, aspect, and action: Empirical and theoretical contributions to language typology (pp. 207–233). Berlin: Mouton de Gruyter. Helasvuo, Marja-Liisa (1996). Ollako vai eikö olla – eksistentiaalilauseen subjektin kohtalonkysymys. Virittäjä, 100, 340–356. Helasvuo, Marja-Liisa (2001). Syntax in the making: he emergence of syntactic units in Finnish conversation. Studies in discourse and grammar 9. Amsterdam: John Benjamins. DOI: 10.1075/sidag.9 Helasvuo, Marja-Liisa, & Laitinen, Lea (2006). Person in Finnish: Paradigmatic and syntagmatic relations in interaction. In Marja-Liisa Helasvuo & Lyle Campbell (Eds.), Grammar from the human perspective: Case, space and person in Finnish. Current Issues in Linguistic heory 277 (pp. 173–207). Amsterdam: John Benjamins. DOI: 10.1075/cilt.277.14hel Huumo, Tuomas (1994). Näkökulmia suomen ja viron sanajärjestyseroihin. In Helena Sulkala & Heli Laanekask (Eds.), Lähivertailuja 8: Suomalais-virolainen kontrastiivinen seminaari Hailuodossa 7–9.5.1994 (pp. 21–39). Research reports from the Department of Finnish and Saami, University of Oulu 40. Huumo, Tuomas (1997). Partitiivisubjekti ja tilajatkumot. Sananjalka, 39, 65–98.

39

40

Tuomas Huumo and Marja-Liisa Helasvuo

Huumo, Tuomas (2003). Incremental existence: he world according to the Finnish existential sentence. Linguistics, 41(3), 461–493. DOI: 10.1515/ling.2003.016 Huumo, Tuomas (2005). How ictive dynamicity motivates aspect marking: he riddle of the Finnish quasi-resultative construction. Cognitive Linguistics, 16(1): 113–144. Huumo, Tuomas (2009). Fictive dynamicity, nominal aspect, and the Finnish copulative construction. Cognitive Linguistics, 20(1), 43–70. DOI: 10.1515/COGL.2009.003 Huumo, Tuomas (2010). Nominal aspect, quantity, and time: he case of the Finnish object. Journal of Linguistics, 46(1), 83–125. DOI: 10.1017/S0022226709990223 Ikola, Osmo (1954). Suomen lauseopin ongelmia I–III. Virittäjä, 58, 209–245. Ikola, Osmo (1961). Lauseopin kysymyksiä. Tutkielmia nykysuomen syntaksin alalta. Tietolipas 26. Helsinki: Finnish Literature Society. Itkonen, Terho (1974). Ergatiivisuutta suomessa I. Virittäjä, 78, 379–398. Itkonen, Terho (1975). Ergatiivisuutta suomessa II. Virittäjä, 79, 31–65. Itkonen, Terho (1976). Erään sijamuodon ongelmia. In Opuscula Instituti linguae Fennicae, Universitas Helsingiensis 53 (pp. 173–217). Helsinki: Helsingin yliopisto. Kangasmaa-Minn, Eeva (1968). Verbi- ja objektikategorioiden keskinäisistä suhteista. Sananjalka, 10, 55–65. Keenan, Edward (1976). Towards a universal deinition of “subject”. In Charles N. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press. Kolehmainen, Leena (2010). Sääverbien syntaksia ja semantiikkaa: semanttiset roolit, osallistujien vaihteleva käsitteistäminen ja sääverbien vaihteleva valenssi. Virittäjä, 1(2010), 5–38. Kotilainen, Lari (2007). Konstruktioiden dynamiikkaa. Unpublished dissertation. University of Helsinki. https://oa.doria.i/bitstream/handle/10024/29208/konstruk.pdf?sequence=1 Langacker, Ronald W. (1987). Foundations of cognitive grammar. Vol. 1: heoretical prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. (1991a). Concept, image, and symbol: he cognitive basis of grammar. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110857733 Langacker, Ronald W. (1991b). Foundations of cognitive grammar. Vol. 2: Descriptive application. Stanford: Stanford University Press. Langacker, Ronald W. (1993). Reference-part constructions. Cognitive Linguistics, 4(1): 1–38. Langacker, Ronald W. (2008). Cognitive grammar: A basic introduction. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780195331967.001.0001 Langacker, Ronald W. (2009). Cognitive (Construction) Grammar. Cognitive Linguistics, 20(1), 167–176. Leino, Pentti (1993). Polysemia – kielen moniselitteisyys. Kieli 7. Helsinki: Helsingin yliopiston suomen kielen laitos. Levin, Beth, & Rappaport Hovav, Malka (2005). Argument realization. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511610479 Penttilä, Aarni (1956). Subjektin totaalisuudesta ja partiaalisuudesta. Virittäjä, 60, 28–49. Pälsi, Marja (2000). Finnish resultative sentences. SKY Journal of Linguistics, 13, 211–250. Sadeniemi, Matti (1950). Totaalisesta ja partiaalisesta predikatiivista. Virittäjä, 54: 46–53. Sands, Kristina, & Campbell, Lyle (2001). Non-canonical subjects and objects in Finnish. In Alexandra Y. Aikhenwald, R. M. W. Dixon, & Masayuki Onishi (Eds.), Non-canonical marking of subjects and objects (pp. 251–305). Amsterdam: John Benjamins. DOI: 10.1075/tsl.46.10san Schlachter, Wolfgang (1958). Partitiv und Inkongruenz beim Subjekt des Finnischen. FinnischUgrische Forschungen, 33.

On the subject of subject in Finnish

Shore, Susanna (1992). Aspects of a grammar of Finnish. Ph.D. thesis. Macquarie University, Sydney. Siro, Paavo (1960). Suomen yksinäislauseen perustavia ongelmia. Sananjalka, 2, 36–51. Siro, Paavo (1964). Suomen kielen lauseoppi. Helsinki: Tietosanakirja Oy. Talmy, Leonard (2000). Toward a cognitive semantics. Volume 1: Concept structuring systems. London: he MIT Press. Tiainen, Outi (1997). Suomen eksistentiaalilause ja päättymätön tarina. Virittäjä, 101, 563–571. Tiainen, Outi (1998). Referenttien kuljettaminen diskurssissa. Virittäjä, 102, 498–528. Toivainen, Jorma (1986). Suomen subjektin piirteitä. Sananjalka, 28, 31–46. Vilkuna, Maria (1989). Free word order in Finnish: Its syntax and discourse functions. Helsinki: Finnish Literature Society. Vilkuna, Maria (1992). Referenssi ja määräisyys suomenkielisten tekstien tulkinnassa. Suomi 163. Helsinki: Finnish Literature Society. Vilkuna, Maria (1996). Suomen lauseopin perusteet. Helsinki: Edita. Wiik, Kalevi (1974). Suomen eksistentiaalilauseiden ‘subjekti’. Turun yliopiston fonetiikan laitoksen julkaisuja 13. Turku: Turun yliopisto.

41

Hidden subjects in conversation Estonian personless verb forms as referential devices* Renate Pajusalu University of Tartu

he chapter deals with implicit unspeciic group referents introduced by verb forms (the conditional and the impersonal) which lack person marking. hey have diferent contexts of use: the personless conditional is preferred when the aim of the conversation is to plan a future activity, while the existential impersonal is irst and foremost used in narratives. he personless conditional is typically inclusive, incorporating one or both of the interlocutors into the group. he impersonal is typically speaker-exclusive and refers to a group that does not include the participants of the conversation. For the analysis, the data from the Corpus of Spoken Estonian (University of Tartu) are used.

1.

On person and reference in Estonian

his paper examines how Estonian verb forms (the conditional and the impersonal) which lack person marking convey referential information about hidden subjects. Every language has devices to create new referents and to maintain the identity of the old referents. Talmy Givón (2005: 135) gives a comprehensive list of grammatical means that are used to create referential coherence. he list includes pronouns (which also include the so-called zero pronouns) as well as syntactic categories such as the grammatical functions of case or the passive. hese indexical expressions operate upon the mental discourse model which the participants are constructing together (Cornish 1999: 5). his model is both dependant on context and, as a part of it, constructs the context. he referents are dynamic (Lambrecht 1994; Cornish 1995; Etelämäki 2006). his means that although the identity of a referent must be maintained throughout the conversation, the referent itself may be subject to changes. Ater all, the speakers *

hanks to Virve-Anneli Vihman for many useful comments and suggestions. doi 10.1075/cal.16.03paj © 2015 John Benjamins Publishing Company

44

Renate Pajusalu

in a conversation do not operate with real entities (although in the real world these are also subject to constant change), but with their mental representations. In the case of such representations, the interlocutors’ knowledge about the referent, their attitudes to and views on it are modiied continually. he status of the referent in a conversation is to a great extent determined by the activity that is performed in the conversation (Etelämäki 2006: 17). he more oten an activity occurs in conversations, the stronger the referential practices related to that activity become (Hanks 1996, 1990). Activities change and referents are modiied internally. his, however, does not prevent us from recognising a referent even if its representation in the conversation has considerably changed since the irst mention. his article makes the assumption that the informational status and inner stability of referents may difer considerably. he article concentrates on the referents picked out by the inite verb forms which lack person marking and mostly represent group referents that are characterised by an extremely vague internal structure but that can still participate in referential chains in a conversation. Another central assumption underlying the analysis presented in this article concerns the referential potential of the verb – when studying referentiality, researchers typically focus on NPs (Gundel, Hedberg & Zacharski 1993; Ariel 1990; Cornish 1999, etc.), yet, verb forms may also carry referential meaning. Although the referents discussed below are far from prototypical, they still have a role in the conversation low. When, in a conversation, it is possible to show that a referent can be maintained even if there is a competing one, this signals that the referent in question is integrated into the framework of the discourse and remains accessible even ater being referred to. Linguistic expressions show diferent degrees of referential potential, i.e. their capability of maintaining boundaries between the referents in a discourse difers. A very high identity-preserving potential, for example, is characteristic of the English pronouns he and she in a discourse context that involves one man and one woman. he verb forms discussed in this article represent the opposite end of the spectrum in that their ability to preserve a referent’s identity is very weak. Still, as will be seen below, they can introduce a referent (usually a group referent) and support it. A characteristic feature of these vague group referents is that they are used in talking about an event irst and foremost from the perspective of the event itself, as opposed to the perspective of the participants of that event. Depending on the informational status of a referent in the low of information, diferent linguistic devices can be used to refer. Several authors have presented these linguistic devices as diferent accessibility scales (Chafe 1994; Gundel, Hedberg & Zacharski 1993; Ariel 1996; Gundel et al. 2010, for an overview, see Laury 1999: 9–10; Cornish 1999: 5–11). he referential power of linguistic devices also difers: short personal pronouns (e.g. ta ‘he, she’ in Estonian) refer to the most salient referents, indeinite NPs (which in spoken Estonian usually include

Hidden subjects in conversation

the determiner üks ‘one’ or mingi ‘some’) to less salient referents. Sometimes new referents are created by means of deinite expressions whose identiiabilty results from their connection with a previously evoked expression (see an overview in Cornish 1999: 10–11; Laury 2005). In addition to direct referential devices there are also indirect referential devices. For the most part, the article at hand focuses on referents that are not identiied by means of an NP or a pronoun but that still play a role in the referential framework of the discourse.

2. Person and how it is concealed in Estonian he grammatical category of person in Estonian has (as traditional Estonian grammar has it) six members (mina/ma ‘I’, sina/sa ‘you, sg.’, tema/ta ‘he, she, it’, meie/me ‘we’, teie/te ‘you, pl.’ and nemad/nad ‘they’). An important feature of this system is the existence of long and short (also called weak and strong, e.g. Kaiser 2010) pronoun forms, which difer in their pragmatic value. Long forms are used to emphasise the person (Pool 1999) or to contrast a person to some other person (Kaiser 2010). However, there are also other aspects which determine the usage of long or short pronoun forms (see Pajusalu 2009 for an overview of minimal referential devices in Estonian). Estonian is a pro-drop language, which means that the pronoun in an utterance can be omitted, leaving only the personal ending of the verb and the context to show us which person is referred to. Negative forms in Estonian lack personal endings. his means that the only explicit indicator of person in the negative, if there is one, is the syntactic subject. Actually, there are many cases in which no person is overtly shown in a clause in the negative and one has to use the context to identify the person referred to. here are also many possibilities to avoid explicit reference to the person in Estonian. We can speak about grammatical and pragmatic ways of ‘hiding’ the person (see also Pajusalu & Pajusalu 2005). Pragmatic ‘hiding’ involves diferent kinds of strategies where the person is expressed in the utterance but the meaning of the pronoun or personal ending is not the prototypical one (we instead of you in baby-talk, for example). Grammatical ‘hiding’ involves, irst of all, the generic and impersonal sentences that allow omitting the grammatical subject (Erelt et al. 1993: 227–228). In addition to the impersonal clauses, there is at least one more regular grammatical way of hiding the person. his is the case of the verb forms in the conditional mood which lack personal endings and may have an overt subject but can also do without it (Pajusalu & Pajusalu 2004; Lindström 2009). his article focuses on impersonal and conditional verb forms without explicit person-marking. here are actually more personless constructions in

45

46

Renate Pajusalu

Estonian (see for example Jokela & Plado, this volume), which sometimes function similarly to impersonal and personless conditional forms in the referential chains. However, they will be mentioned below only if they are part of the referential chains under discussion.

3. he personless conditional in conversations he Estonian conditional marker is -ks(i), which is added to the stem of the verb, e.g. ole + ks ‘(it) would be’. In the 1st and 2nd person singular, and in all persons in the plural, it may be followed by a morpheme indicating the person, e.g. ole + ksi + n ‘I would be’ (-n is the marker of the 1st person). hese variants will be called long forms of the conditional in this study. In actual language use, it is common to omit the person ending, e.g. oleks ‘I/you/he/she/it/we/they would be’; this variant will here be referred to as the short form of the conditional. hese parallel forms are considered equally acceptable in Standard Estonian and they may occur with or without the overt subject. We will call the form without a subject the “personless conditional”. he conditional mood has two tense forms: the present conditional (expresses the present and the future), and the perfect conditional (expresses the past but also perfectivity). Conditional forms can also be used in the impersonal voice but such usage remains rare. he meanings of the conditional form (at least in Finnish and Estonian) have been divided into two that, according to Kauppinen (1998) and Metslang (1999), may be considered two diferent interpretations. he frame interpretation is used when the proposition expressed in the utterance is placed in a conditional situation (Metslang 1999: 49), in which, according to the meaning of the expression, the event described belongs to an unreal world, i.e., it remains hypothetical. In the case of the intentional interpretation the conditional expresses the speaker’s wish and above all conveys pragmatic information about the type of speech act (wish, request, etc.). he most frequent use of conditional verb forms lacking both a personal ending and a subject (the personless conditional) is with modal verbs which convey the meaning of an indirect request/proposal to a person or persons, for example peaks ‘should’ (1a) and võiks ‘could’ (1b) (Pajusalu & Pajusalu 2004). (1)1 a.

pea-ks Jüri käe-st küsi-ma must-cond Jüri hand-ela ask-sup ‘(we/someone) should ask Jüri’

1. All examples in this paper come from the Corpus of Spoken Estonian (University of Tartu), http://www.cl.ut.ee/suuline/Korpus.php?lang=en

Hidden subjects in conversation

b. või-ks tee-d juu-a may-cond tea-par drink-inf ‘let’s drink some tea, lit. could drink tea’

In Estonian the usual reason for concealing the person with a personless conditional probably has to do with politeness considerations: reference to person is avoided in cases in which the speaker is not sure whether the more polite address form teie/te or the informal sina/sa should be used (Keevallik 2004; Lindström 2009; Pajusalu et al. 2010). Politeness is, however, not the only reason why speakers use the personless form. As can be seen below, in some situations the most relevant referent type for the hidden subject of the personless conditional is that of a vague group that, on the one hand, allows one to leave the referent unspeciied but, on the other hand, provides enough certainty to be able to talk about the referent. Du Bois calls it ‘the curiosity principle’: “A reference is only counted as identiiable if it identiies an object closely enough to satisfy the curiosity of the addressee” (Du Bois 1980: 233; see also Laury 1999: 7). he personless conditional occurs most frequently in planning contexts in which the interlocutors discuss future events. In these contexts the link between speech and activity is realised particularly clearly – a speciic type of planning gives rise to a speciic type of referential practice (Hanks 1990). In planning contexts the conditional receives both interpretations: the frame interpretation in the sense that the events discussed have not yet taken place and it is not sure that they will; and the intentional interpretation in the sense that the conditional verb in the present tense (typically a modal verb) infuses the utterance with intentionality. Such intentionality is usually personless – it does not mark grammatical person. In most cases it concerns the speaker himself or herself (especially in case of verbs tahaks or sooviks ‘would like, wish’, see Example 2), the speaker and the hearer together (Example 3); a group of people that includes the speaker or hearer or both (Examples 4, 5). In some cases it is unambiguously clear who is meant (the speaker in 2, the speaker and the doctor in 3), although this need not always be the case (4, 5); the general tendency seems to be that if a group is referred to, the group includes the speaker. (2) (a client entering the hairdresser’s, planning an appointment) 1

2

K: tere (0.5) soovi-ks  ˈlõikus-aega kinni pan-na. hello want-cond cut-time.par shut put-inf ‘hello (0.5) I’d like to make an appointment for a haircut J: millal. when when?

47

48

Renate Pajusalu

3

4

K: kas ˈtäna ole-ks aega peale (.) ˈpool-t q today be-cond time.par ater half-par would you have a slot ater (.) about half ˈnelja  umbes. four.par approximately past three.’

(3) (a patient to the doctor, planning a future treatment of papulas) 1

2

3

4

P: ne-id on mu-l üpris ˈpalju= ja (0.8) nüüd ma 3pl-par be.3sg 1sg-ade quite many and now 1sg ‘I have quite many of them = and (0.8) now I ei ˈoska-gi  öel-da (.)  kas õõ=h (1.0) nad  ˈtule-ksi-d  neg can-ptcl say-inf q 3pl come-cond-3pl can’t say (.) whether mhmh (1.0) whether they will tagasi või ˈei, aga ma mõtle-si-n= et pea-ks back or neg but 1sg think-pst-1sg that must-cond return or not, but I thought that (we) should ikka ˈkontrolli-ma=h. (1.0) always check-sup still check (1.0)’

(4) (a client at the travel agency, planning accommodation) 1

2

me  just  nagu  mõtle-si-m= ka  {selle-s  mõt-s}= et  1pl just like think-pst-1pl also this-ine thought-ine that ‘we were kind of also thinking {I mean that} ole-ks ˈjärv lähedal kus saa-ks ˈuju-da be-cond lake near where can-cond swim-inf it would be nice to have a lake nearby where we could swim’

(5) (everyday conversation, planning a future event) 1

2

3

M: hehe (.) ei ma arva-n et= et või-ks hehe neg 1sg think-1sg that that may-cond ‘hehe (.) no, I think that we could like selle võistluse nagu algata-da, ja mina this.gen contest.gen like start-inf and 1sg.long start this contest and I pane-ks selle pruuni manna-supi-ga kohe kinni. put-cond this.gen brown.gen manna-soup-com at.once shut would tuck it away with my brown wheat-semolina-soup’

In the case of group referents it is oten diicult to decide whether reference is being made to a group or whether the utterance is intended as generic. In Example (4) the conditional relative clause kus saaks ujuda ‘where one could have

Hidden subjects in conversation

a swim’ can be interpreted as referring to a group referent (those who go on the trip can have a swim) or as generic (anyone can swim). In Example (5) the implicit subject of the verb võiks ‘could’ is most probably a certain group of people (including the speaker) who could organise the contest. Personless perfect conditional forms most oten occur in utterances in which the speaker expresses regret about something not taking place. he implicit subject oten includes the speaker but does so less oten than in the planning context. he modal construction oleks pidanud ‘should have done’ is typically used in the perfect conditional. In Example (6) the speaker regrets that there is no ireplace in her room but does not make it clear who should have built it (probably I or we2). In Example (7) the speaker regrets that she did not buy anything “good” from the shop and her interlocutor agrees that she should have bought something (oleks muidugi pidanud). (6) (everyday conversation) 1

2

3

4

tee-b. K: ˈkadeda-ks jealous-trnsl do-3sg ‘it makes (me) jealous A: [kadeda-ks{-} ega ˈtei-l ju eiˈ ole.] jealous-trnsl neg 2pl-ade ptcl neg be jealous but you don’t have one M: [ole-ks  pida-nud ka enda-le] kuskile ˈtuppa be-cond must-ptcp also self-all somewhere room.ill should have built myself ˈkamina ˈtege-ma. ireplace.gen do-sup a ireplace somewhere in the house’

(7) (everyday conversation) 1

2

3

M: [(–-)]ole-ks pida-nud veel mida-gi hea-d be-cond must-ptcp more some-ptcl good-par [(–-)] I should have bought some more goodies ost-ma= aga.(1.0) buy-sup but =but.(1.0) L: ole-ks muidugi pida-nud.(1.0) be-cond naturally must-ptcp of course you should have. (1.0)’

2. he relexive pronoun endale biases interpretation towards the irst person (singular or plural). he author thanks Virve-Anneli Vihman for this comment.

49

50

Renate Pajusalu

To summarise: in the data analysed the personless conditional occurred most oten in contexts involving the planning of future actions and the expression of regret. In all cases the implied referent was human.

4. Personless conditional in the referential framework of the discourse As could be seen in the previous section, personless conditional verb forms can be interpreted as picking out any person available in the context as their referent. Typically, however, reference is made to an inclusive group referent (one that includes the speaker). his section will focus on the efect of such a vague referential act in a referential chain. he question is whether in the case of fuzzy reference a subject can be constituted as a discourse referent and whether the subject as such can participate in a referential chain, or whether such a reference would be incapable of leading to the introduction of a discourse referent whose identity could be maintained through the discourse. In Example (5) above, the irst conditional construction võiks võistluse algatada ‘could start this contest’ refers to a group with fuzzy boundaries. When the speaker then refers to herself exclusively, she uses the conditional paneks (‘would put’, no person marking) with the long pronoun mina. Had the second conditional not been accompanied by a personal pronoun, this would imply reference to the same referent that was evoked earlier (the open ‘we’), which would not correspond to the speaker’s intention. he long pronoun in this case functions similarly to what in other languages are referred to as ‘stressed (or contrastive) pronouns’, which are used when the continuity of the conversation has been ruptured (Givón 2005: 136). In the data for this study, the personless conditional was mostly used to form referential chains in institutional dialogues in which the interlocutors were planning a future event. Since there were relatively few instances of such situations in everyday conversations and those that were found tended to be short, they did not contain extensive chains of personless conditionals. Still, in Example (8) there are ive personless conditional forms: saaks (line 3), jalutaks (line 5) and võiks three consecutive clauses (lines 8, 9, 11). (8) (everyday conversation, planning a party) 1

2

S:

[oi(.) oi(.) ˈmina taha-ks t kohe oh oh 1sg.long want-cond ? just ‘[oh (.) oh (.) ˈI’d like to mida-gi ˈteh-a taha-ks(.) selle-ga] selle ˈkasseti-ga what-ptcl do-inf want-cond this-com this.gen tape-com I’d really like to do something, (.) with this] this tape 

Hidden subjects in conversation

saa-ks teh-a mingi ˈõudselt hea ˈkuulde-mängu may-cond do-inf some horribly good hear-play.gen (I/we?) could make a killer of a ˈradio show  4 võib-olla isegi mingi ˈetenduse.(.) may-be even some play.gen maybe even some kind of play. (.) 5 just tea-d ni-modi(.) ˈturnüüri-ga ˈkleiti-de-ga ˈjaluta-ks. just know-2sg this-way bustle-com dress-pl-com walk-cond just like that y’know (.)(we?’d) walk around in dresses with bustles.’ 6 L:  mhmh ptcl mhmh 7 S: selle(.) tidiridirii saatel. this-gen accompanying ‘to that (.) ((imitation of singing)) music.’ 8 L: mina mõtle-si-n näite-ks välja mis või-ks 1sg.long think-pst-1sg example-trnsl out rel may-cond ‘I’ve come up with what (we) could 9 e rebas-te-le teh-a.(.) rebase-d või-ks ise e fox-pl-all do-inf fox-pl may-cond self do with the freshmen (.) the freshmen could 10 pan-na se-da Elu allika-id ette luge-ma put-inf this-par life source-pl.par for read-sup be made to read ‘Elu allikad’ (Sources of life, a novel) out loud 11 ja sis või-ks rebas-te seast vali-da(.) ä and then may-cond fox-pl.gen from choose-inf and we (?) could choose from freshmen 12 Arvedi-t või Tuuli ja Poliitsia. Arved-par or Tuuli and Poliitsia Arved, or Tuuli and Poliitsia’

3

All these forms (saaks, jalutaks, võiks) refer to probable actors of the planned event (the so-called ‘baptism of freshmen’) who form the fuzzy group ‘we’ that is never mentioned by means of a pronoun or an NP. Line 1 also has a conditional form without person marking (tahaks) which is accompanied by the pronoun mina ‘I’ referring to the speaker. he referent of the second instance of tahaks is unclear – the verb may either belong together with the pronoun mina or it may be used as a way to move on to the chain of the group referent. he verb jalutaks in line 3 clearly refers to the group referent because a single-person referent is unlikely to walk around in ‘dresses’ (kleiti-de-ga, plural). Similarly to Example (5) above, we see that if the speaker wishes to emphasise that he or she is the only subject of a conditional,

51

52

Renate Pajusalu

he is likely to use it with the pronoun mina if the reference takes place in the immediate vicinity of an instance of a conditional chain invoking a group referent. Example (9) presents excerpts from a phone call where the client is calling a car shop. (9) (institutional phone call, a client is calling a car parts shop, planning a purchase) V: Auto-liin= tere car-line hello Autoliin (name of car parts shop) =Hello! 2 H: mt tervist.(.) ee kas tei-l Reˈnoo= auto-de-leˈ mt hello ee q pl2-ade Renault car-pl-all mt Hi! (.) er would you have for Renaults.= 3 ka on juppe.= too be.3sg part.pl.par parts 4 V: =m-midagi ˈikka on. something always be.3sg s-some things certainly. 5 H: ee(.) ole-ks vaja pool-ˈtelgi mõlema-id=hh.(.) ee be-cond need half-axel.pl.par both-pl.par=hh er (.) I/we/one would need both half-axles=hh. (.)  6 parem-poolse-t vasak-pool[se-t.] right-side-par let-side-par the right the let[.] 7 V: [mis] Renoo ˈon. what Renault be.3sg [what] is the Renault ˈmodel. 8 H: Renoo ˈüksteist, kaeksakend=ˈkolm aasta.(.) Renault eleven eighty=three year Renault ˈeleven, built in eighty-three. (.) 9 V: nii. (1.5) kas te ˈeile vist küsi-si-te ˈka so q 2pl yesterday probably ask-pst-2pl too ok. (1.5) did you ask about that yesterday too 10 korra= vä. once q once 11 H: jah, ma ole-n igalt-poolt ˈTartu-st ee ˈküsi-nud= et yes 1sg be-1sg every-side Tartu-ela ee ask-ptcp that yes, I’ve been asking all over Tartu er 12 ma: uuri-n= ne-id nüd 1sg explore-1sg 3sg-par now =so I’m checking these now  1

Hidden subjects in conversation

13 H: ee (.) ole-ks vaja pool-ˈtelgi mõlema-id=hh.(.) ee be-cond need half-axel.pl.par both-pl.par=hh. er (.) I/we/one would need both half-axles=hh. (.)  14 parem-poolse-t vasak-poolse-t. right-side-par let-side-par the right the let[.] /…/ 15 H: jaa ja ˈkui kiiresti need ˈkätte saa-ks.(.) yes and how fast this.3pl hand.ill get-cond yeah and ˈhow soon could I/we/one receive delivery. (.) 16 V: kui on Tallina-s ˈole-ma-s, saa-b kätte if be.3sg Tallinn-ine be-sup-ine get-3sg hand.ill if Tallin ˈhas them, (you) could get them 17 omme ˈomiku= kella ˈkümne-ks. tomorrow morning.gen clock.gen ten-trnsl by ten tomorrow morning

In line 5 the client refers to himself by using a conditional form oleks vaja ‘would need’ for the irst time. he utterance lacks the adverbial mul ‘I-ade’ which could identify the person who needs these things. he conversation seems to be generic but at the same time it is clear that the speaker is talking about himself and his car. In line 7 the shop assistant also avoids personal reference and doesn’t use the pronoun teil (formal you-ade) neither. As the shop assistant recognises the client and turns to him using the second plural pronoun te (line 9), the speaker also becomes more speciic and uses the irst person pronoun ma to refer to himself (line 11, 12). In connection with this the nature of the activity also changes: instead of planning a future purchase the interlocutors turn to comments on the current situation. At the end of the conversation they revert to planning the purchase and the person reference becomes vague again (line 13). In line 15 the personless conditional saaks ‘would get’ is used by the client, referring to himself again. In line 16 the shop assistant uses the indicative verb form saab (‘get’ 3sg) without an explicit subject. By doing this the assistant makes the context more generic and declarative and thereby creates a contrast between the general knowledge implied and the planning discussion (with conditional forms) that preceded it. Example (10) illustrates a typical context for the use of conditional forms: giving advice to one’s interlocutor. he phone conversation takes place between a travel agent (V) and a client (H). (10) (institutional phone call to a tourist oice) 1

H: ise ole-me mõtel-nud kõige rohkem nagu self be-1pl think-ptcp all.gen more like ‘ourselves we’ve been thinking mostly about

53

54

Renate Pajusalu

ˈratta-matka variˈanti= et.hh et ei ole seal cycle-trip variant that that neg be there a cycling trip = (I/we) haven’t been in these places 3 kandi-s ˈkäi-nud, et (.) noh tõesti ei tea place-ine go-ptcp that ptcl really neg know (I/we) really don’t know 4 kus ole-ks võimalik ˈööbi-da ja mis see hea where be-cond possible stay-inf and what this good Where one could stay the night and what’s a good 5 ˈtrajektoor ole-ks kus seal ˈsõit-a= ja(.) trajectory be-cond where there ride-inf and route that one could take ˈ =and (.)  6 ja nii ˈedasi. (.) and so forward and so ˈon. (.) 7 V: no sääl ˈlähikonna-s on nüüd ˈpäris ˈmitu(.) sellis-t= well there nearby-ine be.3pl now quite several dem-par well, in the area, ˈthere’s quite a number of these 8 peatus (.) ˈkohta kus (.) kus ˈvõi-ks nagu ütle-me stop place where where may-cond as say-1pl places (.) ˈto stay where (.) where ˈyou/one could ˈlet’s say 9 ˈol-la. (0.5) ee ˈüks on on (.) kohe Pühajärve-l be-inf ee one be.3sg be.3sg right Pühajärve-ade stay (0.5) er ˈone is is (.) right at Pühajärve  10 ˈenda-l on nüüd hoˈtell, (.) ˈpäris ˈkorralik hoˈtell.(.) self-ade be.3sg now hotel quite decent hotel ˈitself now has a hoˈtel, (.) ˈquite a ˈdecent hotel. (.) 11 H: mhmh mhmh mhmh mhmh 12 V: et ee noh seal võ-ks: või-ks päris päris that ee well there may-cond may-cond quite quite so er well there you/one could: could really really  13 vabalt nagu .hhˈ ööbi-da, (0.5) ee (.) siis seal (.) ˈlähikonna-s freely like hh stay-inf ee then there nearby-ine like stay the night, (0.5) er (.) then nearby (.)  /……/ 14 V: ˈKääriku-l on: (.) no ma vaata-si-n Kääriku-ade be.3sg ptcl 1sg look-pst-1sg ne-il on seal dem-pl.ade be.3sg there ˈKääriku has: (.) now I’ve looked it up 2

Hidden subjects in conversation

ˈrühma-hinna-d, pole-gi kõige ˈhulle-ma-d group-price-pl neg.be-ptcl all.gen mad-cmpr-pl they’ve got ˈgroup prices, which are not really ˈbad 16 kui te-id nüüd ˈpalju ole-ks [vat] if 2pl.par now many be-cond look if you were now ˈa large group [see] 17 H: [kümme]= viisˈteist umbes. ten iteen about [ten]=iteen approximately. 18 V: jah. no ˈsellise-l juhu-l saa-ks rühma ˈhinda yes well dem-ade case-ade can-cond group price.par yes. well ˈin that case (I/we/you) could ask for a group price 19 sealt küsi-da. from.there ask-inf from there /…/ 20 V: ee talve-l muidugi saa-ks(.) saa-ks seal ee winter-ade surely can-cond can-cond there er in winter of course (I/we/you) could (.) you could there nagu (.) ˈsuusata-da like ski-inf like (.) ˈski /…/ 21 V: et pea-ks siis siia nagu ˈkoha-le tule-ma. that must-cond then here like place-all come-sup so (I/we/one) should then then like ˈcome to our oice.’

15

In this example H actually represents a group of people, to whom he has referred using the pronoun me ‘we’ at the beginning of the conversation (not shown in the example excerpt) and about whom he had told the agent that the size of the group was about 10–15 people. his group, on behalf of whom the speaker acts, is the recurring referent of the conversation. One can imagine the group as a vaguely deined body of people from which the client stands out as their representative. In these situations personless forms refer to the group referent. When the group does not it the referent semantically (for example, when an activity that is not suitable for performance by a group is discussed), the personless form may refer to the group’s representative. he excerpt in (10) begins with the caller using the ise oleme ‘(we) ourselves are’ as a referential device (line 1). In line 2 there is a personless negative verb ei ole, which is referentially ambiguous – it does not allow one to say with certainty whether the speaker refers to himself or to the whole group whose representative

55

56

Renate Pajusalu

he is. As for the continuity of the conversation, this does not matter – the vague reference ‘we’ is maintained. here are two conditional forms oleks võimalik ‘would be possible’ and oleks ‘would be’ in lines 4 and 5. Both make their reference to a vague group of people because there is no habitative adverbial to express a speciic agent (to refer to the group of travellers, the habitative adverbial meil or meie jaoks ‘for us’ could have been used). In line 12 there is again a personless conditional form võiks ‘could’. Yet it is clear that the agent of the verb is the same unspeciied group of travellers. he same group is again referred to by means of the 2nd person plural pronoun teid in line 16. he personless conditional saaks ‘could’ in line 18 could in principle refer to the client individually. It is more probable, however, that the travel agent is still talking about the same vague group – later, the travel agent uses the same form saaks to refer to the group (line 20). At the end of the conversation there is one more personless conditional peaks (line 21). his time it appears to refer to the representative of the group rather than the group itself because it is unlikely that the whole group would need to make an appointment at the travel agency. hus, in Example (10) a referential chain is formed in which the referent (a potential and vague group of travellers) are referred to by means of the following verb forms: 1pl – neg (personless form) – cond (personless form – adverbial) – cond (personless form – adverbial) – cond (personless form) – cond (personless form) – 2pl – cond (personless form) – cond (personless form). In Example (10), personless conditional forms preserve the identity of the referent. he situation, however, might be diferent if there are competing referents present. Some examples show that personless conditionals might still suice for maintaining a referent in such a case. In Example (11) there is a conversation between a client (K) and the representative of a building company (T). (11) (institutional conversation in a car) 1

2

3

4

T: no= ma ausalt öel-da jah ma: hõ kui ta well 1sg honestly say-inf yes 1sg hm if 3sg well=I to be frank yes I: mh if it’s on mingi ˈnurga-ke, ega ma siis ee nagu ei be.3sg some corner-dmn neg 1sg then ee like neg only a petty project ˈ, then I really er well ˈ ˈtaha-ks= aga (.) want-cond but wouldn’t like to=but (.) K: ta ˈnurga-kene ˈon jah= jaa ja= ˈmuidugi: (1.0) 3sg corner-dmn be.3sg yes yes and sure it’s ˈa petty project ˈindeed=yes and=ˈof course: (1.0) 

Hidden subjects in conversation

et ˈtee-ks siis juba selle katuse that do-cond then already this roof.gen Well then (I/we/one) would also install the rooing.  6 ˈkatmise ˈka ära. (1.5) siin osa-d ˈosa-d irma-d covering.gen also ptcl here part-pl part-pl irm-pl (1.5) I’ve heard from some ˈirms  7 ˈütle-vad= et [nende] say-3pl that 3pl.gen ˈthat= [their] 8 T: [ˈalla] ei lähe. down neg go won’t go down. 9 K: ˈsiit= siit. ((näitab teed)) (6.0) et ee ˈtaha-ksi-d here here that ee want-cond-3pl ˈfrom here=from here. (6.0) ((showing the way)) well they erˈwould want 10 ikka mingisugus-t (2.0) [ˈtuhande-id.] always some.kind-par thousand-pl.par some (2.0) [ˈthousands.]  5

In lines 2–3 the builder uses the phrase ma ei tahaks ‘I wouldn’t like’, meaning that he would prefer to decline the job if it is too small. he short form of the conditional is used, but the pronoun ma disambiguates the interpretation as 1sg. he client responds by saying that although the project in question is not a big one, there is also the rooing that needs to be installed (teeks siis juba katuse katmise ka ära ‘I/we/one would install the rooing’, lines 5–6). his conditional form is personless and typical of a situation in which interlocutors are planning the provision of a service because it refers to the supplier and the customer as a team. Albeit this referent has not been introduced as such before, the people involved in a conversation are usually present as a ‘we’. he next time the conditional is used in line 9, where it refers to the irms mentioned previously, and in this case using the person-marked long conditional form tahaksid (‘they should want’; instead of the personless tahaks, which would also be grammatically acceptable). It seems that the personless conditional would not have been contextually acceptable here because it would have referred to the already established ‘we’ of the discourse participants. he personless conditional irst and foremost preserves the ‘we’ referent that includes the people present and that does not extend to other grammatically possible referents (e.g. anaphoric 3rd person referents). When competing referents both include persons present, the personless conditional alone does not preserve the referent. In Example (12) there are two closely

57

58

Renate Pajusalu

situated personless conditional verbs (saaks ‘could’ and võiks ‘would’) which refer to diferent referents. However, the reference switch is motivated by the changing activity of the speaker. (12) (institutional conversation, travel agency) 1

2

T: et nagu sellis-t ˈmajutuse= osa saa-ks that like dem-par accommodation.gen part.par can-cond this accommodation part could ise võib-ola ˈinterneti-st teh-a. self may-be internet-ela do-inf be done on the Internet.

[4 lines omitted] 7

ˈmida-gi või-ks ˈuuri-da (1.2) ((otsib arvutist)) some-ptcl may-cond research-inf I/we/one could look into this (1.2)’  ((running a search on the computer))

In line 1 the travel agent says that the client could book accommodation herself on the Internet (majutuseosa saaks ise internetis teha). he relexive pronoun ise shows that the speaker refers to his interlocutor as well as to a general possibility (Lindström & Vihman 2010). his referent is maintained as the imaginary habitive adverbial of the periphrastic verb on vaja sisse toksida ‘one needs to enter (information on the keyboard)’ – a client who wishes to look for accommodation online needs a credit card. In line 7, ofering her help to search for appropriate accommodation on the Internet, the same speaker uses the personless conditional võiks to refer to herself. hus the reference is to 1sg. here is no linguistic element between these two forms to help change the interpretation of the referent; thus, it has to be concluded that the referential relationship of the irst conditional was not suiciently strong to require an explicit discontinuation. However, there is a long pause during which the travel agent starts to search for the information, which allows us to treat the utterance in line 7 to some extent as discontinuous (Givón 2005: 136) with what was said before. In this example referential relationships are formed by means of a change of activity: there is a transfer from planning to information searching. In conclusion one could say that the personless conditional is most frequently used to construct referential chains in planning situations where it functions as a tracking device of vaguely deined speaker-inclusive groups. Although personless conditionals do possess a certain referential ability, this remains closely connected to the ongoing activity.

Hidden subjects in conversation

5. Impersonal in Estonian conversations he Estonian impersonal is a verb form that does not allow overt expression of a grammatical subject. he impersonal is formed with the morpheme –(t/d)akse in the present and the morpheme t/di in the past tense. he perfect and pluperfect tenses of the impersonal are periphrastic and are formed with the auxiliary olema ‘be’ and the impersonal tud-participle (for an overview of the Estonian impersonal, see Torn-Leesik 2009). his article will focus mostly on present and past forms, including comments on the perfect forms and the tud participle where the occasion warrants. here is an ongoing discussion on the question of how many voice paradigms there are in Estonian, because the periphrastic tud participle forms possess a number of features that make them stand apart from simple forms (Erelt et al. 1993: 31; Torn-Leesik 2009; Lindström, this volume; on the same issue in Finnish, see Helasvuo 2006: 236–243). his issue will be let open here, as diferent forms in diferent contexts are bound to have diferent usage anyway. he functions of the impersonal in everyday conversations have recently been studied by Reeli Torn-Leesik and Virve Vihman (2010), who use Susanna Shore’s (1988) analysis of two prototypical meanings of the Finnish passive as one of the points of departure for their study. A similar distinction is also used in the comprehensive descriptive grammar of Estonian, which makes a distinction between generic and impersonal clauses containing impersonal verb forms (Erelt et al. 1993: 227–228). In the case of a universal reading (Shore’s Prototype I3), the impersonal actor has generalised reference and is plural in number. hese clauses typically describe general assumptions or habits and are speaker-exclusive (the speaker is not the agent). In the case of existential readings (Shore’s Prototype II), the unexpressed actor refers to a speciic person or group who performs the action, but the identity of this actor is let unspeciied in the clause. hese cases can be divided according the reference type they express: vague or speciic existential readings. In the case of vague reference the speaker does not know the identity of the actor or, even if the speaker knows the identity, it is let unspeciied. he other type of existential reading is speciic because it refers to speciic referents usually known to both the speaker and reader. Torn-Leesik and Vihman also describe a corporate interpretation, which occurs when the impersonal verb is used to refer to a socially

3. In the Finnish edition of her book (1986) Shore uses the terms K and P prototype. he K-prototype is mostly encountered in written language and corresponds to Prototype I, while the P-prototype is typical of spoken language and corresponds to Prototype II.

59

60

Renate Pajusalu

designated body or group of people (Torn-Leesik & Vihman 2010). he implicit agent in the Estonian impersonal and the Finnish passive is human (Torn 2002; Hakulinen et al. 2004: 1261) and typically refers to a group. Existential readings of impersonals have a speciic referent and so have to participate in referential chains. he question why speakers choose to use the impersonal in cases of speciic and well-known referents has been repeatedly raised in both Estonian and Finnish linguistics. Torn-Leesik and Vihman (2010: 327–328) summarise their indings: “Certainly impersonalisation, or backgrounding the actor, remains an important motivating factor, but the impersonal seems to have other efects as well, such as distancing the speaker from the event, drawing attention to some unexpected element in a situation, and framing the utterance in a pragmatically marked way”. Despite similarities, there also remain considerable diferences between the use of the Estonian impersonal and the Finnish passive: in spoken Finnish the passive form is oten overtly used with the pronoun ‘we’ (see, e.g., Hakulinen et al. 2004: 1256), while in Estonian the 1pl interpretation is possible in exceptional cases and is not prototypical. Torn-Leesik & Vihman (2010) also discuss the behaviour of the impersonal in referential chains. While partly based on their indings, the present analysis will use a wider approach by, on the one hand, including institutional conversations in the data analysed and, on the other hand, concentrating on the identity of the referent in the referential framework.

6. Impersonal verbs in the referential framework of a conversation One-third of the present and past impersonals in the everyday conversations that were analysed as part of this study occurred without any overt mention of the referent either before or ater the impersonal; however, in two-thirds of the cases, the impersonal forms created a referential chain. Based on the data of the study,4 we can claim that impersonal forms oten form referential chains. As the generic function of the impersonal involves presenting general viewpoints, impersonal forms oten occur in conversations or monologues which focus on giving information – the so-called lecture situations. Example (13) below includes utterances from a lecture on potatoes.

4. he quantitative data are drawn from a selection of 3 to 5-page transcribed extracts from 105 conversations (the Corpus of Spoken Estonian at the University of Tartu, see an overview in Hennoste 2003). here were 287 instances of impersonal present and past verb forms (only airmative clauses) from various verb stems.

Hidden subjects in conversation

(13) (lecture, only impersonal utterances are presented) ER: kuidas kartuli-t= ee=h (0.5) e ˈjaota-ta-kse,(.) rühmita-ta-kse, how potato-par ptcl divide-iprs-pre group-iprs-prs how does one classify, (.) ˈgroup potatoes= /…/ Lines 2–5 omitted 6 ER: kuid mis kartuli-ga veel ˈteh-a-kse. but what potato-com else do-iprs-prs and what else does one do with potatoes 7 T: prae-ta-kse./…/ fry-iprs-prs one fries them. /…/ 8 ER: *nii* (.) no teh-a-kse ˈtsipse. (1.2)  so ptcl do-iprs-prs crisp.pl.par *now then* (.) well one makes crisps. 1

All impersonal forms in the excerpt have a generic (universal) reading. While they do not have explicit referents, they still provide a generic background to the story. his repetitive use of the impersonal can also be considered a referential chain, to a degree: every successive generic form strengthens the generic nature of the indexical background. In lines 1–5, there are utterances that represent scientiic classiications (there are 3 impersonals in the omitted part), while in lines 6–8 the lecturer discusses the use of potatoes in everyday life with a student, that is, the conversation leaves the purely scientiic paradigm but still retains its generic background. In the case of an existential reading with a vague referent impersonal forms may consistently refer to a generic/unspeciied referent. his is particularly frequent in narratives, when customs or events are discussed at length. hese referential chains are oten created by alternating impersonal verb forms and pronouns. he general tendency in everyday conversations seems to be that the imperfect impersonal verbs with an existential reading evoke referents who, in a narrative context, function as a group (which normally does not include either the speaker or the listener). Such impersonals oten alternate with third person pronouns in referential chains. his is illustrated in Example (14) in which a student is talking about her job in a shop. (14) (everyday conversation) 1

2

et= ˈtema las-ti nagu ˈkassa taha= ju kohe that 3sg.long let-iprs.pst like cash.register back ptcl at.once hat she was allowed behind the cash till = right away tema ol-i ju nigu noh ˈtema võe-ti 3sg.long be-pst.3sg ptcl like ptcl 3sg.long take-iprs.pst she was right like well, she was taken on

61

62

Renate Pajusalu

ju ˈselle koha nagu ˈeelmise türuku aseme-le ee: ptcl this.gen place like previous.gen girl.gen place-all ee to this position well to replace the previous girl um 4 ja noh mm: ˈmina mina ei ˈsaa-nud ˈkassa-sse and ptcl mm 1sg.long 1sg.long neg get-ptcp cashier-ill and well um I I didn’t get to be the cashier 5 sellepärast= et neid ol-i ˈisegi nii ˈpalju. because that 3pl.par be-3sg.pst even so much because there were too many of them 6 isegi ˈsuve-l kaks tükki ol-i-d nigu= sis ˈsaali= even summer-ade two peace.par be-pst-3sg like then hall.gen (even in summer) two of them were like attending to customers on the loor 7 pial, ja: ˈvaheta-ti= sis= neid= ˈkassa-s, ˈkes on and change-iprs.pst then 3pl.par cashier-ine who and then they were moved to the cash till, 8 nagu ˈlõuna-le ˈläks. parajasti= nii aga noh ˈtagantjärele like lunch-ade go.pst.3sg just so but ptcl aterwards to replace those who went out to lunch but well thinkin’ about it now well 9 võt-te-s= et nei-l ol-i palju ˈlihtsa-m ja palju take-inf-ine that 3pl-ade be-pst.3sg much easy-cmpr and much they had it a lot easier and 10 ˈkerge-m kui nad ˈvõt-si-d ju ˈminu sinna ˈsaali easy-cmpr when 3pl take-pst-3pl ptcl 1sg.gen there hall.gen simpler when they assigned me to watch the loor 11 piale ˈpassi-ma, et= noh. ˈmille-ks nad ole-ks pida-nud on look-sup that ptcl what-trnsl 3pl be-cond must-ptcp because well whyever should they have 12 minu ˈkolme-ks ˈkuu-ks nagu ˈvälja õpeta-ma 1sg.gen.long three-trnsl month-trnsl like out teach-sup like trained me for three months otherwise

3

In lines 1, 2 and 7 the student uses the impersonal verbs lasti ‘was allowed’, võeti ‘was taken’ and vahetati ’was changed’ to refer to the staf of the shop, with a primary emphasis on the management (as opposed to other employees). In line 5 the pronoun neid ‘them’ refers primarily to employees (not management). At this point, one could read the discussion as evoking two diferent referents: the impersonal refers to the management and the 3pl pronoun refers to the cashiers. In lines 9–11, however, the pronouns neil ‘they.ade’ and nad ‘they’ recognisably refer to the management. hus they share a referential chain with impersonal verbs (lines 1, 2 and 7). he vague group referent is created by means of interaction between the impersonal

Hidden subjects in conversation

forms and the 3rd person pronoun: the irst three mentions are made using the impersonal, which in subsequent ones is replaced by the personal pronoun. According to the data, the corporate interpretation of the impersonal is rather prominent and it helps to distinguish between diferent group referents. Example (15) describes a conversation at a bank in which two group referents are mentioned: the sureties of the client and bank representatives. (15) (institutional conversation in a bank about a bank loan) 1

2

3

4

5

6

7

[et] on ˈnäh-a et nen-de (.) mingi sisse- midagi that be.3sg see-inf that 3pl-gen some in something that one can see that they laeku-b ˈpanka. .hh ja tule-vad tee-vad käenduse-d come-3sg bank.ill hh and come-3pl do-3pl suretyship-pl have some income and (they) come and do a contract on suretyship ˈära. ja anna-vad nee-d teie ˈkätte, ja kui away and give-3pl these-pl 2pl hand.ill and if and (they) give them to you, and if tei-l on tudengi-pakett ˈka juba teh-tud,(.) 2pl-ade be.3sg student-package also already do-ptcp your student-package is done already sis peale (.) esimes-t okˈtoobri-t, (.) tule-te uuesti ˈpanka, then on.ill irst-par October-par come-2pl again bank.ill then ater October the 1st you (should) come to the bank again ja=s saa-te ai noh (.) ˈkohe teh-akse teie-ga and get-2pl ptcl ptcl at.once do-iprs.prs 2pl-com and you’ll get, well (they) make laenu-ˈleping. loan.gen-contract a loan contract with you’

First (lines 1–3) the bank clerk talks about sureties in 3pl (nende ‘their’, tulevad ‘they come’, teevad ‘they make’, annavad ‘they give’). In line 6 the clerk uses the impersonal form with corporative reading tehakse ‘one makes’, referring to the bank’s representative(s). Both referents are unspeciic because the discussion revolves around future events and it is not known yet who exactly will stand surety in the case, or which clerk will be there to formalise the agreement. Since it is customary in Estonian to use the polite 2pl pronoun to address an unfamiliar interlocutor, the 2pl verb forms in lines 4–6 (teil ‘you.ade’, tulete ‘you’ll come’, saate ‘you’ll get’, teiega ‘with you’) clearly refer to the client. In Example (16) the participants of the activity are irst referred to by means of impersonal forms and then by the pronoun kõik ‘all’, which is a device typically used to refer to a vague group.

63

64

Renate Pajusalu

(16) (everyday conversation) 1

2

3

4

5

6

ee sinna too-di jalg-ˈratas, ja ja= sis= kui ee to.there bring-iprs.pst leg-wheel and and then when um a bike was brought there and and when see Haaveli kord ol-i ˈsõit-a ratta-ga, siis this Haavel.gen turn be-pst.3sg ride-inf wheel-com then it was Haavel’s turn to ride it, then well noh tõste-ti ta ˈratta selga sest ta ˈise ptcl lit-iprs.pst 3sg wheel.gen back.ill because 3sg self he was lited onto the bike because he was ol-i nii väike et ta ei ˈulata-nud= vist, be-pst.3sg so small that 3sg neg reach-ptcp probably so small that he couldn’t climb on to it I guess, ja= sis noh kõik arva-si-d et ta taha-b ˈära and then ptcl all think-pst-3pl that 3sg want-3sg away and then well everyone thought that he wanted sõit-a sealt ride- inf from.there to ride the bike away from there’

In this example, the groups evoked by the impersonal verb forms (toodi ‘was brought’ and tõsteti ‘was lited’, lines 1 and 3) and the pronoun kõik ‘all’ (line 5) are probably not identical but still play the same role in the conversation. he same applies in the case of diferent instances of kõik in referential chains generally: even though the group that the next kõik refers to may difer in terms of the individuals it is composed of, the group’s prominent role in the events of the narrative is maintained (Pajusalu 2008). As in the previous Example (14), in (16) the impersonal verb is used for the irst mention of the chain, which is then made more speciic by switching to pronouns. Yet while in Example (14) the referent is drawn into sharper focus by the use of the pronoun nad ‘they’, in Example (16) it is maintained by means of the pronoun kõik, which informs the listener that the referent is the whole group. Torn-Leesik and Vihman (2010) also report that the impersonal verb may refer to a single agent. It is, indeed, possible – yet by default the reference made by impersonal verbs points to a plural referent. A case in point is Example (17), which initially gives the reader the impression of evoking a group referent. (17) (everyday conversation) 1

T: ˈtäna mei-le ˈteh-ti (.) sis ku me ˈReeda-ga today 1pl-ade do-iprs.pst then when 1pl Reet-com ‘today we were given … when we were singing together with Reet

Hidden subjects in conversation

2

3

4

5

6

7

kaheksi l-ˈlaul-si-me sis ˈteh-ti mei-le ˈnaeru two.of.us sing-pst-1pl then do-iprs.pst 1pl-ade laugh.gen then we were given a laughing ˈproov. (1.3) test test P: nae[ru-proov] laugh.gen-test laughing test? T: [et= see] ee ˈpuldi-poiss (0.4) -s= ˈtema: that this ee board.gen-boy then 3sg.long this um sound board guy then he took võtt-is mingi sellise (0.7) selle (0.5) ˈpõse take-pst.3sg some.gen dem.gen this.gen cheek.gen some kind of microphone that is held against the cheek juurde käi-va ˈmigrofoni siis ˈlaul-is mei-le to go-ptcp.gen microphone.gen then sing-pst.3sg 1pl-ade and then (he) sang to us.’

By using the impersonal verb tehti ‘one did’ twice on lines 1 and 2, the speaker seems to create a group referent. When, in line 4, P asks for more detailed information by repeating the new word naeruproov ‘laughing test’, in response (line 5) T already ills in the referent with the NP puldipoiss ‘the sound board guy’, following that with the 3sg pronoun tema ‘he’ and, in line 5, with the 3sg verb laulis ‘he sang’. he vague referent that was created by the impersonals in lines 1–2 turns out to be a speciic individual. Yet this transformation only occurs ater other devices (the NP puldipoiss) have been used to point to the referent. Examples (14), (16) and (17) in fact represent the same type of referential chain: one in which a vague referent is introduced into the narrative by an impersonal verb; ater that, the referent is speciied further by the pronoun kõik (as in Example 16) or by the plural or singular 3rd person pronoun (as in Examples 14 and 17 respectively). Sometimes – although, judging by the data analysed as part of this study, relatively rarely – the impersonal may refer to a group that also includes one of the participants of the conversation. In the case of such instances the referential model is the opposite of the one described in relation to Examples (14), (16) and (17). Unlike in the latter, the speaker-inclusive-group model starts with the use of a personal pronoun (in the examples given, the pronouns meie ‘we’ or teie ‘you’) ater which an impersonal verb is used as the referential device. hus, we see that in order to refer to a group that includes a participant of the speech act the impersonal verb must be assisted by a more potent referential device – for instance, a personal pronoun.

65

66

Renate Pajusalu

Torn-Leesik and Vihman (2010: 325–326) also provide examples in which, apparently out of pragmatic consideration, the speaker alternates the use of the impersonal form and the personal 1pl verb. Having started to tell a story in the 1pl form, the speaker then appears to wish to distance herself from the activity and starts referring to the same group (that also includes herself) with the impersonal form of the verb. In the conversation at the travel agency (Example 18), the impersonal alternates with the 2pl – the referent is created with the 2pl pronoun te, ater which the same hypothetical group is evoked by the use of the impersonal. In this case, too, the relevant model appears to be one that is opposite to the model use in the case of the 3rd person referents in Examples (14), (16), and (17). (18) (institutional conversation, travel agency) T: kõigepealt on siis=ää(0.9)ä< ˈkolmanda ˈpäeva> irst.of all be.3sg then=ee third.gen day.gen First of all then er on the third day 2 kolmanda-l ˈpäeva-l, s= kui te ole-te juba ˈkoha-le third-ade day-ade then when 2pl be-2pl already place-ade on the third day when you have arrived there 3 jõud-nud juba ˈöö [seal] Asireaale hotelli-s ˈära reach-ptcp already night there Asireal.gen hotel-ine ptcl already spent a night at the Asireale hotel 4 maga-nud, sleep-ptcp resting 5 K: [mhmh] mhmh 6 T: .hh ä ˈsiiss külasta-takse Taurˈmiina-t. hh then visit-iprs.prs Taormina-par hen you will visit Taormina 7 see on ˈlinn Etna ˈmäe (.) ˈläheda-l. this be.3sg town Etna mountain.gen near-ade that’s a city near Mount Etna 8 K: mhmh mhmh 9 T: kus on siss hästi säili-nud vana-kreeka ˈteater, where be.3sg then well preserve-ptcp old-Greek theater Where there’s a well-preserved Ancient Greek theatre, 10 ˈkatedraal, katedraali ˈväljak, vana-d ˈkiriku-d, cathedral cathedral.gen square old-pl church-pl cathedral, cathedral square, old churches,

1

Hidden subjects in conversation

11

12

paˈlee-d, mt=.hh ja ja ˈtutvu-takse ka Etna palace-pl and and look-iprs-prs also Etna.gen palaces … and and you will also see the volcano, Mount Etna, vulˈkaani-ga, .hh ˈmil ˈmoel või kui palju ˈtutvu-takse volcano-com hh how way or how much look-iprs.prs as to how exactly of how much you will see

In Example (18)5 the client has asked the travel agent to give him or her more details about a future trip to Sicily. In line 2, the pronoun te refers to the group the client will probably be travelling with. In line 6, however, the impersonal külastatakse ‘one visits / will visit’ is used to refer to the same hypothetical group, followed by tutvutakse ‘one will familiarise oneself ’ in lines 11 and 12. he conversation then carries on, featuring a number of other impersonal forms. In conclusion one could say that the impersonal is most frequently used to construct referential chains in narratives where it functions as a tracking device of vaguely deined speaker-exclusive groups. Combined with other referential forms, impersonals may participate in creating speaker-inclusive and/or less vague group referents. Although impersonal verb forms do possess a certain referential ability, this remains closely connected to the ongoing activity.

7.

Personless conditional and impersonal in the same referential frame

As can be seen in the sections above, the personless conditional and the impersonal have diferent contexts of use: the conditional is used to plan, the impersonal to tell a story. Although sometimes the same conversation may involve both a planning and a story-telling activity, these activities tend to form their own selfcontained conversational units and they do not, as a rule, create shared indexical ields. here are, however, also situations in which these activities create a common indexical ield. Example (13) above contains utterances from an extensive lecture on potatoes. he majority of utterances contain impersonal verbs and the indexical ground for the whole lecture is generic. he same lecture also contains an utterance (19) in which the personless conditional is used.

5. When Example (18) is compared to Marja-Liisa Helasvuo’s Finnish example (Helasvuo 2006: 243–244, Example 9), it becomes clear that in both languages the speakers talk about a future activity, but the referent is diferent: in Estonian the impersonal is by default exclusive, in Finnish the passive may also be inclusive.

67

68

Renate Pajusalu

(19) noh= et ei pea-ks kartuli-koorimis-noa-ga seal ptcl that neg must-cond potato-peeling-knife-com there ‘without having to stick the knife in’

In this turn, a personless conditional verb is used to express the reason why one type of potato should be considered better than the other – the better one is the one that one could peel more easily. he referent is as generic as in the whole conversation (13) because the reference is to any consumer of potatoes. Personless conditionals and the impersonals can successfully co-exist also in other types of context. Example (20) is about a meeting to discuss insurance fraud. (20) (institutional conversation, a business meeting) 1

2

3

4

S: pea-ks oma dedektiivi palka-ma ne-id ˈsaa-b must-cond own detective hire-sup 3pl-par can-3sg I/we/one should hire my/our/one’s own private detective, it is possible juba palga-ta. (1.0) already hire-inf to hire them already A: noh eks nee-d siin ˈteh-akse ju koos-töö-d ptcl ptcl this-pl here do-iprs.prs ptcl together-work-par Well now, these here, one cooperates nii ˈmõne-gi-ga= ja. so some-ptcl-com and with a number of them and

One of the speakers suggests hiring a private detective (peaks oma detektiivi palkama ‘schould hire own detective’). his is a typical personless suggestion that leaves unspeciied who should implement the suggestion, but implies that it should be someone from the irm whose representatives are engaged in the discussion. In line 3, another participant of the meeting says that the irm is already cooperating with detectives (tehakse koostööd ‘one cooperates’), probably referring to the same vague corporate group referent as the suggestion itself. he conditional and the impersonal both support the same corporate referent, leaving it unspeciied on the individual level. Personless conditionals and impersonals may be used together to support an unspeciic group referent of a narrative, as in Example (21). (21) (everyday conversation) 1

esimese-l purustamise-l jä-i siiski sammas nii irst-ade wrecking-ade stay-pst still column so ‘ater the irst wrecking the monument still remained suiciently

Hidden subjects in conversation

2

3

4

5

6

7

8

’terve-ks= et saksa= aja-ga ˈtaasta-ti, ja ta whole-trnsl that german time-com restore-iprs.pst and 3sg intact for it to be restored during the German period and it ol-i ˈüleva-l veel mitu aasta-t, aga ˈviiekümnenda-l be-pst up-ade more several year-par but ity.gen-ade remained there for several more years, but in 1950 aasta-l lõha-ti ta ˈnii, et jä-i-d ainult year-ade blow-iprs.pst 3sg so that stay-pst-3pl only it was blown up such that only ˈtüki-d järele. me ˈleid-si-me= ned ˈtüki-d siit peace-pl let 1pl ind-pst-1pl this.pl piece-pl from.here pieces of it were let. We found these pieces here surnu-aia ˈtagant, aga ˈhoopis-ki ei= ol-nd nii dead-garden from.back but quite-ptcl neg be-ptcp so behind the cemetery. But there weren’t at all ne-id ˈpiisavalt, et ole-ks saa-nd se-da this-pl.par enough that be-cond can-ptcp this-par enough peaces that one could have uuesti ˈkokku pan-na. again together put-inf reassembled it again’

Example (21) presents an excerpt of a conversation about how a monument was restored (taastati, iprs, line 2) and then blown up (lõhati, iprs, line 4) such that it could not be reassembled again (ei … oleks saand uuesti kokku panna, cond, line 7–8). Although there is a 1pl pronoun me ‘we’ in line 3, it does not create a strong speciic referent and the conditional form in line 7 continues to support the unspeciic group referent of the narrative.

8. Conclusion he Estonian personless conditional and impersonal resemble each other in terms of their referential function: they both introduce an unspeciic group referent into the referential frame of the conversation and they both lack an explicit subject. Although the unspeciic group referent is a weak constituent of the referential frame in both cases, it can still be used to support a referent. here are contexts in which the low of conversation shows the personless conditional and the impersonal to possess the ability to introduce a discourse referent and to maintain it. While in the case of the personless conditional the event itself is unreal (as is generally the case with the conditional in Estonian), in the case of the impersonal

69

70

Renate Pajusalu

it is, as a rule, real and speciic or generic. his diference in meaning accounts for the fact that they are typically used in diferent referential practices. Although it is not the aim of this article to give a quantitative analysis, one cannot help noticing that in the corpus used for this study, the impersonal occurs rather frequently in everyday conversations and is relatively rare in institutional conversations which involve service contexts. In this respect the situation is complementary to that of the conditional forms discussed above, which occur more frequently in service conversations. hus, despite the fact that both forms are personless, they have diferent contexts of use: the personless conditional is preferred when the aim of the conversation is to plan a future activity, while the existential impersonal is irst and foremost used in narratives. Situations in which the impersonal form refers to an institutional group that has an existence independent of the conversation – i.e., impersonals with the so-called corporate interpretation – should probably be regarded as a separate group. It is likely that, regardless of the aim of the conversation, this type of impersonal occurs invariably whenever corporate referents are evoked. Although the personless conditional and the impersonal may be used to point to an identical group referent, their typical referents still difer. he personless conditional is typically inclusive, incorporating one or both of the interlocutors into the group. he impersonal is typically speaker-exclusive and refers to a group that does not include the participants of the conversation. his diference is also relected in the respective referential chains. he group referent of the personless conditional is oten created on the basis of the 1st person reference. A typical referential chain is as follows: irst, one of the participants is mentioned by a personal form, ater which the personless conditional is used to expand the referent into a group. In the case of the impersonal, the referent is in general created with a verb in the impersonal and later speciied (or just maintained) by means of a pronoun. When, however, an impersonal form is intended as inclusive, it is usually preceded by an inclusion-marking pronoun and the form then elaborates some pragmatic aspect of the referent that has been created with the personal reference. hus, personless verbs and their imaginary agents are an integral part of the referential mechanism similarly to pronouns and NPs. hey are not as strong in terms of their referential power as pronouns and NPs are, and they do not create as clear a referent as pronouns do. Yet, they have a distinct role in contributing to the continuity of conversations.

Hidden subjects in conversation

References Ariel, Mira (1996). Accessing noun-phrase antecedents. London and New York: Routledge. Chafe, Wallace (1994). Discourse, consciousness and time. he low and displacement of conscious experience in speaking and writing. Chicago: he University of Chicago Press. Cornish, Francis (1999). Anaphora, discourse, and understanding. Evidence from English and French. Oxford: Clarendon Press. Du Bois, John W. (1980). Beyond deiniteness: he trace of identity in discourse. In Wallace L. Chafe (Ed.), he pear stories. Cognitive, cultural, and linguistic aspects of narrative production (pp. 203–274). Norwood: Ablex. Erelt, Mati, Kasik, Reet, Metslang, Helle, Rajandi, Henno, Ross, Kristiina, Saari, Henn, Tael, Kaja, & Vare, Silvi (1993). Eesti keele grammatika II. Tallinn: Eesti Teaduste Akadeemia Eesti Keele Instituut. Etelämäki, Marja (2006). Toiminta ja tarkoite. Helsinki: Finnish Literature Society. Givón, T. (2005). Context as other minds. he pragmatics of sociality, cognition and communication. Amsterdam: John Benjamins. DOI: 10.1075/z.130 Gundel, Jeanette, Hedberg, Nancy & Zacharski, Ron (1993). Cognitive status and the form of referring expressions in discourse. Language, 69, 274–307. DOI: 10.2307/416535 Gundel, Jeanette, Bassene, Mamadou, Gordon, Bryan, Humnick, Linda, & Khalfaoui, Amel (2010). Testing predictions of the Givenness Hierarchy framework: A crosslinguistic investigation. Journal of Pragmatics, 42, 1770–1785. DOI: 10.1016/j.pragma.2009.09.010 Hanks, William F. (1990). Referential practice: Language and lived space among the Maya. Chicago and London: he University of Chicago Press. Hanks, William F. (1996). Language & communicative practices. Oxford and Boulder: Westview Press. Hakulinen, Auli, Vilkuna, Maria, Korhonen, Riitta, Koivisto, Vesa, Heinonen, Tarja-Riitta, & Alho, Irja (2004). Iso suomen kielioppi. Helsinki: Finnish Literature Society. Helasvuo, Marja-Liisa (2006). Passive personal or impersonal? A Finnish perspective. In MarjaLiisa Helasvuo, & Lyle Campbell (Eds.), Grammar from the human perspective: Case space and person in Finnish (pp. 233–255). Amsterdam: John Benjamins. DOI: 10.1075/cilt.277.16hel Hennoste, Tiit (2003). Suulise eesti keele uurimine: Korpus. Keel ja Kirjandus, 7, 481–500. Kaiser, Elsi (2010). Efects of contrast on referential form: Investigating the distinction of strong and weak pronouns. Discourse Processes, 47, 480–509. DOI: 10.1080/01638530903347643 Kauppinen, Anneli (1998). Puhekuviot: Tilanteen ja rakenteen liitto. Helsinki: Finnish Literature Society. Keevallik, Leelo (2004). Politeness in Estonia: A matter of fact style. In Leo Hickley (Ed.), Politeness in Europe (pp. 203–217). Clevedon, England: Multilingual matters. Lambrecht, Knud (1994). Information structure and sentence form. Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511620607 Laury, Ritva (1999). Deiniteness. In Jef Verschueren, Jan-Ola Östman, Jan Blommaert, & Chris Bulcaen (Eds.), Handbook of pragmatics. 1999 installment. Amsterdam: John Benjamins. Laury, Ritva (2005). First and only: Single mention pronouns in spoken Finnish. In Ritva Laury (Ed.), Minimal reference. he use of pronouns in Finnish and Estonian discourse (pp. 56–74). Helsinki: Finnish Literature Society.

71

72

Renate Pajusalu

Lindström, Liina (2009). Kõnelejale ja kuulajale viitamise vältimise strateegiaid eesti keeles. Emakeele Seltsi aastaraamat, 55, 88–118. Lindström, Liina, & Vihman, Virve (2010). Ise ise. Journal of Estonian and Finno-Ugric Linguistics, 1, 219–241. Metslang, Helle (1999). Is the Estonian and Finnish conditional actually a conditional? In Mati Erelt (Ed.), Estonian: Typological studies III (pp. 97–127). Publications of the Department of Estonian of the University of Tartu. Tartu: Tartu University Press. Pajusalu, Renate (2008). Pragmatics of quantiiers: he case of Estonian kõik ’all. Journal of Pragmatics, 40(11), 1950–1965. DOI: 10.1016/j.pragma.2008.04.006 Pajusalu, Renate (2009). Pronouns and reference in Estonian. Sprachtypologie und Universalienforschung, 62(1/2), 122–139. Pajusalu, Renate, & Pajusalu, Karl (2004). he conditional in everyday Estonian: Its form and functions. Linguistica Uralica, 4, 257–269. Pajusalu, Renate, & Pajusalu, Karl (2005). Ways of revealing and concealing person: he conditional in Estonian conversation. In D. Monticelli, R. Pajusalu, & A. Treikelder (Eds.), Regards multidisciplinaires sur la deixis (pp. 67–79). Tartu: Tartu University Press. Pajusalu, Renate, Vihman, Virve, Birute, Klaas, & Pajusalu, Karl (2010). Forms of address across languages: Formal and informal second person pronoun usage among Estonia’s linguistic communities. Intercultural Pragmatics, 7(1), 75–101. DOI: 10.1515/iprg.2010.004 Pool, Raili (1999). About the use of diferent forms of the irst and second person singular pronouns in Estonian cases. In Mati Erelt (Ed.), Estonian typological studies III, 11 (pp. 158–184). Publications of the Department of Estonian of the University of Tartu. Tartu: University of Tartu. Shore, Susanna (1986). Onko suomessa passiivia. Helsinki: Finnish Literature Society. Shore, Susanna (1988). On the so-called Finnish Passive. Word, 39, 151–176. Torn, Reeli (2002). he status of the passive in English and Estonian. In H. Hendriks (Ed.), RCEAL working papers in English and applied linguistics 7 (pp. 81–106). Cambridge: Research Centre for English and Applied Linguistics. Torn-Leesik, Reeli (2009). he voice system of Estonian. Sprachtypologie und Universalienforschung, 62(1/2), 71–90. Torn-Leesik, Reeli, & Vihman, Virve (2010). he uses of impersonals in spoken Estonian. SKY Journal of Linguistics, 23 ed. by Mark Kaunisto, Rea Peltola, Erika Sandman, Heli Tissari, & Katja Västi, 301–343.

Subjects under generic conditions Implied subjects in Finnish and Estonian if-clauses Hanna Jokela and Helen Plado University of Tartu / University of Turku

he chapter discusses implied subjects creating open or generic reference in Finnish and Estonian if-clauses from a contrastive perspective. We focus on the so-called zero person constructions, passives/impersonals, and the Estonian da-ininitive which acts as a predicate of an if-clause. hese construction types lack subject; in addition they can create open or generic reference. he constructions are closely interrelated; however, each of them has some restrictions in use and diferences in interpretation. he passive in Finnish and the impersonal in Estonian if-clauses have a tendency to refer to a group, whereas the zero person and Estonian da-ininitive refer to one individual. he interpretation of implied subject depends on the conditional clause type, the context, and the verb form that is used.

1.

Introduction

his paper discusses open, non-speciic, and generic reference forms in conditional clauses in Finnish and in Estonian.1 We approach this question from the perspective of the particular constructions used to convey open and non-speciic reference. hese constructions include the impersonal (Estonian), the passive (Finnish), and the zero person construction (both languages). he ininitival construction (the da-ininitive) that Estonian allows as a predicate of conditional clauses is also included. In addition, other possibilities to express generic and open reference in Finnish and in Estonian are investigated. We assume that the zero construction, the da-ininitive, and the impersonal forms can express open and generic reference. We are particularly interested in ifclauses with implied subjects and their possibilities to express generic conditions. 1. his study has been supported by the Academy of Finland (Subject expression in Finnish and other Finnic languages: Interactional and Cognitive Perspectives). doi 10.1075/cal.16.04jok © 2015 John Benjamins Publishing Company

74

Hanna Jokela and Helen Plado

In Finnish, the zero construction can typically occur in if-clauses (Example 12). If-clauses are also suitable contexts for the Estonian da-ininitive as a predicate (Example 2). Neither the zero construction nor the da-ininitive construction has prototypical subjects (see 4.2 and 4.3). (1) Finnish:

Jos syö liikaa, voi lihoa. If eat.3sg too.much may.3sg gain.weight:inf ‘If one eats too much, one may gain weight.’

(2) Estonian: Kui süüa liiga palju, võib juurde võtta. If eat:inf too much, may:3sg weight gain:inf ‘If one eats too much, one may gain weight.’

Implicit personal reference and subjectless expressions are considered typical in Finnish (Laitinen 1995). Interestingly, for instance, translation studies show that texts of translated Finnish express person more explicitly, and personal pronouns are used, whereas in non-translated Finnish, implicit personal reference and nonspeciic forms, such as the zero person constructions, are preferred (Mauranen & Tiittula 2005). he person systems of Finnish and Estonian are almost identical, but persons might be expressed diferently in diferent contexts and constructions. Diferent means may be used when referring to speech-act participants, or alternately, to open referents. Comparing the Finnish and Estonian systems will shed light on possibilities to express generic or open reference in both languages. he zero construction, the passive, and the da-ininitive are closely interrelated, and yet each of them seems to have some restrictions in use and diferences in interpretation. Both the zero person and the passive may be translated into Estonian with the da-ininitive or the impersonal. Our aim is to contrast diferent possibilities for expressing open and generic reference and to analyze the possible motivations for choosing diferent constructions and forms in certain contexts. Our focus is on the following questions: What is the diference between the zero person and the passive in Finnish, and is there a similar diference between the da-ininitive and the impersonal in Estonian? What afects the degree of generality in expressions containing diferent implied subjects? 1.1

Data

he article is based on written language. For the analysis we use two diferent data sets – corpus data and translated data. Corpus data is drawn from the 1990s subcorpora of the Corpus of Written Estonian3 (totaling 1,467,000 textual words). 2. Examples (1) and (2) have been composed by the authors of this article. 3. http://www.cl.ut.ee/korpused/

Subjects under generic conditions

.

Altogether there were 3,319 conditional clauses marked with the conjunction kui; in 258 conditional clauses the predicate was the da-ininitive, and there were 113 generic conditional clauses in which the predicate was in the impersonal form. he Finnish and translated data come from articles that have been published in Finnish magazines and translated into Estonian. Articles concern cooking, gardening, and similar subjects. Authors of the articles give readers instructions and advice, and texts of this kind seem dialogical: writers refer to (non-speciic) readers, themselves, or anyone. he translated data consist of 139 Finnish if-clauses4 and their Estonian translations. he translated data are based on 1,280 Finnish clauses with the zero person construction and their Estonian counterparts. Comparing the translations reveals diferent correspondences in similar languages and shows how similar languages use their resources and possibilities diferently. Because the interpretation of diferent forms depends greatly on the context, the immediate context of the zero person clause or its counterparts is included. 1.2

Conditional markers

In Estonian, conditional clauses can be formed either by using the conjunction kui ‘if ’, as in (3), or by placing the predicate in an initial position, as in (4) (see, e.g., EKG II: 308). Conditional clauses with the conjunction kui occur more frequently than conditional clauses that begin with a verb. In type 2, the predicate is always in a inite form. hat is, the da-ininitive is not possible in this type. (3) Eile rääkis ta mulle, et hoiab alles mehe kallima saadetud kirju, et kui lahutuseks läheb, jääb mees süüdi. when/if divorce:trnsl go:3sg stay:3sg man guilty ‘Yesterday she told me that she would keep the letters sent by her husband’s sweetheart so that if they had to divorce, then the husband would take the blame.’ (Est; ILU1990) (4) Lähed vasakule, saad surematuks. go:2sg let:all get:2sg immortal:trnsl ‘If you turn let, you will become immortal.’

(Est; ILU1980)

4. In this translated data, there are 84 if-clauses with the Finnish zero person construction and their Estonian translations. All the if-clauses with the Finnish zero person and their Estonian counterparts that occurred in the articles were included, and additional 55 generic if-clauses with implied subjects were collected from the same articles. In these 55 clauses there were constructions with the generic or open reference other than zero person (for example, passive). he data collected from the magazines is not extensive, but it gives us an opportunity to compare diferent means of expressing open and generic reference and diferences between them.

75

76

Hanna Jokela and Helen Plado

In Finnish, conditional clauses are typically marked with the conjunction jos (5), but also with the conjunction kun (6), which is prototypically a temporal conjunction. In Finnish, there are also the negative conditional conjunctions ellei and jollei, which have the same personal endings as the negation word (see Section 4), e.g. ellen ‘if I do not’, ellet ‘if you do not’ (VISK § 1134). (5) Jos neuvottelut onnistuvat on jalkamme If negotiation:pl be.successful:3pl be.3sg foot:1pl.poss eurooppalaisen kaupankäynnin European:gen trade:gen ja yhteistoiminnan oven välissä. and cooperation:gen door:gen between ‘If the negotiations are successful, we’ll get a foot in the door with European trade and cooperation.’ (Fin; LAY) (6) Maidonjalostajat säästäisivät, kun tavaraa dairy.manufacturer:pl save:cond:3pl when product:par ei tarvitsisi ajaa kauppaan yhtä usein kuin nyt. not have.to:cond drive:inf shop:ill same oten as now ‘Dairy manufacturers would make a savings if they didn’t have to distribute products to shops as oten as they do now.’ (Fin; LAY)

Estonian kui and Finnish kun share the same origin – the pronominal stem ku(Herlin 1998: 21). Kui and kun are both polysemous conjunctions.

2. Conditional clauses Traditional descriptions of Estonian adverbial clauses usually classify them according to their semantic function (for example, conditional clauses, concessive clauses, and causal clauses). In addition, an alternative approach is taken here based on their function in the communication situation and the function they have regarding the main clause. According to Sweetser (1990), there are three types of conditional clauses: content, epistemic, and speech-act conditionals. Content conditionals are prototypical conditional clauses. hey specify the content of the main clause; the conditional clause sets conditions necessary for the realization of the state of afairs or event described in the main clause. Epistemic conditional clauses (7) deliver reasons for the conclusion made in the main clause; if one knows that the content of the conditional clause is true, then (s)he can conclude (based on her/his knowledge) that the content of the main clause is also true. We add metalinguistic conditionals to the speech-act conditionals and call them conversational conditionals. Conversational conditional clauses express conditions

Subjects under generic conditions

of appropriateness and do not involve any real-world dependence between the contents of their antecedents and consequents (Dancygier 2006: 105). Speech-act conditionals (8) make the utterance of the main clause relevant in the communicative situation. Metalinguistic conditional clauses (9) comment on the appropriateness of some expression or word used in the main clause. (7) Aken peab lahti olema, kui nii vali müra kuulda on. window must:3sg open be:inf when/if so loud noise hear:inf is ‘If you can hear the noise so well, the window must be open.’ (Est; ILU1990) (8) See on alatu, kui teada tahad! this is mean when/if know:inf want:2sg ‘hat is mean, if you want to know my opinion.’

(Est; ILU1990)

(9) Ma olen lugejale võlgu autobiograaia ja I be:1sg reader:all debt autobiography.gen and autokarakteristika (kui sellist sõna on). autocharacteristics.gen when/if such:par word.par be.3sg ‘I owe the reader my autobiography and autocharacteristics (if there is such a word).’ (Est; ILU1990)

3. Genericity he use of the zero construction is considered a conventional way in Finnish to construe generic statements concerning human beings (Laitinen 2006). he Estonian da-ininitive does not indicate any person, tense, or mood, and therefore it could easily be regarded as a natural choice to build generic statements. In this section, we introduce the main features of genericity. here can be genericity on the diferent levels of language, as Behrens (2005) clearly demonstrates. Behrens (2005: 288–289) discusses the distinction between “generic noun phrase” and “generic sentence” and also introduces a third level of linguistic description as being relevant for genericity, the text level. In this article, our focus is not on generic noun phrases, but we deal mainly with the implied subject of the generic if-clauses. According to Behrens (2005: 275) a classic generic sentence contains a kindreferring noun phrase as its topic and a characterizing predicate that expresses a time-stable and prototypical (but not necessarily essential) property of the topic. Similarly, Langacker (1997: 191) describes generic expression as the one that ascribes a general property to all members of a class. Behrens focuses mainly on generic noun phrases, but we ind some of her comments also applicable to generic person reference. Generic interpretation typically results from the interaction

77

78

Hanna Jokela and Helen Plado

of a number of variable factors such as the lexical semantics of the constituting elements, pragmatic knowledge, and discourse situation, as well as grammatical marking of tense, aspect, and mood on the predicates, and so on (Behrens 2005: 276). However, Dahl (1975: 99–100) claims that the common semantic property of all generic expressions is that they are used to express law-like, or gnomic, statements. He inds it important to diferentiate between two types of statements: accidental generalizations and gnomic or law-like statements. In an accidental generalization, one speaks only about actual cases, whereas gnomic statements also concern possible, non-actual cases. According to Cardinaletti and Starke (1999: 9), generic reference precludes speciic time reference. Bybee et al. (1994: 126) point out that the present is what includes ongoing activities and habitual situations, and it is typically used for gnomic situations, those that apply to generic subjects and basically hold for all time, such as Dogs pant to cool of. It is also important to point out that generics are closely related to habituals. Langacker (1997: 194) demonstrates that it is artiicial to draw a sharp distinction between generics and habituals. For one thing, generics vary in their degree of generality, and generic normally implies habituality on the part of any single individual. For that reason we do not draw a sharp distinction between generics and habituals either. In short, genericity may be deined in many diferent ways. For us, the most important feature of generic clauses is that they do not refer to any concrete event, person, or time. We do not draw a sharp distinction between generic and open ifclauses with implied subjects studied here. We are interested in the use of implied subjects in if-clauses, as the if-clause seems to provide a frame that puts implied subjects under generic conditions. However, implied subjects may be interpreted as not only generic and open but also more speciic. Implied subjects may refer to the speaker, the addressee, or anyone in general, and their interpretation is construed in a particular context. herefore, implied subjects might be generic, but they do not automatically make expressions generic. We argue that generic or non-generic interpretation oten depends on the context. he interaction between the implied subject and its context is bidirectional. It can be argued that some contexts are inherently more generic than others. For example, our data includes magazine articles with recipes. hese recipes provide instructions and directions that apply to anyone (who is interested in reading them).

Subjects under generic conditions

4. he zero person, the da-ininitive, and the passive as constructions In both Finnish and Estonian, the category of person is expressed with personal pronouns (see also Pajusalu’s chapter in this volume) and personal endings on verbs. In Finnish, person may also be expressed with possessive suixes (for example, autoni ‘my car’). In Estonian, there are no possessive suixes. Both Finnish and Estonian may be considered “pro-drop” languages: personal pronouns as subjects are not obligatory, as the personal ending of the verb alone conveys the person referred to (see more about pro-drop in Estonian and Finnish in Duvallon & Chalvin 2004). In both languages, the predicate verb agrees with the subject in person and number. In standard language, there is more agreement in Finnish than in Estonian (see Metslang 1994: 216). In the negative forms of verbs, the Finnish negation word ei has personal endings depending on the person of the subject (en, et, ei, emme, ette, eivät). he negative forms in Estonian lack personal endings, and the negation word ei is used with all persons. It is quite common in Estonian negation that there is no overt expression of person, and the implied subject has to be construed solely on the basis of the context (see also Lindström 2010). Compared to Finnish, Estonian has additional forms lacking personal endings. Namely, in the conditional mood, verbs can be used with conditional markers only, without personal endings. In the case of conditional verb forms without personal ending, known as short conditional forms, sometimes the referent can be inferred only from the context. Pajusalu (Pajusalu’s chapter in this volume) introduces this phenomenon. Technically, the short conditional form and the zero person constructions in conditional mood overlap, and each has a context-based interpretation. 4.1

he zero person construction

Along with the passive, the zero person has been treated as a non-speciic member of the personal system in Finnish (cf. Hakulinen 1987; Helasvuo & Laitinen 2006). Helasvuo and Laitinen (2006) discuss the zero person construction and the passive in Finnish as grammatical forms that create open reference. he most extensive grammar description of Estonian, Eesti keele grammatika (referred to as Estonian Grammar or EKG), does not represent the impersonal or the zero construction as members of the personal system, but instead discusses them as generic and impersonal forms (EKG II: 227–228). Estonian Grammar (EKG II: 227–228) introduces generic clauses (üldisikulised laused) and impersonal clauses (umbisikulised laused), which all lack explicit subjects (see also Pajusalu’s chapter). Generic clauses describe situations in which the

79

80

Hanna Jokela and Helen Plado

potential agent may be any person. In a generic clause, the predicate verb is usually in the singular third person, impersonal, or periphrastic passive5 (EKG II: 227). Estonian Grammar (EKG II: 227) notes that the singular second person may also occur in generic clauses. In the zero person construction, there is no overt subject, and the verb is in the third person singular form: in Finnish VV (lukee ‘reads’) and in Estonian Vb (loeb ‘reads’). he Finnish zero person is translated into English using the generic pronouns you or one. Lölund (1998) has compared the zero person construction in Finnish to the man-construction in Swedish. Helasvuo and Johansson (2008) contrast Finnish zero person construction and passive to French on-construction and passive; both the zero person construction and on-construction are third person forms, but in the Finnish construction there is no overt subject. Helasvuo and Johansson argue that the French passive and the Finnish zero person are both used in cases where some kind of generalization is being made. However, with the Finnish zero person construction, this generalization is represented as something relating to the personal experience of the participants (Helasvuo & Johansson 2008: 48). Jokela (2012) studies and compares the zero person in Finnish and Estonian. Her study sheds light on the zero person construction, its use, functions, and interpretation. he meaning of the zero is oten considered generic, although the zero may be interpreted as open so that the reference can be construed in the context. Laitinen (2006) shows that in Finnish the zero person construction is used in contexts where the reference of the zero is construed as a speech-act person. According to Helasvuo and Vilkuna (2008: 233) the zero construction combines genericity and the perspective of the speech-act participant. Vihman and Kaiser (2006) point out that in Estonian, the zero person construction is less studied than the impersonal. According to Vihman and Kaiser (2006: 129), the implicit argument is present on the semantic level in the zero person constructions and in the impersonal and passive constructions in both languages. he zero person construction is used in both Finnish and Estonian, but there are interesting and signiicant grammatical diferences and contextual constraints (for more discussion, see Jokela 2006, 2012).

5. In addition to impersonal, which is regarded as the basic voice construction in Estonian, one can also ind in Estonian the periphrastic stative passive. As the personal passive uses the same elements as the periphrastic tenses of the impersonal, it is oten hard, if not even impossible to draw a sharp distinction between the impersonal and passive. (See Torn-Leesik 2009.)

Subjects under generic conditions

4.2

he Finnish passive and Estonian impersonal

It is noteworthy that the terminology used for passive and impersonal reference forms is diferent in Estonian and in Finnish linguistics. he Finnish passive and the Estonian impersonal are treated diferently in grammatical studies and descriptions. Estonian studies see the impersonal merely as a device to conceal or hide the agent, whereas that is not the case in recent Finnish studies dealing with the passive. In Estonian, both transitive and intransitive verbs can be impersonalized. hus impersonalization does not change the valence of a main verb, and the object retains its object properties in the impersonal clause (Torn-Leesik 2009: 73–74). In Estonian, impersonalization is mainly restricted to verbs that take human actors as their subjects (with some exceptions) (Torn-Leesik 2009: 77–78). According to Estonian Grammar (EKG II: 227–228), in impersonal clauses there is a concrete (not potential) actor, but it is not expressed. However, Torn-Leesik and Vihman (2010) show based on spoken Estonian that impersonals can also have a universal and even hypothetical reading. In Finnish linguistics, the passive is generally labeled as passive, not impersonal. here are also other suggestions. In some studies the Finnish passive is treated as impersonal, and Blevins (2003) considers it non-passive. However, for example, Helasvuo and Johansson (2008: 31) ind it “rather an unfortunate terminological choice to call the Finnish passive impersonal.” According to them, passive belongs to a system that serves to express the discourse roles of the participants of the speech situation, and they prefer to speak of the non-speciic reference of the passive personal marker (for further discussion, see Helasvuo 2006). Helasvuo and Vilkuna (2008: 230) point out that the reference of the implied agent is let open and has to be inferred from the context. Tuomikoski (1971) describes Finnish passive as the fourth person in the personal system of three singular and three plural forms. Helasvuo and Laitinen (2006) emphasize the importance of the context where the passive occurs when interpreting its meaning and reference. hey illustrate that the passive is the most lexible member of the verbal person paradigm. In non-standard Finnish, the passive form is oten used with the irst person plural pronoun me ‘we’. 4.3

he da-ininitive

In the traditional sense ininitives, as well as the other non-inite forms, are not marked by categories such as person, number, tense, or mood, and they cannot function as the only predicate of an independent sentence (Koptevskaja-Tamm 1999: 146–147); typically ininitives are used as the argument of the predicate (Noonan 1985; Ylikoski 2003: 197). However, under some conditions an ininitive

81

82

Hanna Jokela and Helen Plado

can be the only predicate of the clause, even the only predicate of an independent clause (Erelt 1984; Nikolaeva 2007; Visapää 2008). In Estonian, the da-ininitive is marked with -a-, -da-, or -ta-suix. Ininitives mainly serve as subjects or objects in Estonian (Erelt 2003: 65), but on a par with these functions, they can also be used as the only predicate in some subordinate clauses, including conditional, purpose, and adversative comparison clauses. In the purpose6 and adversative comparison clause, the da-ininitive has the same actor as the main clause, but the actor of the conditional clause does not have to refer to the subject of the main clause (EKG II: 244–246). At least in if-clauses, the da-ininitive seems to have functions similar to those of the Finnish zero person constructions. In Finnish, the A-ininitive, which is equivalent to the Estonian da-ininitive, cannot occur as an independent predicate of the if-clause. However, there are some constructions that may have the A-ininitive as an independent predicate. In a study of the independent A-ininitive constructions in Finnish, Visapää (2008: 120) suggests that the ininitive marker “atemporalizes” the verbal process referred to with the A-ininitive construction.

5. Diferent possibilities for expressing generic conditionals he Estonian non-inite da-ininitive is formally not marked for person or tense, and it seems suitable for the generic conditionals. However, there are also other possibilities to express open person reference in generic conditionals. We contrast the da-ininitive with the zero person, the passive, and the impersonal. Actually, it is possible to express generic meaning by all person forms both in the singular and the plural. Even the irst and the second person forms and generically used noun phrases may be used in both Finnish and Estonian to create statements with generic or open reference. In this section, there are some examples drawn from previous research and/or our data. 5.1

Generic person forms

5.1.1 First person Helasvuo (2008) illustrates that both the irst and second person singular forms can be used to create open reference in Finnish. In her article, Helasvuo demonstrates through examples how the interpretation of a reference as open or anchored

6. For a more detailed discussion about the actor of the Estonian purpose clause, see Plado (2013).

Subjects under generic conditions

to a speciic referent occurs in context. Helasvuo’s examples (2008: 192–201) show that the interpretation may be triggered by contextual cues, such as grammatical forms, syntactical relations, and lexical cues. When contrasted to other means of open reference, the reference of the irst person singular is distributive, whereas the passive is interpreted as collective (Helasvuo 2008: 198). here seems to be a similar diference between the open irst person singular and the irst person plural; the open irst person is distributive, and the irst person plural creates reference which implies a group, us (Helasvuo 2008: 201). Our data does not provide any examples of Finnish generic irst person singular, but in Example (10) there is generic irst person singular in Estonian. (10) Eestlane tahab kindla peale välja minna kui tõmban marihuaanat, when/if smoke:1sg marijuana:par siis peab kindla peale lugu tulema. then must:3sg for.certain story come:inf ‘Estonians want to play it safe – so if you smoke pot, you’ll deinitely make up a song.’ (Est; AJA1990)

In Example (10) the generic interpretation of the pronoun of the irst person singular is conirmed by the previous clause – it shows that the following holds for all Estonians. Obviously, the irst person plural may represent a open and non-speciic reference in both languages: it implies a group that includes the speaker, but other members of the group may remain unknown. In Finnish, the passive form is also frequently used when talking about us, and in colloquial Finnish the passive verb form may be used even with the plural irst person pronoun me ‘we’ (see VISK § 1326). In Estonian the irst person plural is used instead.

Second person 5.1.2 he generic use of the second person singular is more widespread in Estonian7 than in Finnish (see also 6.1). In Finnish, the generic use of the second person singular is mostly considered a foreign-language inluence, but it has also been used in dialects (Seppänen 2000). Helasvuo (2008: 190) argues that the second person singular typically creates open reference. Example (11) represents Estonian and Example (12) the Finnish if-clause with the generic second person singular.

7. However, Estonian schools teach that second person singular with generic reference should not be used. It is considered a part of narrative texts and is treated as a mistake in all other kinds of texts (EKRE).

83

84

Hanna Jokela and Helen Plado

(11) Kui lähed mööda rabaäärset teed kaugemale when/if go:2sg along bogside:par road:par further:all maa poole, land.gen towards näed võsa vahel kerkimas uusi müüre. see:2sg bush between rise:inf:ine new:pl.par wall.pl.par ‘If you follow the path along the bog farther inland, you’ll see new walls rising in the bushes.’ (Est; ILU1990) (12) Jos päätät automatkailla enemmänkin, if decide:2sg make.car.trips:inf even.more pitemmät automatkat kannattaa minimoida long:cmpr.pl car.trip:pl be.worth:3sg minimize:inf varaamalla puolimatkan majoituksia book:inf:ade halfway:gen accommodation:pl:par pikkukaupungeista. small town:pl.ela ‘If you decide to make more car trips, it’s worth minimizing longer car trips by booking accommodation in small towns at the halfway point.’ (Fin; TR2/04/120)

5.1.3 hird person In the zero person construction, there is no overt subject, and the verb is in the third person singular form. he zero person is discussed in Section 4.1 (along with the passive/impersonal). In both languages, the third person plural forms can be used to refer to a non-speciic group. he Estonian Grammar labels this kind of use of the person plural form a subjectless passive. (Räägivad, et tänavu tuleb külm talv ‘hey say that winter is going to be cold this year’) (EKG II: 30–31). It seems possible to use the pronoun keegi (Est: ‘someone’) and joku (Fin: ‘someone’) generically, but the generic use of third person singular pronouns is rare. Generically interpreted clauses with the third person predicates rarely have pronouns as their subjects, whereas noun phrases can occur as subjects in generic clauses. 5.2

Generic noun phrases

Noun phrases may be used to build generic statements. In the singular, the interpretation of the particular noun phrase is more generic; plural implies merely a non-speciic group. In Examples (13) and (14) there are generically used nouns in the singular. In (13) the Estonian noun is laps ‘child’ and the generic statement applies to any child. Likewise, in (14) the eating habits of Portugese are discussed in a generalizing manner.

Subjects under generic conditions

(13) Kui laps saab maiustusi vajadusest rohkem, when/if child get:3sg sweet:pl.par need:ela more keeldub ta muust toidust, milles on organismi refuse:3sg s/he other:ela food:ela arenguks tähtsamaid aineid (valku, mineraale, vitamiine). ‘If children are given more sweets than they need, they’ll refuse other food containing the nutrients that are important for the body (li pids, minerals, vitamins).’ (Est; AJA1990) (14) Jos portugalilaiselta jää keitto väliin, käytännössä if Portuguese:abl miss.3sg soup in.practice:ine hän ei ole mielestään syönyt mitään. s/he neg.3sg have opinion:3sg.poss eat:ptcp anything ‘If a Portuguese misses her/his soup, s/he feels like s/he has had nothing to eat at all.’ (Fin; TR2/04/121)

6. Interpretation of the implied subjects In Section 6.1 the zero person construction in Finnish is contrasted with its typical counterparts in Estonian, namely the da-ininitive, the zero person, the generic second person singular, and the impersonal. In 6.2 the Finnish zero person is compared to the passive, and the possible counterparts in Estonian are considered. Section 6.3 focuses on the diferences and similarities of the implied “subjects” in the da-ininitive and the impersonal in non-translated Estonian. 6.1

he zero person and its possible counterparts

In this section we compare diferent possibilities to translate the Finnish zero person into Estonian. In our translated data, there are 84 if-clauses with the Finnish zero person. According to our translated data, the Finnish zero person is translated into Estonian with the da-ininitive (25 examples), the zero person (21), the generic second person singular (14), the impersonal (11), the explicit pronoun subject (2), or with some other construction, such as an explicit personal form of a predicate or a noun phrase (11). Practically all Estonian zero person constructions occur with a modal verb (for more details, see below).

he zero person 6.1.1 he zero person construction typically contains a verb expressing changes of state (for example, nukahtaa ‘to fall asleep’) or a modal verb (Laitinen 1995, 2006). he most common verbs used in the zero person construction in Finnish are modal verbs, such as saada ‘be allowed to’, voida ‘may’ and täytyä ‘must, have to’. Also in

85

86

Hanna Jokela and Helen Plado

Estonian, zero subjects typically occur with modal verbs (see Jokela 2006, 2012). he implied subject of zero person construction is usually in the role of an experiencer or a patient, not an agent. Modal expressions of this kind tell what one may do, must (not) do, and so on, under certain conditions. In if-clauses with the zero person, the meaning is usually generic. However, the zero may be interpreted as indexically open so that the reference can be construed in the context. In Finnish, however, the if-then frame ofers a context where zero person subjects can be found together with any kind of verb: not only modal verbs, but also agentive verbs expressing mental or concrete action (see also Laitinen 1995; Jokela 2012). Estonian if-clauses with agentive verbs are not suitable contexts for the zero person, and in the Estonian counterparts for Finnish if-clauses with zero person, there is usually some other form, such as the da-ininitive, the generic second person singular, or the passive instead of the zero person. Example (15) demonstrates a typical zero person construction in Finnish and in Estonian. he zero person in Finnish is used with the modal verb predicate voi ‘can’, and in Estonian with the modal verb predicate võib ‘can’. Here the zero occurs in the if-then frame, albeit not in the if-clause but the main clause starting with the predicate verb of the zero person. he if-then frame contributes to creating a generic and open statement – the same conditions apply to anyone, and the reference can be considered both generic and open. (15) Kun tilaa on vähän, when space:par be:3sg only.a.little voi rinteenkin hyödyntää. [–] can.3sg hillside:gen make.use.of:inf ‘When/if there is only a little space, one can make use of a hillside.’ (Fin; KV5/01/61) Kui ruumi on vähe, when/if space:par be:3sg only.a.little siis võib lilleaia rajada ka künkanõlvale. then can:3sg garden.gen start:inf also hillside:all ‘When/if there is only a little space, one can start a garden on a hillside.’ (Est; KA4/02/85)

It is noteworthy that in our translated data there are two non-modal predicates with the Estonian zero persons, but they are in the negative form, such as in Example (16). In (16), again, instructions and recommendations apply to anyone, and the reference is let open. We need to recall that the negative form is a special case in Estonian: all negative forms lack personal endings (see above Section 4). herefore the interpretation of person remains open. Uuspõld (1989: 476) argues that the negative form of third person singular is a grammatical means to express dynamic possibility: Selle palgaga autot ei osta ‘One does not buy a car with that salary’.

Subjects under generic conditions

(16) Ihan heti lamppua ei tarvitse vaihtaa, right away the light.bulb:par neg:3sg need replace:inf ellei riko sitä omalla varomattomuudellaan. if-neg:3sg break it:par own:ade carelessness:ade.3poss ‘he light bulb doesn’t need to be replaced right away if you do not break it through carelessness.’ (Fin; TM3/04/59) Nii et lampi ei pea kohe so that light.bulb:par neg need right.away vahetama hakkama, kui seda just omaenese replace:inf start:inf when/if it:par just own.gen hoolimatusega ära ei riku. carelessness:com away neg break ‘he light bulb doesn’t need to be replaced right away if you do not break it carelessness.’ (Est; TME9/04/28)

6.1.2 da-ininitive Both the Finnish zero person and the Estonian da-ininitive may be interpreted as generic and open. Oten they refer to anyone, including the speaker and/or the addressee. he interpretation of open expressions is based on several factors, above all context. Context may include the main clause, or it may be wider, even including the whole text and genre. Example (17) from the translated data describes a situation in which following the given instructions and advice will lead to the desired outcome. his is valid for anyone who is interested in ixing the wooden furniture. In the Finnish if-clause, there is the zero person construction, and in Estonian translation there is the daininitive. In both Finnish and Estonian, there is the zero person construction in the main clause in Example (17). he interpretation of the zero persons and the da-ininitive is generic. (17) Kuluneet puupinnat saa helposti alkuperäisen Worn wooden.surface:pl get.3sg easily genuine:gen näköiseksi, kun käsittelee ne puuöljyllä. looking:trnsl when/if process:3sg they wood oil:ade ‘One can make worn wooden surfaces genuine-looking when/if [one] treats them with wood oil.’ (Fin; TR01/05/94) Kulunud puitpinnad saab kergesti endiseks, Worn wooden.surface:pl get:3sg easily genuine.looking:trnsl kui need puiduõliga üle tõmmata. when/if they wood.oil:com over brush:inf ‘One can make worn wooden surfaces genuine-looking when/if they are polished with wood oil.’ (Est; TRE02-03/06/105)

87

88

Hanna Jokela and Helen Plado

hese sentences represent an if-then frame that describes actions and consequences. One can say that they express generic conditions (involving certain consequences) that apply to anyone who positions her/himself in that conditional frame. Many of the generic clauses of our data have the same properties as directives and suggestions. hey involve actions, conditions, and consequences. Additionally, Langacker has discussed some properties of directives and instructions. According to Langacker (1991: 266), the sentence To turn of the dishwasher, you move the lever all the way to the let does not designate any speciic instance of moving the lever, but indicates that an action of this type is part of the normal course of events constituting the canonical way of operating the device. By using the present tense in a recipe or in giving directions, for instance, one is portraying the world as being structured so that the speciied sequence of events regularly leads to the desired outcome. Plado (2010) studies Estonian conditional clauses with the predicate in the da-ininitive form. She notes that conditional clauses with the predicate in the da-ininitive form are either content or conversational conditional clauses. In the corpus, there were no epistemic conditional clauses with the da-ininitive as their predicate (however, Plado [2010] shows that this type of clause is not impossible). In content conditional clauses, the predicate verb is usually a verb that expresses a concrete action, state, or event. In conversational conditional clauses, the verb is typically a verb that expresses a mental action, state, or event, although there may also be a concrete verb. he Estonian corpus of if-clauses with da-ininitives as independent predicates illustrates that the da-ininitives are used mostly in conversational conditional clauses and that they are oten mental verbs. In our translated data, many of the zero person constructions with the Estonian da-ininitive as their counterpart have an agentive and concrete verb as their predicate, such as käsitellä ‘to process’ in Example (17). his is probably because in Finnish the if-clause enables the use of all verbs, including agentive ones, as a predicate in the zero person construction. he main clause describes the result that is reached when the action referred to in the if-clause – expressed in Finnish by the zero person construction, and in Estonian by the da-ininitive – is completed. Mental verbs, such as verbs of perception or cognitive processes, can be found as the predicates of the Finnish zero person construction and the Estonian daininitive in translated data as well. In Example (18), the Finnish predicate verb is katsella ‘to watch’. In Finnish it is used in the zero person construction, and in the Estonian translation there is the da-ininitive. he zero person in Finnish and the da-ininitive in Estonian leave the person open. However, the speaker is interpreted as one of the speech-act participants – the speaker is commenting on what is said in the main clause. It is likely that the statement in the if-clause is based at least partly on his/her own experience.

Subjects under generic conditions

(18) Arkinen visailuohjelmakin saa aivan uudet Everyday entertainment.program get.3sg totally new:pl mittasuhteet, kun sitä katselee tavallista proportions:pl when/if it:par watch:3sg regular:ela isommasta televisiosta. bigger:ela TV screen:ela ‘Any ordinary entertainment program becomes more interesting when/if [one] watches it on a bigger TV screen.’ (Fin; TM4/03/10) Tavaline meelelahutussaade saab uued mõõtmed, Everyday entertainment.program get:3sg new:pl dimension:pl kui seda vaadata hoopis suuremast telerist. when/if it:par watch:inf bigger:ela TV screen:ela ‘Any ordinary entertainment program becomes more interesting when/if [one] watches it on a bigger TV screen.’ (Est; TME7/03/26)

6.1.3 he second person singular We assume that the generic second person form is more widely used in Estonian than in Finnish (see also Jokela 2006). In Estonian the generic second person form and the zero person form are more intertwined and optional than in Finnish, and there are more restrictions for the zero person use. In the if-clause of Example (19), the Estonian counterpart for the Finnish zero person is the generic second person. he predicate verb is tekee ‘one does’ in Finnish and teed ‘you do’ in Estonian. In the if-clause, the concrete and agentive predicate tekee is possible with the zero person in Finnish, but not in Estonian. It seems, also based on the corpus of Estonian, that in Estonian, the singular second person is preferred instead of the zero person when the speaker is referring to his/ her own experience, which may apply to anyone. (19) Yleensä riittää, kun tekee työnsä usually be.enough:3sg when/if do:3sg work:gen.1poss tarpeeksi hyvin, täydellisyyttä ei tarvitse tavoitella. enough well perfection:par neg:3sg need seek:inf ‘It is usually good enough when [one] does [one’s] own work well enough; there is no need to seek perfection.’ (Fin; TR5/04/92) Enamasti piisab sellest, kui teed oma usually be.enough:3sg that:ela when/if do:2sg own tööd lihtsalt hästi – täiuslikkust ei ole vaja püüda. work:par simply well ‘It is usually good enough when you do your work well – there is no need to seek perfection.’ (Est; TRE4-5/05/92)

89

90

Hanna Jokela and Helen Plado

6.1.4 he impersonal he interpretation of the zero person may be very similar to the passive/impersonal in both Estonian and Finnish. In Example (20), the Finnish zero person construction is translated into Estonian with the impersonal predicate. he predicate verb is a concrete verb: käyttää ‘use’. he interpretation is generic and open – not using one’s brain cells would be disadvantageous to anyone. It is noteworthy that agentive or concrete verbs are not used in the zero person construction in Estonian, whereas the zero person construction in Finnish can typically occur in if-clauses together with any kind of verb (see also 6.1.1 and 6.1.2). However, it is possible to use agentive and concrete verbs in both the da-ininitive and the impersonal in Estonian. (20) Aivosolut alkavat surkastua, jos niitä ei käytä. brain.cell:pl begin:3pl shrink:inf if they:par neg.3sg use ‘Brain cells begin to shrink if one does not use them.’ (Fin; TR10/05/113) Kui ajurakke ei kasutata, hakkavad when/if brain.cell.pl.par neg use:iprs begin:3pl need känguma. they shrink:inf ‘If brain cells are not used, they begin to shrink.’ (Est; TRE10/06/88)

6.2

Finnish passive and its possible Estonian counterparts

In our translated data, the focus is on the if-clauses with the zero person construction and its Estonian translations. Our data also includes some additional if-clauses (see 1.1 Data) with constructions other than the zero person, in order to demonstrate diferent possibilities to express generic and open reference, and one of these possibilities is passive. Our data is not extensive enough to draw conclusions about their frequency compared to the zero persons, for instance. However, we may speculate about the diferences between if-clauses with the zero person constructions and the passive and their possible Estonian counterparts. Not only the zero person but also the Finnish passive may be translated into Estonian with the da-ininitive, as in Example (21). In (21), Finnish passive katsotaan ‘is looked’ is translated into Estonian with the da-ininitive vaadata ‘to watch’. In this case the reference of the Finnish passive includes the speaker along with anyone else. It seems that the da-ininitive is preferred in Estonian instead of the impersonal, as the Estonian impersonal is more likely to be interpreted as speaker-exclusive than the da-ininitive, which primarily expresses open reference.

Subjects under generic conditions

(21) Jos katsotaan sotaa ainoastaan teknologian if look:pass war:par only technology:gen näkökulmasta, se on usein suuresti vauhdittanut siihen viewpoint:ela it be:3sg oten greatly speed:ptcp it:ill liittyvien tekniikoiden kehitystä, attach:pl.gen technique:pl.gen development:par raha ei silloin ole ollut esteenä. ‘If we look at the war only from the viewpoint of technology, it has speeded up the development of diferent techniques; money has not been a problem.’ (Fin; TM10/03/130) Kui vaadata sõdu ainult tehnoloogia when/if look:inf war:pl.par only technology.gen seisukohalt, siis on inimeste tapmiseks viewpoint:abl then be.3sg human:pl.gen killing:trnsl mõeldud tehnikat alati hoogsalt arendatud, think:ptcp techniques:par always dynamically develop:ptcp sageli küsimata, kas muuks raha jätkub. ‘If we look at the war only from the viewpoint of technology, it has sped up the development of diferent techniques; money has not been a problem.’ (Est; TME7/03/26)

Helasvuo and Vilkuna (2008: 233) argue that the passive can refer to a group that does not include a speech-act participant, and the passive can describe a particular event without generic meaning. hey also point out that oten the more generic reference to “anyone” remains as a possibility on a par with the implied reference to the speech-act participants. In Example (21), the Finnish passive and the Estonian da-ininitive refer to a perspective that is shared by the speaker and anyone who wants to share the speaker’s opinion. In Example  (22) the Finnish passive is translated into Estonian with the impersonal, and the past tense is used in both languages. It seems obvious that tense afects the interpretation between generic/open or speciic. Present tense is preferred in generic expressions, whereas the past-tense form is less likely to be generic. In Example (22) the use of passive in both languages implies a speakerexclusive group that was responsible for building this device, and that the actual building took place in the past. Neither the writer nor the addressee/reader is assumed to belong to the group. (22) Kun sotilaskoneisiin rakennettiin paikannuslaite, when military.plane:pl.ill build:pass.pst tracking.device, ‘When a tracking device was placed into military planes,’

91

92

Hanna Jokela and Helen Plado

joka mittasi nopeudet, lentokulmat, korkeuden, ‘that measured speed, lying angles, elevation,’ tuulennopeudet ja niin edelleen, ‘wind speed, etc.,’ tätä painavaa kojetta ei aluksi voitu this heavy:par device:par neg.3sg tart:trnsl be.able:pcp ‘this heavy device could not sijoittaa kuin; suuriin pommikoneisiin. place:inf than big:pl.ill bomber:pl.ill be placed anywhere but into the big bombers at irst.’ (Fin; TM14/03/57) Kui sõjalennukitele hakati paigutama when/if military.plane:pl.all start:iprs.pst place:inf asukoha kindlaks tegemise seadmeid, location.gen ascertain:gen device:pl.par ‘When a tracking device was placed into military planes,’ mis mõõtsid lennuki kiirust, lennusuunda,kõrgust, tuule kiirust, jne, ‘that measured speed, lying angles, wind speed, elevation, etc.,’ siis sai see aparatuur nii suur ja raske, then get:pst.3sg this device so big and heavy ‘the device was so heavy’ et seda oli võimalik paigutada üksnes pommitajatele. ‘that it was possible to place it only into bombers’ (Est; TME7/04/65)

In Finnish, both the zero person and the passive may refer to the speaker, but generally speaking, the passive is more likely to refer to a group instead of anyone or an individual, and this group may or may not include the speaker. 6.3

he implied “subject” of the impersonal and the da-ininitive in non-translated Estonian

In our Estonian corpora there were content and conversational conditional clauses with the predicate in the da-ininitive form, but only content conditional clauses with the predicate in the impersonal form. his might be inluenced by the fact (among other factors) that the reference of the conversational conditionals includes the speaker, and that is why the da-ininitive is chosen. he reference of the Estonian impersonal includes the speaker less oten than the da-ininitive, and therefore it is not used in conversational conditionals. Furthermore, there seems to be a connection between diferent functional types of conditionals and genericity: the usage of the da-ininitive predicate in content conditionals makes the whole sentence generic, while in the case of

Subjects under generic conditions

conversational conditional clauses, the sentence is still interpreted as non-generic. he conversational conditional clause connects the main clause to the concrete communicative situation and explains why the main clause is uttered. his may be the reason why the generic interpretation is blocked. But the implied “subject” of the da-ininitive form remains an indeinite human who is probably a participant of the speech act (23). Visapää (2008) shows that in Finnish as well, the implied subject of the independent A-ininitive construction is always generic and that the use of the ininitive makes the whole process atemporal. (23) Ja kui arvestada siia juurde veel töötajate and when/if add:inf here also worker:pl.gen perekonnaliikmed, siis elab Kohtla-Järvel tänu family.member:pl then live:3sg Kohtla-Järve:ade thanks to “Põlevkivikeemiale” hästi umbes 20 000 inimest. oil.shale.chemistry:all well about 20,000 people:par ‘And if you add workers’ family members, then about 20,000 people are well of, all thanks to the ‘Oil Shale Chemistry’ in Kohtla-Järve.’ (Est; AJA1990)

he use of the da-ininitive form in the content conditional clause makes the whole sentence generic. he “subject” of the da-ininitive is also an indeinite human or a group, but no information is given about the participation of the “subject” in the speech act (24). (24) Kui kunstlik sülem raamile paigutada, siis when/if artiicial bee.swarm frame:all put:inf then tagab loomulik töö kärje ehitamisel ensure:3pl natural work honeycomb.gen building:adess ja paljunemine pere edasise elujõulisuse. and reproduction family.gen further.gen viability.gen ‘If you put the artiicial bee swarm on a frame, the bees can continue their daily work of building honeycombs and reproduce, which will ensure the swarm’s viability.’ (Est; AJA1990)

Many of the generic conditional sentences with the predicate in the impersonal form are law-like sentences (25). here are also directives expressed by the same construction (26). Both of these constructions apply to anyone; the “subject” of the impersonal form is also an indeinite human or a group who mostly is not a participant in the communication situation. Pajusalu (this volume) reaches the same conclusion based on conversation data. However, she demonstrates that sometimes – albeit rarely – the impersonal may refer to a group that includes one of the participants.

93

94

Hanna Jokela and Helen Plado

(25) Kui ettekirjutust ei täideta, rakendatakse when/if order:par neg follow:iprs impose:iprs ameti poolt sanktsioone. oice.gen by sanction.pl.par ‘If the order is not followed, the oice will impose sanctions.’ (Est; AJA1990) (26) Kui pannilt tahetakse kalalõhna eemaldada, siis when/if pan:abl want:iprs ishy.smell.par remove:inf then puhastatakse seda soolaga ja loputatakse hästi clean:iprs that:par salt:com and rinse:iprs very külma veega. cold.gen water:com ‘If you want to get rid of the ishy smell on your frying pan, clean it with salt and rinse with freezing water.’ (Est; AJA1990)

7.

Conclusion

he if-then frame alone does not automatically create generic statement, but implied subjects in this frame can have a generic and open interpretation. herefore, the if-then frame can provide conditions that are valid for anyone who its or wants to it into that frame in a particular context. he degree of genericity depends on the predicate verb, the participants of the main clause, and the wider context. here are various possibilities for implied subjects in if-clauses. hese implied subjects can refer to the speaker and/or the addressee, or they can be interpreted as open, and oten they refer to anyone in the world in general. Our focus has been on the zero person construction, the da-ininitive, and the passive in Finnish and impersonal in Estonian, and diferences in their use and interpretation. he implied subject of the da-ininitive and the impersonal is an indeinite human being or a group. In the case of conversational conditionals, the implied “subject” tends to be the participant(s) of the speech act. he content conditional clause does not provide any information on the subject’s participation in the speech act, and the “subject” of the conditional clause with the predicate in impersonal form mostly is not the participant of the speech act. he interpretation of implied subject depends on the conditional clause type, the context, and the verb form that is used. he passive in Finnish and the impersonal in Estonian in if-clauses have a tendency to refer to a group, either speaker-inclusive or speaker-exclusive, whereas the zero person refers to one individual at a time. hus the da-ininitive in Estonian is similar to the zero person construction (both in Finnish and Estonian): it refers to anyone and may include the speaker personally.

Subjects under generic conditions

Translated data (magazines) KV = Kotivinkki. Helsinki: Kustannus Oy Forma. (home, gardening, cooking) K&A = Kodu & Aed. Tallinn: ICA Media AS. TM = Tekniikan maailma. Helsinki: Yhtyneet Kuvalehdet Oy. (cars and technology) TME = Tehnikamaailm. Tallinn: Ühinenud Ajakirjad AS. TR = Trendi. Helsinki: Forma Publishing Group. (fashion, health, beauty) TRE = Trend. Tallinn: Forma Media AS.

Non-translated data AJA1990  = Sub-corpus of 1990s newspapers of the Corpus of Estonian Literary Language (http://www.cl.ut.ee/korpused/baaskorpus/) ILU1990 = Sub-corpus of 1990s iction of the Corpus of Estonian Literary Language (http:// www.cl.ut.ee/korpused/baaskorpus/) LAY = Sub-corpus of standard language of Syntax Archive Data of Finnish at the University of Turku and Research Institute for the Languages of Finland (http://syntaxarchives.suo.utu. i/~lauseopin_arkisto/)

References Behrens, Leila (2005). Genericity from a cross-linguistic perspective. Linguistics, 43(2), 275–344. DOI: 10.1515/ling.2005.43.2.275 Blevins, James P. (2003). Passive and impersonal constructions. Journal of Linguistics, 39, 473–520. DOI: 10.1017/S0022226703002081 Bybee, Joan, Perkins, Revere, & Pagliuca, William (1994). he evolution of grammar. Tense, aspect, and modality on the languages of the world. Chicago: he University of Chicago Press. Cardinaletti, Anna, & Starke, Michal (1999). he typology of structural deiciency: A case study of the three classes of pronouns. In Henk van Riemsdijk (Ed.), Clitics in the languages of Europe (pp. 145–233). Berlin, New York: Mouton de Gruyter. Dahl, Östen (1975). On generics. In Edward Keenan (Ed.), Formal semantics of natural languages (pp. 9–111). Oxford: Cambridge University Press. Dancygier, Barbara (2006). Conditionals and prediction. Time, knowledge and causation in conditional constructions. Cambridge: Cambridge University Press. Duvallon, Outi, & Chalvin, Antoine (2004). La réalisation zéro du pronom sujet de première et de deuxième personne du singulier en innois et en estonien parlés. Linguistica Uralica, 4, 270–286. EKG II = Erelt, Mati, Kasik, Reet, Metslang, Helle, Rajandi, Henno, Ross, Kristiina, Saari, Henn, Tael, Kaja, & Vare, Silvi (1993). Eesti keele grammatika II. Süntaks. Lisa: Kiri. Tallinn: Eesti TA Keele ja Kirjanduse Instituut. EKRE = Eesti keele riigieksami eristuskiri. http://www.innove.ee/UserFiles/ Riigieksamid/2013/ Eristuskiri_kodulehele.pdf. [Cited 01/21/2013.]

95

96

Hanna Jokela and Helen Plado

Erelt, Mati (1984). Da-ininitiivi süntaktilised funktsioonid tänapäeva eesti keeles. Soome-eesti kontrastiivseminar (Helsingi, 17.–20.10.84). Preprint KKI-24. Tallinn: Eesti NSV Teaduste Akadeemia. Erelt, Mati (Ed.) (2003). Estonian language. Linguistica Uralica (Vol. 1). Supplementary Series. Tallinn: Estonian Academy Publishers. Hakulinen, Auli (1987). Avoiding personal reference in Finnish. In Jef Verschueren & Marcella Bertucelli-Papi (Eds.), he pragmatic perspective. Pragmatics & Beyond (pp. 140–153). Amsterdam: John Benjamins. Helasvuo, Marja-Liisa (2006). Passive: Personal or impersonal? A Finnish perspective. In MarjaLiisa Helasvuo & Lyle Campbell (Eds.), Grammar from the human perspective: Case, space, and person in Finnish. Current Issues in Linguistic heory (pp. 233–255). Amsterdam: John Benjamins. DOI: 10.1075/cilt.277.16hel Helasvuo, Marja-Liisa (2008). Minä ja muut. Puhujaviitteisyys ja konteksti. Virittäjä, 112, 186–206. Helasvuo, Marja-Liisa, & Laitinen, Lea (2006). Person in Finnish: Paradigmatic and syntagmatic perspectives. In Marja-Liisa Helasvuo & Lyle Campbell (Eds.), Grammar from the human perspective: Case, space, and person in Finnish. Current Issues in Linguistic heory (pp. 173–209). Amsterdam: John Benjamins. DOI: 10.1075/cilt.277.14hel Helasvuo, Marja-Liisa, & Johansson, Marjut (2008). Construing reference in context: Nonspeciic refernce forms in Finnish and French discussion groups. In María de los Ángeles Gómez González, J. Lachlan Mackenzie & Elsa M. González Álvarez (Eds.), Current trends in contrastive linguistics. Functional and cognitive perspectives. Studies in functional and structural linguistics 60 (pp. 27–57). Amsterdam: John Benjamins. DOI: 10.1075/sfsl.60.04hel Helasvuo, Marja-Liisa, & Vilkuna, Maria (2008). Impersonal is personal: Finnish perspectives. Special issue on impersonals, guest editor Anna Siewierska. Transactions of the Philological Society, 106(2), 216–245. DOI: 10.1111/j.1467-968X.2008.00208.x Herlin, Ilona (1998). Suomen kun. Helsinki: Finnish Literary Society. Jokela, Hanna (2006). Geneerinen nollasubjektilause suomessa ja virossa. In Krista Kerge & Maria-Maren Sepper (Eds.), Finest Linguistics. Proceedings of the annual Finnish and Estonian conference of linguistics (pp. 59–72). Tallinn, May 6–7, 2004. Tallinn: Tallinn University Press. Jokela, Hanna (2012). Nollapersoonalause suomessa ja virossa. Tutkimus kirjoitetun kielen aineistosta. Annales Universitatis Turkuensis. C 334. University of Turku. Koptjevskaja-Tamm, Maria (1999). Finiteness. In Keith Brown & Jim Miller (Eds.), Concise encyclopedia of grammatical categories (pp. 146–149). New York: Elsevier. Laitinen, Lea (1995). Nollapersoona. Virittäjä, 99, 337–358. Laitinen, Lea (2006). Zero person in Finnish: A grammatical resource for construing human reference. In Marja-Liisa Helasvuo & Lyle Campbell (Eds.), Grammar from the human perspective: Case, space, and person in Finnish. Current Issues in Linguistic heory (pp. 209–233). Amsterdam: John Benjamins. DOI: 10.1075/cilt.277.15lai Langacker, Ronald (1991). Foundations of cognitive grammar, vol. 2, Descriptive Application. Stanford: Stanford University Press. Langacker, Ronald (1997). Generics and habituals. In Angeliki Athanasiadou & René Driven (Eds.), On conditionals again (pp. 191–222). Amsterdam: John Benjamins. DOI: 10.1075/cilt.143.11lan

Subjects under generic conditions

Lindström, Liina (2010). Kõnelejale ja kuulajale viitamise vältimise strateegiad eesti keeles. In Mati Erelt (Ed.), Emakeele Seltsi aastaraamat, 55 (2009), 88–118. Lölund, Juhani (1998). Suomen kirjoitetun yleiskielen passiivi. Turku: Åbo Akademis förlag. Mauranen, Anna, & Tiittula, Liisa (2005). Minä käännössuomessa ja supisuomessa. In Anna Mauranen & Jarmo H. Jantunen (Eds.), Käännössuomeksi. Tutkimuksia suomennosten kielestä. Tampere Studies in Language, Translation and Culture A1 (pp. 35–69). Tampere: Tampere University Press. Metslang, Helle (1994). Kielet ja kontrastit. Virittäjä, 98, 203–226. Nikolaeva, Irina (2007). Constructional economy and noninite independent clauses. In Irina Nikolaeva (Ed.), Finiteness: heoretical and empirical foundations (pp. 138–180). Oxford: Oxford University Press. Noonan, Michael (1985). Complementation. In Timothy Shopen (Ed.), Language typology and syntactic ieldwork (pp. 42–140). Cambridge: Cambridge University Press. Plado, Helen (2010). Eesti keele da-ininitiivis öeldisverbiga tingimuslaused. Eesti ja soomeugri keeleteaduse ajakiri ESUKA / Journal of Estonian and Finno-Ugric Linguistics JEF, 2, 255–272. Plado, Helen (2013). Eesti keele otstarbelausest. Keel ja Kirjandus, 4, 269–284. Seppänen, Eeva-Leena (2000). Sinä ja suomalaiset: yksikön toisen persoonan yleistävästä käytöstä.Kielikello, 3/2000, 16–18. Sweetser, Eve E. (1990). From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511620904 Torn-Leesik, Reeli (2009). he voice system of Estonian. Sprachtypologie und Universalienforschung, 62(1–2), 71–90. Torn-Leesik, Reeli, & Vihman, Virve (2010). he uses of impersonals in spoken Estonian. SKY Journal of Linguistics, 2, 301–343. Tuomikoski, Risto (1971). Persoona, tekijä ja henkilö. Virittäjä, 75,146–152. Uuspõld, Ellen (1989). Modaalsusest ja modaalsest predikaatist eesti keeles. Keel ja Kirjandus, 8, 468–477. Vihman, Virve-Anneli, & Kaiser, Elsi (2006). Efects of Demotion in Estonian and Finnish. In Torgrim Solstad (Ed.), Demoting the agent: Passive, middle, and other voice phenomena (pp. 111–141). Amsterdam: John Benjamins. VISK. Hakulinen, Auli, Vilkuna, Maria, Korhonen, Riitta, Koivisto, Vesa, Heinonen, Tarja Riitta, & Alho, Irja (2004). Iso suomen kielioppi. Helsinki: Finnish Literature Society. www.scripta. kotus.i/visk/etusivu.php Visapää, Laura (2008). Ininitiivi ja sen ininiittisyys. Tutkimus suomen kielen itsenäisistä A-ininitiivikonstruktioista. Suomalaisen Kirjallisuuden Seuran Toimituksia 1181. Helsinki: Finnish Literature Society. Ylikoski, Jussi (2003). Deining non-inites: Action nominals, converbs and ininitives. SKY Journal of Linguistics, 16, 185–237.

97

section ii

Stretching the limits of subjecthood

Abstract locational subjects Field and settings in French and English Michel Achard Rice University

his chapter analyzes the English and French abstract locational subjects il, ça, there, and it which refer to an ‘abstract region’ within which the entity in the complement is identiied. here, it, and il describe the ‘ield’, namely the conceptualizer’s scope of awareness relative to the complement content, but they operate at diferent levels of reality. Existential there and simple il (followed by a nominal) describe a section of basic reality, while it and complex il (followed by a clause) operate at a more analytical level of reality which includes the conceptualizer’s efort to assess or evaluate the complement content. Ça proiles the abstract setting composed of the immediate context from which the complement content is extracted for expressive purposes.

1.

Introduction

his chapter proposes a preliminary investigation of the class of non-canonical subjects the Cognitive Linguistics literature has called “abstract locationals” (Achard 1998; Langacker 1987b; Lakof 1987; Smith 1985). his class is illustrated in (1)–(4) by the English forms it and (existential) there and their French counterparts il and ça (ce, c’):1 (1) It seems that Durning’s stepdaughter, as aspiring actress, had seen a copy of the script that had somehow never reached Durning and insisted he read it. (LATWP)

1. hese three French neuter demonstrative pronouns will be considered together in this chapter. First, they are diachronically related (Brunot 1936; Wilmet 1997). Secondly, they all participate in the constructions under investigation with a very speciic distribution (Olsson 1986). Ce and ça (along with cela not considered here) occur preceding a consonant initial form of the être ‘be’ copula, while c’ occurs before a vowel initial form. doi 10.1075/cal.16.05ach © 2015 John Benjamins Publishing Company

102 Michel Achard

(2) here is some great children’s literature out there. It’s an untapped natural resource. (LATWP) (3) Mais il reste des dictatures dans le monde, des systèmes de parti unique par exemple dans cette grande zone d’ombre qu’est l’Afrique. (AFP) ‘But there still are dictatorships out there, systems with a single part y, for example in this large area of shadows that is Africa’ (4) Vous pensez bien que nous sommes soulagés et heureux, a airmé un membre de la direction générale, c’était vraiment très dur pour nous de savoir un homme comme M. Pineau-Valencienne en prison. (AFP) ‘As you can imagine, we are relieved and happy, a member of the management team said, it was really hard for us to know that a man like Mr. PineauValencienne was in prison’

he cognitive analysis of these forms is organized around two main points. First, every one of these four pronouns is meaningful, and secondly, their meaning depicts some kind of “abstract region” (Achard 1998; Langacker 1987b; Lakof 1987; Smith 1985) within which the entity (thing or event) described in the complement clause can be identiied and described. As it currently stands, however, this abstract region is only described in the most general terms applicable to all pronouns, and therefore incapable of capturing the diferences in distribution which exist among them. he analysis developed in this chapter reines the abstract locational proposal by showing that the description of the distribution of the English and French pronouns (respectively it / there and il / ça) requires the further speciication of diferent kinds of abstract regions, and that the nature of the region each pronoun proiles largely explains its usage in particular situations. he goal of the analysis is thus twofold. First, to precisely describe the diferent regions which it, there, il and ça respectively proile, and secondly, to point out the need for further investigation of this class of abstract locational subjects, so that, in time, we might be able to provide a typology of the diferent types of regions they code in the various languages of the world. his chapter is organized in the following fashion: Section 2 argues that the four pronouns illustrated in (1)–(4) constitute a valid natural class of abstract locational subjects. Section 3 introduces the Cognitive Grammar (henceforth CG) (Langacker 1987a, 1991, 2008) analysis of these subjects. Section 4 investigates the speciic kinds of abstract locations each pronoun codes. Section 5 recapitulates the results and concludes the chapter. he data on which the analysis is based come from single language corpora [FRANTEXT database of 20th century literature, Agence France Press (AFP) for French, Los Angeles Times Washington Post (LATWP) for English], as well as a parallel corpus of European Union texts in their English and French versions.

Abstract locational subjects

2. A natural class of subjects he analysis of the four forms illustrated in (1)–(4) as a natural class appropriately captures the functional overlap that exists between them. In a parallel corpus of EU parliament deliberations, English it and French il most frequently correspond to each other, as the examples in (5) and (6) illustrate:2 (5) It is therefore necessary that the Commission should now do some tightening up. (5′) Il est dès lors nécessaire que la Commission fasse des eforts de son côté. (6) It is important to identify the skills and potential of our regions in the hi-tech sector. (6′) Il est important d’identiier les capacités et potentialités de nos régions dans le secteur de la haute technologie.

English existential there is also oten attested as the most natural translation of il, as the examples in (7) and (8) show:3 (7) Il y a bien deux chaînes innoises et une chaîne portugaise, mais il n’y a toujours aucune chaîne néerlandaise. (7′) Although there are now two Finnish channels and one Portuguese one, there is still no Dutch channel… (8) Comment se fait-il qu’il n’y ait pas de consignes en cas d’incendie? (8′) Why are there no ire instructions?4

2. Since the data come from a parallel corpus of EU parliament deliberations, it is hard to know with precision which instance in each pair constitutes the original. he use of prime numbers and the absence of italics indicate that each individual example is treated as an independent utterance which corresponds to similar conceptual content in the other language, rather than a translation per se. 3. Il y a ‘there is’ has been suggested as the French equivalent of the existential there construction (Bergen and Plauché 2005). 4. A possible challenge to the analysis of there as the subject of the existential construction comes from the agreement pattern illustrated in (7) and (8). Unlike the three other forms considered in this chapter, there does not always determine agreement on the following predicate. In (8) for example, the French predicate is in the third person singular following the pronoun il, while the third person plural of the English predicate in (8′) indicates its agreement with the post verbal expression. Despite these agreement facts, there behaves as a subject in every other respect because of its preverbal position, and its participation in tag questions and raising respectively illustrated in (i) and (ii) (from Lakof 1987: 547): (i) here was a dog in the yard wasn’t there? (ii) here is believed to be a dog in the yard.

103

104 Michel Achard

Finally, ça /ce /c’ also provide well attested correspondences to it, as in the examples presented in (9) and (10): (9) It is true that consultation with Parliament has only come about at a late date. (9′) C’est vrai que la consultation de votre Assemblée est intervenue tardivement. (10) It is true that the problem of the forestry sector is an extremely complex one. (10′) C’est vrai que le problème du secteur du bois est terriblement complexe.

Despite the high degree of correlation these four forms exhibit, they have seldom been considered together in the literature. It and il, in their uses illustrated in (1) and (3) are uncontroversially recognized as impersonal pronouns in their respective languages and hence meaningless “pleonastic” elements (Chomsky 1986: 26) whose presence in the clause solely satisies the language’s syntactic requirement for a subject.5 here is oten presented as an “existential” (Jespersen 1924), and the French demonstratives ça, ce, c’ are predominantly analyzed as “referential expressions which refer forward to the inite or ininitival clause” (Jones 1996: 128).6

Even the agreement facts do not constitute a serious challenge to the subject status of there, once they are taken in their broader context. First, in a number of cases, the pronoun does determine verb agreement, as illustrated in (iii), and a thorough account of the agreement patterns there exhibits needs to factor in several sociolinguistic variables (Crawford 2005; Breivik & MartínezInsua 2008; Martínez-Insua & Palacios Martinez 2003). (iii) “It’s obvious there remains substantial disagreements over how to best proceed,” Majority Leader George Mitchell, D-Maine, said ater the session. (LATWP) Secondly, even when agreement follows the post verbal entity, there is evidence of “indirect agreement” (Lakof 1987: 548), whereby there agrees with the following noun phrase, and the noun phrase agrees with the verb. In this analysis, “all agreement occurs via there” (Lakof 1987: 548). For the arguments in favor of indirect agreement see Lakof 1987: 546–549). he account presented in this chapter is perfectly compatible with Lakof ’s proposal. 5. For example, Jones (1996: 120) claims that: “Intuitively, we know that il (…) does not refer to anything – it does not represent an argument of the verb which can be characterized in terms of a theta-role such as Agent, Experiencer, etc. Rather, its function appears to be purely syntactic, satisfying a requirement that all inite clauses must have a subject”. Similar positions are also proposed in Radford (2004: 291f; Gledhill 2003: 131; Rowlett 2007: 133). 6. he analysis of il as the subject of the impersonal construction in a construction such as (3) became standard following Brunot (1936 [1922]) In earlier analyses which oten confuse the concepts of “subject” and “agent”, a majority of grammarians attempted to distinguish the sujet grammatical, formel, logique, or morphologique ‘grammatical, formal, logical, or morphological subject’ from the sujet réel ‘real subject’ of the predicate coded by the post verbal expression in what is traditionally called the ‘two subjects solution’ (Lauwers 2004: 293). Brunot clearly points out that the post-verbal expression is not a subject but a complement: “Dans l’expression actuelle – fort ancienne – il faut de la vertu, il est impossible de considérer de la vertu autrement que

Abstract locational subjects

As briely mentioned in the introduction, in the CG literature, it and il are not meaningless, but describe a kind of abstract region within which the entity coded in the complement can be identiied. his section argues that the other two forms should be analyzed in a similar manner. Although speciic details will be presented in Section 3, it is intuitively clear that existential there is most readily accommodated because it most closely resembles the two universally recognized impersonal pronouns. First, there constructions are not usually called impersonals explicitly because their existential label, as well as the speciic agreement issues they raise, have justiied their separate analysis, but the view of existentials as “impersonalizing devices” (Alonso 2008) is well attested in the literature. Secondly, despite the pronoun’s analysis as a meaningless “expletive” (Williams 1984; Postal & Pullum 1988; Hazout 2004 among others), existential there has also been argued to resemble it and il, in coding a “mental space” (a sort of abstract region) within which the following entity can be accessed (Lakof 1987). he addition of demonstrative pronouns ce / ça / c’ to the class of abstract locational subjects requires a little more explanation. In the syntactic literature, the semantic distinction between pleonastic il and its cataphoric counterparts motivates the structural diferences between the constructions they respectively introduce. For example, Jones (1996: 120) analyzes structures such as the one illustrated in (11) as impersonals, namely: “constructions in which the subject position is occupied by a dummy pronoun il, which does not refer to anything”, but he treats demonstratives structures illustrated in (12) very diferently: “It is postulated that the constructions with ce or ça are not impersonal constructions, but dislocated constructions analogous to elle est arrivée, Marie” [‘she arrived Mary’] (Jones 1996: 128, insertions into brackets mine). he argument in favor of the structural distinction between impersonal and dislocated constructions therefore rests on (i) il’s semantic vacuity and (ii) ça’s cataphoric value.7

comme un complément et non point un sujet.” ‘In the current – very old – expression il faut de la vertu, it is impossible to consider de la vertu as anything else than a complement, and not a subject’ (Brunot 1936: 290, translation mine). 7. Although certainly the most popular, this structural view of impersonals is not the only one attested in the literature. For example, Olsson (1986) uses morphological criteria to deine the impersonal category. He writes: “les verbes et les locutions précédés d’un pronom sujet neutre (il, ce, ou cela/ça) qui n’a aucun rapport avec ce qui précède dans le contexte, ni avec un mot particulier, ni avec le contenu total.” ‘the verbs and locutions preceded by a neuter subject (il ce, or cela/ça) which has no connection with the preceding context, either with a speciic word or the global content.’ Olsson (1986: 29, my translation). In Olsson’s analysis, the construction in (12) is indeed an impersonal.

105

106 Michel Achard

(11) Pour la première fois je sentais qu’il était possible que ma mère vécût sans moi (Proust, M. A l’ombre des jeunes illes en leurs: 648) ‘For the irst time I felt it was possible for my mother to live without me’ (12) bien sûr que la journée ne se passera pas sans pluie. Ce n’était pas possible que ça reste comme ça, il faisait trop chaud (Proust, M. Du côté de chez Swann: 101) ‘Of course the day will not inish without rain. It was not possible it would stay that way, it was too hot’

he meaning of impersonal pronouns will not be discussed here, as it constitutes the main topic of the next section. For now, the brief statement in the introduction that impersonal pronouns code an abstract region within which the entity described in the complement can be identiied suices to challenge their purported vacuity. With respect to (ii), Achard (2010) argues that ça cannot be considered a cataphoric pronoun because it refers in equal parts to the preceding and following context. A comparison with ceci ‘this’ reveals that ça lacks the “strictly cataphoric” sense illustrated in (13) other demonstratives display: (13) Revenant l’autre soir du théâtre avec Robert, nous passons près d’un groupe d’agents qui causent entre eux à mi-voix, et ceci parvient jusqu’à nous dans le grand silence de la rue déserte: “il lui a ilé un coup de lame.” (Green, J. Journal. T. 5. 1946–1950: 210) ‘On the way back from the theater the other night with Robert, we walked past a group of policemen talking to each other in low voices, and this comes to us in the silence of the deserted street: “he cut him with a blade.” ’

In the example in (13), the presence of the punctuation marks between the pronoun and the entity it refers to (the referred entity) underscores the high level of semantic independence this entity enjoys with respect to its surroundings. he previous context only allows the hearer to expect an utterance from one of the participants in the observed group, but provides absolutely no insight concerning its speciic content. he pronoun therefore does not directly refer to any element of the preceding context. Its exclusive and unambiguous function is to announce the upcoming entity. Ça difers from ceci in that it does not have a strict cataphoric sense. As a general rule, the referred entity is very closely integrated in the surrounding context semantically. he most extreme case of semantic integration is illustrated in (14) where the referred entity is lexically identical to a segment of the preceding discourse (both entities are underlined in the example). In this case, the referred entity could be viewed as lexically redundant, and thus omitted without greatly impacting the meaning of the passage:

Abstract locational subjects 107

(14) Promenons-nous. Amusons-nous tant qu’il nous reste de la chair sur les os.” Il haussa les épaules: “tu sais bien que ça n’est pas si facile de s’amuser. (Beauvoir, S. de. Les mandarins: 93) ‘Let’s go for a walk. Let’s have fun while we still have lesh on our bones.” He shrugged: “you know that it is not so easy to have fun.’

In other instances, the referred entity expresses a reformulation of the previous context, which oten involves reanalysis, as in (15), where le péché par omission ‘a sin by omission’ constitutes a recapitulation and generalization of the situation described in the previous context:8 (15) Scriassine vida son verre d’un trait: “et c’était si simple de lui dire de s’asseoir; quand on pense à tout ce qu’on pourrait faire et qu’on ne fait pas! Toutes les occasions qu’on laisse échapper! On n’a pas l’idée, pas l’élan; au lieu d’être ouvert on est fermé; c’est ça le plus grand péché: le péché par omission.” (Beauvoir, S. de. Les mandarins: 70) ‘Scriassine inished his glass in one swallow: “and it was so simple to tell him to sit down; when you think about all we could do and don’t do! All these opportunities we waste! We lack the idea, the momentum; we are closed rather than being open; it is the greatest sin: the sin by omission’

he examples in (14) and (15) clearly indicate that the high level of semantic integration between the entity which follows the pronoun and the context that precedes it makes it impossible to consider ça’s cataphoric sense independently of its anaphoric function. his dual reference characteristic of ça is particularly clear in the example in (16) where the pronoun could be analyzed either as referring forward to the following entity, or back to some aspect of the immediate context: (16) Une morale de l’universel, on peut tâcher de l’imposer. Mais le sens qu’on donne à sa vie, c’est une autre afaire. Impossible de s’en expliquer en quatre phrases: il faudrait amener Lambert à voir le monde avec mes yeux. Henri soupira. C’est à ça que ça sert la littérature: montrer aux autres le monde comme on le voit… (Beauvoir, S. de. Les mandarins: 255) ‘One can try to impose a universal morality. But the meaning one gives to one’s life, it is something else. Impossible to explain this in four sentences. I would have to bring Lambert to see the world through my eyes. Henri sighed. It is what literature is for: showing others the world as one sees it.’ 8. he shared information between the referred entity and the previous context can also be provided by the set of inferences made available by well-established schemas, or the larger context of the sentence (Achard 2010).

108 Michel Achard

In (16), ça is traditionally analyzed as cataphorically referring to the following clause montrer aux autres le monde comme on le voit ‘showing others the world as one sees it’. However, the pronoun could just as easily anaphorically refer to the previous context, because that following clause represents the generalized reformulation of the preceding amener Lambert à voir le monde avec ses yeux ‘bring Lambert to see the world through his eyes’. Examples such as (16) clearly show that the boundary between ça’s anaphoric and cataphoric senses is blurry at best, and that the recognition of this dual reference provides crucial insight into the pronoun’s meaning.9 he examples in (13)–(16) underscore the need for an analysis that accounts for the diferences between ceci and ça. Smith (2000) provides an insightful starting point by proposing that in English examples such as ‘I despise (it) that John voted for the governor’, cataphoric pronouns serve a “space designating function” (Smith 2000: 486) by which they “anticipate the mental spaces set up by space builders” by “designating the spaces themselves in the grammar” (Smith 2000: 487; see also Fauconnier 1985).10 Smith’s analysis is directly applicable to ceci’s strict cataphoric sense, as we saw in the example in (13), where ceci codes the abstract nominal entity which announces the utterance to follow, and Il lui a mis un coup de lame ‘he cut him with a blade’ elaborates this abstract setting. Such strict cataphorics constitute a limiting case, because the mental space which announces the upcoming entity is maximally abstract and devoid of speciic contextual content. In a slight elaboration of Smith’s proposal, Achard (2010) notes that there are no limits as to the possible internal structure of the space the pronoun codes and that it may be contextually elaborated at diferent degrees. Rather than empty and maximally abstract, it may contain a wide range of elements already present in the context or easily inferable from it. Consequently, in the examples illustrated in this section, ça resembles strict cataphorics in coding a mental space to be elaborated by its following antecedent, but it difers from them in the level of contextual elaboration that space contains. At the opposite end of the continuum from the strict cataphorics, the mental space ça designates is fully inclusive of the discourse context within which the referred entity is extracted. he reasons for that extraction essentially pertain to the author’s narrative strategies, and may include the need to highlight the general value of a previously expressed position as in (16), or to provide a summarizing statement of a character’s emotional state following the description of the conditions that induced that state as in (17): 9. he abundant literature on ça has emphasized the lexibility of pronoun’s anaphoric reference (Cadiot 1988; Carlier 1996; Achard 2000, 2010 among others). 10. Smith also analyses German and Russian cases which will not be considered here.

Abstract locational subjects 109

(17) il gagnait du terrain en province; et ce qu’il y avait de réconfortant, c’est que les communistes ne l’attaquaient plus: l’espoir d’une union durable se réveil lait. C’est à l’unanimité que le comité décida en novembre de soutenir horez contre De Gaulle.“Ça facilite bien la vie de se sentir en accord avec ses amis, ses alliés, avec soi-même”, pensait Henri … (Beauvoir, S. de. Les mandarins: 231) ‘he was gaining ground outside Paris; and what comforted him the most was that the communists weren’t attacking him any more: the hope of a lasting alliance was awakening. he committee unanimously decided to support horez against De Gaulle in November. “It makes life easier to feel in agreement with one’s friends, one’s allies, and oneself ”, Henri thought…’

In (17), the mental space ça designates contains the abstract region composed of all the positive elements relative to Henry’s relationships with his friends and his enemies which contribute to his emotional stability. he ininitival clause which follows the pronoun represents their reformulation in a more synthetic form. A purely cataphoric analysis of ça which would obscure the role of the preceding context would therefore fail to adequately describe the pronoun’s semantic import. Ça does not exclusively designate the ininitival clause that follows, but also the abstract region composed of the preceding context from which the proposition is extracted. he examples presented in this section clearly indicate that il and ça cannot be respectively analyzed as pleonastic and cataphoric pronouns. On the contrary, the meaning of demonstratives strongly resembles that of it, il, and there in describing an abstract region within which the entity coded in the complement can be extracted. he next two sections examine the natural class these four pronouns form by respectively presenting the CG analysis of abstract locational subjects, and illustrating the need for further elaboration of the diferent kinds of regions the four pronouns respectively code. he recognition of a class of subjects which proile a speciic kind of abstract location necessarily intersects with the treatment of impersonal constructions, because the pronouns which code these regions (most notably it and il) unquestionably perform impersonal functions. However, the internal coherence of that class needs to be kept separate from the possible function of the forms which compose it as impersonals, and thus, by extension, from any attempt at deining the impersonal category. In fact, the implication of the analysis of it, there, il, and ça as a natural class to our understanding of impersonals is limited to the claim that because of their semantic and syntactic similarities, the four constructions illustrated in (1)–(4) should be considered impersonals (Achard 2010). More generally, however, the importance of this claim cannot be evaluated outside of the speciic view of the category diferent theoretical models espouse. It is perhaps most farreaching for purely syntactic accounts which restrict the scope of impersonals

110 Michel Achard

to constructions with pleonastic subjects, because it challenges the cornerstone of such analyses in French, namely the structural distinction between il and the demonstrative pronouns. On the other hand, the analysis developed in this chapter is certainly compatible with the more inclusive treatment of impersonals developed in the Functional literature (Siewierska 2008a, 2008b; Helasvuo & Vilkuna 2008; Divjak & Janda 2008; Sansó 2005; Slón 2007), because abstract locational subjects illustrate a speciic way of defocusing the agent (Shibatani 1985).11 It is therefore important to remember that despite its relevance to more global concerns, this chapter is primarily concerned with it, there, il, and ça as abstract locational subjects, and will therefore necessarily overlook the more general issues which pertain to impersonals at large.

3. he abstract locational subject analysis We are now in a position to examine the CG analysis of abstract locational subjects. his section begins with a brief exposition of the tenets of the model most directly relevant to the concerns of this chapter. 3.1

Language and human experience: Clause structure

In Cognitive Grammar, language forms are grounded in human experience. As a result, the conceptual base relative to which clause structure can be described consists of a relatively small number of very general models and archetypes which represent speciic aspects of our conceptual organization, and can be exploited for purposes of linguistic expression (Langacker 2008: Chapter 11). he irst archetype pertains to “the organization of a scene into a global setting and any number of smaller, more mobile participants” (Langacker 2008: 355, emphasis in the original) which interact with each other, and may occupy diferent locations at diferent times. he notion of interaction between the participants necessarily involves

11. Siewierska (2008a) notes that in the syntactic literature, “impersonal constructions are seen to either lack a grammatical subject altogether or alternatively feature only a pleonastic (semantically empty) subject, be it an overt one or potentially a covert one (Siewierska 2008a: 4). By contrast, functional models are more prominently concerned with the semantic value of impersonals and thus tend to consider as such constructions that “lack a deinite human agent as subject” (Siewierska 2008a: Note 3), but it is diicult to distinguish them from other constructions such as anticausatives or unaccusatives which perform similar functions. For an attempt to combine the functional perspective with some selectional criteria to select an impersonal class in French, see Achard (2010).

Abstract locational subjects

reference to the “billiard-ball model” which represents “our conception of objects moving through space and impacting one another through forceful physical contact” (Langacker 2008: 355), as well as our knowledge that some objects possess the resources necessary to provide the required energy, while others can only transmit or absorb it. he archetypal conception of an “action chain” represents “a series of forceful interactions, each involving the transmission of energy…from one participant to the next.” (Langacker 2008: 355–356). In addition, the roles of the participants in the described event are established with respect to diferent conceptual “archetypal roles” available for linguistic purposes. For instance, “an agent is an individual who willfully initiates and carries out an action, typically a physical action afecting other entities. It is thus an “energy source” and the initial participant in an action chain.” (Langacker 2008: 356). At the opposite end of the action chain, the patient role is deined as “something that undergoes an internal change of state…Typically inanimate and nonvolitional, a patient usually changes as the result of being afected by outside forces. It is then an “energy sink” and the inal participant in an action chain” (Langacker 2008: 356, emphasis in the original). Another archetype relevant to the characterization of clausal structure pertains to the manner in which we perceive the diferent facets of the world around us. he “stage model” likens the conceptualizers’ apprehension of a scene for purposes of linguistic expression to that of spectators watching a play unfold. Langacker (2008: 356) describes the analogy in the following manner: “the maximal ield of view, the onstage region, and the focus of attention correspond respectively to an expression’s maximal scope, immediate scope, and proile.” Closely related is a group of archetypes which pertain to the speech event itself, and involves models of speaking, listening, and engaging in social interaction, as well as diferent models of viewing arrangement which describe the perceptual asymmetry that exists between the subject of perception (conception) and the perceived (conceptualized) object. Speciic conigurations may arise as the interaction unfolds, but the default arrangement consists of “two interlocutors being together in a ixed location, using a shared language to describe occurrences in the world around them” (Langacker 2008: 357). he canonical event model integrates these interconnected archetypes into “what is arguably the most typical kind of occurrence” (Langacker 2008: 357), or more precisely “… a bounded, forceful event in which an agent (AG) acts on a patient (PAT) to induce a change of state. his event is the focus of attention within the immediate scope or onstage region (IS), being apprehended from ofstage by a viewer (V) not otherwise involved in it. All of this unfolds within some global setting” (Langacker 2008: 357). he canonical event model is illustrated in Figure 1, where MS stands for maximal scope of predication and IS for immediate scope of predication.

111

112

Michel Achard

MS

setting IS AG

PAT

V

Figure 1. Canonical event model (from Langacker 2008: 357).

Finally, the “control cycle” (Langacker 2002, 2004) describes a large number of perceptual, physical, social, or mental control events. At any moment (the baseline), an actor (A) has control over a certain number of entities which collectively constitute her dominion (D). In the next phase, a target (T) enters A’s ield of potential interaction (F), thereby creating a state of tension. In order to resolve the tension, A may choose to exert force onto T to bring it into her dominion. he result, of this action phase is a modiied static dominion which incorporates the newly acquired element. he four phases of the model are illustrated in Figure 2. Potential

Baseline

A

T

A

D

D

F

F



Action

Result

A

A

T

T

D

D F

F



Figure 2. he control cycle (from Langacker 2004: 536).

Abstract locational subjects

Although diferent manifestations of this cycle constantly unfold in various domains of human experience, we will pay particular attention to epistemic and social control for the purposes of this chapter. Epistemic control pertains to the “acquisition of propositional knowledge” (Langacker 2009: 131). he actor is a conceptualizer, the target a proposition which represents a facet of the world around her, and the dominion is her view of reality, composed of the propositions she holds true (Achard 1998, 2002). Social control describes the manipulation of other individuals’ behavior, according to a set of expectations and obligations. Certain types of clauses are particularly well suited to express certain types of events. Among all the possible types, the transitive clause, which codes the interaction between an agent designated as the trajector (tr) of the proiled relation (the most focal igure) and a patient coded as its landmark (lm) (the second most focal igure), is perhaps the most common, because it allows the participants with the highest degree of cognitive salience (the agent and patient), to also be treated as the two most focal igures in the linguistic representation of the conceptualized event. his alignment between focal prominence and semantic roles is illustrated in (18) in both the French construction and its English translation, and represented in Figure 3: (18) En quinze jours, la 1ère armée a tué 10000 allemands, fait 18000 prisonniers, enlevé 120 canons. (Gaulle, C. de. Mémoires de guerre, le salut: 136) ‘Within two weeks, the irst army killed 1000 Germans, captured 18000 prisoners, removed 120 guns.’ MS

setting IS tr

lm

G

Figure 3. Coding of a transitive clause (from Langacker 2008: 357).

A transitive clause might represent the unmarked, possibly even prototypical way of coding events such as the one presented in (18), but it does not constitute the only alternative. he coding of the conceptualized scene is a matter of construal, and speakers thus have the possibility of giving focal prominence to any entity they choose. he multiple clause types available in various languages illustrate the lexibility of construal because they provide conventionalized alternatives for the

113

114 Michel Achard

description of scenes with similar content. he impersonal constructions considered in the course of this chapter depart from the transitive construal illustrated in Figure 3 with respect to subject selection, or in other words the choice of the most focal igure in the proiled relation. 3.2

Toward setting subject constructions

Speakers may decline to select the agent as the linguistically most salient igure in the proiled relation despite its inherent cognitive salience for a number of reasons. hey may want to protect it from undesirable consequences, ind it irrelevant to the description of their conceptualization, or simply do not know its real nature. Consequently, languages provide a variety of ways of defocusing the agent (Shibatani 1985) by removing the spotlight that its focal role in the proiled process inherently focuses on it. Cross-linguistically, passive and middle constructions represent some of the most frequently attested ways of defocusing the agent by selecting the patient as the most focal igure in the proiled relation (Langacker 1982, 2006). Any entity which composes the conceptualized scene is available for selection as most focally prominent for linguistic purposes. In particular, the location within which the proiled event takes place can be selected as the focal igure, and thus marked as subject. Two diferent construals which result in the selection of alternative entities as the trajector of the proiled relation are illustrated in the examples in (19) and (20), respectively represented in Figures 4a and 4b. (19) …des ombres douteuses grouillaient sur les voûtes souillées. (Gracq, J. Le rivage des Syrtes: 299) ‘…dubious shadows were crawling on the soiled vaulted ceilings.’ (20) La chambre, ou cellule, où il se trouvait, grouillait d’hommes et de femmes de blanc vêtus. (Beckett, S. Malone meurt: 137) ‘he bedroom, or cell, where he was located was crawling with men and women in white.’

he examples in (19) and (20) describe similar scenes where a group of participants is perceived as crawling within a certain location. However, their linguistic representations relect diferent structurations of that scene by selecting alternative entities as the trajector of the proiled relation. In the intransitive clause in (19), the participants (P) are selected as the most focal igure (tr) and thus marked as subject. he location within which the crawling takes place is not treated as a focal participant and coded by the oblique complement introduced by the preposition sur ‘on’. In (20) by contrast, the location within which the

Abstract locational subjects

crawling takes place (la chambre, ou cellule, où il se trouvait ‘the room or the cell where he was located’) is selected as the most focal igure of the proiled relation and therefore coded as subject. In this structuration, the participants are not focal igures, and thus coded as an oblique introduced by the preposition de. In the locational subject construction represented in (4b), the (solid) line between the location and the process indicates that the proiled process is somehow crucially associated with that location. In (20), the location (the bedroom) is a relatively restricted area. Compared to a location, a setting is “a global expanse within which events unfold (the diference is one of degree).” (Langacker 2009: 118). Settings can also function as clausal trajectors, as illustrated in (21)–(23), and represented in Figure 4c. Whereas voir ‘see’ normally selects an experiencer as its subject, the focal participant in (23) is the spatial setting (France) within which the event coded by the object nominal can be experienced. Such “setting subject constructions” do invoke a conceptualizer, but a “generalized one” (Langacker 2009: 118), or more precisely, in the case of (23), anyone with suicient knowledge of art history to identify the rebirth of the interrupted trend.12 (21) “Our barrios are crawling with drugs,’’ said Bethancour, who has reared six children and helps raise a dozen grandchildren in El Chorrillo.‘Young children are using drugs like never before.’ (LATWP) (22) Forty-ive Georgia counties have been placed in a state of emergency. hirtyone counties have been declared federal disaster areas. “Over the past six days Georgia witnessed a natural disaster like no one has ever seen before in our history,’’ Gov. Zell Miller said in a news conference at the capitol in Atlanta. (LATWP) (23) À partir du xviiie siècle, un sourd travail devient perceptible… La France voit une école bourgeoise renouer avec la tradition interrompue de ses propres “peintres de la réalité”… (Huyghe, R. Dialogue avec le visible: 154) ‘Starting in the 18th century, an obscure current begins to surface…France sees a bourgeois school reconnect with the interrupted tradition of its own “reality painters”…’

12. he setting can also be temporal, as in the example in (i): (i) Last year witnessed a dramatic surge in E. coli 0157:H7 cases, the most striking of which involved 700 Jack-in-the-Box hamburger consumers in California, Washington, Idaho and Nevada: 178 were hospitalized, 56 sufered severe illnesses that led to lasting ailments and four children died. It was one of 16 outbreaks in 1993. (LATWP)

115

116 Michel Achard

tr

tr E

P

E location

a. Intransitive clause

tr

b. Locational subject

setting

lm

c. Setting subject

Figure 4. Shit in construal from participant to setting (4b and 4c are from Langacker 2009: 118).13

3.3

Impersonals

As indicated earlier, impersonals are closely related to setting subject constructions because, just like these constructions, they do not select the participants as the focal igure in the relation the predicate describes.14 However, the entity impersonals select is considerably more abstract than the temporal or spatial setting within which the process is carried out. In order to characterize it precisely, we need to recall that in the Control cycle model represented in Figure 2, agents can only exert their dominion over their target if the latter is within their reach, or in other words, contained within the ield that permits potential interaction with it. In the perceptual and physical domains, that ield can be respectively characterized as the perceptual and kinesthetic range within which agents can exert the necessary force to capture their target. In more abstract domains which are much more inclusive, it is more diicult to outline precisely, and its deinition is understandably much more general: “More abstractly, it might be interpreted as 13. In Figure 4, P stands for participant and E for the experiencer who perceives the stimulus the location or setting provides. In 4c, the experiencer is represented inside the setting to indicate the latter’s global nature. 14. Consequently, just like setting subject constructions, impersonals do not passivize since passivizability correlates with transitivity, and the constructions which passivize the most felicitously are those that involve the interaction between the participants in an action chain (or analog) respectively coded as trajector and landmark. his is the case in English and French, as illustrated in (i)–(iv): (i) (ii) (iii) (iv)

*Big drops are being rained (by it) *A natural disaster was witnessed by Georgia *Une école bourgeoise a été vue par la France *De grosses gouttes sont plues (par il)

he parallel behavior of setting and impersonals to passivization has led to the treatment of impersonals as “abstract setting” constructions.

Abstract locational subjects

indicating the relevant scope of awareness, i.e. everything invoked in apprehending the situation described.” (Langacker 2009: 138). his scope of awareness obviously depends on the precise nature of the domain considered. For example, in (24), it is mainly restricted to the perception of the ground, but the emphasis may be placed on more experiential factors: (24) We can’t go through here, it’s all wet.

In (25) and (26), the relevant scope of awareness is composed of all the factors which allow a generalized conceptualizer to experience the scene described. hese factors may involve perceptual, mental, social, or psychological elements which all contribute to the experience the scene evokes. (25) It’s quiet in the countryside

(From Langacker 2009: 138)

(26) It’s embarrassing when you can’t remember somebody’s name15

One kind of mental experience consists in making the kind of epistemic judgments by which we continuously enrich our conception of reality, as illustrated in (27) and (28): (27) It is true that there was a joint decision by the Council and Parliament seeking to discontinue emergency lines. (28) It is perfectly correct that we need to apply this in certain cases.

In cases such as those in (27) and (28), the scope of awareness is maximally inclusive and incorporates everything that the conceptualizer takes into account to make the epistemic judgment the predicate codes. By analogy with the perceptual and physical domains, Langacker (2009: 139) describes the ield in abstract domains as a conceptualizer’s “mental reach” with respect to a certain entity: “At higher levels of cognition, the ield is much harder to delineate, given our extraordinary mental capabilities. But by analogy to the physical and perceptual levels, we can describe the ield for higher-level cognitive processes as comprising everything a conceptualizer is capable of apprehending at a given moment, or everything apprehended for a given purpose. Metaphorically, it is the conceptualizer’s “mental reach”” (Langacker 2009: 139). his most inclusive conception of the ield constitutes the cornerstone of the CG analysis of impersonals. In a proposal which recapitulates and synthesizes the previous research in the area, Langacker suggests that impersonal constructions select the ield as the focal igure in the proiled relation, and that ield is

15. It is worth noting that the French equivalent of it in examples such as (24)–(26) is c’ and not il since il is no longer a neuter demonstrative in Contemporary French (see note 17).

117

118

Michel Achard

represented by the impersonal pronoun: “I thus propose, as a general characterization, that impersonal ‘it’ proiles the relevant ield, i.e. the conceptualizer’s scope of awareness for the issue at hand” (Langacker 2009: 139). Achard (2009, 2010) adopts this analysis to French, and claims that il in (29) codes the speaker’s mental reach relative to the existence of the alternative edition, or in other words the knowledge base composed of all the facts which allow him to assert the edition’s existence.16 (29) le désespéré que vous avez eu tant de peine à vous procurer, dites- vous, est, sans doute, l’édition Soirat. Il en existe une autre qui vient de paraître, à mon insu et sans mon autorisation… (Bloy, L. Journal: 74) ‘he desperate [person] you say you had such trouble inding is most likely the Soirat Edition. here exists another one which was just released unbeknownst to me and without my permission.’

he speciic structure impersonal pronouns such as it and il code is represented in Figure 5:

tr P F setting

Figure 5. Impersonal pronouns.

In Figure 5, the ield is marked F and indicated with a broken rectangle. It is selected as the trajector of the proiled process and thus marked as subject. he process [exister ‘exist’ in (29)] is also proiled. By contrast, the only participant marked as P [the new edition in (29)] is let unproiled at this level of organization. As previously indicated for setting constructions, the (broken) line between the process and the ield indicates a tight association between the two entities. he nature of that association, however, deserves further attention. On the one hand, it seems fairly straightforward. he very deinition of the ield confers to the latter a high level of responsibility with respect to the process coded by the predicate, because that process could only be uncovered against its background. On the other 16. In the French tradition, the CL analysis of the subject of impersonals is anticipated in the works of Galichet (1947) and Séchehaye (1950) for example.

Abstract locational subjects

hand, the ield was described as a given conceptualizer’s scope of awareness with respect to the proiled process, and it is therefore not immediately obvious how one individual’s knowledge base concerning a process might be responsible for its occurrence. In order to reconcile these seemingly diverging observations, we need to remember that even though the conceptualizing experience which leads to the identiication of the alternative edition is indeed that of the speaker (or any relevant conceptualizer), it is not presented as such. In fact, one of the roles of impersonal constructions is precisely to present the conceptualized scene as the product of a “generalized conceptualizer” (Langacker 2009: 115). he basic idea is that any person in a similar position would invariably reach the same conclusion. For example, in (29), anyone with suicient knowledge of the author’s life and works would unfailingly identify the alternative edition. In this sense, the missing address can be considered a property of the ield.17 he notion of abstract locational subjects introduced in this chapter subsumes at the same time subjects that proile either the setting or the ield, or in other words, place focal emphasis on some kind of region (at diferent levels of abstraction) within which the proiled relation can be apprehended. In order to distinguish these constructions from those which code the participant in the 17. A possible objection to adopting Langacker’s analysis to French, and thus treating it and il in a similar manner, comes from the distribution of these pronouns in other contexts. Synchronically, it also functions as a neuter 3rd person singular pronoun. In Contemporary French, a demonstrative (ce / ça, c’ perform that function, while il in addition to its impersonal sense also marks a 3rd person singular masculine agreement on the predicate. his apparent diiculty, however, is quickly solved diachronically. he pronoun il comes from the Latin neuter third person demonstrative illud which might be translated in French by cela ‘this, that’. In Old French, the most common impersonal marker is Ø, but beginning in the 12th century, il gradually becomes more and more grammaticalized, to become generalized by the 16th century (Brunot 1936: 285). During that time, il was a neuter pronoun which didn’t exclusively mark masculine referents, but also more general propositional referents, as illustrated in (i): (i) C’estoit jadis chose bien rare, Que de veoir un abbé ignare: Aujourd’ huy il est si commun, Que cent mille, aussi bien comme un, Se trouveront. (Marot, C. [1526]. Les traductions: 171) ‘It was in the past a very rare thing to see an uncultured abbot: Nowadays it is so common, that you can ind one hundred thousand as easily as one’ Since il was a neuter pronoun at an earlier stage of the language (and still exhibits frozen traces of this usage), it is possible to suggest that the pronoun became specialized in two opposite directions. On the one hand it became a masculine 3rd person subject pronoun, and on the other hand, it became exclusively specialized as an impersonal. From this standpoint, even though the French and English systems synchronically diverge, their diachronic similarity, especially their shared characteristic of having a neuter pronoun used as an impersonal at some point in their history justiies treating il in a way parallel to ‘it’.

119

120 Michel Achard

main relation as the focal igure, and further emphasize the subtle distinction between the setting and the ield, compare the three examples in (30) (adapted from Langacker 2009: 143): (30) a. I am cold here in Chicago b. It’s cold in Chicago c. Chicago is cold

he three constructions in (30) and illustrated in Figure 6 present essentially the same conceptual content, namely an experiencer (E), who (by deinition) possesses a scope of awareness (F) which includes the ambient environment where the sensation (cold) is experienced, within a speciic locational setting (Chicago). In each case, the trajector is speciied by the nominal which therefore functions as the subject. he three constructions “difer primarily in which facets of this situation they highlight through proiling and choice of trajector.” (Langacker 2009: 143). In (30a), the experiencer and the experience itself is emphasized by the choice of the experiencer as subject, but the cause is let implicit. he constructions in (30b) and (30c) shit the primary emphasis away from the experiencer and place it either on the scope of awareness (ield) the pronoun ‘it’ codes, or on the spatial setting expressed by ‘Chicago’. In both cases, the experiencer is generalized and defocused. he shit in focus emphasizes the trajector’s responsibility in causing the cold sensation. he contrast between (30b) and (30c) “is a matter of whether the trajector is identiied as the experiential ield per se or as the spatial setting with which it is largely co-extensive” (Langacker 2009: 144). Achard (2010) argues that the distinction between ield and setting plays a crucial part in the selection of il or ça constructions in French, and this distinction will be considered in further detail in Section 4.3. tr tr

tr

E

E

setting

setting (a)

F

F

F setting

E

(b)

(c)

Figure 6. Personal, ield, and setting constructions (adapted from Langacker 2009: 143).

Abstract locational subjects

3.4

Need to further specify the notion of abstract locational setting

he treatment of impersonal pronouns as meaningful structures which code some abstract mental construct or location within which the event/proposition in the complement can be conceptualized represents a crucial insight which has been expressed in various (compatible) forms in a variety of languages. For example, over the years, they have been described as designating: “a deinite nominal with almost the greatest possible generality of meaning (Bolinger 1977 [English it]), a “general presence or availability” (Kirsner 1979 [Dutch]), “a mental space” (Lakof 1987 [there]; Smith 2000 [German es], both following Fauconnier’s (1985) model), “an abstract setting” (Langacker 1993 [English it]), and “the abstract setting identiiable as the immediate scope for the existential predication (Achard 1998 [French il]). Although I believe the basic insight represented in these various formulations is fundamentally correct, the current CG analysis is not, in its current state, capable of accounting for the distribution of the French and English pronouns because it cannot adequately capture the ine grained distinctions between the different forms. Language internally, it is impossible to precisely distinguish between the possible forms which perform similar functions. In English for instance, Lakof ’s proposal that existential there “designates a mental space in which a conceptual entity is to be located” (Lakof 1987: 542) is not distinctive enough from Langacker’s suggestion that it proiles “the conceptualizer’s scope of awareness for the issue at hand” (Langacker 2009: 139) to convincingly capture the meaning diferences between the two pronouns. Cross-linguistically, the assumption made in Achard (2010) that il is similar to it in that it also proiles the ield within which the event or proposition coded in the complement can be located is diicult to maintain without further precision, given the diferent distributions the two forms exhibit. he remainder of this chapter irst shows that in order to be speciic enough to distinguish between it and there, the ield, or mental reach within which the impersonal process is conceptualized needs to be characterized with respect to diferent levels of reality (Achard 1998); it and there both correspond to French il because the latter operates at both levels of reality. Secondly, the distinction between il and ça relects a subtle distinction between a ield and a setting construction, where the setting is more readily deined by the current discourse context than the conceptualizer’s mental efort to assess the epistemic status of the event / proposition in the complement.

121

122 Michel Achard

4. Diferent kinds of abstract regions his section argues that the four pronouns presented in this chapter code different kinds of abstract regions. In a nutshell, there designates a region of “basic reality”, while it exclusively describes a region of “elaborated reality” (Langacker 1991; Achard 1998). Il is compatible with both English pronouns in diferent circumstances because it operates at both levels of reality (Achard 1998). Finally, ça designates an “abstract setting” composed of the circumstances of the immediate context (Achard 2010). 4.1

Two levels of reality

he analysis of French impersonals proposed in Achard (1998: Chapter 7, see also Achard 2009) distinguishes simple impersonals, namely constructions where the post verbal expression is a nominal, from complex ones, where the post verbal expression takes the form of a inite or ininitival clause. Simple impersonals essentially pertain to the coming into existence (or awareness) of a certain entity (Achard 1998: 283). his entails asserting the existence of the post verbal entity as in (31), its location (32), arrival onto the conceptualized scene (33), or the overall impression which emanates from that scene (34). he examples in (31), (33) and (34) are from Achard (2009). (31) … Et puis, j’incarne un modèle de femme qu’on voit rarement au petit écran. Il existe de nombreux modèles masculins (Dechavanne, Ardisson, Rapp), mais peu de femmes… (AFP) ‘….Plus the fact that I represent a type of women you don’t oten see on TV. here exist many such males (Dechavanne, Ardisson, Rapp), but few women’ (32) Je suis montée avec elle et pendant qu’elle était dans le cabinet de toilette, j’ai ouvert le sac à fermeture éclair: au fond il y avait une petite iole brunâtre que j’ai enfouie dans ma poche. (Beauvoir, S. de. Les mandarins: 420) ‘I went up with her and while she was in the bathroom, I opened the bag with the zipper: At the bottom there was a little brownish bottle which I put in my pocket.’ (33) Entends- tu ce que je te dis? Joseph hocha la tête. – tu attendras dans les bois que la nuit tombe. À ce moment, tu descendras dans le ravin et tu rejoindras la route. Tu attendras encore, une heure s’ il le faut. Il passera une voiture qui ralentira et s’arrêtera à la hauteur du ravin pour te laisser monter. (Green J. Moira, roman: 244)

Abstract locational subjects

‘Can you hear what I am telling you? Joseph nodded. You will wait in the woods until nightfall. At that time, you will go down the ravine toward the road. You will wait again, one hour if you have to. here will pass a car that will slow down and stop by the ravine to let you in.’ (34) Sur sa table de travail un poudrier d’argent était ouvert, laissant voir une houppe ronde et blanche, pareille à un petit nuage. Il lottait entre ces murs une odeur afreusement douce et grisante qu’il s’eforça de ne pas respirer, une odeur de lilas. (Green J. Moira, roman: 158) ‘On his desk, a silver powder box was open, revealing a puf round and white like a little cloud. ‘here loated between these walls a horribly sweet and enticing odor that he tried not to breathe in, an odor of lilac.’

As was mentioned earlier, the ield the pronoun il codes can be described as the mental reach which allows the conceptualizer to experience the entity coded in the postverbal expression. In purely existential cases such as (31), the latter is identiied with respect to the highly abstract concept of reality. Speciic instances of the category under investigation (a particular kind of male TV presenter) are identiied within the French TV programs currently available, without any consideration of their speciic location. he ield the pronoun il proiles is therefore best described as the knowledge of the French TV landscape necessary for the identiication of speciic presenters as instances of the relevant category. It is considerably less abstract in the case of location because it is inherently tied to a speciic physical area. For instance, in (32), the bottle can only be located within the expanse of the bag, and in (34), the enticing smell is conined to the physical scene from which it emanates. However, even in these more concrete cases, the ield cannot be restricted to the physical location which contains the entity expressed in the post verbal expression, because it also crucially codes the mental constructs which enable the conceptualizer to search the area of interest and assess the importance of the located entity. Complex impersonals by contrast express the deontic, epistemic, or evaluative interpretation of the event / proposition the complement clause describes, as illustrated in (35)–(37):18 (35) Oui, j’ai honte. J’ai visé trop haut. Il faut que je travaille à ma place. Une toute petite place. La seule dont je sois digne. (Camus, A. Les justes: 346) ‘Yes, I am ashamed. I aimed too high. I have to work on keeping my place. A very small place. he only one I am worthy of.’ 18. French impersonal can contain single predicates as illustrated in (35), or copular predicates composed of the copular être ‘be’ followed by an adjectival form as illustrated in the other examples. Because English impersonals are copular, only copular forms will be considered in the remainder of this chapter.

123

124 Michel Achard

(36) Nuançant néanmoins son propos, M. Carlot a tenu à préciser que cela ne signiiait pas pour autant qu’il se rangeait derrière des organisations comme Greenpeace. “Il est certain que le monde libre a besoin du nucléaire, mais Vanuatu ne signera pas le traité de non-prolifération. (AFP) ‘Sotening his stance somewhat, Mr. Carlot speciied that this didn’t mean he was siding with organizations such as Greenpeace. “It is certain that the free world needs nuclear power, but Vanuatu will not sign the non-proliferation treaty.’ (37) “Je sentis alors, écrit-il, qu’il n’est pas toujours aussi aisé qu’on se l’imagine d’être pauvre et indépendant. (Guéhenno, J. Jean-Jacques. T. 2.: 23) ‘I then felt, he wrote, that it is not always as easy as one imagines to be poor and independent.’

Complex impersonals difer from their simple counterparts in that the event / proposition in the complement is oten already considered true. his is particularly clear with evaluative predicates illustrated in (37) where the proposition être pauvre et indépendent ‘be poor and independent’ has already been accepted as real. In order to describe the diference between the simple and complex cases, two separate (yet integrated) levels of reality need to be recognized (Langacker 1987a; Achard 1998). A basic level of reality (R) is composed of objects, the existence of which can be questioned or assessed. At this level, events also occur, and we can question or observe their occurrence. However, the report of the events of basic reality for the purposes of thought or communication necessarily entails their epistemic evaluation, or in other words, their assessment with respect to a more abstract level of reality (R′) where their occurrence can be characterized. he difference between R and R′ and its ramiications for the analysis presented in this chapter can easily be illustrated by comparing the two examples in (29) repeated here for convenience and (38). In (29), the speaker introduces the presence of a speciic edition in reality. he ield the pronoun proiles includes the immediate circumstances that pertain to his knowledge of that fact. here is no explicit epistemic evaluation. he speaker’s scope of awareness is contained within the level of simple reality (R) and essentially pertains to the objects whose presence is assessed, and the events whose occurrence is observed. In (38) by contrast, the report of a fact as a proposition involves its explicit epistemic assessment. he proposition je suis avec des montagnards, écartés des centres industriels et très fatalistes ‘I am with mountain men, remote from the industrial areas and very fatalistic’ is accepted for integration into the speaker’s conception of reality. In order to perform this conceptual task, the conceptualizer’s scope of awareness needs to expand beyond R (where things happen or fail to do so) to encompass notions of assessment, evaluation, comparison, epistemic location, all notions of R′. he ield il codes therefore designates a region of R′.

Abstract locational subjects

(29) le désespéré que vous avez eu tant de peine à vous procurer, dites-vous, est, sans doute, l’édition Soirat. Il en existe une autre qui vient de paraître, à mon insu et sans mon autorisation… (Bloy, L. Journal: 74) ‘he desperate [person] you say you had such trouble inding is most likely the Soirat Edition. here exists another one which was just released unbeknownst to me and without my permission.’ (38) À voir cela, il me semble que la révolte est plus loin de nous que je ne croyais d’abord. Il est vrai que je suis avec des montagnards, écartés des centres industriels et très fatalistes. (Alain-Fournier, H. Correspondance avec J. Rivière: 120) ‘When I see this, it seems to me that the rebellion is furtherfrom us than I irst thought. It is true that I am with mountain men, remote from the industrial areas and very fatalistic’

he recognition of the two levels of reality is of critical importance for the understanding of the English and French impersonal systems, and more speciically of the respective distributions of there, it and il. As the previous examples in this section illustrated, French, il codes existence at both R and R′. With simple (nominal) expressions, il codes the ield at R, with complex (verbal) expressions, il codes the ield at R′, that is the subsection of R′ within which the assessment relative to the event or proposition coded in the complement can be formulated. In the complex cases, the presence of the impersonal pronoun highlights the separation between R and R′. he reminder of this section argues that English difers from French in that it codes the existence of an entity at R and R′ with diferent forms. More speciically, there codes the ield at R (scope of observation), and it codes the ield at R′ (explicit mental efort of deontic, epistemic, or evaluative assessment). 4.2

here and il: A region of R

his section does not pretend to provide an exhaustive comparison of all aspects of there and il impersonals, but merely to establish their functional equivalence and the fact that they both proile a subsection of R as their subject. In particular, the distributional diferences between the two constructions will be let for further research.19

19. Although both there and il proile the presence of the postverbal entity in reality (existence), they do so in collaboration with diferent predicates. In French, il oten combines with exister, as illustrated in (29), but English there, although attested with exist as shown in (39) much more oten occurs with be in this case. hese distributional issues will not be considered in this chapter.

125

126 Michel Achard

As it has largely been documented in the literature (Lakof 1987; Bergen and Plauché 2005), existential there proiles the presence of the post verbal entity in reality, as illustrated in (39)–(41): (39) In the Foreign Ministry iles, researchers have discovered a 1946 note from the U.S. Embassy that speaks of large-scale eforts to sneak Nazis into Argentina. “here exists a concerted plan to arrange the clandestine departure from Spain and entry into Argentina of former German agents,” the note says. (LATWP) (40) “his is a tremendous breakthrough,” said Dr. Daniel Weedman, NASA’s director of astrophysics. A longtime black hole skeptic, he is now a convert. “I do believe there is a supermassive black hole at the center of (M87),” he said. (LATWP) (41) he word bimbo reinforces the stereotype of a woman using sexuality to promote herself. hat there is no equivalent male term is signiicant, but not as important as the fact that we seem willing to use the word bimbo unquestioningly. (LATWP)

he functional equivalence between the English and French constructions is best illustrated by the parallel corpus examples in (7) repeated here, which both describe the TV landscape in the European Union. Both forms proile the mental construct within which the presence of the channels can be ascertained. In both cases, this construction represents the subsection of reality R where knowledge about TV stations in the relevant countries is available. Because the abstract region the pronoun designates is part of basic reality R, the mere existence of the relevant entities is at issue (i.e. their presence in reality), rather than their epistemic evaluation: (7) Il y a bien deux chaînes innoises et une chaîne portugaise, mais il n’y a toujours aucune chaîne néerlandaise. (7′) Although there are now two Finnish channels and one Portuguese one, there is still no Dutch channel…

Just like il, there also proiles the presence of the post verbal entity within a speciic physical setting as in (42)–(44): (42) Residents said that at times over the weekend there were nine tanker trucks on the dock getting fuel from tanks on visiting ships. (LATWP) (43) Crayish are found on every continent except Africa and (surprise!) Antarctica. But there are crayish in Tasmania, which isn’t so darn far from Antarctica, and they can weigh up to eight pounds apiece. (LATWP) (44) Zagreb is pretty and culturally interesting, and there are scenic drives north of the city into countryside that still retains the look of old Europe. (LATWP)

Abstract locational subjects

In these examples, the search area (the setting within which the postverbal entity can be observed) is coded by a prepositional phrase or an adverbial, and the pronoun there proiles the mental constructs the conceptualizer needs to focus on the search area and locate the relevant entity. For instance, if the area around Zagreb represents the search area in (44), there proiles the mental structures required to scan it, as well as the knowledge of the various trajectories involved in deining the drives and their scenic interest. Finally, in a way very reminiscent of il, there codes the arrival of an entity onto a scene (45), or the overall impression which permeates a scene, as in (46), and the semantic import of the pronoun can be analyzed in a parallel fashion in both languages. (45) “In the life of every gay and lesbian there comes a critical point when they decide whether to live openly and honestly or whether to continue to hide,” said Tim McFeeley, president of the Human Rights Campaign Fund in Washington. (LATWP) (46) “From the sheer sensuous beauty of a genuine Jan van Eyck,” agreed Erwin Panofsky, the Princeton iconographer, “there emanates a strange fascination not unlike that which we experience when we permit ourselves to be hypnotized by precious stones or when looking into deep water.” (LATWP)

As indicated in the introduction and presented in (7) and (8), the large semantic overlap between il and there is relected by overwhelming correspondence between the two forms in parallel corpora, as the examples in (47) and (48) further illustrate: (47) Il y a une leçon fondamentale que je voudrais que nous tirions. (47′) here is one basic lesson I would like us to learn from this. (48) Dans l’ensemble, il est question de 27 cas pour l’année 1998. (48′) here were a total of 27 cases in 1998.

he analysis proposed in this chapter also accounts for the senses that there does not systematically share with its French counterpart, including its “list” sense (Lakof 1987: 561) illustrated in (49) and (50), where the semantic import of the construction is not to assert the existence of the entity which follows the pronoun, but its illustrative value: (49) Charles Kaiser, author of “1968 in America,’’ calls 1968 “the year of greatest trauma in America in the last 40 years. here were the assassinations of Bobby Kennedy and Martin Luther King Jr., the worst riots ever in 130 cities, with the hideous climax of Nixon’s e lection ater all the hope that had been invested in Bobby Kennedy’s and Gene McCarthy’s campaigns. (LATWP)

127

128 Michel Achard

(50) “Jerusalem can bring up deep things from the unconscious,” says Bar-El. here was the 36-year-old British teacher who burst into an emergency room, panicked that she was miscarrying baby Jesus. here was Catherine, 19, from France, who had an urgent message from Heaven and minced into a posh hotel to deliver it; nude. here were two men who coincidentally proclaimed themselves to be the Messiah. Bar-El put them together in a room, each growling, “Impostor!” (LATWP)

Unlike the examples introduced up to this point, the entities which follow there in (49) and (50) are not presented as new information, as relected by the presence of the deinite article. he speaker is not concerned with their existence per se, but by their illustrative value with respect to a speciic point. For example, the assassinations of Bobby Kennedy and Martin Luther King Jr. constitute well known facts of American history which everyone knows. he presence of the pronoun in (49) therefore does not seek to establish their existence, but their illustrative value to the main communicative purpose of the sentence, namely a description of the traumatic impact of the year 1968 on the American psyche. Here again, there proiles the ield, in this case the speaker’s knowledge of American history, (with speciic emphasis on all that occurred in 1968), within which speciic events can be selected for their representative value. In a similar fashion, there in (50) proiles the speaker’s knowledge of Jerusalem history which incorporates the anecdotes in which tourists believe they are channeling biblical characters. Unlike the American historical facts presented in (49), these episodes probably constitute new information to the hearer, but each one is presented as a speciic instance of the category “deep things from the unconscious” witnessed in Jerusalem, and thus preceded by a deinite article. 4.3

It and il: A region of R′

he parallel between it and il is straightforward. he same semantic categories of deontic, epistemic modality, as well as evaluation are coded by the two pronouns in their respective languages, as illustrated in (51)–(53): (51) It was therefore necessary to facilitate this approach. (51′) Il était donc nécessaire de faciliter cette voie. (52) It is absolutely certain that all three let us with an ecological disaster. (52′) Il est absolument certain que ces trois navires ont disparu laissant derrière eux une catastrophe écologique. (53) It is very hard to understand why we do not have better cooperation from all the Member States.

Abstract locational subjects

(53′) Il est très diicile de comprendre pourquoi la coopération avec l’ensemble des États membres n’est pas meilleure.

In these examples, in both English and French, it and il proile the mental range which makes the conceptualization of the event or proposition in the complement possible. Because this mental range includes the explicitly mentioned efort of necessity (51), epistemic (52), and evaluative (53) assessment, all constructs of the more inclusive level of reality R′, the ield can be characterized as the relevant portion of R′ with respect to which the entity coded in the complement is characterized. he similarity of the proiles between the two languages relects the very high level of correspondence between il and it in this function. 4.4

R and R′: A second look

he labeling of the two levels of reality discussed in this section as R and R′, as well as the statement that R is somehow more basic than R′, is potentially misleading because it might carry the unwanted assumption that R′ is necessarily more abstract than R. While it is true that R is the level at which objects can be observed and events occur, there is no limit to the level of abstraction the observed object itself may possess. Consider for instance the example in (54): (54) he analysts also say there is a chance that Mandela’s moral suasion could break the impasse in the Angolan peace talks. (LATWP)

(54) conforms to the deinition of existential there construction because it proiles the “conceptual existence” (Lakof 1987: 543) of the object chance. he object that chance represents, however, is highly abstract. First, it is a metaphorical object, a product of the general metaphor IDEAS ARE OBJECTS. Secondly, it internally contains the epistemic assessment of the proposition coded in the complement. In fact, in terms of the range of cognitive abilities necessary for the conceptualizer to apprehend the postverbal entity, the example in (54) is comparable to the one in (55): (55) he Pentagon and Lederberg’s panel have said there is no evidence that American troops were exposed to chemical weapons during Desert Storm, but it is possible that the companies named in the suit, rather than endure expensive litigation and negative publicity, will settle for a smaller but still prodigious sum. (LATWP)

Conceptually, the examples in (54) and (55) are remarkably similar. In both cases, reality is examined and the likelihood of occurrence of a given situation is assessed as possible. he only diference pertains to the speciic viewpoint from which the

129

130 Michel Achard

conceptualizer’s epistemic efort is presented. In (55), it is put onstage and explicitly coded by the predicate possible. In (54) by contrast, even though the speaker’s epistemic stance is just as present, it is backgrounded, implicit, subjectively construed, and subsumed within the lexical semantic organization of the noun chance. he reiication of these highly abstract and complex abilities and assessments in a single lexical item allow the latter to be presented as a complex object which can be observed as a constitutive part of the basic model of reality R the speaker constructs. In cases such as (54) and (55), the diference between R and R′ essentially pertains to a diference in the explicit mention of the conceptualizer’s epistemic efort. It and there thus present the mental construct which makes it possible for the post verbal entity to be localized (the ield) in diferent fashions. It proiles all the elements which the conceptualizer takes into account to come to the epistemic conclusion the following predicate codes (what was earlier described as her mental reach). here proiles the relevant subsection of reality R the conceptualizer creates within which a given object can be identiied. When this object is highly abstract as in (54), the conceptualizer’s conceptual manipulations that lead to her epistemic assessment of the situation coded in the complement are backgrounded. his results in the presentation of that situation as the way things are, rather than the outcome of the conceptualizer’s mental reasoning. he conceptual proximity that exists between the there construction which incorporates the epistemic assessment of the situation coded in the complement and the more explicit mention of that assessment in a predicate adjective construction is clearly relected in the functional equivalence these two structures exhibit in translation, as the parallel examples in (56)–(58) illustrate: (56) I believe there is also a need to raise the de minimis regulation. (56′) Je pense qu’il est aussi nécessaire de relever le règlement minimal. (57) Without registration, careful statistics cannot be kept and there are therefore, of course, opportunities for a series of errors regarding payments. (57′) Sans identiication, pas de statistiques précises -; de plus, il est évidemment également possible de commettre toute une série d’erreurs dans les paiements. (58) Food safety, risk assessment and the very concept of safety are such important areas that there is no avoiding them nowadays. (58′) La sécurité alimentaire, l’évaluation des risques, la notion même de sécurité sont des questions tellement importantes qu’il n’est aujourd’hui plus possible de les éluder.

he reasons for the systematic parallels observed in the data between the English there construction and the French copular (complex) construction are diicult to pin down precisely because they pertain to the general ecology of the two

Abstract locational subjects

languages. In (56) for example, the literal translation of the English there is a need as *il y a une nécessité would be very awkward if at all possible, despite the felicity of the noun nécessité. hese matters of constructional preference in speciic situations will not be considered further in this chapter. One issue remains to explain, namely the reason why the scope of there is restricted to observation (R), whereas il codes existence at both levels of R and R′. I believe the answer lies, at least partially, in the deictic origin of the there construction. Lakof (1987: 541) claims that “the central existential construction is based on the central deictic construction” via the metaphor “existence is understood as location in a conceptual space” (Lakof 1987: 543). In this analysis, the adverb there in (59) outlines the speciic location within which Mary becomes visible. In the existential construction in (60), there doesn’t delineate a speciic location, but the mental space within which the scene in the complement can be observed. (59) here goes Mary with her funny hat on (60) here is a funny hat on the table

Even though the mental space there proiles in (60) is abstract, it nonetheless inherits the characteristics of its physical deictic origin. One can hypothesize that existential there is restricted to observation (R) because its deictic counterpart is necessarily limited to the physical space within which the observed entity is identiied. It is not (as is the case in French) used with emotional and epistemic assessment because the cognitive operations of evaluation and judgment are not part of its deictic heritage. In other words, existential there remains faithful to its deictic origin by restraining its scope to the strict description of location in reality (existence).20 Interestingly, there are no constraints on the kind of nominal which can occur as a complement in the there construction. Gerunds, in particular, are perfectly felicitous in this position, as illustrated in (61)–(65). his is not the case in French, and consequently, existential there cannot be translated by its most frequently attested counterpart il y a in those examples. Here again, this diference between the two languages is imputable to general conditions on reiication and nominalization rather than to the meaning of the two constructions, and will thus not be considered any further here. (61) Earlier this year, Rostenkowski reimbursed the government $82,095 for the supply store items. But he insists there was no intentional wrongdoing. (LATWP)

20. In French, the il impersonal construction does not have a deictic origin. he deictic domain is covered by the pronouns voici and voila (Bergen and Plauché 2005).

131

132 Michel Achard

(62) here is no smoking on domestic lights. here is no smoking in the White House; irst lady Hillary Rodham Clinton will not tolerate it. here is no smoking with your Big Mac; McDonald’s recently banned tobacco in its corporate-owned restaurants. (LATWP) (63) Sore throats and lu-like symptoms should be monitored. If there is a worsening of symptoms, a sudden rash, a spreading infection, swollen lymph nodes or a rise in fever, especially above 102 degrees Fahrenheit; seek medical attention. (LATWP) (64) “Journalists are traveling a very dangerous road.” Steele said while most mainstream news organizations do not pay for stories, if there is any undermining of the judicial system all journalists will be blamed. (LATWP) (65) As Tom Hanks once said, there is no crying in baseball.

(LATWP)

his section has shown that even though il, there, and it all proile abstract locations within which the postverbal entity can be identiied, the speciic nature of that region can be further speciied. Existential pronouns (there and il) proile the relevant part of basic reality R with respect to which the entity described in the complement can be identiied. he epistemic and emotion reaction pronouns (il and it) proile the subsection of R′ which critically contains all the elements which the conceptualizer manipulates in order to reach her epistemic or evaluative conclusion. In many instances, the diference between the two levels of reality, and hence between the use of there and it, pertains to the explicit mention of the epistemic / emotional assessment of the postverbal entity. 4.5

Field versus setting: Il and ça

We can now return to the distinction between the French forms il and ça (ce/c’) which both correspond to English it, as was illustrated by the examples in (9) and (10) repeated here: (9) It is true that consultation with Parliament has only come about at a late date. (9′) C’est vrai que la consultation de votre Assemblée est intervenue tardivement. (10) It is true that the problem of the forestry sector is an extremely complex one. (10′) C’est vrai que le problème du secteur du bois est terriblement complexe.

It was mentioned earlier that the two pronouns il and ça both designate abstract regions within which the entity described in the complement can be isolated. While il codes the ield, or in other words the conceptualizer’s mental reach with respect to the conceptualization of the entity coded in the complement, ça designates the abstract region (mental space) composed of the immediate discourse

Abstract locational subjects

context from which that entity is extracted. he distinction between il and ça therefore relects a subtle distinction between a ield and a setting construction, where the setting is more readily deined by the current discourse context than the conceptualizer’s mental efort to assess the epistemic status of the entity the complement describes. he mental space the pronoun designates contains the segment of context from which a more speciic entity is extracted. It is thus analogous (to a certain extent) to the more physical setting considered in (30c) and illustrated in Figure 6c Chicago is cold. Just as Chicago provides the boundaries within which the sensation of cold can be experienced, the mental space which ça designates outlines the section of discourse context (the literature related memories) from which the following entity is extracted. he complement clause represents the expression of a speciic part of that scene singled out for expressive reasons. In this speciic discourse context, demonstrative impersonals are therefore true abstract setting constructions, where the setting, i.e. the abstract region the pronoun refers to, is composed of interconnected entities from the current discourse context. his analysis captures the subtle distinction between il and ça, and therefore the compatibility of both forms with English it. Il places maximum emphasis on the mental efort necessary to assess the event or proposition described in the complement, while ça is more concerned with the immediate circumstances from which that event or proposition is extracted than the mental act itself. Il is therefore somewhat larger in scope, because it includes the conceptualizer-centered considerations of assessment and analysis among others, while ça is more speciically context bound. his distribution between the two forms relects this diference in meaning. An exhaustive account cannot be attempted here, because the choice of pronoun depends on too many factors including register and text genre which cannot be adequately investigated within the conines of this chapter.21 his section concentrates on the single copular predicate être vrai ‘be true’. he copular construction represents the structural environment within which the competition between il and ça is the iercest because the mental efort required in locating or evaluating the event / proposition in the complement is largely based on the examination of the circumstances which surround it. In this context, il and ça exhibit such a large amount of conceptual overlap that they oten can be used almost interchangeably, as illustrated in (66), where the pound sign indicates a manufactured form:22 21. he distribution of il and ça is also very sensitive to the lexical semantics of individual predicates, and therefore requires the examination of a large number of speciic cases (Achard 2010). 22. he examples in the remainder of this section are from Achard (2010).

133

134 Michel Achard

(66) Il me suira de rappeler comment M. Klein, dans une question relative aux surfaces de Riemann, a eu recours aux propriétés des courants électriques. Il / #c’est vrai que les raisonnements de ce genre ne sont pas rigoureux… (Poincaré H. La valeur de la science: 154) ‘It will be suicient to remind you how M. Klein, in a question relative to Rieman surfaces, used the properties of electrical current. It is true that such arguments are not rigorous’

However, consistent with c’ context-bound nature, the pronoun is most frequently attested when the predicate expresses agreement with a previously made statement, as illustrated in (67) and (68): (67) je sais: il a tué un pauvre vieil homme sans défense: Farnese était seul, – pas un laquais, – et le coup de revolver a été tiré par derrière. Je sais tout ça…. mais écoutez un peu: ce n’est pas vrai que Farnese était seul. (Farrère C. L’homme qui assassina: 280) ‘I know: he killed a poor defenseless man: Farnese was alone, – not aservant, – and the shot was ired from behind. I know all that…but listen for a minute: It is not true that Farnese was alone’ (68) S’il te faut une coniance perpétuelle sache que tu l’as et que c’est elle qui s’inquiétait quand j’écrivais ma dernière lettre. Mais sache aussi que cette coniance est exigeante et demande qu’on la satisfasse. C’est vrai que je suis “près de mes intérêst”. Plus je vais, plus je veux acquérir. (Alain-Fournier, H. Correspondance avec J. Rivière: 207) ‘If you require everlasting trust, know that you have it and it was that trust getting worried when I wrote my last letter. Be also aware, however, that this trust is demanding and expects to be satisied. It is true that I am “close to my interests”. I want to acquire more and more as time passes’

In (67) and (68), the entity coded in the complement is deeply semantically integrated in the immediate context of the passage. In (67), the content of the proposition that follows the predicate Farnese était seul ‘Farnese was alone’ repeats a section of the preceding discourse verbatim. In (68), the quotes that surround “près de mes intérêts” ‘close to my interests’ indicate that this very expression was used in a previous letter. Conversely, il is most frequently attested when the proposition it introduces serves to temper a previously made statement by presenting a piece of information that challenges its force, as previously illustrated in (38) repeated here: (38) À voir cela, il me semble que la révolte est plus loin de nous que je ne croyais d’ abord. Il est vrai que je suis avec des montagnards, écartés des centres industriels et très fatalistes. (Alain-Fournier, H. Correspondance avec J. Rivière: 120)

Abstract locational subjects

‘When I see this, it seems to me that the rebellion is furtherfrom us than I irst thought. It is true that I am with mountain men, remote from the industrial centres and very fatalistic’

In (38), the proposition il introduces presents a piece of information that challenges the generalizing force of the previous statement. he author’s earlier position about the state of the rebellion is nuanced by his further consideration of the fatalistic nature of his companions. Consistent with il’s meaning, the pronoun selection emphasizes the conceptualizer’s deductive powers which led to the discovery of the fact that tempers his initial stance. he pronoun is ideally suited to express the critical importance of the epistemic assessment itself, since the mere existence of a fact which runs counter to the overall argument suices to weaken the latter’s scope and power. he epistemic efort responsible for the assessment is highlighted, to the detriment of the context from which it is extracted. If the content expressed in the complement proposition has already been previously established, and the communicative purpose of the predicate is merely to conirm it as in the example in (69), c’ alone is possible: (69) Mais je ne prendrai pas un coup, Maria, pas un seul! Il hésita un peu et demanda abruptement, les yeux à terre: – peut- être… vous a- t- on dit quelque chose contre moi? – non. – c’est vrai que j’avais coutume de prendre un coup pas mal, quand je revenais des chantiers et de la drave; mais c’est ini. (Hémon L. Maria Chapdelaine: 93) ‘But I won’t have a drink, Maria, not a single one! He hesitated a little and asked suddenly, his eyes downcast: maybe someone told you something against me didn’t they? – no. – it is true that I used to drink quite a bit when I came back from working or cutting wood; but it’s over.’

In (69), the speaker conirms a rumor about himself. Ç’s presence is thus expected. Il would be extremely awkward if at all possible, because it would imply the speaker is using his epistemic faculty to uncover the existence of a proposition which is not only already irmly established in reality, but also part of the immediate context.

5. Recapitulation and conclusion his chapter argued that there, it, il and ça constitute a natural class of abstract locational subjects because each pronoun proiles an abstract region within which the entity coded by the post verbal complement is localized or assessed. Individual forms proile slightly diferent kinds of regions, whose speciic nature largely determines pronoun distribution. he analysis presented here follows the standard

135

136 Michel Achard

CG account of these constructions in arguing that there, it, and il proile the ield, i.e. the conceptualizer’s scope of awareness for the issue at hand, but the distributions of these three forms in English and French reveals that the conceptualizer’s mental reach needs to be considered at diferent levels of reality. Existential there and simple (nominal) il impersonals proile the subpart of basic reality R which is restricted to the observation of things that are or are perceived as missing, while it and complex impersonal il operate at a more analytical level of reality R′ which prominently includes the epistemic or evaluative efort the conceptualizer exercises toward the event or proposition coded in the complement. Ça difers from il in that it proiles the abstract setting from which the postverbal entity has been extracted for expressive purposes. his abstract setting is distinguished from the ield in that it places the emphasis on the immediate circumstances which include the event or proposition coded in the complement rather than on the mental efort required to evaluate its epistemic status or its impact. he subtlety of the distinction between this kind of abstract setting and the ield explains the large amount of semantic overlap between il and ça, especially in the copular construction. he recognition of a class of abstract locational subjects is relevant to several very general lines of inquiry, even though each one requires much more thorough investigation than the scope of this chapter allows. First, it makes it possible for a typological assessment of the category to be conducted, which would allow us to determine if the kinds of abstract locations observed in French and English are common in the languages of the world, what other kinds are used for the purposes of linguistic expression, and which constructs are most frequently attested. Secondly, the treatment of impersonal pronouns and demonstratives as part of the same class has interesting implications for the study of impersonal constructions. At the very least, it challenges the strictly syntactic criteria by which a meaningless class of dummy pronouns is kept separate, and forces the consideration of a larger range of constructions. Finally, it provides an insightful look into large sections of English and French grammar by showing how similar functions are achieved in diferent ways. For example, the data observed in this chapter reveal that English has no specialized impersonal construction per se, but that the impersonalizing function is shared between the existential there, itself an extension of the deictic construction, and the neuter anaphoric it. French is organized diferently. One could argue that neuter (as opposed to personal) il is a specialized impersonal with a primarily existential role. Crucially, its function is restricted to providing the presentational frame for the entity coded in the complement, which explains why it cannot be used (as English it is) for simple anaphoric reference where demonstrative pronouns are required (You can’t do this to me, it’s not fair! / Tu ne peux pas me faire ça, *il/c’est pas juste). As shown in this chapter, demonstratives only enter

Abstract locational subjects

in competition with impersonals when a speciic element of the referred scene is extracted from that scene for expressive reasons. By focusing on there, it, il, and ça, the account presented in this chapter highlights the intrinsic connection between the meaning of individual items and the global ecologies in which they participate.

References Achard, Michel (1998). Representation of cognitive structures: Syntax and semantics of French sentential complements. Berlin and New-York: Mouton de Gruyter. DOI: 10.1515/9783110805956 Achard, Michel (2000). French ça and the dynamics of reference. LACUS Forum, 1–12. Achard, Michel (2002). he meaning and distribution of French mood inlections. In Frank Brisard (Ed.), Grounding: he epistemic footing of deixis and reference (pp. 197–249). Berlin: Mouton de Gruyter. Achard, Michel (2009). he distribution of French intransitive predicates. Linguistics, 47(3), 513–558. DOI: 10.1515/LING.2009.018 Achard, Michel (2010). Fields and settings: French il and ça impersonals in copular complement constructions. Cognitive Linguistics, 21(3), 441–498. DOI: 10.1515/COGL.2010.016 Alonso, Susana (2008). Existentials as impersonalizing devices. Transactions of the Philological Society, 106, 180–215. DOI: 10.1111/j.1467-968X.2008.00192.x Bergen, Benjamin, & Plauché, Madelaine (2005). he convergent evolution of radial constructions: French and English deictics and existentials. Cognitive Linguistics, 16, 1–42. DOI: 10.1515/cogl.2005.16.1.1 Bolinger, Dwight (1977). Meaning and form. London and New York: Longman. Breivik, Leiv, & Martínez-Insua, Ana (2008). Grammaticalization, subjectiication and nonconcord in English existential sentences. English Studies, 89, 351–362. DOI: 10.1080/00138380802011321 Brunot, Ferdinand (1936) [1922]. La Pensée et la Langue. Paris: Masson. Cadiot, Pierre (1988). De quoi ça parle? A propos de la référence de ça pronom-sujet. Le français Moderne, 65, 174–92. Carlier, Anne (1996). Les Gosses ça se lève tôt le matin’: L’interprétation générique du syntagme nominal disloqué au moyen de ce ou ça. Journal of French Language Studies, 6, 133–62. DOI: 10.1017/S0959269500003045 Chomsky, Noam (1981). Lectures on government and binding. Dordrecht: Foris. Crawford, William (2005). Verb agreement and disagreement: A corpus investigation in existential here + Be constructions. Journal of English Linguistics, 33, 35–61. DOI: 10.1177/0075424204274001 Divjak, Dagmar, & Janda, Laura (2008). Ways of attenuating agency in Russian. Transactions of the Philological Society, 106, 138–179. DOI: 10.1111/j.1467-968X.2008.00207.x Fauconnier, Gilles (1985). Mental spaces: Aspects of meaning construction in natural language. Cambridge, Mass.: MIT Press, and London: Bradford. Galichet, Georges (1947). Essai de Grammaire Psychologique de la Langue Française. Paris: Presses Universitaires de France. Gledhill, Christopher (2003). Fundamentals of French syntax. LINCOM EUROPA.

137

138 Michel Achard

Hazout, Ilan (2004). he syntax of existential constructions. Linguistic Inquiry, 35, 393–430. DOI: 10.1162/0024389041402616 Helasvuo, Marja-Liisa, & Vilkuna, Maria (2008). Impersonal is personal: Finnish perspectives. Transactions of the Philological Society, 106, 216–245. DOI: 10.1111/j.1467-968X.2008.00208.x Jespersen, Otto (1924). he philosophy of grammar. London: Routledge. Jones, Michael A. (1996). Foundations of French syntax. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511620591 Kirsner, Robert (1979). he problem of presentative sentences in modern dutch. North-Holland Linguistic Series 43. Amsterdam: North-Holland. Lakof, George (1987). Women, ire, and dangerous things: What categories reveal about the mind. Chicago and London: University of Chicago Press. DOI: 10.7208/chicago/9780226471013.001.0001 Langacker, Ronald W. (1982). Space grammar, analyzability, and the English passive. Language, 58, 22–80. DOI: 10.2307/413531 Langacker, Ronald W. (1987a). Foundations of cognitive grammar. Vol. l. heoretical prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. (1987b). Nouns and verbs. Language, 63, 53–94. DOI: 10.2307/415384 Langacker, Ronald W. (1990). Subjectiication. Cognitive Linguistics, 1, 5–38. DOI: 10.1515/cogl.1990.1.1.5 Langacker, Ronald W. (1991). Foundations of cognitive grammar. Vol. 2: Descriptive application. Stanford: Stanford University Press. Langacker, Ronald W. (1993). Grammatical traces of some “Invisible” semantic constructs. Language Sciences, 15, 323–355. DOI: 10.1016/0388-0001(93)90008-G Langacker, Ronald W. (2002). he control cycle: Why grammar is a matter of life and death. Proceedings of the Annual Meeting of the Japanese Cognitive Linguistics Association, 2, 193–220. Langacker, Ronald W. (2004). Aspect of the grammar of inite clauses. In Michel Achard & Suzanne Kemmer (Eds.), Language, culture, and mind (pp. 535–577). Stanford: CSLI Publications. Langacker, Ronald W. (2006). Dimensions of defocusing. In Masayoshi Shibatani & Taro Kageyama (Eds.), Voice and grammatical relations; In honor of Masayoshi Shibatani (pp. 115–137). Amsterdam: John Benjamins. DOI: 10.1075/tsl.65.08lan Langacker, Ronald W. (2008). Cognitive grammar: A basic introduction. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780195331967.001.0001 Langacker, Ronald W. (2009). Investigations in cognitive grammar. Berlin and New York: Mouton de Gruyter. DOI: 10.1515/9783110214369 Lauwers, Peter (2004). La Description du Français entre la Tradition Grammaticale et la Modernité Linguistique: Etude Historiographique et Epistémologique de la Grammaire Française entre 1907 et 1948. Leuven: Peeters. Martínez-Insua, Ana, & Palacios Martinez, Ignacio (2003). A corpus-based approach to nonconcord in present day English existential there-constructions. English Studies, 84, 262– 283. DOI: 10.1076/enst.84.3.262.16852 Olsson, Hugo (1986). La Concurrence entre il, ce et cela (ça) comme Sujet d’ Expressions Impersonnelles en Français Contemporain. Stockholm: Almqvist and Wiksell. Postal, Paul, & Pullum, Geofrey (1988). Expletive noun phrases in subcategorized positions. Linguistic Inquiry, 19, 635–670. Radford, Andrew (2004). Minimalist syntax. Exploring the structure of English. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511811319

Abstract locational subjects

Rowlett, Paul (2007). he syntax of French. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511618642 Sansó, Andrea (2005). Semantic maps in action, a discourse-based approach to passive and impersonal constructions. In Annalisa Baicchi, Cristiano Broccias, & Andrea Sansó (Eds.), Modelling thought and constructing meaning (pp. 89–106). Milan: Angeli. Séchehaye, Albert (1950). Essai sur la Structure Logique de la Langue. Paris: Champion. Shibatani, Masayoshi (1985). Passive and related constructions: A prototype analysis. Language, 61, 821–848. DOI: 10.2307/414491 Siewierska, Anna (2008a). Ways of impersonalizing. Pronominal vs. verbal strategies. In Maria de los Angeles Gómez González, J. Lachlan MacKenzie, & Elsa M. González Alvarez (Eds.), Current trends in contrastive linguistics, functional and cognitive perspectives (pp. 3–26). Amsterdam: John Benjamins. DOI: 10.1075/sfsl.60.03sie Siewierska, Anna (2008b). Impersonalization from a subject-centered vs. agent-centered perspective. Transactions of the Philological Society, 106, 115–137. DOI: 10.1111/j.1467-968X.2008.00211.x Sloń, Anna (2007). he ‘impersonal’ impersonal construction in Polish. A Cognitive Grammar analysis. In Dagmar Divjak & Agata Kochańska (Eds.), Cognitive paths through the slavic domain (pp. 257–287). Berlin: Mouton de Gruyter. DOI: 10.1515/9783110198799.3.257 Smith, Michael (1985). An analysis of German dummy subject constructions in Cognitive Grammar. In Scott DeLancey & Russell Tomlin (Eds.), Proceedings of the First Annual Meeting of the Paciic Linguistics Conference (pp. 412–425). Department of Linguistics: University of Oregon. Smith, Michael (2000). Cataphors, spaces, propositions: Cataphoric pronouns and their function. Proceedings from the Meeting of the Chicago Linguistic Society, 36(1), 483–500. Williams, Edwin (1984). here-insertion. Linguistic Inquiry, 15, 131–153. Wilmet, Marc (1997). Grammaire critique du français. Louvain-la-Neuve: Duculot.

139

Subjecthood of the agent argument in Estonian passive constructions Liina Lindström University of Tartu / University of Turku

his paper tackles diachronic changes in the choice between elative and adessive case for marking the agent in Estonian periphrastic passive constructions in two time periods. In 1800–1850, the main agent-marking device was the elative case, whereas in the 1990s the elative was limited to inanimate actors, and the use of the adessive had increased considerably. However, the adessive can only be used for marking volitional, animate agents. Changes are observed with regard to semantic constraints, subject properties of the adessive and elative agents, and language contacts. Adessive arguments behave like non-canonical subjects in many constructions in Estonian, and the use of the adessive for marking agents in passives is strengthened by the possessive perfect construction in Eastern Circum-Baltic languages.

1.

Introduction

he passive is one of the most oten described and discussed issues in the ield of syntax. In Estonian linguistics, the most intriguing and discussed question concerning the category of voice is the distinction between the impersonal compound tenses and periphrastic passive, as both constructions share basically the same morphosyntactic means. While the Estonian impersonal voice typically demotes the human agent, the periphrastic passive, additionally, promotes the patient to subject (cf. Rajandi 1999 [1968]; Erelt 1979; Pihlak 1993; Holvoet 2001; Blevins 2003; Torn 2002, 2006; Torn-Leesik 2009; Lindström & Tragel 2007, 2010). hus, the impersonal is used mainly to background the subject, whereas the periphrastic passive is used to foreground the patient and background the agent. Less attention has been paid to diferent agent marking devices and conditions of using these devices in Estonian passive and impersonal clauses (see Section 2.2 for references). he Estonian passive is typically agentless, but it is still possible to express the agent in the passive clause by various means, e.g. postpositional phrases, the genitive, adessive and elative case (cf. Rajandi 1999[1968]). In this paper, two doi 10.1075/cal.16.06lin © 2015 John Benjamins Publishing Company

142 Liina Lindström

types of agent phrases are compared: those marked with the elative and the adessive case. he use of elative and adessive agents in Estonian has changed considerably; in the early 19th century the main agent-marking device in the passive construction was the elative case, whereas by the end of the 20th century the use of the elative was limited to inanimate actors and the use of the adessive had increased noticeably. his article discusses this change and the possible reasons for it.1 Note that the term agent is used in a wide sense here as it refers not only to volitional agents, but also to non-volitional and sometimes even inanimate actors. To avoid further confusion, I use the abbreviation A for the most agent-like argument in this paper and P for the most patient-like argument (similarly to Comrie 2008: 4). hus the term A is used for the most agent-like argument which is coded as a subject of an active clause and is demoted to an oblique or omitted in the corresponding passive clause. he available options for expressing adessive and elative As are investigated during two time periods: from 1800 to 1850, and during the 1990s. Between these two time periods, the use of the elative case as an agent-marking device decreased and the adessive case spread. Possible explanations for these changes can be summoned from three diferent areas: (1) the semantics of the respective cases, (2) the subject properties of the adessive and elative arguments, and (3) their use in different constructions within the Circum-Baltic language area. he properties of the subject have been much discussed since E. Keenan’s pioneering article in 1976 (cf. Keenan 1976; Andrews 1985; Onishi 2001; Barðdal 2006, among others); these subject properties have been tested on particular constructions in particular languages (Timberlake 1976; Beck 2000; Barðdal & Eythórsson 2003; Barðdal 2006 among others, regarding Estonian: Erelt 2004; Koks 2004; Lindström 2012; Lindström 2013, Metslang 2013). his article discusses some subject properties that can be used to test the A of a passive construction – agentivity, topicality, and control over relexivization. Since the passive sentence has a syntactic subject (P), the competition between adessive or elative As and the P for the position of subject in these constructions is discussed. he data for this study was obtained primarily from the Estonian language corpora. he data from 1800 to 1850s was taken from the Corpus of Old Written Estonian2; contemporary Estonian is represented by the 1990s iction texts subcorpus of the Corpus of Literary Estonian,3 and to a lesser extent the texts from the Corpus of Spoken Estonian4 (Corpus of Spoken Estonian). 1.

his study was supported by the Estonian Research Council (PUT90) and the Academy of Finland.

2. http://www.murre.ut.ee/vakkur/Korpused/korpused.htm 3. http://www.cl.ut.ee/korpused/grammatikakorpus/ 4. http://www.cl.ut.ee/suuline/

Subjecthood of the agent argument in Estonian passive constructions

All passive sentences present in the data were included in the analysis. Estonian passive constructions contain the verbs ole- ‘be’ (1) or saa- ‘get, become’ (2) as auxiliaries and the past passive participle (ppp). he promoted patient phrase shows agreement with the auxiliary (as canonical subjects do). he A can be oblique or – more typically – is absent. (1) P A Tibu-d ol-i-d prae-tud (tun-tud koka poolt). chick-pl.nom be-pst-3pl fry-ppp (know-ppp chef.gen by) ‘he chickens were fried (by a well-known chef)’ (2) P A Sibula-d sa-i-d (mei-l) lõpuks söö-dud. onion-pl.nom get-pst-3pl (we-ade) inally eat-ppp ‘he onions were inally eaten by us’

he paper is organized as follows: Section 2 gives a brief overview of the category of passive in Estonian. Section 3 compares the dynamics of the use of agent-marking means in two time periods: 1800–1850 and the 1990s. Section 4 compares the use of the adessive and elative As in the two time periods, taking into account their semantic constraints, topicality and control over the use of the possessive relexive pronoun. In Sections 5, 6 and 7, possible explanations for the increase of the use of the adessive and decrease of the use of the elative are considered: Section 5 takes into account semantics of the adessive and elative case, Section 6 takes a look at their functions and subject-like behavior in other constructions, and in Section 7, the spread of the possessive perfect construction in Eastern Circum-Baltic languages as a possible explanation for these dynamic tendencies in Estonian is taken under consideration.

2. Estonian impersonal and passive constructions 2.1

Impersonal and passive: An overview

Estonian, like other Baltic-Finnic languages, has historically only distinguished between personal and impersonal voice (Viitso 2003: 216). he Estonian impersonal is subjectless; the A of the event is covert. Syntactically, the P is an object of the impersonal clause and is marked with either the partitive (partial object, see Example 3) or the nominative (total object, Example 4). he impersonal is formed with the verb suixes -ta(kse)- (present), -ti (past), or with compound verb forms: the verb olema and the past passive participle ending -tud (perfect and pluperfect). he impersonal can be derived from both transitive and intransitive clauses.

143

144 Liina Lindström

(3) Se-da raamatu-t loe-ti suure huvi-ga. this-par book-par read-iprs.pst big.gen interest-com ‘(People) read this book with great interest.’ (4) See raamat loe-ti suure huvi-ga läbi. this book read-iprs.pst big.gen interest-com through ‘(People) read this (whole) book with great interest.’

he A of the impersonal construction is typically let uncoded. he referent of the implicit argument is always human, mostly a general or plural participant (Rajandi 1999[1968]; Pihlak 1993; Torn 2002; Blevins 2003; Erelt 2003; Vihman 2008; TornLeesik 2009; Torn-Leesik & Vihman 2010). he Estonian impersonal is used mainly speaker-exclusively (Torn-Leesik & Vihman 2010; Pajusalu, this volume). he Estonian personal passive is also referred to as a resultative or stative passive. It has an overt subject in the nominative case and expresses a state into which the referent of the subject (semantically the patient) has entered as a result of the action. he Estonian passive is more stativizing and resultative than the English passive (Vihman 2008). he passive can be formed only from transitive clauses. In the passive construction, the P is promoted to subject and agrees with the verb olema ‘be’ (5) or saama ‘get’ (2). Raamat ol-i läbi loe-tud. book.sg.nom be-pst.3sg through read-ppp ‘he book was read (all the way through).’ b. Raamatu-d ol-i-d läbi loe-tud. book-pl.nom be-pst-3pl through read-ppp ‘he books were read (all the way through).’

(5) a.

here has been a great deal of discussion about the relationship between impersonal and passive categories in Estonian (Wiedemann 1875; Erelt 1979; Pihlak 1993; Rajandi 1999 [1968]; Torn 2002, 2006; Vihman 2007; Torn-Leesik 2009; Lindström & Tragel 2007, 2010, among others). he source of disagreement is that the Estonian passive and impersonal paradigms are not always distinguishable. here is a notable overlap between impersonal compound tenses (present and past perfect) and passive simple present and past: both constructions use the auxiliary verb olema ‘be’, sometimes also saama ‘get, become’, the past passive participle (which historically belongs to the impersonal paradigm) and the NP in the nominative case, which can be interpreted as a total object (in the impersonal) or subject (in the passive construction); in both constructions the nominative NP can trigger agreement with the verb. Some researchers have distinguished impersonal compound tenses with a nominative object from passive simple present and past

Subjecthood of the agent argument in Estonian passive constructions

mainly on the basis of semantics: the impersonal is more dynamic and expresses the event of change, which brings about a new state, while the passive is stative and expresses being in the state (Rajandi 1999 [1968]: 94). However, in analyzing corpus data, this distinction is hardly applicable as a criterion, as most of the clauses are ambiguous between the two. In this article, I distinguish impersonal compound tenses with NPnom from passive clauses by formal criteria; all clauses with olema ‘be’ or saama ‘get, become’ + ppp + NPnom with verb agreement are categorized as passive. he passive voice in Estonian typically does not express the A. In cases where the A is expressed, many diferent means can be used. As there is more than one auxiliary verb and more than one possible way to express the agent, we are dealing with a bundle of constructions rather than one single passive construction. Hence I prefer to talk about passive constructions in this paper. My main focus is on describing historical changes in marking A in these constructions during the past 200 years and inding explanations for these changes. 2.2

Agent marking in Estonian passive constructions: Previous studies

he irst text about expressing the A in Estonian is a manuscript, in which Ney (1928) names the following possibilities for expressing the A in the passive construction: the genitive, elative, and (in South Estonian) ablative cases, and the postpositions poolt, poolest ‘by, from’, läbi ‘through’. Ney also claims that the elative construction was used most in earlier times, but that in the contemporary language, it was giving ground to the poolt ‘by’-construction. Uuspõld (1970) only discusses attributive ppp-constructions which cannot be considered passive. However, the past passive participle (ppp) also occurs in the passive construction, thus the agent-marking devices are largely similar in both of these. Uuspõld lists the genitive, the postposition poolt ‘by, from’, and the elative as the main devices for marking the A. Rajandi (1999 [1968]) mentions the same agent-marking devices as Uuspõld does, adding the adessive to the list. Ross (1997) has mentioned in connection with the use of local cases in 17thcentury texts that the elative is oten used to mark the agent. Although her analysis concerns also the adessive, she has not mentioned it in this function. Torn (2006) has closely studied the use of genitive + poolt ‘by’-constructions in impersonal and passive sentences based on the corpus of literary Estonian and has found that the poolt ‘by’-construction can be used in impersonal sentences with some restrictions: the noun in the poolt ‘by’-construction usually refers to a group or an institution (e.g. by the police) but not an individual. However, in the stative passive the choice of the noun is not as limited and refers more oten to an individual.

145

146 Liina Lindström

Lindström & Tragel (2007) have discussed the use of the adessive argument in impersonal and passive sentences and have found that in an impersonal sentence the adessive argument expresses non-agentive roles (place, possessor etc.), whereas in the passive and the compound tenses of the impersonal it mainly expresses the volitional agent. Some of the sentences with adessive topical agents form a distinctive construction – the possessive perfect construction (cf. Section 7). Sahkai (2011) has discussed the functions of the genitive A in participial and ininitival constructions more closely. he genitive deviates from other agent marking means by its syntactic dependency – it cannot be an argument on the clause level but belongs syntactically to the participle.

3. Expression of the agent in the passive construction in the 19th and the late 20th century In this section, I take a brief look at the agent-marking devices during two time periods, from 1800–1850 and during the 1990s, based on the corpora. 3.1

Expression of the A in the passive construction in the 19th century

In May 2010, when this corpus query was performed, the corpus of 19th century texts contained approximately 409,000 words. his material contained a total of 876 passive sentences. Table 1 shows the agent-marking devices in the early 19th century texts and the number of occurrences for each of them. In the 19th century, passive constructions were used frequently, possibly because of the German inluence which was very strong before the 20th century. From the point of view of modern Estonian usage, the amount of passive sentences in the 19th-century texts seems much greater than it is nowadays, and compared to modern Estonian the saama ‘get’-passive is much more common. Using the verb saama as an auxiliary in passive constructions is at least partially a German inluence (Raun & Saareste 1965: 67). As can be seen in Table 1, the A in the passive sentences has been let out altogether in the overwhelming majority of examples – the agentless passive is characteristic of Estonian. If the A is expressed, it is most typically coded with the elative case (56% of clauses with explicit A). he occurrence of all other devices is much less frequent. he comitative case and the postposition läbi ‘through’ are rather exceptional among A-marking devices, as they are more oten used for instrument-like arguments (alanduse õitega ‘with the blossoms of humility’ in Example 6a). However, since they can be expressed as subjects in corresponding active sentences (alanduse õied in 6b), they behave like other As.

Subjecthood of the agent argument in Estonian passive constructions

Table 1. Agent-marking devices from 1800–1850. Marking device No.

Examples

Adessive

9 (+ 4*) Nattukesse aea-ga oll-i temma-l üks raske kimp koggo-tud …. short.gen time-com be-pst. 3sg he/she-ade one heavy bunch gather-ppp ‘In a short time, a heavy bunch was gathered by him/her and he/she was binding it up tight’ (1844-Schwelle3_3)

Elative

70

Et nemmad kül sepärrast teis-te-st sa-wad teo-tud ja naer-tud, ei panne nemmad sedda naero siiski ennestele raskeks. that they ptcl therefore other-pl-ela get-3pl disgrace-ppp and laugh-ppp ‘hough they will therefore be disgraced and derided by others, they do not ind this laughter hard for themselves’ (1847-Kersten_87)

Genitive

8

Nemmad ep olle mitte innimesse teh-tud, waid kaswawad ja sünniwad issi+ +suggu konna+ +karpides, mis Perlmusliks nimmetakse. they not be not man-gen make-ppp ‘hey are not made by a person but grow and are born in particular mussels which are called pearlmussels’ (1818-Masing_60)

Comitative

14*

Kui üks usklik hing nenda allandusse öi-te-ga ehhi-tud on, siis paistab se ka temmast wälja, sest allandusse-+ +emma on waimo waesus. if/when one faithful soul so humility-gen blossom-pl-com decorate-ppp be.3sg ‘If/when a faithful soul is decorated with the lowers of humility, it can be seen in him/her because the mother of humiliation is the poverty of the mind’ (1847-Kersten_97)

Pärrast juttustas temma keikidele, kuidas ta omma paggana genitive + läbi 6 (+ 10*) pimmedusse-st ja waimolikko surma-st lapse läbbi oll-i ülles ärra-tud … ‘through, by’ … how he/she own devil.gen darkness-ela and spiritual death-ela child.gen through be-pst. 3sg up awake-ppp … ‘Later he told everyone how he was awakened from his devilish darkness and spiritual death by a child…’ (1839-Masing_13) genitive + käest 2 ‘from’, lit. ‘from the hand’

Lambrine peab wägga murret piddama et ka umbest need järad lammaste jure sawad, mis wannematte käest temma-le näide-tud ja käs-tud on. what parent.pl.gen from he/she-allshow-ppp and order-ppp be.3sg ‘he shepherd must take care that these rams also get to the lambs which are shown to him and ordered by the parents’ (1840-Jordan_31)

genitive + poolt, 2 poolest ‘from, by’, lit. ‘from the side of…’

Selle peäle sa-i temma-le kohto poolt, temma üllekohto pärrast, surma nuhtlus moiste-tud ja sedda ka temmale kulutud. this ater get-pst. 3sg he/she-all court.gen from he/she-gen injustice.gen because death.gen penalty assign-ppp ‘Ater this, the death penalty was assigned to him by the court for his injustice, and this was announced to him.’ (1847-Kersten_117)

Agentless

Härjepä ei tohhi mitte söde-tud sa-da, kui saddo pöllu pehmeks teind… clover.nom not must.conneg not feed-ppp get-inf… ‘he clover must not be fed when the rain has made the ield sot…’ (1840-Jordan_9)

751

(*interpretation as A is ambiguous)

147

148 Liina Lindström

Kui üks usklik hing nenda allandusse if/when one faithful soul so humility.gen öi-te-ga ehhi-tud on, blossom-pl-com decorate-ppp be:3sg ‘If/when a faithful soul is decorated with the blossoms of humility’ (1847-Kersten_97) b. Obj Subj kui usklikku hinge ehi-vad alanduse if/when faithful.par soul-par decorate-3pl humility.gen õie-d blossom-pl:nom ‘if/when the blossoms of humility decorate the faithful soul’

(6) a.

he genitive is diferent from other agent-marking devices since it belongs syntactically together with the ppp and thus its position in the sentence is ixed – always before the ppp (Sahkai 2011; see the example in the table). 3.2

Agent-marking options in modern Estonian

In the 1990s corpus of iction texts, a selection of data approximately equal in size to that of the corpus of 19th century texts (roughly 2/3 of the total 1990s iction text corpus) was included in the analysis. A total of 4636 instances of the ppp were found, of which 84 were used in passive sentences where the A was expressed in some way (see Table 2). Compared to the texts from the early 19th century, both the total number of passive sentences and passive sentences with explicit As had decreased considerably. In Table 2, agentless passive sentences have been omitted. For comparison, data from the corpus of spoken Estonian, which contains everyday, face-to-face conversations have also been added to the table. Since the total amount of spoken data is considerably lower, the numerical data are not comparable. Nevertheless, some trends can be observed. First, we can see from Table 2 that compared to the early 19th century, the modern time period has considerably fewer passive constructions even though the texts included in the analysis were comparable. It is possible that this is a result of the nature of iction texts, as the passive is not a common construction in iction. However, it is likely that the decrease in German inluence plays a more important role. In the early 19th century the number of native Estonian authors writing in Estonian was not very high, and the whole written culture was under strong inluence from German. Also, many of the authors whose texts can be found in the corpus were not Estonians but Germans who had learnt Estonian; thus they do not represent spoken Estonian at that time. However, by the end of the 20th

Subjecthood of the agent argument in Estonian passive constructions 149

Table 2. Agent-marking devices in texts from the 1990s. Marking device Fiction Spoken Examples corpus Adessive

21

8

Ja mamma oli see, kes tõi veel oma pisikese hõbedakohvri välja, mis ta-l ol-i paki-tud. what she-ade be-pst. 3sg pack-ppp ‘And grandma was the one who brought out her small trunk of silver which was already packed by her’ (ILU1990\ilu0273)

Elative

40

1

ja kask majaotsas kohises endiselt, nagu poleks ta oksa-d ja tüvi saa-nud-ki riku-tud raudse-st vihma-st… like be-cond.conneg it.gen branch-pl and trunk get-app-ptcl ruin-ppp iron-ela rain-ela ‘and the birch was still soughing, as if its branches and trunk had never been ruined by the hard rain’ (ILU1990\ilu0028)

Genitive

6

2

Ta ju saab aru, et need lille-d on sinu oste-tud. … that these lower-pl be.3pl you.gen buy-ppp ‘He understands that these lowers have been bought by you’ (ILU1990\ilu0166)

Comitative

9

1

Ann oli koju jõudes kohe kirjutanud ning korranud oma kutset, mis ol-i ka teis-te kodakondse-te arvamus-te-ga kinnita-tud. what be-pst. 3sg too other-pl.gen member_of _household-pl.gen opinion-pl-com irm-ppp ‘Ater coming home, Ann had written at once and repeated her invitation, which was airmed also by the opinions of the other members of the household’ (ILU1990\ilu0262)

genitive + poolt 8 ‘from, by’

0

Eilsed tunnistuse-d sa-i-d vahialuse-le ette loe-tud ja tema poolt kinnita-tud. yesterday testimony-pl.nom get-pst-3pl arrestant-all ahead read-ppp and he/she.gen side_of airm-ppp ‘Yesterday’s testimonies were read to the arrestant and airmed by him/her’ (ILU1990\ilu0224)

century the direct inluence of German had been absent for a long time and the Estonian written culture had developed, with a strong tradition and a diverse body of native authors. As we can see in the table, by the end of the 20th century the most frequent agent-marking device in iction texts had become the elative (49% of clauses where A is expressed). However, compared to the earlier period, its use has strong constraints which will be discussed in Section 4.1.

150 Liina Lindström

he use of the adessive has broadened, as has the use of the postposition poolt ‘by’. Some of the agent-marking devices – the postpositions läbi ‘through’ and käest ‘from, lit. from the hand’ – have disappeared completely, as witnessed by the current material. he comitative case is still used, but mostly in sentences where the main verb is katma ‘to cover’ (Example 7), i.e. mainly in instrument-like functions. (7) /…/ pruss ol-i kae-tud rohelis-te vetika-te-ga, baulk be-pst.3sg cover-ppp green-pl.gen seaweed-pl-com libe nagu saarmanahk. slippery as otter_skin ‘/…/ the baulk was covered with green seaweed, slippery like the skin of an otter.’ (ILU1990\ilu0167)

In comparison with the language of iction texts, the spoken language corpus very rarely shows the elative and the poolt ‘by’-construction used, and the main agentmarking device in spoken language is the adessive case.

4. Comparison of the use of elative and adessive As in 1800–1850 and the 1990s: Subject properties In this section, the subject properties of elative and adessive As are compared with regard to semantic properties, topicality and control over relexivization. 4.1

Semantic constraints on the use of adessive and elative as agent-marking devices

While the frequency of the adessive and elative as agent-marking devices has not changed much – at least based on the corpora used for this study – the semantic constraints show a greater change. In this subsection I will concentrate on the constraints for the use of elative and adessive As during the two time periods in question. he main characteristic under observation will be the animacy of the A, following the animacy hierarchy described irst in Silverstein (1976), later adjusted by others (cf. Crot 2003: 130): 1st & 2nd person > 3rd person > proper names > humans > non-human animates > inanimates. Elative In the early 19th century, the elative was widely used for denoting the volitional agent as well as inanimate As. Using the elative to refer to animate As was more frequent (49 instances) than using it for inanimate As (21 instances). Following the animacy hierarchy it can be assumed that in the 19th century, the elative could

Subjecthood of the agent argument in Estonian passive constructions

be used to refer to nearly all levels of the hierarchy. While there are no examples of the 1st and 2nd person pronouns (probably due to the nature of the texts), there are examples of 3rd person pronouns (see Example 8), proper names, humans (Example 9), non-human animates (10), and inanimates (11) which can also be abstract (12). he elative can also be used to refer to a collective A (institution) (13). By far the most frequent A referred to with the elative was God (also Creator, Lord, Father, Holy Ghost). Jumal ‘God’ alone was used as an elative A 20 times in the material (14). (8) Et nemmad kül se++pärrast teis-te-st sa-wad that they part therefore other-pl-ela get-3pl teo-tud ja naer-tud, disgrace-ppp and laugh-ppp ei panne nemmad sedda naero siiski ennestele raskeks. ‘hough they will be therefore disgraced and derided by others, they do not ind this laughter hard for themselves’ (1847-Kersten_87) (9) Keik nee-d Sanna-d sa-i-d, senni kui ta räki-s, all these-pl word-pl get-pst-3pl until then he/she talk-pst.3pl Sikkertari-st ülles kirjo-tud. secretary-ela up+ write-ppp ‘All these words were written down by the secretary, while he was speaking.’ (1817-Holtz_9) (10) Kui naad agga sui ussi-de-st wägga ärra++ If they but summer worm-pl-ela very rikku-tud on, neid ei sünni raio-da. blemish-ppp be:pl3, they.pl.par not beit.conneg chop_down-inf ‘But if they are very spoiled by worms, then they are not it for chopping down’ (1840-Jordan_25) (11) Noor-t härjapea-d ei pea ka mitte söde-ta-ma, young-par clover-par not must.conneg ptcl not feed-iprs-sup kui pöld wihma-st on ärraleu-tud … when ield.nom rain-ela be.3sg soak-ppp ‘Young clover must not be fed [to animals] when the ield is soaked by rain.’ (1837-Knüpfer_9) (12) Ei mahtu-nud rööm mis senne-st temma-l oll-i, not it-app joy what it-ela he/she-ade be-pst.3sg temma süddame-sse, mis ligarmu-st hopis täide-tud. he/she.gen heart-ill what overmuch_love-ela ptcl ill-ppp ‘he joy he had because of this did not it into his heart, which was illed with overlowing love.’ (1816-Masing_25)

151

152 Liina Lindström

(13) Agga keik nisugguse-d kauba-d but all such-pl.nom agreement-pl.nom pea-wad Kohtu-st sa-ma kinni-tud. must-3pl court-ela get-sup airm-ppp ‘But all such agreements must be airmed by the court’

(1817-Holtz_79)

(14) Olgo need kül sure-d, siis õmmetige selle wasto ei be-jus.pl3 these ptcl great-pl then however it:gen against not seisa, kes Jummala-st nende ülle on stand.conneg who god-ela they.pl.gen over be.3sg pan-tud wallitse-ma. put-ppp govern-sup ‘hough they are great, still [they] can’t stand against the one who is appointed to govern them by God’ (1818-Masing_108)

hus, the elative has no considerable semantic constraints as a marker of A in the early 19th-century texts; it is the most ordinary and frequent way to denote the A. In the texts from the second period (the 1990s), however, the situation is quite diferent. Only ive sentences had an animate A marked with elative case, while 34 sentences had an inanimate A, thus the elative is used mainly to denote inanimate As. Moreover, the elative may express animate As only in a few constructions, mainly in the passive experiencer construction. Namely, four out of ive instances of the animate As marked with elative occurred in the passive experiencer construction as a stimulus argument (minust ‘I-ela’ in Example 15, sinust ‘you-ela’ in Example 16). In its active counterpart the stimulus is marked with the nominative and agrees with the verb (as ma ‘I’ in Example 15b) whereas the experiencer is marked with the partitive (kedagi in 15b). he construction as in Example (15b) represents the experiencer-object construction where the experiencer is marked as a typical object (Lindström 2013). he experiencer-object construction is non-canonical in the sense that the participant roles and grammatical relations are inverted compared to typical or canonical transitive clauses: the most likely A (= experiencer) is marked as the object and the (typically non-human) stimulus is marked as the subject of the clause. (15) a.

Viieteistkümnesele on see traagika: keegi ei märka mind, Keegi pole minu-st huvita-tud,/–-/ Experiencer stimulus anybody not_be.conneg I-ela interest-ppp ‘For a 15-year-old, it is a tragedy: nobody notices me,nobody is interested in me, nobody cares about me’ (ILU1990\ilu0002)

Subjecthood of the agent argument in Estonian passive constructions

b. Ma ei huvita kedagi. Stimulus experiencer I.nom not interest.conneg anybody.par ‘Nobody is interested in me’ (16) Ah, kuidas ma sinu-st huvita-tud ole-n, Algi! oh how I.nom you-ela interest-ppp be-1sg name ‘Oh, how I am interested in you, Algi!’

hus, the elative stimulus is used oten in the passive counterpart of the experiencer-object construction. Interestingly, in this passive counterpart the relations of semantic roles and syntactic relations resemble a canonical (transitive or intransitive) clause, usually described in Estonian grammars as a normal clause (EKG II: 14, Erelt 2003: 93) or unmarked basic clause (Erelt & Metslang 2006) – the experiencer (the more agentive participant in the event) is the grammatical subject in this passive sentence, as it is case-marked by the nominative and agrees with the verb (16) and thus resembles the canonical subject more closely than the partitive experiencer in the corresponding active sentence does. he stimulus is in the elative case as in other types of experiencer constructions (sinust ‘you-ela’ in Example 17). (17) Mu-l on sinu-st kahju. I-ade be.3sg you-ela sorry ‘I am sorry for you’

However, the use of the elative A (= stimulus) is not restricted to human or animate stimuli, but also to clearly non-agentive stimuli, as millestki ‘something’ in Example (18). (18) Henri pid-i enne töö-le tuleku-t ole-ma Henri must-pst.3sg before work-all coming-par be-sup millestki häiri-tud. something.ela disturb-ppp ‘Henry must have been disturbed by something before coming to work’ ILU1990\ilu0159)

he number of predicates that allow the use of elative As like these is limited to those itting into the experiencer-object construction [np(experiencer)-par v np(Stimulus)-nom], for example huvitama ‘to interest’, häirima ‘to disturb’, rõõmustama ‘to gladden’, vaimustama ‘to inspire, to enrapture’, ehmatama ‘to startle’, erutama ‘to excite’, lõbustama ‘to amuse’ etc. (cf. Lindström 2012, 2013).

153

154 Liina Lindström

Nevertheless, it seems that the passivized equivalents of this construction type have developed into an independent construction [NP-nom(Experiencer) be NP-ela(Stimulus) V-ppp], since the same construction is used with some predicates which cannot be used in the experiencer-object construction (compare 19a and b). (19) a.

Passive: Ma ole-n Esa-st sisse võe-tud. I be-1sg Esa-ela in take-ppp ‘I am fond of Esa’ b. Active: *Min-d võtti-s Esa sisse. I-par take-pst.3sg Esa.nom in ≠‘Esa is dear to me’ (grammatical under literal interp.: ‘Esa took me in’)

Although it was not observed from the corpus of iction of the 1990s, Uuspõld (1970) has mentioned another type of predicate which allows the use of elative animate As: ümbritsema, piirama (‘to border, to surround’, Example 20a). In this case it is a collective A in an instrumental-like use; this is indicated by the possibility of paraphrasing it with the comitative case (20b), a typical instrument-marking device in Estonian. It is interesting to note that the active equivalent of this construction is somewhat similar to the experiencer-object construction – the clause begins typically with a partitive NP [NP-par V NP-nom] (20c), cf. Rätsep 1978: 129). Nevertheless, the use of elative As cannot be generalized to all passive equivalents of sentences containing the [NP-par V NP-nom] construction, as some of these can also be used with the comitative (e.g. katma ‘to cover’, Example 7). Juubilar ol-i sõpra-de-st ümbritse-tud. jubilarian.nom be-pst.3sg friend-pl-ela surround-ppp ‘he jubilarian was surrounded by friends’ b. Juubilar ol-i sõpra-de-ga ümbritse-tud. jubilarian.nom be-pst.3sg friend-pl-com surround-ppp ‘he jubilarian was surrounded with friends’ c. Juubilari ümbritse-sid sõbra-d ja kolleegi-d. jubilarian.par surround-pst.3pl friend-pl and colleage-pl ‘Friends and colleagues surrounded the jubilarian’

(20) a.

On the basis of the corpus of the 1990s, it appears that elative As denoting animate participants can be used also while referring to Jumal ‘God’. he frequent use of this word in the elative case in the earlier texts gives reason to assume that in current usage its elative use has become ixed, expecially in connection with the predicate looma ‘create’ (21).

Subjecthood of the agent argument in Estonian passive constructions

(21) Inimese-d olla ju jumala-st loo-dud kahekesi people-pl be.kvt part god-ela create-ppp by_two koos ela-ma, teineteise-st rõõmu tund-ma. together live-sup each_other-ela joy.par feel-sup ‘People are created by God to live together and enjoy each other’ (ILU1990\ilu0205)

Compared to animate As, the elative is used considerably more to mark inanimate As, both concrete (22) and abstract (23). Some of them could be described semantically as force (22). his is the most usual and the most frequent usage of elative As in passive constructions in modern Estonian. (22) ja kask maja otsas kohises endiselt, nagu pole-ks ta oksa-d ja tüvi like not_be-cond.conneg it-gen branch-pl and bole saa-nud-ki riku-tud raudse-st vihma-st… get-app-ptcl ruin-ppp iron-ela rain-ela ‘and the birch was still rustling, as if its branches and trunk had never been ruined by the hard rain’ (ILU1990\ilu0028) (23) See või-s olla tingi-tud ta mineviku-st. it can-pst.3sg be.inf condition-ppp he/she.gen past-ela ‘It could have been conditioned by his past’ (ILU1990\ilu0216)

hus, adding animate human As to the passive construction is possible in modern Estonian only with a limited number of verbs or construction types. Most verbs do not allow the use of elative to mark animate As; the elative argument will then be interpreted as something else. For example, sentence (24) is only acceptable if the elative kokast ‘chef-ela’ refers to the ingredients of the soup. (24) ?Supp on koka-st keede-tud. soup be.3sg chef-ela cook-ppp ≠‘he soup is cooked by the chef ’ / =‘he soup is made of the chef ’

he elative case cannot be used to mark higher animals as As either (25). However, using the elative to express smaller animals and insects as As is plausible if there is an unspeciied number of them, just as in Example (26) the number of mice is unspeciied and irrelevant. (25) *Kont on koera-st näri-tud. bone be.3sg dog-ela chew-ppp ‘he bone has been chewed by the dog’ (26) Vaip ol-i hiir-te-st näri-tud. carpet be-pst.3sg mouse-pl-ela chew-ppp ‘he carpet was chewed by mice.’

155

156 Liina Lindström

hus, there has been a considerable change in the use of the elative between the early 19th century and late 20th century: the elative, which was once used to express all kinds of As, can now only be used to mark inanimate As. In modern Estonian, the elative case can be used to mark animate As only in certain constructions (the passive experiencer construction, with a limited number of predicates) and to mark smaller animals and insects in groups of unspeciied numbers. Adessive he constraints on the use of adessive to mark As are quite diferent from the elative. I found a total of 13 instances of the adessive case being used to express As in the 19th-century texts. All of these referred to an animate human agent acting volitionally. Most of the instances were pronouns (27); NPs referring to humans were rarer (28). (27) Mei-l on jo hea koggo walmis wolista-tud. we-ade be.3sg ptcl good collection.nom ready carve-ppp ‘We have a good amount carved already’ (1843-Schüdlöfel_1106) (28) “Seal ta on!” ütle-s märra, jalga üllestöstes, there it be.3sg say-pst.3sg broodmare foot-par up_lit.ger, mis eile seppa-l uest oll-i rauta-tud. what.nom yesterday blacksmith-ade again be-pst.3sg shoe-ppp ‘“here it is!” said the broodmare, liting up the foot which had been shod again by the blacksmith yesterday.’ (1850a-Kreutzwald_lisa2_19)

During the second period under observation (the 1990s) the situation is the same: the adessive case is only used to refer to animate human volitional agents, which are usually expressed by a pronoun (29), on a few occasions also by an NP referring to a human (30). (29) einoh, ta-l on juhtme-d pan-tud noh part he/she-ade be.3sg wire-pl.nom put-ppp part ‘Well, he has done the wiring’ (30) See on tudengi-l salaja teh-tud. it.nom be.3sg student-ade secretly make-ppp ‘his has been made by the student secretly’

(spoken)

(ILU1990\ilu0214)

In modern Estonian, the adessive argument can only be used to refer to humans, exceptionally also to animals depicted as having some human characteristics such as volitional action, premeditation (31). Sentence (31) is more acceptable if the adessive argument is located at the beginning of the sentence in a typical subject (topic) position. he adessive case can never be used to refer to smaller animals and inanimate agents (32)–(33).

Subjecthood of the agent argument in Estonian passive constructions

(31) Koera-l on kont näri-tud. / ?Kont on koeral näritud. dog-ade be.3sg bone.nom bite-ppp ‘he bone has been chewed by the dog’ (32) *Pluus on koi-de-l söö-dud. / shirt nom be.3sg moth-pl-ela eat-ppp *Koidel on pluus söödud. ‘he shirt has been eaten by moths.’ (33) *Puu on külma-l võe-tud. tree.nom be.3sg cold-ade take-ppp ‘he tree is frozen’, lit. ‘he tree has been taken by cold’

In conclusion, it appears that during the last 200 years, a specialization of the means of marking A has taken place. In modern Estonian, the elative is mainly used to express inanimate As such as stimulus, instrument, force, etc., located at the end of the animacy hierarchy, whereas the adessive is used to express volitional, animate As. Comparing the two time periods, it is clear that in the early 19th century, this division of labor did not yet exist as both the elative and adessive could be used with animate volitional As. It developed at some time between the two periods. Agentivity has been considered one of the properties of (canonical) subjecthood – subjects are typically agents or initiators of events (Keenan 1976). herefore, in modern Estonian the adessive A can be attributed at least one subject-like property that the elative A does not have – agentivity, i.e. the ability to be an energy source and act volitionally. In the next sections I take a closer look at some other subject properties of adessive and elative As in the passive construction such as topicality (Section 4.2.1) and control over relexivization (4.2.2). 4.2

Other subject properties of adessive and elative As

4.2.1 Word order and topicality Syntactic position can shed light on the possible behavior of the argument as a non-canonical subject (see e.g. Andrews 1985; Onishi 2001; Haspelmath 2001; Barðdal & Eythórsson 2003; Barðdal 2006). he basic word order of Estonian is SVO (Tael 1988). In interrogatives and in some embedded clauses, the verb tends to occur in the inal position (Lindström 2005, 2006). In main clauses starting with the adverbial, the verb usually takes the second position and the subject occurs in the 3rd position (Remmel 1963; Tael 1988; Lindström 2005) – the V2-rule, well known from Germanic languages,

157

158 Liina Lindström

operates in Estonian as a result of long-lasting contacts with German. In spite of the restriction, Estonian word order is relatively free and very sensitive to the information structure (Lindström 2005). here is a strong tendency in languages for subjects to also be topics (Lambrecht 1994: 132; Chafe 1976). his is also the case in Estonian: the nominative subject is typically in the beginning of the clause and functions as a topic in basic non-inverted clause types. Passivization is the syntactic process through which the P (normally the object of the active clause) is foregrounded and the A (subject of the active clause) is backgrounded. Foregrounding can be done syntactically by promoting the P argument to the subject and backgrounding by demoting the A argument to oblique or eliminating it from the clause. he P argument is also foregrounded pragmatically: the passive prefers P to be the topic, while the active prefers A to be the topic (see Comrie 2008: 9; Keenan & Dryer 2007). hus the most typical position of the P in passive clauses is in the beginning of the clause – in the topic position. If the A acquires the features of the subject (non-canonical subject), we can presume that it also tends to take the topic position. herefore in the passive clause it may happen that two arguments – the P marked by the nominative (syntactic subject) and the A marked by the oblique – may compete for the topic position. In the next paragraphs, the order of the A and P is discussed (see Table 3). he position of the verb is not taken into account, as it varies also because of V2 (Tael 1988; Lindström 2005). Table 3. he ordering of A and P arguments in passive clauses. Time period

1800–1850

1990s

Order of A and P

PA

AP

Total

PA

AP

Total

Elative A Adessive A

63 6

5 6

68 12

37 12

3 15

40 27

Table 3 shows that in sentences with an elative argument the P is overwhelmingly located before the A, even if the P is inanimate and the A is animate. he P argument is typically located at the beginning of the sentence (Example 34). hus, sentences with the elative exhibit the word order characteristic of passive sentences. his is common to both the 19th-century and late 20th-century texts; no signiicant changes have taken place. Elative A can occur at the beginning of the sentence before P (35), but this is rare. It is important to note that all examples of AP word order with elative A were found in sentences where the A denotes an animate human participant (or God, as in 35).

Subjecthood of the agent argument in Estonian passive constructions

(34) Kiskja-d on Jummala armu-st teis-te predator-pl be.3pl god.gen grace-ela other-pl.gen loma-de hea-ks lo-dud. animal-pl.gen good-trnsl create-ppp ‘Predators are created by God’s grace for the good of other animals’ (1818_Masing_163) (35) /…/ Jssanda-st oll-i temma-l üks suur koorm god-ela be-pst.3sg he/she-ade one big burden peäle pan-dud, mis temma omma ello otsa-ni on put-ppp, what he/she own life.gen end-ter kannatlikkult ja hea mele-ga kand-is. patiently and good mind-com carry-pst.3sg ‘A great burden had been put onto him/her by God, which he/she bore patiently and happily until the end of his/her life.’ (1847-Kersten_1719)

he adessive A can be located equally before or ater the P in both time periods; there have been no relative changes in frequency of position of adessive arguments during the past 150–200 years. In sentences with adessive arguments (AdA), there is a diference in interpretation between sentences beginning with A and with P: if the AdA is located at the beginning of the sentence (before P), the sentence is more likely to describe the (past) action of the A (27, 36). If P is located before A, the sentence describes the state of P (which is the result of the past action of the A) rather than A’s past action itself (37). hus, the adessive A at the beginning of the sentence clearly acts as the topic of the sentence, like a canonical subject in a basic sentence (Keenan 1976: 318), and has an efect on the interpretation of the whole sentence as an A-oriented event. herefore, a passive sentence beginning with AdA is similar to an active sentence – AdA behaves like a “igure within the proiled relationship” (Langacker 1991: 316). (36) Ruiso-l, muide, ol-i alati raamat Ruiso-ade by_the_way be-pst.3sg always book.nom kaasa võe-tud. with take-ppp ‘Ruiso, by the way, had always taken the book with him’ (ILU1990\ilu0021) (37) See kena kahetoaline ol-i ju this nice two_room.nom be-pst.3sg ptcl nei-l koos oste-tud /…/ they-ade together buy-ppp ‘his nice two-room apartment was bought by them together’ (ILU1990\ilu0172)

159

160 Liina Lindström

he adessive A also behaves like a subject in a passive sentence by following the V2 rule. If the adverbial is placed at the beginning of the sentence, the subject moves from the beginning of the sentence to the position ater the verb (subjectpredicate inversion). he same happens to the adessive A (but not to the elative A): AdA can be inverted and moved to the 3rd position in the clause (38). At the same time the AdA continues to act as a topic (just like adessive A in the beginning of the sentence): in Example (38) the focus is on the action of A, not the state of P. Moreover, P is a new referent in the discourse, it is afected by the action of the A (like a typical object of a transitive sentence) and is located in a typical object position in the sentence, although it is a syntactic subject (by nominative case marking and agreement). (38) Aga Romeo kõrvalegi on mu-l oma inimene sea-tud. but Romeo.gen next be.3sg I-ade own person set-ppp ‘But I have also set my own person beside Romeo.’ (ILU1990\ilu0012)

4.2.2 Relexivization A well-known feature which is usually considered to be a subject property is control over relexivization – the relexive pronoun is coreferential with the subject rather than any other elements of the clause (cf. Keenan 1976; Andrews 1985; Barðdal 2006). his seems to be one of the most universal subjecthood tests – no language has been found where the object argument only (and no A argument) controls relexivization (Onishi 2001: 15). hus, it is the subject which controls the use of the relexive pronoun oma ‘own’ or enese~enda ‘self ’ in Estonian. Relexivization is also suitable for testing the subjecthood of the A argument in passive constructions where two subject candidates appear: a nominative P argument and an elative or adessive A argument. As the corpus data is not suicient to draw any conclusions here, in the following section the relexivization test is conducted mainly on the basis of the author’s own intuitions. hus the results represent only the contemporary situation and say nothing about the situation in the 19th century. In agentless passives, the foregrounded P argument (syntactic subject, lapsed in Example 39) controls the use of the possessive relexive pronoun oma ‘own, self ’. (39) Lapse-di ol-i-d omai laua taha istu-ma pan-dud. child-pl.nom be-pst-3pl own desk.gen behind sit-sup put-ppp ‘Children were seated at their desk’

Subjecthood of the agent argument in Estonian passive constructions

Elative In passive clauses with an elative A, the only controller over relexivization seems to be the nominative P argument, even if the A argument refers to animate human As, as happens mainly with experiential predicates (40). Even the topicalization of A does not change the interpretation of the relexive pronoun (41). (40) Tai pole Etheli-st omai isekuse he/she.nom not_be.conneg Ethel-ela own selishness.gen tõttu kuigivõrd huvita-tud. because very interest-ppp ‘He is not very interested in Ethel because of his selishness.’ (41) Etheli-st pole tai omai isekuse Ethel-ela not_be.conneg he/she own selishness.gen tõttu kuigivõrd huvita-tud. because very interest-ppp ‘He is not very interested in Ethel because of his selishness.’

Adessive Generally, the adessive A seems to control the use of the relexive pronoun, which marks the possessor of P (42). P, on the contrary, cannot control the relexive pronoun which syntactically belongs to AdA (43), thus, AdA is higher on the hierarchy of grammatical relations than the P argument in this construction. (42) Aga Romeo kõrvalegi on mu-li omai inimene sea-tud. but Romeo.gen next be.3sg I-ade own person set-ppp ‘But I have also set my own person beside Romeo. / But next to Romeo, my own person is set by me’ (ADE ILU1990\ilu0012) (43) *See lapsi ol-i omai ema-l alati hästi riieta-tud. this child be-pst.3sg own mother-ade always well dress-ppp ‘his child has always been dressed well by his/her mother’

If we compare clauses where both the AdA and P arguments might be potential antecedents of the possessive relexive pronoun, the AdA controls the use of the relexive pronoun (44). However, this is not without exceptions. Example (45) is ambiguous: the most likely interpretation is (a), but (b) is possible as well. Even the topicality of P does not prevent this ambiguity (46). Context seems to play a crucial role here; with the help of context, it is also clear that in Example (47), oma refers back to the table, and not to the mother, because typically the table has its own place, but not the mother.

161

162 Liina Lindström

(44) Ema-li ol-i leib juba omai kotti piste-tud. motheri-ade be-pst.3sg bread already owni bag.ill put-ppp ‘Mother had put the bread into her bag’ (45) Ta-li ol-i-d lapse-dj omaij laua taha he/she-ade be-pst.3pl child-pl own desk.gen behind istu-ma pan-dud. sit-sup put-ppp a. ‘She had seated the children at her desk.’ b. ‘She had seated the children at their desk.’ (46) Lapse-di ol-i-d pan-dud ta-lj omaij child-pl be-pst.3pl put-ppp he/she-ade own laua taha istu-ma. desk.gen behind sit-sup ‘he children were seated by her at her/their desk.’ (47) Ema-l ol-i laudi omai koha peale tõste-tud. mother-ade be-pst.3sg table own place on lit-ppp ‘Mother has moved the table to its place.’

hus, AdA generally controls the reference of the relexive pronoun, but this depends on the context; sometimes also the P argument may control relexivization. 4.3

Conclusion: Properties of the subject

During the period between the early 19th and the late 20th century there was a change in the marking of As in passive sentences in Estonian. While in the 19th century the main agent-marking device was the elative, by the end of the 20th century considerable restrictions had arisen in its use; now the use of the elative is restricted to inanimate As. At the same time, usage of the adessive argument, with animate volitional As only, has increased. In addition to animacy, the adessive A has other properties of the subject that the elative argument does not have in modern Estonian. One of the most important of these is topicality: AdA is located in the topic position in a sentence considerably more oten than the elative argument, and similarly to nominative (canonical) subjects, it follows the word-order changes caused by the V2 principle. he topical AdA afects the interpretation of the whole sentence as an A-oriented event, despite its passive morphology. Such a sentence is thus similar to the basic sentence with a nominative subject. Elative A – even if it occasionally occurs in a topic position – does not have this property. In addition, AdA controls relexivization, which is another property that the elative does not have. We may ask why the use of the adessive as an agent-marking device has increased and the use of

Subjecthood of the agent argument in Estonian passive constructions

the elative has decreased; why has the adessive argument in a passive construction acquired more subject properties than the elative? In the next sections, I will briely look at two points that may explain this: (1) the semantics of the adessive and the elative and their use in diferent constructions and (2) the rise and spread of the possessive perfect construction in Estonian and neighbouring languages.

5. Semantics of the adessive and the elative case One possible explanation for the increase in the use of the adessive and decrease of the elative as agent-marking devices might come from their semantics. Both the adessive and the elative are very polysemous cases and can be used to mark diferent semantic functions; both are also originally locative cases. he elative is one of the three internal locative cases, marking separation (‘away from X’) or source (‘from’). Related meanings where this component can easily be seen are source (48), material (49), partitive (part-whole relations, 50), time (51), and cause (52), among others. A more abstract use is marking the stimulus with the elative – the connection to the ‘source’ meaning is not so obvious (53). (48) source: Poiss tuleb kooli-st. ‘he boy comes from school’ (49) material: puu-st laud ‘the table made of wood’ (50) part-whole relations: üks poiste-st ‘one of the boys’ (51) temporal: Esmaspäeva-st hakkab sadama. ‘It will rain starting from Monday’ (52) cause: Käed läksid külma-st kangeks. ‘he hands were going numb with cold’ (53) Stimulus: Ma sain selle-st aru. ‘I understood it’ [lit. ‘I got understanding from it’] (54) Inanimate A: Puu on külma-st võe-tud. ‘he tree is frozen’, lit. ‘taken by the cold’

Cause is especially closely related with inanimate As and sometimes it is even dificult to distinguish cause and inanimate A supericially (compare 52 and 54). In both cases, the elative marks the source of the energy which causes the change, the main diference being that arguments with the role of cause typically do not occur as subjects in the active clause.

163

164 Liina Lindström

hus, the elative case should be an appropriate means for marking the A (including volitional agent), as it marks the source – and also the source of energy. Other agent-marking devices in Estonian (the poolt-construction, the ablative in South Estonian) also have the same meaning component (source, directionality). However, the meaning ‘source’ is present also in the German von-construction (55), hence it has probably also inluenced A marking in Estonian passive constructions. (55) Der Brief wird von mir geschrieben.5 ‘he letter is being written by me.’

As German had a strong inluence on Estonian during the irst observed period (1800–1850), there is a reason to assume that the use of the elative as a marker of A in Estonian passive constructions is a contact-induced phenomenon which is based on the source-metaphor and is strengthened by the German von-construction. Also, Ross (1997) has explained the use of the elative as a marker of agent in 17th-century texts with German inluence. As the German von-construction typically refers to animate As (as in 55), it obviously inluenced the widespread use of the elative in referring to animate As during the irst observed period. Since then the German inluence on Estonian has decreased, and also the use of the elative has become restricted only to inanimate As. he adessive, on the contrary, has a stative meaning which does not include directionality. It is hard to ind any semantic motivation why the adessive is used to mark a volitional agent as it typically does in passive constructions. he adessive is an external locative case, meaning ‘on the surface/top of X’ (56) which can also be used to mark event time (57). However, compared to the elative it has more grammaticalized functions, e.g. possessor (58), experiencer (59), deontic agent (60), etc. (cf. Vainik 1995; Lindström & Tragel 2006; Klavan et al. 2011 etc.) As we can easily see from these examples, all the constructions used in Examples (58)–(60) are based on Location schema (56). (56) Location Riiuli-l on tolmu. Shelf-ade be.3sg dust.par ‘here is some dust on the shelf ’ (57) Temporal (event time): Esmaspäeva-l saja-b. Monday-ade rain-3sg ‘It is/will be raining on Monday’ 5. Source of the example: http://german.about.com/library/weekly/aa012901a.htm

Subjecthood of the agent argument in Estonian passive constructions

(58) Possessor Jaani-l on väike õde. Jaan-ade be.3sg little sister ‘Jaan has a little sister’ (59) Experiencer Jaani-l on paha. Jaan-ade be.3sg bad ‘Jaan feels sick’ (60) Deontic agent in modal constructions Mu-l on vaja tööta-da. I-ade be.3sg need work-inf ‘I have to work’

hus, we can conclude that, semantically, the elative seems to be an appropriate means for marking the A, as it has the meaning component ‘source’, which can also be used as a source of energy (agent). he adessive is a polysemous static case and its diferent uses in diferent constructions are more grammaticalized. As it also marks the deontic agent (in modal constructions) and experiencer, the agent-like use can motivate its spread to the passive constructions. As we can see in the next section, the adessive argument has some subject properties also in other constructions.

6. Subject properties of the adessive argument in other constructions he adessive argument has subject properties in many other constructions as well. For example, the adessive argument controls the use of the relexive pronoun in the possessive construction (61, Erelt & Metslang 2006) and in the experiencer construction (62). (61) Mu-l on oma venna-ga ühine maja. I-ade be.3sg own brother-com common house ‘I share a house with my brother’ (62) Mu-l on enda-st kahju. I-ade be.3sg self-ela pity ‘I feel sorry for myself ’

According to Seržant (to appear), the use of the adessive argument in Estonian experiencer clauses (as in Example 62) is similar to what is found in Latvian and Lithuanian, where the same constructions exploit the dative to mark the experiencer; at the same time, Russian uses either the dative or (in some constructions)

165

166 Liina Lindström

the spatial schema (Russian u menja is a counterpart of the Estonian adessive argument mul ‘on me’). he dative-like experiencers have spread widely in the Eastern Circum-Baltic area; predicate classes where dative-like experiencers are used show a signiicant structural parallelism in Eastern Circum-Baltic languages (Seržant, to appear). At the same time, dative-like experiencers are not characteristic of Standard Average European (SAE) but are used more oten on the periphery of SAE; in central SAE languages, agent-like (nominative) experiencers dominate (Haspelmath 2001). hus, the use of adessive non-canonical subjects seems to be an areal feature. In addition to the experiencer constructions, the adessive agent argument is used in many modal constructions where it controls subject deletion in ininitives (in 63, deletion of mina ‘I’ from the ininitival construction endast kirjutada) and relexivization (endast ‘from (my)self ’ in 63). (63) Mu-li on vaja enda-sti kirjuta-da. I-ade be.3sg need self-ela write-inf ‘I have to write about myself ’

hus, adessive arguments are used as non-canonical subjects in many constructions, and this may be the reason why the AdA has obtained the behavioral properties of the subject in passive constructions as well, though still in the early 19th century, the elative A was more common and more frequent.

7.

he spread of the possessive perfect

As we can see, a passive sentence with adessive A behaves in many ways diferently from other passive sentences: the adessive is typically topicalized and bears some subject properties, and in this case the overall construction designates the completion of the action of A rather than the state of the P (which is typical of passives). herefore, we have also called it the possessive perfect construction (PPC; cf. Lindström & Tragel 2007, 2010). he PPC is in many ways similar to the have-perfect found in several European languages – it is based on the possessor construction (cf. Heine & Kuteva 2006) but uses diferent morphosyntactic means from the SAE languages. he possible development of this construction in Estonian has been discussed by Lindström & Tragel (2010). he PPC has evolved from a passive-like construction where the topical (external) possessor has been reinterpreted as an agent, e.g. in (64) the adessive argument sul may be interpreted as a possessor (older interpretation) or an agent (newer interpretation). From there the interpretation as an agent has been extended to intransitive verbs, which cannot be used in the passive construction

Subjecthood of the agent argument in Estonian passive constructions

(Examples 65–66). hus, the possessive perfect construction can also be used with intransitive verbs and it cannot be simply considered a subcategory of the passive construction. (64) Kas su-l on pilet oste-tud? q you-ade be.3sg ticket.nom buy-ppp ‘Have you bought a ticket?’ (65) Mu-l on poe-s käi-dud. I-ade be.3sg shop-ine go-ppp ‘I have done the shopping’ / ‘I have been to the store’ (66) Su-l on juba aasta otsa seal ol-dud, you-ade be.3sg already year long there be-ppp peaks olema küll hollandi keel suus. ‘As you have been there for a year already, you should speak Dutch.’

In Estonian linguistics, sentences like in Example (64) are usually regarded as belonging to the passive paradigm (Pihlak 1993: 81; Rajandi 1999; Erelt (ed.) 2003: 102–103). Some studies have treated it as a separate construction (Meerwein 1994; Holvoet 2001; Lindström & Tragel 2007, 2010). Lindström & Tragel (2010) have shown that the construction is not always clearly distinguishable from the passive. he main diference concerns word order and information structure: in passive clauses, typically the P argument is in the topic position, while in the possessive perfect construction, the AdA is topical (and P can be missing). As the distinction of the PPC and passive is not clear-cut and using information structure as a main distinguisher may be problematic (Erelt 2013: 195), the distinction is not made in the paper. Nevertheless, it is rather obvious that the use of AdA in passive and passivelike constructions in Estonian has been inluenced by the PPC in surrounding languages. As has been claimed, similar constructions exist in e.g. North-Russian, Latvian, Votic, etc., i.e. are characteristic to Eastern Circum-Baltic languages (Seržant 2012). According to Seržant, the epicenter of the PPC in this area has been North-Russian, where the construction is the most grammaticalized (Seržant 2012: 381). As Timberlake has shown (1976), in North-Russian dialects the locative construction u + gen is used to mark the agent in a possessive perfect construction – it is the semantic equivalent of Estonian adessive arguments (‘on; on the top of)’. Additionally, the North-Russian A has many subject properties (Timberlake 1976; Seržant, this volume), similarly to Estonian. he possessive perfect construction exists and exploits the adessive as an Agent-marking device in some other Baltic-Finnic languages as well, for example in Ingrian (67) and Veps (68), but not in Finnish.

167

168 Liina Lindström

(67) Ingrian6: Minu-n tüdö-ll oppi on lobete-ttu. I-gen daughter-ade school be.3sg inish-ppp ‘My daughter has inished school’ (68) Veps: Ol’-i tsar’i-l sauba-tud t’ürm-ha zmei ulan. be-pst.3sg tsar-ade bolt-ppp prison-ill snake Ulan ‘he snake Ulan had been imprisoned by the tsar’ (Grünthal 2003: 145)

he construction is also present in Latvian (a Baltic language) and in Livonian (a Baltic-Finnic language which has been under the strong inluence of Latvian), but in these languages, the dative case has acted as the counterpart of the Estonian adessive. (he dative is also used to mark the possessor in the possessive construction.) (69) Latvian Viņam viss jau bija izteikts. he:dat all:nom:sg:masc be:pst:3sg say:ppp:sg:masc ‘He had already said everything’ (Holvoet 2001: 375) (70) Livonian kouv mä’ddõn vo’ĺ gōstõt mōdõ tä ̄dõks well we.dat be.3sg belt_down land.par full.trnsl ‘We have fulilled the well by the land’ (Meerwein 1994: 172)

It is notable that in Veps and Livonian, the elative has also been used as a marker of A (71–72), but at least in Veps (according to Grünthal 2003: 145) the use of the elative in this function is quite infrequent and atypical. hus, the adessive dominates also in Veps. (71) Veps leśk om mužika-spei jät-tut widow.nom be.3sg husband-ela leave-ppp ‘he widow has been let by her husband’

(Grünthal 2003: 146)

(72) Livonian sǟl um võ-nd ikš kurēst apsǟs-tõd neitst there is be-app one devil.ela possess-ppp maid ‘here was a maid who was possessed by the devil’ (Grünthal 2003: 147)

6. Example is obtained from a translation test, on the Soikkola peninsula (Valjanitsa, informant Aleksandr Sergejev, 23.07.2009).

Subjecthood of the agent argument in Estonian passive constructions 169

hus, it seems that the elative is or has been a means of A marking also in Veps and Livonian and it has been replaced by the adessive (in Veps) or dative (Livonian), but there is no evidence in these languages of the kind of semantics-based division of functions that has taken place in Estonian during the past 200 years. here is a reason to assume that in both Veps and Livonian the increase of the use of adessive A has happened at least partially because of the contacts with Russian (in the case of Veps) or Latvian (Livonian), i.e. with Eastern Circum-Baltic languages. It seems possible that the reason why the adessive dominates and the elative has lost ground as an agent-marking device in passive constructions in Estonian (and probably also in Veps and Livonian) is the spread of the PPC construction in Eastern Circum-Baltic languages. On the one hand, Estonian adessive A is a semantic counterpart of the same construction in Russian and other languages in the Eastern Circum-Baltic area; on the other hand, AdA has some subject behavioral properties already in these languages.

8. Conclusion A is not commonly expressed in the Estonian passive, but when it is expressed there are several ways to do it. he two agent-marking devices discussed in this paper – the elative and adessive – have gone through a great deal of change since the early 19th century: while the elative was the main agent-marking device in the passive sentence in the 19th century, by the end of the 20th century it could only be used to refer to inanimate As. In the early 19th century, the adessive was rarely used to express the A, but by the end of the 20th century its use had become more frequent. he use of the adessive A is especially profuse in spoken Estonian. However, the adessive is used only for marking animate volitional agents. In addition to the specialization in animate volitional agents, adessive A has gained some other subject properties. Above all, these include topicality (a tendency to occur in the topic position and thus afect the interpretation of the whole sentence) and control over the use of the relexive pronoun. he elative A does not have these properties. here are several reasons for these changes in marking the A. Firstly, this change is likely to have been caused by the subject-like properties of the adessive argument in other constructions – the adessive also tends to behave as a non-canonical subject in possessive constructions, experiencer constructions, and modal constructions. he use of the adessive in these constructions is highly grammaticalized, and its meaning has shited away from the original locative meaning ‘on, on the top of ’. Although the elative is also a polysemous case, its

170 Liina Lindström

meanings in diferent constructions are more closely connected to the original ‘source, separation from’ meaning. Secondly, the elative is a semantic counterpart of the German von-construction and this is a fact which may have afected the large use of the elative as an A marking device in the beginning of the 19th century. At that time, the Estonian literary language was heavily inluenced by German. By the end of the 20th century, German inluence on Estonian had weakened remarkably, and this may serve as one of the reasons why the use of the elative has become restricted such that nowadays it may refer mainly to inanimate As, low in the animacy hierarchy, i.e. untypical As. hirdly, the spread of the adessive argument is undoubtedly connected to the rise and spread of the possessive perfect construction in Eastern Circum-Baltic languages. hus, the elative and adessive as markers of the A in Estonian passive sentences have undergone a change which can be described as a specialization of functions. However, this change cannot be explained with a single factor but rather with a bundle of factors, above all language contacts. Which is, of course, not surprising at all.

References Andrews, Avery (1985). he major functions of the noun phrase. In T. Shopen (Ed.), Language typology and syntactic description. Vol. 1: Clause structure (pp. 62–154). Cambridge: Cambridge University Press. Barðdal, Jóhanna (2006). Construction-speciic properties of syntactic subjects in Icelandic and German. Cognitive Linguistics, 17(1), 39–106. DOI: 10.1515/COG.2006.002 Barðdal, Jóhanna, & Eythórsson, hórhallur (2003). he change that never happened: he story of oblique subjects. Journal of Linguistics, 39, 439–472. DOI: 10.1017/S002222670300207X Beck, David (2000). Semantic agents, syntactic subjects, and discourse topics: How to locate Lushootseed sentences in space and time. Studies in Language, 24(2), 277–317. DOI: 10.1075/sl.24.2.03bec Blevins, James P. (2003). Passives and impersonals. Journal of Linguistics, 39, 473–520. DOI: 10.1017/S0022226703002081 Chafe, Wallace L. (1976). Givenness, contrastiveness, deiniteness, subjects, topics, and point of view. In Charles N. Li (Ed.), Subject and topic (pp. 25–55). New York: Academic Press. Comrie, Bernard (2008).What is a passive? In Z. Estrada Fernández, S.Wichmann, C. Chamoreau, & A. Álvarez González (Eds.), Studies in voice and transitivity (pp. 1–18). München: Lincom. Crot, William (2003). Typology and universals (2nd ed.). Cambridge: Cambridge University Press. EKG II = Erelt, Mati, Kasik, Reet, Metslang, Helle, Rajandi, Henno, Ross, Kristiina, Saari, Henn, Tael, Kaja, & Vare, Silvi (1993). Eesti keele grammatika II: Süntaks. Lisa: Kiri. Tallinn, Estonia: Eesti Teaduste Akadeemia Keele ja Kirjanduse Instituut. Erelt, Mati (1979). Eesti lihtlause probleeme. Tallinn: Eesti Raamat.

Subjecthood of the agent argument in Estonian passive constructions

Erelt, Mati (2003). Syntax. In M. Erelt (Ed.), Estonian language. Linguistica Uralica Supplementary Series 1 (pp. 93–129). Tallinn: Estonian Academy Publishers. Erelt, Mati (2004). Lauseliigendusprobleeme eesti grammatikas. In L. Lindström (Ed.), Lauseliikmeist eesti keeles (pp. 7–15). Tartu Ülikooli eesti keele õppetooli preprindid 1. Tartu. Erelt, Mati, (2013). Eesti keele lauseõpetus. Sissejuhatus. Öeldis. Tartu Ülikooli eesti keele osakonna preprindid 4. Tartu. Erelt, Mati, & Metslang, Helle (2006). Estonian clause patterns from Finno-Ugric to Standard Average European. Linguistica Uralica, 4, 254–266. Grünthal, Riho (2003). Finnic adpositions and cases in change. Suomalais-ugrilaisen Seuran toimituksia 244. Helsinki: he Finno-Ugric Society. Haspelmath, Martin (2001). Non-canonical marking of core arguments in European languages. In A. Y. Aikhenvald, R. M. W. Dixon, & M. Onishi (Eds.), Non-canonical marking of subjects and objects (pp. 53–83). Amsterdam: John Benjamins. DOI: 10.1075/tsl.46.04has Heine, Bernd, & Tania Kuteva (2006). he changing languages of Europe. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780199297337.001.0001 Holvoet, Axel (2001). Impersonals and passives in Baltic and Finnic. In Ö. Dahl & M. KoptjevskajaTamm (Eds.), he Circum-Baltic languages. Grammar and Typology Vol. 2 (pp. 363–389). Amsterdam: John Benjamins. Keenan, Edward L. (1976). Towards a universal deinition of “subject”. In Charles N. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press. Keenan, Edward L., & Dryer, Matthew S. (2007). Passive in the world’s languages. In Timothy Shopen (Ed.), Language typology and syntactic description. (2nd ed.), Vol. I: Clause Structure (pp. 325–361). Cambridge: Cambridge University Press. Klavan, Jane, Kesküla, Kaisa, & Ojava, Laura (2011). he division of labour between synonymous locative cases and adpositions: the Estonian adessive and the adposition peal ‘on’. In Seppo Kittilä, Katja Västi, & Jussi Ylikoski (Eds.), Studies on case, animacy and semantic roles (pp. 111–134). Amsterdam: John Benjamins. DOI: 10.1075/tsl.99.04kla Koks, Helen (2004). Subjekti ja objekti käitumisreeglid komplekslauses. In L. Lindström (Ed.), Lauseliikmeist eesti keeles (pp. 34–39). Tartu: Tartu Ülikooli eesti keele õppetooli preprindid 1. Lambrecht, Knud (1994). Information structure and sentence form. Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511620607 Langacker, Ronald W. (1991). Foundations of cognitive grammar, Vol. 2: Descriptive Application. Stanford: Stanford University Press. Lindström, Liina (2005). Finiitverbi asend lauses. Sõnajärg ja seda mõjutavad tegurid suulises eesti keeles. Dissertationes philologiae estonicae universitatis tartuensis 16. Tartu: Tartu Ülikooli Kirjastus. Lindström, Liina (2006). Infostruktuuri osast eesti sõnajärje muutumisel. Keel ja Kirjandus, 11, 875–888. Lindström, Liina (2012). Tundekausatiivikonstruktsioon eesti moodi. Keel ja Kirjandus, 1, 30–47. Lindström, Liina, (2013). Between Finnic and Indo-European: Variation and change in the Estonian experiencer-object construction. In Ilja A. Seržant & Leonid Kulikov (Eds.), he Diachronic typology of non-canonical subjects (pp. 139–162). Amsterdam: John Benjamins.

171

172 Liina Lindström

Lindström, Liina, & Tragel, Ilona (2006). Mul on saba. Adessiivsest omajakonstruktsioonist. In K. Kerge & M. M. Sepper (Eds.), Finest linguistics. Proceedings of the Annual Finnish and Estonian Conference of Linguistics (pp. 386–399). Lindström, Liina, & Tragel, Ilona (2007). Eesti keele impersonaali ja seisundipassiivi vahekorrast adessiivargumendi kasutuse põhjal. Keel ja Kirjandus, 7, 532–553. Lindström, Liina, & Tragel, Ilona (2010). Possessive perfect construction in Estonian. Folia Linguistica, 44/2, 371–400. Meerwein, Georg (1994). Einige Anmerkungen zu Gemeinsamkeiten in den Tempussystemen der Sprachen des Ostseeraums. Linguistica Uralica, 3, 168–176. Metslang, Helena, (2013). Coding and behaviour of Estonian subjects. Journal of Estonian and Finno-Ugric Linguistics 4(2): 217–293. DOI: 10.12697/jeful.2013.4.2.12 Ney, A. (1928). Kentütlev tehtaviku tegijanimetuseks! Manuscript in the Archive of the Estonian Dialect and Kindred Languages, University of Tartu. Onishi, Masayuki (2001). Non-canonically marked subjects and objects: Parameters and properties. In A. Y. Aikhenvald, R. M. W. Dixon, & M. Onishi (Eds.), Non-canonical marking of subjects and objects (pp. 1–51). Amsterdam: John Benjamins. DOI: 10.1075/tsl.46.03oni Pihlak, Ants (1993). A comparative study of voice in Estonian. Eesti Sisekaitse Akadeemia toimetised. Tallinn. Rajandi, Henno (1999) [1968]. Eesti impersonaali ja passiivi süntaks. Tallinn: Eesti Keele Instituudi toimetised 3. Raun, Alo, & Saareste, Andrus (1965). Introduction to Estonian linguistics. Ural-Altaische Bibliothek XII. Wiesbaden: Otto Harrassowitz. Remmel, Nikolai (1963). Sõnajärjestus eesti lauses. In Eesti keele süntaksi küsimusi. Keele ja Kirjanduse Instituudi uurimused VIII (pp. 216–389). Tallinn: Eesti Riiklik Kirjastus. Ross, Kristiina (1997). Kohakäänded Georg Mülleri ja Heinrich Stahli eesti keeles. In M. Erelt, M. Sedrik, & E. Uuspõld (Eds.), Pühendusteos Huno Rätsepale. Tartu Ülikooli eesti keele õppetooli toimetised 7 (pp. 184–201). Tartu. Rätsep, Huno (1978). Eesti keele lihtlausete tüübid. Tallinn: Valgus. Sahkai, Heete (2011). Eesti keele genitiivse agendifraasi süntaks. Keel ja Kirjandus, 1, 12–30. Seržant, Ilja A. (2012). he so-called possessive perfect in North Russian and the Circum-Baltic area. A diachronic and areal account. Lingua, 122, 356–385. DOI: 10.1016/j.lingua.2011.12.003 Seržant, Ilja A. to appear. Dative experiencer constructions as a Circum-Baltic isogloss. In P. Arkadiev, A. Holvoet & B. Wiemer (Eds.), Contemporary approaches to Baltic linguistics. Berlin: Mouton de Gruyter. Silverstein, Michael (1976). Hierarchy of features and ergativity. In R. M. W. Dixon (Ed.), Grammatical categories in Australian languages (pp. 112–171). Canberra: Australian Institute of Aboriginal Studies. Tael, Kaja (1988). Sõnajärjemallid eesti keeles (võrrelduna soome keelega). Tallinn: Eesti NSV Teaduste Akadeemia Keele ja Kirjanduse Instituut. Preprint KKI-56. Timberlake, A. (1976). Subject properties in the North Russian passive. In Charles N. Li (Ed.), Subject and topic (pp. 545–570). New York: Academic Press. Torn, Reeli (2002). he status of the passive in English and Estonian. In H. Hendriks (Ed.), RCEAL working papers in English and applied linguistics 7 (pp. 81–106). Cambridge: Research Centre for English and Applied Linguistics. Torn, Reeli (2006). poolt-tarind eesti keele impersonaalis ja passiivis. In P. Penjam (Ed.), Lause argumentstruktuur (pp. 108–121). Tartu Ülikooli eesti keele õppetooli preprindid 2. Tartu.

Subjecthood of the agent argument in Estonian passive constructions

Torn-Leesik, Reeli (2009). he voice system of Estonian. Language Typology and Universals (Sprachtypologie und Universalienforschung), 62(1/2), 72–90. DOI: 10.1524/stuf.2009.0005 Torn-Leesik, Reeli, & Vihman, Virve-Anneli (2010). he uses of impersonals in spoken Estonian. SKY Journal of Linguistics, 23, 301–343. Uuspõld, Ellen (1970). On the agent adverbial in the tud-construction. Annual meeting of the research group for generative grammar. Abstracts (pp. 38–43). Tartu State University. Department of Estonian. Vainik, Ene (1995). Eesti keele väliskohakäänete semantika kognitiivse grammatika vaatenurgast. (Ars grammatica.) Tallinn: Eesti Teaduste Akadeemia Eesti Keele Instituut. Vihman, Virve-Anneli (2007). Impersonaliseeritud impersonaal kui konstruktsioonitasandi grammatisatsioon. Emakeele Seltsi aastaraamat, 52, 158–178. Vihman, Virve-Anneli (2008). Construction-based interpretation of implicit arguments. In N. Adams, A. Cooper, F. Parrill, & T. Wier (Eds.), CLS 40: he Panels (2004): Annual Meeting of the Chicago Linguistic Society (pp. 225–239). Chicago: Chicago University Press. Viitso, Tiit-Rein (2003). Rise and development of the Estonian language. In M. Erelt (Ed.), Estonian Language. Linguistica Uralica Supplementary Series (Vol. 1, pp. 130–230) Tallinn: Estonian Academy Publishers. Wiedemann, Ferdinand Johann (1875). Grammatik der ehstnischen Sprache, zunächst wie sie in Mittelehstland gesprochen wird, mit Berücksichtigung der anderen Dialekte. St. Pétersbourg: Imprimerie de l’Académie Impériale des sciences.

173

Categorization and semantics of subject-like obliques A cross-linguistic perspective Ilja A. Seržant Institute of Lithuanian Language

his chapter’s underlying framework is one of functionalist cognitive linguistics. It suggest a categorization of non-prototypical trajector (subject) constructions into syntax-, gram- and lexeme-driven ones depending on the nature of the domain that triggers the oblique case-marking on the trajector argument. Additionally, a uniied semantic account is proposed, which is based on the comparison with causative events. he structure of these events consists of an antecedent subevent (typically implicit) and a subsequent subevent. his study argues that constructions with non-prototypical trajectors (subjects) refer to consequent events. hat is, all three types of constructions exhibit an invariant semantic core; they conceptualize the event as being a (e.g. causally) consequent event and imply the existence of a (e.g. causally) antecedent event. he diferences between the three types pertain mainly to the referential properties of the antecedent event and its main participant: while with the syntax-driven type the antecedent event is explicit, referential and conceptualized onstage, with the gram-driven type it is implicit, non-referential and ofstage, though conined to a particular concept. he lexeme-driven type only implies the existence of an antecedent event; it does not, however, commit any assessment on the concept of this event.

1.

Introduction

Extensive research on non-canonical subjects and their relation to canonical ones (cf., inter alia, Aikhenvald et al. 2001; Bhaskararao & Subbarao 2004; Barðdal 2008, 2009; Seržant & Kulikov 2013) as well as research on alternations in the case marking of the highest ranked argument (Kittilä 2002; de Hoop & de Swart 2008, inter alia) currently being undertaken. he main body of previous investigations is concerned with the non-canonical subjects resulting from the lexical entailments of doi 10.1075/cal.16.07ser © 2015 John Benjamins Publishing Company

176 Ilja A. Seržant

their verbs, whereas other types of non-canonical subjects have received much less attention. In this paper, I argue that diferent kinds of constructions with subjectlike obliques or non-canonical subjects tend to exhibit the same semantic core. he present paper adheres to the framework of functionalist cognitive linguistics as instantiated in, inter alia, Langacker (2008). his model provides the concepts of trajector (TR) and landmark (LM) that will be used in the paper instead of the traditional notions of subject and object. TR status implies that the given argument is endowed with the primary focal prominence as compared to the other arguments (Langacker 2008: 72, 366f). he trajectorhood is thus deined relatively and within the semantic-pragmatic domain (focus of attention) as opposed to subjecthood that is primarily a syntactic notion. Within the semantic-pragmatic domain, the selection of the trajector participant is rather a matter of construal and less, or secondarily, of conceptual content. As Langacker notes (1997: 66), the trajector is, thus, not tied to a semantic role in a straightforward way, though acknowledging certain preferred but defeasible correlations. Both subject and trajector oten coincide in one and the same NP as trajectors prototypically are encoded with subjecthood (Langacker 2008: 365). he advantage in rather employing semantic-pragmatic dimension is motivated by the fact that it much more precisely mirrors the synchronic conceptualization of an event, whereas formal properties may rather be conservative and represent traces of a historically previous conceptualization and/or construal that have already been overridden by a new conceptualization/construal (e.g., manifested in the change from subject to object). Additionally, certain types of predicates are predisposed to difuse subjecthood (cf. Holvoet 2009), i.e. the symmetric predicates in Langacker’s terms (2008: 369) and, as a consequence, apt to some uncertainty with regard to trajectorhood. hus the trajector is subject to prototype efects as regards its syntactic and morphosyntactic encoding: it is the prototypical trajector when it is encoded by the canonical subject (i.e. by means of both behavior and coding properties as deined in Keenan 1976), and it is less of a prototypical trajector when it is only encoded by some of the behavioral and not by the respective coding properties. his is the type of trajectorhood the present paper is devoted to. A construction with a trajector argument endowed with only some subject properties will be referred to here as a non-prototypical trajector construction, abbreviated NTC, while the endowment with morphologically driven properties (e.g. the nominative case and verbal agreement) alone will not be taken as indicative of trajectorhood. I adopt the assumption adhered to by Construction Grammar that constructions are grammatical units which themselves have meanings (Goldberg 1995; Crot 2001). hus the constructions under investigation encode non-prototypical transitivity that is typically substantiated in the lack of full control on the part of the

Categorization and semantics of subject-like obliques

subject-like participant. his meaning is, however, not marked on the predicate but rather by assigning an oblique case to its highest ranked argument. Constructions of this type are less frequent in accusative languages (Malchukov 2006) but nevertheless can be encountered, e.g., in a number of languages around the Eastern part of the Baltic Sea (East of Circum-Baltic area as deined in Koptjevskaja-Tamm & Wälchli 2001), cf. Seržant (2012, forthcoming). Notably, there is great cross- and intralinguistic variation as to how languages may syntactically treat this kind of oblique case-marked arguments (Langacker 2008: 359) stretching from Icelandic with all syntactic properties to English (e.g. in it occurs to me that…) with no syntactic subject properties at all. Apart from these two extremes, various intermediate stages seem to be very frequent cross- and intralinguistically. he aim of this paper is to provide a categorization of construction types (Section 2) and to approach the invariant of their meanings so that the parallelism in their encoding can be explained. In Section 3, I will try to establish the semantics of diferent kinds of NTCs as well as an invariant semantic core and account for the deviations from this core by postulating a radial category (cf., inter alia, Lakof 1987; Janda 1993; Luraghi 2009).

2. Categorization of the NTC he irst approach in categorizing diferent kinds of NTCs is Haspelmath (2001: 56) who suggests the following three types thereof: “(i) reference-related conditions, (ii) clause-related conditions, and (iii) predicate-related conditions”. In what follows, I will suggest a somewhat modiied categorization. As noted above, the NTCs encode situations or events that considerably deviate from the transitive prototype (cf., inter alia, Lazard 1998) primarily due to the lack of proto-agent properties (as in Dowty 1991) on the part of the subjectlike argument. he lack of proto-agent entailments can be anchored diferently in the event structure. First, it might be due to entailments that the lexical verb itself imposes on its most salient argument (e.g. an experiencer verb), i.e. due to the predicate-related conditions in Haspelmath (2001: 56). Secondly, diferent kinds of grammatical categories such as tense, aspect, mood, negation etc. may override the lexical entailments of the underlying verb (the clause-related conditions in Haspelmath 2001: 56). herefore, John killed Bill and John did not kill Bill considerably difer in terms of their proto-agent and proto-patient properties. 1 1. E.g., there is no endowment with causing an event or change of state in another participant (as per Dowty 1991) on the part of the subject argument with the latter; equally, the object participant is not afected by the event.

177

178 Ilja A. Seržant

hese diferences are induced by the negation operator and do not inhere in the lexical verb kill in any straightforward way. Certain languages adjust the morphosyntactic encoding of the irst participant in such cases accordingly. Finally, the clause-external context may also detach proto-agent properties from a participant. Correspondingly, I will introduce the following categorization of constructions with non-prototypically encoded trajectors: 1. lexeme-driven non-prototypical trajector constructions (Subsection  2.1), i.e. those constructions in which the non-canonicity of the irst argument is driven by the lexical entailments; 2. gram-driven non-prototypical trajector constructions (Subsection 2.2), i.e. those constructions in which the non-canonicity of the irst argument is driven by the corresponding grammatical category in that the verb occurs; 3. syntax-driven non-prototypical trajector constructions (Subsection  2.3), i.e. those constructions in which the non-canonicity of the irst argument is driven by the clause linking mechanisms. Categorizations of this kind – even if presupposed in several previous investigations – have not been explicitly formulated except for by Haspelmath (2001). Needless to say, this categorization rather aims at pointing to the relevant prototypes, and there are obviously predicates where several aspects (i.e. the syntactic, grammatical and lexical ields of the “grammar-lexicon” continuum) interact. I will, however, try to avoid the complex types in this paper in order to remain within a certain level of simplicity since, irst of all, I am trying to establish the very general semantic motivations which in later research may be improved by taking more complex situations into account. 2.1

Lexeme-driven NTCs

A lexeme-driven NTC consists of a lexical predicate that shows up with nonprototypical valence for its most salient argument throughout its paradigm. he non-prototypical morphosyntactic interface is the consequence of the lexical entailments that this lexical predicate imposes on its argument. his type of NTC corresponds to the “predicate-related condition” for the non-canonical case assignment in Haspelmath (2001). A cross-linguistically frequent, lexeme-driven NTC is typically an experience denoting predicate, cf. (1), though not only, cf. (2) and (3):

Categorization and semantics of subject-like obliques

(1) Icelandic Okku fellur þessi bók we.dat like.3sg this.nom book.nom ‘We like that book.’ (2) Lithuanian Kalbininkui trūksta idėjų linguist:dat lack:prs.3 idea:gen.pl ‘he linguist is short of ideas.’ (3) Latvian Man piestāv šis uzvalks I.dat it:3sg this:nom suit:nom ‘his suit its me well.’ (4) Russian Rebenka vyrvalo kašej child:acc vomit:pst.3sg.n porridge:instr ‘he child vomited the porridge.’

hese lexical verbs subcategorize for non-prototypically encoded trajectors only, at least, in relevant meaning; other uses of these verbs are instances of homonymy and never of synonymy. As I noted above, trajectorhood is deined in terms of the relative focal prominence which is primarily subject to construal. Since the Examples (1)–(4) all exhibit unmarked word order, I take the initial position of the oblique argument in the unmarked word order as indicative of its primary focal prominence and, hence, trajectorhood, relative to the second argument in these languages. Alongside diferent kinds of experiencer predicates, there are also beneiciary/ maleiciary predicates (such as lacking and succeeding in Example 2) or the mihiest possessive construction (cf. Clancy 2010 for Slavic). he non-prototypical or non-canonical case marking of the trajector argument is motivated by its lacking of the most of Proto-Agent properties, such as volitionality (“volitional involvement in the event or state”), causation (“causing an event or change of state in another participant”) or movement (“relative to the position of another participant”) in terms of Dowty (1991). 2.2

Gram-driven NTCs

Gram-driven NTCs (G-NTC) refer here to those predicates where the nonprototypical case assignment to the trajector argument is not triggered by the lexical semantics of the given verb (as in 2.1 above) and not clause-externally

179

180 Ilja A. Seržant

(as in 2.3 below) but rather clause-internally by the grammatical category within which it occurs (the clause-related condition in Haspelmath 2001: 56). hat is to say, the case frame and, subsequently, the lexical entailments of the verb are overridden or inverted by the entailments of the grammatical category within which the verb occurs. With lexical NTCs, case is assigned by the verb on the basis of the argument’s thematic role, while, with gram-driven NTCs, the verb’s case frame is overridden by the imposed case frame of the given gram (e.g. TAM, negation, quantiication etc.). he grammatical categories that can have such diathetic efects on a lexical verb’s case frame are typically the following: evidentiality (2.2.1), (deontic) necessity modality (2.2.2), P-lability anticausative with no morphosyntactic promotion of the Patient argument, perfect/resultative predicates (2.2.3) or negated genitive subject predicates (2.2.4).2 2.2.1 Evidentiality In Lithuanian, the evidential mood can be formed by the non-agreeing form of the (formally) passive participle with the genitive case-marked logical subject and nominative (and dialectally accusative) case-marked logical object, cf. indicative past in (5a) and the corresponding evidential construction in (5b): (5) a.

Lithuanian Senieji miškus mylėjo old:nom.pl forest:acc.pl love:pst.3 ‘he elders loved the forests.’ b. Lithuanian Senų miškai mylė-t-a old:gen.pl forest:nom.pl love-ppp-sg.n ‘he elders [apparently] have loved the forests’ (adopted from Jablonskis 1922: 141)

he evidential construction has, in the past, been regarded by some scholars as passive (cf., inter alia, Ambrazas et al. 1997).3 However, apart from the participle’s morphology, there is no other reason to regard it as passive: there is no syntactic, and dialectally, no morphological object promotion,4 nor is there subject 2. In this paper, I will not discuss voice, e.g. passive or middle, which also has this function but usually dooes not result in a non-canonical case assignment. 3. Unfortunately, this view has been reiterated in some typological works. 4. Even from a purely morphological perspective, unequivocal object promotion is found in Lithuanian only if both the nominative case-marking and the verbal agreement co-occur. he nominative case-marking alone does not suice since, in this language, (non-agreeing) nominative objects and nominative adverbs exist.

Categorization and semantics of subject-like obliques

demotion in terms of discourse prominence or word order. he construction patterns rather with an active construction in any other TAM forms of the given verb. Furthermore, it is usually formed out of intransitive verbs (including such unaccusatives as ‘to be’) which are highly unlikely to occur in passives (Blevins 2003: 495–9; Holvoet 2007: 90f; Seržant 2012). he predicate in the evidential construction encodes not only two core participants (the elders and the forests as in (5b) above), but also an event-external and syntactically implicit participant, namely, the “inferer” (which need not be co-referential with the conceptualizer/speaker). he inferer is inherently present in the overall semantic structure and cannot be ruled out. hus there are two acting or controlling participants in the overall semantic structure of the event encoded by the evidential construction: the event-internal participant, the elders, and the event-external participant, the inferer. he inferer consciously makes the inference, (s)he is inherently endowed with sentence and perception. 2.2.2 Necessity modals he Latvian debitive mood is formed by the verbal preix jā-, dative case-marking of the logical subject and nominative case-marking of the logical object (colloquially also accusative). he meaning is one of necessity, in most cases deontic necessity, cf. the indicative in (6a) and debitive in (6b): (6) a.

Latvian Es lasu šo grāmatu I:nom read:prs.1sg this:acc.sg.f book:acc.sg.f ‘I read/am reading this book.’ b. Latvian Man (ir) jā-lasa šī grāmata I:dat copula deb-read.invar this:nom.sg.f book:nom.sg.f ‘I have to read this book.’

I assume that the reasoning here must be parallel to the evidential construction of Lithuanian. Necessity denotes an obligation subevent that is typically not under the (full) control of the obligee. In this way, debitive mood also invokes a second acting participant that is not a core participant of the event ‘to read this book’. It is the “obliger” participant that imposes the obligation on the agent of the verb lasīt ‘to read’ detaching and attracting, thereby, Proto-Agent properties from the agent of the underlying event ‘read’. his third participant can be animate (e.g., a professor that obliged his student to read this book) or inanimate circumstances that are not under control of the obligee. he obliger and the very obligation subevent (as with the inferer above) are also inherently implicit.

181

182 Ilja A. Seržant

Other predicates expressing deontic modality with a non-prototypically marked logical subject can be adduced from Russian (7) (deontic possibility): (7) Russian Mne ne uvidet’ Pariža I:dat not see:inf Paris:gen ‘I will not be able to see Paris’ [e.g., because I don’t have enough money to travel]

In the Russian example, the agent of the subevent logically embedded under the modal operator is demoted because most of its controlling properties were overtaken by the main participant (e.g. the circumstances that prohibit travel to Paris) of the implicit “matrix” event. 2.2.3 Resultative/perfect he North Russian perfect has a fully grammaticalized, valence-bound nonprototypical subject that is marked with the PP u ‘at’ + genitive.5 It fulils most of the subject tests such as equi-NP deletion, control of relexivization and topicality / irst position in the unmarked word order (Timberlake 1975), but it cannot trigger verbal agreement, cf. (8): (8) North Russian U menja ruka porane-n-o at me:gen hand:nom.sg.fem injure-ppp-nom.n.sg=invar ‘I have injured my hand.’

A resultative or perfect predicate as in (8) proiles only the ater-state/-situation subsequent to the event referred to by the lexical verb (cf. Crot 1998: 56). However, a perfect or resultative predicate is semantically more complex than a simple stative as in, for example, ‘He is clever’. he perfect/resultative predicate proiles an (ater-) subevent of the base event, presupposing preceding action. Both the preceding action event and the resultative/perfect ater-event share the main participant. Yet, even though the agent-like main participant of the preceding action is coreferential with the main participant of the resultant ater-event, it has quite different proto-agent entailments in Dowty’s (1991) sense at both stages. While the agent-like participant of the preceding action is a prototypical agent that carries out an action, the main participant of the resultant ater-event is not so. he latter is experiencing a resultant state that (s)he cannot control (anymore). In the same way, the patient-like participant cannot afect it either. In other words, the

5. here is some subdialectal variation (Roduner and Privitelli 2006: 417; Seržant 2012).

Categorization and semantics of subject-like obliques

agent-like participant had the opportunity to control the event at the stage of the preceding action but no longer has it at the resultant stage (Seržant 2012). he only diference between this perfect construction, on the one hand, and the evidential or debitive construction, on the other hand, is that the event-external and the event-internal main participants are co-referential. hus, similarly to the evidential construction in Lithuanian and debitive construction in Latvian, the North Russian perfect implies the existence of a third participant that is endowed with control properties. By comparing the semantics of these diferent kinds of gram-driven NTCs, one already gets an idea of what can be semantically common to all these – at irst glance – diferent constructions. 2.2.4 Genitive-under-negation, a special case of G-NTC in Russian Russian has a so-called genitive-under-negation rule entailing that the objects of transitive predicates and, crucially, the subjects of some non-agentive (unaccusative) intransitive predicates may turn from accusative/nominative into genitive if the predicate is negated (G-NTC) and some additional conditions apply. It has been claimed that reference-related conditions (cf. Haspelmath 2001: 56), such as the referent’s being in the scope of the negation, select genitive, cf., inter alia, Babby (2001). However, this explanation does not account for all cases. It is especially at odds with those cases, in which the genitive case-marked subjects are deinite and wide-scope (e.g. with personal pronouns or proper names). Instead, the rule that accounts for all instances has been put forward in Padučeva (1997, 2005). She claims that, under negation, the choice between nominative (i.e. the canonical trajector) and genitive (i.e. G-NTC) is governed by whether or not the absence (negated presence) of the subject NP has been experienced by an implicit inferer. Contrast the nominative case-marking in (9b) and the genitive case-marking in (9a): (9) a.

Russian Otca ne bylo na more father:gen not be:pst.3.n on sea ‘Father was not on the seashore’ [while I was there and Ihaven’t seen him] b. Russian Otec ne byl na more father:nom not be:pst.3.masc on sea ‘Father was not on the seashore’ [he stayed home]

While (9b) asserts that the father has not been at a certain place/location, the utterance in (9a) implies that the father has not been at a certain place/location in the perceptual world of the inferer (Padučeva 2005: 103). In this example, the inferer

183

184 Ilja A. Seržant

enters the event structure as an implicit additional participant having certain control properties over the overall event alongside the subject participant, as has been noted by Padučeva (2005). he utterance in (9a) not only implies the absence of the subject referent at the given location (‘seashore’), but also the presence of the inferer and his mental activity. If the inferer had not been at the location, (s)he would not have been able to make the inference, part of which is the subject referent (‘the father’). In other words, the very absence of the father at the location can be interpreted as logically dependent on the implied inference event. hus Padučeva (2005) states that the existence at the location is denied in the inferer’s mind when the subject is case-marked genitive. I also conclude that the genitiveunder-negation constructions such as in (9a) entail the existence of a main event, which the event they encode is logically dependent on. he controlling properties of the implicit main event and its main participant may even increase from an inferer to an agent, cf.: (10) Russian Na šedevrax naročno ne bylo on masterworks on.purpose:adv not be:pst.sg.n imen ix sozdatelej, name:gen.pl their authors no vse znali, čto sredi eksponatov but all knew, that among showpieces Džakometti, Xerst i Kuns Giacometti Hirst and Koons ‘[It was] on purpose [that] there were no author names on the masterworks; but everyone knew that there were Giacometti, Hirst and Koons among the showpieces.’6

In contrast to Polish (cf. Dziwirek 1994: 173–4; Blaszczak 2008: 125–34), Russian marginally allows negated, genitive-subject sentences with agentive adverbials such as naročno ‘on purpose’ indicative of the presence of a controlling agent, as in (10). Crucially, the agentivity properties of the adverb cannot be attributed to the genitive subject referent since the latter inherently lacks them in this construction. he grammaticality of the sentence in (10) highlights an event structure that, in addition to the absence event encoded overtly, imply presence of an another, implicit event with an implicit participant. It is this implicit participant that can be attributed the agentivity properties of the adverb. hat is to say, in (10), it was an implicit agent that arranged the exhibition in such a way that there were no author names on the showpieces. his implicit agent is endowed with the following 6. http://www.izvestia.ru/chronicle/article3120761/

Categorization and semantics of subject-like obliques

proto-agent entailments (Dowty 1991: 572): (a) volitional involvement in the event or state, (b) sentience, (c) it causes a situation not to take place, and (d) it exists independently of the event named by the verb. To summarize, the third implicit participant of this type of G-NTC inferes or arranges the situation in such a way that the event encoded by the underlying lexical verb (restricted to inagentives/unaccusatives) does not take place. 2.3

Syntax-driven NTC

With syntax-driven NTCs (S-NTC), the oblique-case assignment to the logical subject is triggered by the syntactic (subordinate) status of its clause. he syntaxdriven NTCs difer from both the lexeme- and gram-driven ones formally, namely, in that they do not constitute syntactically independent (main) clauses themselves. Nevertheless, these constructions also employ the morphological demotion strategy on the part of their most salient participant (trajector). I will discuss only some subtypes of syntax-driven NTCs: AcI (accusativus cum ininitivo) (2.3.1) and some absolutive constructions (2.3.2).

Accusativus cum ininitivo (AcI) 2.3.1 Syntax-driven NTCs are well known, for example, from the raising-to-object- or control-constructions such as in English I see him going in which the subject of the subclause him is encoded with an oblique, typically direct-object case. Below, I discuss a subtype of subclauses traditionally referred to as accusativus cum ininitivo (AcI), cf. (11) with the main clause légetai ‘it is said’ and the AcI …aphikésthai tòn stratón [lit.] ‘the army to have come’: (11) Classical Greek Es mèn dḕ toûton tòn chôron- légetaiapikésthai in prt prt this the place say:pass.prs.3sg arrive:aor.inf tòn stratón the:acc army:acc [lit.] ‘Until this place, it is said, the army came.’ (Hdt. 3.26.8)

he logical subject tòn stratón, ‘the army’, is marked with the accusative case. Apart from the formal subordination of this type of NTCs, there is also a logical dependence in terms of semantics: the encoded event is interpreted as being in the scope of the matrix event. hus the accusative case-marked logical subject (‘army’) in (11) and its event encoded by the ininitival predicate (‘to have arrived’) are construed as being dependent on the main clause event (‘it is said’) with a generic, indeinite main participant. he whole event of arrival is asserted for the world

185

186 Ilja A. Seržant

that is spoken about and that need not be the same as the real world. Diferently from the G-NTCs, with S-NTCs the logically main event is overtly expressed (‘it is said’) while the event encoded by the S-NTC is syntactically dependent on it. he embeddedness of S-NTCs, however, does not always have to be in terms of logic; it may also shit into the domain of pragmatics as in (12) below. 2.3.2 Absolutive/Adverbial clauses here are diferent kinds of absolutive/adverbial constructions that involve diferent cases, e.g., in the ancient Indo-European languages, such as ablativus absolutus in Latin, genitivus absolutus in Classical Greek or locativus absolutus in Sanskrit (cf. Keydana 1997). his type of construction also requires a morphosyntactic “subject-demotion”, cf. the genitivus absolutus constructionsin in Ancient Greek: (12) Ancient Greek Élpeto gàr katà thymòn apostrépsontas hope:3.sg prt at spirit turn.back:ptcp.acc.pl hetaírous ek Trṓōn iénai, friend:acc.pl from Trojans go:inf pálin Héktoros otrúnantos back:adv Hector:gen.sg stir.up:ptcp.gen.sg ‘For in his heart he hoped that friends were coming from the Trojans to turn [him] back, [since] Hector ordered [him] back.’ (Il. 10.356)

he absolutive adverbial clauses – diferent to the dependent constructions like AcI – are logically independent from the situation coded by the respective main clause, cf. the Lithuanian dative abolutive construction: (13) Lithuanian Jam atėjus pas karalių, šis paklausė he:dat come:pst.conv to king this:nom ask:pst.3 ‘When he came to the king, the [king] asked …’

Logically, the embedded event, ‘When he came to the king’, in (13) is not in the scope of the main event. It is presupposed as it does not turn into non-existence if the whole sentence is negated. As Cristofaro (2003: 30) points out, one has to distinguish between two dimensions: (i) logical semantics that has to do with the truth conditions of a sentence and (ii) the pragmatic dimension. “he pragmatic sense … crucially refers to the speakers’ assumption concerning the information status of the sentences they utter” (Cristofaro 2003: 30). hus the proiles of the subordinate events ‘he came to the king’ (in 13) and ‘Hector ordered him back’ (in 12) are overridden by the proiles of the respective main events (in terms of Langacker 1987: Chapter 7). Notably, the utterances in (12) and (13) are about the

Categorization and semantics of subject-like obliques

main events while the respective embedded events are neither topics nor foci; they are not construed as onstage events. he proile of (13) is one of an act of asking, not coming. While the main clauses’s event has an autonomous proile, the embedded event lacks its own proile being a part of the asking event (cf. Langacker 1991: 498–501; Cristofaro 2003: 30). It is intended to provide (in Langacker’s terms) a piece of ofstage information. Furthermore, it is not pragmatically assertive, which can be tested by sentential questions and negations (cf. Cristofaro 2003: 30–3). hus, while in (11) there is a logical asymmetry in terms of scope, in (12) and (13) there is an asymmetry in terms of the construal. To summarize, the event encoded by S-NTCs entails its dependence on the matrix event, either in terms of semantics or pragmatics. In other words, both the semantically non-prototypical trajectors (cf. Example 11) and the pragmatically non-prototypical trajectors (Examples 12 and 13) are coded with an oblique case. Crucially, both these subtypes of S-NTCs entail the existence of another, prototypical, main trajector and its event (the main event).

3. A uniied semantic account of the non-prototypically realized trajector In the previous sections I have presented diferent types of NTCs. Some similarities across the three types of NTCs have already been indicated. In this section, an attempt will be made to provide a systematic account of the denotational facet of NTCs. his account will consist of (i) an invariant semantic core that is inherent to diferent types of NTCs and (ii) those denotational aspects that are subject to variation. I adhere to the constructionist approach to grammar (as substantiated in, inter alia, Goldberg 1995; Crot 2001) and claim that there is an invariant meaning found with various NTCs. I will start with taking a closer look at gramdriven NTCs that are, in a way, in an intermediate position between syntax- and lexeme-driven NTCs. he conceptual semantics of events encoded by the grammatical categories such as evidentiality, necessity modality or perfect is complex. Even though one clause typically refers to only one event, the semantics of gram-driven NTCs implies an additional (grounding) event parallel to the main event referred to by the lexical verb. One can describe this implicit event as the speaker’s inferenceevent in the case of the evidential construction, or as the presence of a participant or some event-external circumstances that create an obligation with the debitive construction, or as the preceding action with the North Russian perfect construction. In order to demonstrate the common semantic core, I will compare these constructions with causatives.

187

188 Ilja A. Seržant

Generally, a causative predication implies two subevents: an antecedent subevent and the consequent subevent (Nedjalkov & Sil’nickij 1969). According to Nedjalkov & Sil’nickij (1969), there are diferent kinds of causatives depending on the semantic type of the antecedent subevent: from the permissive causation (cf. 14), i.e. just permitting the consequent event to come about, to the coercive causation, i.e. forcing the consequent event to come about. In the former case, the controlling properties are shared by the trajector of the antecedent subevent and by the trajector of the consequent subevent, whereas, in the latter case, it is rather only the trajector of the antecedent subevent that is endowed with the control over the whole complex event. hus, generally, a whole range of antecedent and consequent relations in terms of logical causality is available. Consider an example with permissive causation: (14) French Jean m’ a laissé partir John me aux let:ptcp leave:inf ‘John let me leave.’ [lit. ‘John admitted me to leave’]

In this example, me (the trajector of the verb partir) is encoded by oblique case to signal that it has redirected some of its control properties to the causer, i.e. to Jean: both participants are endowed with control over the subevent encoded by the lexical verb ‘to leave’, both participants are sentient, etc. In other words, only the cooperation of both participants has allowed the subevent encoded by ‘to leave’ to occur. he main participant of the matrix event intrudes into the event encoded by the underlying lexical verb and, subsequently, reorders the force-dynamic relations (in terms of the causal approach, cf. Talmy 1976; DeLancey 1985; Langacker 1987; Crot 1993, 1994, 1998a, b). Both subevents merge, thereby, into one complex event. Now, turning back to NTCs, we observe that the only diference between the causative construction as in (14), on the one hand, and the gram-driven NTCs (G-NTCs), on the other hand, is the presence in the former but absence in the latter of an overtly realized main participant of the antecedent event. Otherwise, both constructions are semantically parallel: both denote a consequent subevent and an implicit antecedent subevent. hus, as is the case with the Lithuanian evidential mood, the construction implies an antecedent event (the inference) and a consequent event (the event encoded by the verb). he same is true for the permissive causation in (14) except for the main participant of the antecedent event. he G-NTCs are construed from the perspective of the trajector of the consequent event while the permissive causative in (14) is construed from the perspective of the antecedent-event trajector. Analogically, in the case of the Latvian debitive mood, its construction implies an antecedent event (the obligation) and

Categorization and semantics of subject-like obliques 189

a consequent event (the event encoded by the verb); other G-NTCs may be analyzed analogically. One also inds some sort of cooperation between both trajectors, the implicit trajector of the antecedent event (the inferer, the obliger) having priority over the overt, non-prototypical (consequent-event) trajector (the inferee, the obligee, etc.). his necessarily leads to an oblique case-marking on the inferee or the obligee, that is, the consequent event trajector. Essentially, G-NTCs do not provide speciic or referential information about the antecedent subevent but rather only specify the concept of the antecedent subevent. hus the evidential construction implies that the antecedent subevent must be an act of inference while the necessity modals imply an obligating subevent (the existence of obligating circumstances or persons). However, there is neither exact identiication of nor reference to a particular subevent. he comparison is particularly obvious in those cases where the causee and the obligee are illed with animate NPs. hus Divjak and Janda (2008) argue that the dative arguments with modal predicates in Russian are agentive experiencers as they are agents of the embedded event but experiencers of the modal (antecedent) subevent. Fully parallel to this, the causee of the permissive causative, as in (14), is also a sort of experiencer or beneiciary of the antecedent admission subevent but (less prototypical) agent of the consequent (admitted) subevent. I argue that the internal structure of G-NTCs consists of two subevents: an antecedent and consequent subevent, somewhat reminiscent of causatives. I summarize: G-NTC: he event is construed as an inherently consequent subevent: apart from the onstage information it entails, it provides information about the antecedent subevent with a controlling participant; the antecedent subevent is described only in terms of a concept; the antecedent subevent and its trajector remain inherently implicit and covert. Semantically, the consequent subevent has inherently narrow scope in respect to the antecedent subevent.

Note that the temporal sequence such as, e.g., preceding action vs. resultant state (as with the North Russian perfect) can also be subsumed under the causal sequence: obviously, there is also a causal relation between the preceding action and the resultant state: the resultant state can only take place if there has been a preceding action. As  I mentioned above, the antecedent subevent with its main, controlling participant is inherently covert and, hence, inherently discursively backgrounded. It “is quite schematic semantically” and “lacks the speciicity and rich detail” (in terms of Langacker 1990: 13). Nevertheless, the meaning of the gram determines the concept of the antecedent subevent which, therefore, is not arbitrary. he antecedent subevent is a grounding element that inheres in the verbal

190 Ilja A. Seržant

morphology and non-prototypical case assignment to the trajector-like argument; it is construed with maximal subjectivity (as deined in Langacker 1985, 1990: 13, 1997, 2008: 259f). his is only somewhat diferent with syntax-driven NTCs that tend to express the same semantic structure as G-NTCs. With S-NTCs, the antecedent subevent and its main participant can and must be exactly speciied. S-NTCs require reference to a particular action and to a particular, main controlling participant, respectively. hus the antecedent subevent is not only conceptually (as with the G-NTC), but also referentially determined. he antecedent subevent is construed onstage with the S-NTC while it is construed ofstage with the G-NTC (in the sense of Langacker 1990, 2008), but both the S-NTC and G-NTC require it. I summarize: S-NTC: he event is construed as an inherently consequent subevent; it entails an antecedent subevent with the main participant; the antecedent subevent and its main participant must be realized by means of a matrix clause.

here are some gram-driven NTCs that almost exclusively imply the speaker (the subject of conceptualization Langacker, passim) as the main participant of the antecedent subevent which may be cancelled only by embedding them into indirect speech. hese are mainly epistemic G-NTCs, such as the evidential construction in Lithuanian. It has to be stressed, however, that the main controlling participant of the implicit antecedent subevent by no means has to be coreferential with the subject of conceptualization, it can equally be a third entity, as, for example, in the Latvian debitive construction. It can have epistemic meaning in which, as I noted, the subject of conceptualization will be the controlling participant anchored in the antecedent subevent, but it can also have dynamic or deontic necessity meanings which do not impose such restrictions on the controlling participant of the antecedent subevent. In other words, the antecedent subevent provides a grounding situation that may but needs not coincide with the vantage point (as deined by Langacker 2008) or with the subject of conceptualization. he diference in construals between the G- and S-NTCs lies in how much information about the antecedent subevent is provided and how this information is integrated into discourse. As I noted, while G-NTCs imply the existence of the logically antecedent subevent and provide its concept only, the S-NTCs presuppose an explicit reference to a particular antecedent subevent. he latter is inherently backgrounded with the G-NTCs and foregrounded with the S-NTCs. Now, bearing in mind that the conceptual prototype behind S- and G-NTCs is their logical consequence or “logical embeddedness”, we may now turn to lexeme-driven NTCs. I assume that generally the same concept is found with L-NTCs, too: the event is generally conceptualized as being not controlled by its main participant. Recall that, with S-NTCs, the antecedent subevent with the

Categorization and semantics of subject-like obliques

controlling participant/trajector is onstage, referential, explicit and discursively prominent and, with G-NTCs, is ofstage, non-referential, implicit and discursively inherently backgrounded. Yet, with the lexeme-driven NTC (L-NTC), it is only the existence of an antecedent subevent and of an event-external controlling participant that is implied. hus there is not even commitment as to what kind of concept the antecedent subevent may belong to. he L-NTCs encode the very fact that there is something or that something has occurred resulting in the event-internal main participant (e.g. the experiencer) not having full control over the situation. hus, as regards Example (1) ‘We like this book’ above, for example, the antecedent subevent may describe some properties of the book that are responsible for the consequent event, e.g.: his book is so interesting (and so we like it). he concept of the antecedent subevent is in no way pre-determined with the L-NTCs and is compatible with any kind of situation that is pragmatically coherent. By implying that something has occurred to X over which X did not have a full control, one automatically implies that there was a Y that has been (co-)responsible for what has occurred to X. It is natural that we sometimes cannot pinpoint Y, and its conceptual content may remain unknown to us and/or discursively irrelevant. he encoding of an event as a logically consequent event allows for the signalling of the existence of such a Y with no requirement to specify it somehow. In some languages (as Icelandic, West- and East Slavic, Baltic, Finnic, Old Scandinavian etc.), this strategy to conceptualize the experiencer events is productive and became the main means of encoding the experiencer events, whereas in other languages, the Transitive Construction is used instead, which difers regarding its conceptualization. Indeed, a whole body of empirical evidence exists for the fact that L-NTCs may conceptually be related to “weak” causative events like permissive causativity. hus both the caused subevent (e.g. in 14) and the experiencer event are sometimes encoded by morphologically invariant or defective predicates. One inds a number of experiencer predicates that consist of an adverb and a light verb, or of a verb in the sustained third person singular form. Analogically, the consequent subevents of the causatives are typically encoded by ininitives or participial forms. Furthermore, many languages encode the causee of a causative construction and the experiencer of some L-NTCs alike. A piece of evidence comes from the Mehweb dialect of the East Caucasian language, Dargwa (Magometov 1982; Ganenkov 2013). In this dialect, the subject/experiencer marking of the verbs ‘to ind’, ‘to forget’, ‘to see’, ‘to hear’, ‘to know’, and ‘to understand’ changed from the original dative common in the East Caucasian languages to encode experiencers, to a locative form (denoting movement into a mass or attachment to a landmark), cf. the following examples (Magometov 1982: 80–81):

191

192 Ilja A. Seržant

(15) Dargwa di-ze нu ǧ-ub-ra. I-in you:sg(abs) see:pf-pst-1 ‘I saw you.’ (16) Dargwa di-ze b-arg-i-ra kung. I-in n-ind:pfv-pst-1 book(abs) ‘I’ve found a book.’

Ganenkov (2013) argues that the experiencer marking here is based on the casemarking used to encode the causee in causative constructions, as in (17) from (Magometov 1982: 108): (17) Dargwa nu-ni b-ic-aq-i-ra ixi-ze urči. I-erg n-sell:pfv-caus-pst-1 3sg-in horse(abs) ‘I made him sell the horse.’

In this dialect we thus observe an analogical spread of the case-marking from the causee (over the involuntary agents) to the experiencers (Ganenkov et al. 2008; Ganenkov 2013). his morphosyntactic process calls for explanation, and such an explanation may indeed assume that there is a semantic link between the conceptualization of the caused/consequent subevent and the experiencer event, and eventually between the causee and the experiencer. In the same vein as the examples from Dargwa above, there are some languages that signal the link between causatives and L-NTCs not by means of casemarking but by verb morphology as in some Lithuanian or Finnic L-NTCs (cf. Pörn 2008 for an overview on Finnish): (18) Lithuanian Akvilij-ą pyk-in-o dėl toksikozės Akvilija feel.nauseous-caus-pst.3 due toxicosis ‘Akvilija feels nauseous due to toxicosis.’ (19) Veps Mindei säreidii-ta-b I:part cold-caus-prs.3sg ‘I feel cold.’

(adapted from Zaitseva 2001: 81)

(20) Finnish Minua pelotti ja olin aivan paniikissa. I:par get.scared-caus-pst.3sg and be:pst.1sg fully in panic  ‘I got scared and was fully in panic.’7 7. http://www.freewebs.com/one-life/tarinat/yksin.htm

Categorization and semantics of subject-like obliques

Notably, also within the formal approach, it has been suggested that such examples can be treated as causative events without an external argument, the instigator (Pylkkänen 1999), i.e., in our terms, assuming an antecedent subevent but without naming it. I summarize: L-NTC: he event is conceptualized as an inherently consequent event: it implies the existence of an antecedent event but makes no assumptions as to the concept of that antecedent event and, consequently, its trajector.

I repeat here the semantic deinition of the G-NTC and S-NTC for convenience and comparison: G-NTC: he event is construed as an inherently consequent subevent: apart from the onstage information, it entails and provides information about the antecedent subevent with a controlling participant; the antecedent subevent is described only in terms of a concept; the antecedent subevent and its trajector remain inherently implicit and covert. Semantically, the consequent subevent has inherently narrow scope in respect to the antecedent subevent. S-NTC: he event is construed as an inherently consequent subevent; it entails an antecedent subevent with the main participant; the antecedent subevent and its main participant must be realized by means of a matrix clause and be referential.

Since it is the conceptualization of an event as a consequent one (i.e. the implication of an antecedent event) that unites the three NTC types, I will refer to the NTC as the Consequency Construction, i.e. a construction that encodes logical embeddedness/consequence of an event. I adhere, thereby, to the view put forward in the framework of the Construction Grammar that constructions are meaning-bearing units of grammar (as per, Goldberg 1995; Crot 2001). Notably, this assumption of the core meaning does not, of course, exclude diferent constructional subtypes each with diferent nuances existing. he assumed embeddedness/consequence conceptualization that is common to all three NTC types accounts for the similarity in the main argument’s morphological encoding. Additionally, it also accounts for several other common formal properties, strengthening the suggested analysis. hus diferent kinds of the Consequence-Construction predicates are typically, in some respect, defective with regard to their TAM forms, patterning somewhat with the deranked (cf. Stassen 1985; Koptjevskaja-Tamm 1993; Cristofaro 2003; Crot 2001) or non-inite subordinate clauses. hus, apart from S-NTCs that are evidently defective crosslinguistically (cf., inter alia, Crot 2001: 321; Evans 2007), G-NTCs and L-NTCs also show certain degree of defectiveness: the Lithuanian evidential and the North Russian perfect construction do not form the future tense, the Lithuanian evidential construction cannot form the simple past, the Russian independent-ininitive-predicate almost does not inlect for tense (no future, rarely past) and mood. Common to

193

194 Ilja A. Seržant

all three types is a certain degree of non-initeness: all gram-driven NTCs mentioned above do not inlect for person and are not capable of having person agreement. Additionally, most of them are incapable of having any kind of agreement, including most of the lexeme-driven NTCs. An exception is only found in several lexeme-driven NTCs which may have gender and number agreement with the less subject-like nominative (only). hough, even these oten do not inlect for person (consistently exhibiting third person only), e.g. Icelandic lika ‘to like’. he conclusions drawn so far can be summarized as follows: all subtypes of the Consequency Construction discussed here encode the following conceptualization, and many of them have the following morphosyntactic properties: Table 1. Conceptualization and morfosyntactic properties of the Consequency Construction. Conceptualization

the event is conceptualized as a consequent (conceptually embedded) event

Morphosyntactic properties

– defectiveness as regards TAM; – lack of person-number forms or generally – ininite

he three types of NTCs are diferentiated, however, by how much information on the antecedent event they entail and by the discursive prominence thereof. hus the antecedent event has the following properties encoded with each type of NTC: Table 2. Properties of the Antecedent event.8 Properties of the Antecedent event Concept

Syntax-driven NTC Gram-driven NTC a particular concept of the antecedent event is implied

Lexeme-driven NTC no commitment as to a particular concept

Domain

superclausal domain clausal domain

lexeme domain

Explicitness

explicit

implicit

–8

Deiniteness

deinite, referring

generic, non-referring

–

Discourse

fore-grounded

back-grounded

–

As a preliminary result, I suggest that the hypocategories of the Consequency Construction discussed here have an invariant semantic core (cf. Table 1) which implies the existence of another, logically antecedent event anchoring yet another, main participant external to the event encoded by the Consequency Construction. Note that the antecedent event here is taken very broadly to also include the 8. “ – ” means not applicable.

Categorization and semantics of subject-like obliques

concept of stative situations such as, for example, the existence of an external controlling participant that is in some way responsible for what the event encoded by the Consequency Construction denotes. While the entailment of an antecedent event is the common semantic core of all NTCs, deviations from this core are conined to how much information about the antecedent event they provide and the discursive status of this information, see Table 1 above. A variation like this, which exhibits an invariant semantic and formal core while having a number of varying parameters, can be captured by postulating a radial category (Lakof 1987; Janda 1993; Luraghi 2009). Radial categories presuppose a central, prototype category that is responsible for linking the network of its partial deviations. he linking rules represent the semantic proximity between the prototype and its quasi-synonyms as well as mirror the semantic relationship with regard to other quasi-synonyms, thereby providing the internal structure of the overall category (Janda 1993). As Nesset et al. (2011) note, there is no uniqueness requirement, that is to say, a particular radial subcategory of a category may enter radial relationship with several prototypes at the same time. Originally, prototype categories have only been applied to lexical concepts (extending beyond the pure linguistic dimension, cf. Hudson 1980). Subsequently, the prototype category has also been applied to linguistic units at the grammatical edge of the language realm, cf. Bybee and Moder (1983), Taylor (1995). In this paper, I will apply the notion of the prototype and the radial category approach to illustrate the correlation of both semantics and formal expression, i.e., the more distant a category is compared to the prototype, the less overlap it shows in both its formal organization and semantics. Figure 1 represents the semantico-syntactic organization of diferent extensions of the Consequency Construction in terms of a radial category:

Consequency Construction

G-NTC

S-NTC

Transitive Construction

L-NTC

Figure 1. Consequency Construction as a Radial Category. (prototypes are squares; their radial subcategories are round)

195

196 Ilja A. Seržant

As noted in Nesset et al. (2011), subcategories may have several prototypes (as indicated in Figure 1). his is the case with both G-NTCs and L-NTCs. I assume that several particular G-NTCs and L-NTCs are conceptually linked to the transitive-construction prototype as deined in Lazard (1998), as well (this is rendered by the arrow in Figure 1). Within this conceptualization, the less salient participant metaphorically acts on the more salient one, cf. English his made me angry. Here, his encodes an inanimate stimulus and me an animate experiencer. his construal implies that the transmission of the force dynamics runs from the stimulus his to the experiencer me (cf. Talmy 1988; Crot 1998: 55; Crot 1994). Obviously, this use is a metaphorical extension of the transitive construction in English. However, such metaphorical extensions can approach the Consequency Construction with its properties in diachronic perspective. hus cf. (21) from German that patterns in very much the same way as the English example above but, in contrast, exhibits some deviations towards Consequency Construction: (21) German Mich ergrif I:acc grasp:pst.3sg ‘I got scared.’

die Angst the fear:nom

Evidently, (21) also represents a metaphorical extension of the transitive construction. However, the unmarked word order, in contrast to the English example, is reversed. In (21), it is rather the accusative-case-marked argument that is the trajector and the nominative-case-marked argument that is the landmark (in terms of Langacker 1987: 217f. and 2008: 70f.). By virtue of the word order, the construal of Example (21) approaches the consequency construction and diverges from its original transitive construction. here are some other diferences that group L-NTCs and G-NTCs together. he G-NTCs and L-NTCs types of the consequency construction instantiate independent, main clauses, whereas the consequence construction of the S-NTC type always constitutes a syntactically dependent, subordinate clause. he syntactic status of a clause, in turn, imposes certain requirements on its internal morphosyntactic organization. It has oten been noticed that main clauses lacking a nominative subject are generally dispreferred (in accusative languages), cf. Tsunoda’s Unmarked-Case-Constraint which predicts that, in a non-elliptical sentence, at least one NP must be in the unmarked case, i.e. nominative or absolutive, (Tsunoda 1981), “Obligatory NOM Requirement” in Primus (1999), “Default Linking” in Wunderlich & Lakämper (2001), Malchukov (2005: 95). At the same time, there is no such restriction for the ininite subclauses, which typically lack nominative. hus the L-NTCs and G-NTCs also deviate from the S-NTCs in terms of their internal morphosyntactic organization.

Categorization and semantics of subject-like obliques

4. Some diachronic evidence (insubordination) In this Section, I will provide some diachronic evidence for the uniied semantic analysis of all three types of the NTCs suggested above. I have suggested that all three types of NTCs are construed as consequent events, the diference between them being in how the antecedent event is construed and realized. I will show that diachronic developments from the S-NTC to the G-NTC conirm this analysis since the changes, which take place in these developments, are changes in the construal and realization of the antecedent event only while the consequent event remains persistent throughout the development. Consider (11) repeated as (22) for convenience: (22) Classical Greek Es mèn dḕ toûton tòn chôron- légetaiapikésthai in prt prt this the place say:pass.prs.3sg arrive:aor.inf tòn stratón the:acc army:acc [lit.] ‘Until this place, people say, the army came.’ (Hdt. 3.26.8)

he construction in (22) represents a non-grammaticalized evidential meaning induced by the matrix verb légetai ‘it is being said’. However, the embedded S-NTC does have the potential to shit from a S-NTC into a G-NTC, cf. (23): (23) Classical Greek toútous dè éphasan oikeîn anà tà órē … these:acc.pl prt say:aor.3pl live:inf on the mountains kaì basiléōs ouk akoúein and king:gen.sg not hear:inf allà kaì embaleîn pote eis autoùs basilikḗn but and invade:inf once in they king’s:adj.acc.sg stratiàn dṓdeka myriádas. army:acc.sg twelve 10- thousand toútōn d’ oudén’ aponostêsai dià tḕn dyschōrían. these:gen.pl prt none:acc.sg come.back:inf due the bad.place ‘hese, – they said, – dwelt up among the mountains, …, and were not subjects of the King; in fact, a royal army of one hundred and twenty thousand men had once invaded them, and, by reason of the ruggedness of the country, not a man of all that number came back.’ (X. Anab. 3.5.16 17, translated by Brownson 1922)

he last (sub)clause, oudéna aponostêsai [none:acc come.back:inf], with the aorist ininitive is already several clauses distant from the matrix predicate éphasan

197

198 Ilja A. Seržant

‘they said’; it also shows a change in its subject referent [lit.] ‘none of the men’ instead of ‘these’ as in the ininitive clauses before. Note also the punctuation (that is orginal). hese indications suggest a certain degree of emancipation of the AcI construction from its matrix predicate and change into a reportative mood. Even though Greek has not conventionalized and grammaticalized this use as a reportative mood, Example (23) still shows that the potential for a full emancipation of the AcI has existed. his is a less frequently discussed subtype of insubordination, which is a process widely attested cross-linguistically whereby an original subordinate clause is reinterpreted as a main clause retaining and conventionalizing the concept of the former main clause (see Evans 2007 for an extensive typological overview). Such a development is also attested in Classical Latin even to a higher degree. In this language, the AcI can emancipate itself and be used without a matrix predicate that is then only conceptually (as with the G-NTCs) implied. As such, the exclamatory use of sīcine ‘thus (negatively)’ is oten used with the AcI (Lewis & Short, sub verbo), cf. (24) from Plautus Pers. 45: (24) Latin sīcine hoc tē mihī facere thus this:acc(=nom) you:acc I:dat do:inf ‘[Ought you] to be treating me in this fashion?’

In the construction sīcine + AcI, a modal event denoting necessity in very general terms, i.e. only conceptually, is implied. hus the original S-NTC form becomes a G-NTC here, representing another instance of insubordination. he loss of the antecedent clause might be discursively motivated. As Langacker (2008: 418) points out, the real news is oten the consequent event while the antecedent event only introduces and frames it. Crucially, nothing happens to the very consequent event during this change: it remains the same throughout the development. he changes only afect the set of information regarding the antecedent event that must be minimally present at a particular utterance: originally, the antecedent event must have been speciied, but in examples such as (24), only its concept is provided.

5. Conclusions To summarize the main conclusions, I have irst put forward a categorization of the NTCs into lexeme-, gram- and syntax-driven NTCs according to the source of the entailments on the trajector-like argument that lead to the assignment of an oblique case to that argument. Notably, this disivion is an approximation glossing over more speciic subtypes (like the genitive-under-negation construction which

Categorization and semantics of subject-like obliques 199

respresents a special subtype of G-NTC combined with the type referrd to as “reference-related conditions” in Haspelmath 2001: 56). Secondly, I have suggested a radial category, which I labelled Consequency Construction and which has an invariant semantic core with all three types of NTCs, namely, the implication of an (logically) antecedent event anchoring an external participant that (logically) (co-)controls the event carried out by the main participant of the event referred to. he diference between the three diferent NTC subtypes lies in how the antecedent event is construed and what kind of information the given NTC provides and requires about it. While L-NTCs only imply the existence of an antecedent event, making no further commitments, G-NTCs also provide the concept of the antecedent event. Finally, S-NTCs are even more speciic and require an exact reference to the antecedent event and its main participant. hirdly, I have tried to provide some diachronic evidence illustrating the change from syntax- to gram-driven NTCs, a subtype of the process of insubordination described in Evans (2007). he transition from S-NTC to G-NTC represents an instance of subjectiication (cf., inter alia, Langacker 1990, 2006; Traugott & Dasher 2003).

Acknowledgements Many ideas in this paper were inspired by the PhD course in Cognitive Linguistics at the University of Tromsø, April 26–29, 2011, organized by Laura Janda and Tore Nesset. I also thank the editors of the present volume for their comments. All disclaimers apply.

References Aikhenvald, A., Dixon, R. M. W., & Onishi, M. (2001). Non-canonical marking of subjects and objects. Typological Studies in Language 46. Amsterdam: John Benjamins. DOI: 10.1075/tsl.46 Ambrazas, V., Genušienė, E., Girdenis, A., Sližienė, N., & Tekorienė, D. (1997). Lithuanian Grammar. Lietuvių kalbos gramatika, ed. by V. Ambrazas. Vilnius: Baltos lankos. Babby, L. H. (2001). he genitive of negation: a uniied analysis. In Steven Franks, Tracy Holloway King & Michael Yadrof (Eds.), Annual workshop on formal approaches to slavic linguistics: he Bloomington meeting 2000 (FASL 9) (pp. 39–55). Ann Arbor: Michigan Slavic Publications. Barðdal, J. (2008). Productivity: Evidence from case and argument structure in Icelandic. Amsterdam: John Benjamins. DOI: 10.1075/cal.8 Barðdal, J. (2009). he development of case in Germanic. In Jóhanna Barðdal & Shobhana Chelliah (Eds.), he role of semantic, pragmatic and discourse factors in the development of case (pp. 123–159). Amsterdam: John Benjamins. DOI: 10.1075/slcs.108.09bar

200 Ilja A. Seržant

Bhaskararao, K., & Subbarao, V. (2004). Non-nominative subjects (Vol. 1–2). Amsterdam: John Benjamins. Blaszczak, J. (2008). Diferential subject Marking in Polish. In Helen de Hoop & Peter Swart (Eds.), Diferential subject marking. Studies in Natural Language and Linguistic heory 72 (pp. 113–150). Dordrecht: Springer Netherlands. Blevins, J. (2003). Passives and impersonals. Journal of Linguistics, 39, 473–520. DOI: 10.1017/S0022226703002081 Brownson, C. L. (1922). Xenophon. Xenophon in Seven Volumes, 3. Cambridge, MA: Harvard University Press, London: William Heinemann, Ltd. (Quoted from the PDL). Bybee, J. L., & Moder, C. L. (1983). Morphological classes as natural categories. Language, 59, 251–270. DOI: 10.2307/413574 Clancy, S. J. (2010). he chain of being and having in slavic. SLCS 122. Amsterdam: John Benjamins. DOI: 10.1075/slcs.122 Cristofaro, S. (2003). Subordination. Oxford studies in typology and linguistic theory. Oxford: Oxford University Press Crot, W. (1993). he semantics of mental verbs. In J. Pustejovsky (Ed.), Semantics and the Lexicon (pp. 55–72). Dordrecht/Boston/London: Kluwer Academic Publishers. DOI: 10.1007/978-94-011-1972-6_5 Crot, W. (1994). he semantics of subjecthood. In M. Yaguello (Ed.), Subjecthood and Subjectivity. he status of the subject in linguistic theory (pp. 29–76). Proceedings of the Colloquium “he Status of the Subject in Linguistic heory” London, 19–20 March 1993. Paris: OPHRYS. Crot, W. (1998a). Event structure in argument linking. In M. Butt & W. Geuder (Eds.), he projection of arguments. Lexical and compositional factors (pp. 21–64). Stanford: CSLI Publications. Crot, W. (1998b). he structure of events and the structure of language. In Michael Tomasello (Ed.), he new psychology of language: Cognitive and functional approaches to language structure (pp. 67–92). Mahwah, N.J.: Lawrence Erlbaum Associates. Crot, W. (2001). Radical construction grammar. Syntactic theory in typological perspective. Oxford/ New York: Oxford University Press. DOI: 10.1093/acprof:oso/9780198299554.001.0001 DeLancey, S. (1985). Agentivity and syntax. In W. H. Eilfort, P. D. Kroeber & K. L. Peterson (Eds.), Papers from the parasession on causativity and agentivity at the twenty-irst regional meeting. Chichago Linguistic Society (CLS 21), Part 2 (pp. 1–12) Chicago: Chicago Linguistic Society. Divjak, Dagmar, & Janda, Laura A. (2008). Ways of attenuating agency in Russian. Transactions of the Philological Society, 106(2), 138–179. DOI: 10.1111/j.1467-968X.2008.00207.x Dowty, D. R. (1991). hematic proto-roles and argument selection. Language, 67, 547–619. DOI: 10.1353/lan.1991.0021 Dziwirek, K. (1994). Polish subjects. New York/London: Garland Publishing. Evans, N. (2007). Insubordination and its uses. Chapter 11. In Irina Nikolaeva (Ed.), Finiteness. heoretical and empirical foundations (pp. 366–432) Oxford: OUP. Ganenkov, D., Maisak, T., & Merdanova, Solmaz R. (2008). Involuntary agent as non-canonical subject in Agul. In Helen de Hoop & Peter de Swart (Eds.), Diferential subject marking (pp. 173–198). Dordrecht: Springer. Ganenkov, D. (2013). Diachrony of experiencer subject marking: Evidence from east caucasian. In I. A. Seržant & L. Kulikov (Eds.), he diachronic typology of non-canonical subjects (pp. 231–256). SLCS Series. Amsterdam: John Benjamins. DOI: 10.1075/slcs.140.11gan

Categorization and semantics of subject-like obliques 201

Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press. Haspelmath, M. (2001). Non-canonical marking of core arguments in European languages. In A. Aikhenvald, R. M. W. Dixon, & M. Onishi (Eds.), Non-canonical marking of subjects and objects (pp. 53–83). Amsterdam: John Benjamins. DOI: 10.1075/tsl.46.04has Holvoet, A. (2007). Mood and modality in Baltic. Baltica Varsoviensia VI. Kraków. Holvoet, A. (2009). Difuziniai subjektai ir objektai. In A. Holvoet & R. Mikulskas (Eds.), Gramatinių funkcijų prigimtis ir raiška (pp. 37–68). Vilnius: Vilniaus universitetas & Asociacija ‘Academia Salensis’. [Difused subjects and objects] de Hoop, H., & Swart, P. (2008). Diferential subject marking. Studies in Natural Language and Linguistic heory 72. Dordrecht: Springer Netherlands. Hudson, R. (1980). Sociolinguistics. Cambridge: Cambridge University Press. Jablonskis, J. (1922). Lietuvių kalbos gramatika. Etimologija. Kaunas, Vilnius: Švyturio. [Grammar of Lithuanian. Etymology.] Janda, L. (1993). A geography of case semantics. he czech dative and the Russian instrumental. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110867930 Keenan, E. L. (1976). Towards a universal deinition of ‘subject’. In Ch. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press. Keydana, G. (1997). Absolute Konstruktionen in altindogermanischen Sprachen. Göttingen: Vandenhoeck & Ruprecht. Kittilä, S. (2002). Transitivity: Toward a comprehensive typology. Publications in General Linguistics 5. Turku: University of Turku. Koptjevskaja-Tamm, M. (1993). Nominalizations. London: Routledge. Koptjevskaja-Tamm, Maria, & Wälchli, Bernhard (2001). he Circum-Baltic languages: An areal-typological approach. In Östen Dahl & Maria Koptjevskaja-Tamm (Eds.), CircumBaltic languages (Vol. 2, pp. 615–750). Amsterdam: John Benjamins. Lakof, G. (1987). Women, ire and dangerous things. Chicago: Chicago University Press. DOI: 10.7208/chicago/9780226471013.001.0001 Langacker, R. W. (1985). Observations and speculations on subjectivity. In J. Haiman (Ed.), Iconicity in syntax (pp. 109–150). Amsterdam: John Benjamins. DOI: 10.1075/tsl.6.07lan Langacker, Ronald W. (1987). Foundations of cognitive grammar. Vol. 1: heoretical Prerequisites. Stanford: Stanford University Press. Langacker, R. W. (1990). Subjectiication. Cognitive Linguistics, 1, 5–38. DOI: 10.1515/cogl.1990.1.1.5 Langacker, Ronald W. (1991). Foundations of cognitive grammar. Vol. 2: Descriptive Applications. Stanford: Stanford University Press. Langacker, R. W. (1997). Consciousness, construal and subjectivity. In M. I. Stamenov (Ed.), Language structure, discourse and the access to consciousness (pp. 49–77). Amsterdam: John Benjamins. DOI: 10.1075/aicr.12.05lan Langacker, R. W. (2006). Subjectiication, grammaticalization, and conceptual archetypes. In A. Athanasiadou, C. Canakis, & B. Cornille (Eds.), Subjectiication. Various paths to subjectivity (pp. 17–40). Berlin/New York: Mouton de Gruyter. Langacker, R. W. (2008). Cognitive grammar. A basic introduction. New York: Oxford University Press. DOI: 10.1093/acprof:oso/9780195331967.001.0001 Lazard, G. (1998). Transitivity revisited as an example of a more strict approach in typological research. Folia Linguistica, XXXVI(3–4), 141–190.

202 Ilja A. Seržant

Lewis & Short. A Latin dictionary: Founded on Andrews’ edition of Freund’s Latin dictionary / revised, enlarged and in great part rewritten by Charlton T. Lewis and Charles Short. First published in 1879. [Quoted from PDL] Luraghi, S. (2009). Case in cognitive grammar. Chapter 9. In A. Malchukov & A. Spencer (Eds.), he oxford handbook of case (pp. 135–150). Oxford/New York: Oxford University Press. Magometov, A. A. (1982). Megebskij dialect darginskogo jazyka (issledovanie i teksty). Tbilisi: Mecniereba. Malchukov, A. (2005). Case pattern splits, verb types and construction competition. In M. Amberber & H. de Hoop (Eds.), Competition and variation in natural languages. he case for case (pp. 73–118). Amsterdam/Boston: Elsevier. DOI: 10.1016/B978-008044651-6/50006-9 Malchukov, A. (2006). Transitivity parameters and transitivity alternations. In L. Kulikov, A. Malchukov & P. de Swart (Eds.), Case, valency and transitivity (pp. 329–358). Amsterdam: John Benjamins. DOI: 10.1075/slcs.77.21mal Nedjalkov, V. P., & Sil’nickij, G. G. (1969). Tipologija kauzativnych konstrukcij. In A. Xolodovič (Ed.), Tipologija kauzativnych konstrukcij. Morfonologičeskij kauzativ (pp. 5–20). Akademija Nauk SSSR, Institut Jazykoznanija. Leningrad: Izdatel’stvo Nauka. Nesset, T., Endresen, A., & Janda, L. (2011). Two ways to get out: Radial category proiling and the Russian preixes vy- and iz-. Zeitschrit für Slawistik, 56(4), 377–402. DOI: 10.1524/slaw.2011.0039 Padučeva, Je. V. (1997). Roditel’nyj subjekta v otricatel’nom predloženii: Sintaksis ili semantika? Voprosy Jazykoznanija, 2, 101–116. [he genitive case of subject in a negated sentence: syntax or semantics?] Padučeva, Je. V. (2005). Eščo raz o genitive subjekta pri otricanii. Voprosy Jazykoznanija, 2005(5), 84–99. [Once again about the genitive under negation] PDL = Perseus Digital Library. Gregory R. Crane, Editor-in-Chief, Tuts University. http://www. perseus.tuts.edu/ Pörn, M. (2008). Psychophysical and physical causative emotion verbs in Finnish: he temporal Structure of causative emotion verb + ininitive 1 – Constructions within conceptual semantics. SKY Journal of Linguistics, 21, 201–218. Primus, B. (1999). Cases and thematic roles. Tübingen: Niemeyer. DOI: 10.1515/9783110912463 Pylkkänen, L. (1999). Causation and external arguments. In L. Pylkkänen, A. van Hout, & H. Harley (Eds.), Papers from the second Penn/MIT roundtable on the lexicon. MIT Working Papers in Linguistics 35 (pp. 161–183). Cambridge, MA: MIT. RNC = Russian National Corpus at http://www.ruscorpora.ru/search-main.html Roduner, M., & Privitelli, T. (2006). Der Genitiv des Agens / Experiencers in Litauischen und Russischen Dialekten. Acta Balto-Slavica, 30, 403–425. Seržant, I. A. (forthcoming). Dative experiencer constructions as a circum-baltic isogloss. In P. Arkadiev, A. Holvoet, & B. Wiemer (Eds.), Contemporary approaches to Baltic linguistics. Berlin: De Gruyter. Seržant, I. A. (2012). he so-called possessive perfect in North Russian and the Circum-Baltic area. A diachronic and areal approach. Lingua, 122, 356–385. DOI: 10.1016/j.lingua.2011.12.003 Seržant, I. A., & Kulikov, L. (2013). Diachronic typology of non-canonical subjects, ed. by I. A. Seržant & L. Kulikov. Amsterdam: John Benjamins. DOI: 10.1075/slcs.140 Stassen, L. (1985). Comparison and universal grammar. Oxford: Basil Blackwell.

Categorization and semantics of subject-like obliques 203

Talmy, L. (1976). Semantic causative types. In Masayoshi Shibatani (Ed.), he grammar of causative constructions. Syntax and Semantics (Vol. 6, pp. 43–116). New York: Academic Press. Talmy, L. (1988). Force dynamics in language and cognition. Cognitive Science, 12, 49–100. DOI: 10.1207/s15516709cog1201_2 Taylor, J. (1995). Grammatical categories, ch. 10. Linguistic categorization: Prototypes in linguistic theory (2nd ed.). Oxford: Oxford University Press, 293–308. Timberlake, A. (1975). Subject properties in the North Russian passive. In Li & hompson (Eds.), Subject and topic (pp. 545–594). New York: Academic Press. Traugott, E. C., & Dasher, R. B. (2002). Regularity in semantic change. Cambridge: Cambridge University Press. Tsunoda, T. (1981). Split case-marking in verb types and tense/aspect/mood. Linguistics, 19, 389–438. Wunderlich, D., & Lakämper, R. (2001). On the interation of structural and semantic case. Lingua 111 (4–7): 377-417. (Special Issue on Efects of morphological case, ed. by H. de Hoop, O. Koeneman, J. Mulders & F. Weerman.) Zaitseva, Maria (2001). Vepsän kielen lauseoppia. Suomalais-ugrilaisen Seuran toimituksia 241. Helsinki: Suomalais-ugrilainen Seura.

he world is raining Meteorological predicates and their subjects in a typological perspective Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen University of Oslo / University of Helsinki / University of Eastern Finland

he peculiar semantic properties of meteorological predicates have linguistic consequences. Some of these consequences concern their subjects, which do not behave like prototypical subjects do. Meteorological predicates oten lack a subject, and in the languages where they do have subjects they are non-referential, non-topical, indeinite, inanimate or non-agentive entities. hey are typically either expletive elements or lexical elements referring to the spatial, temporal or atmospheric background, i.e. elements which would otherwise have been encoded as adverbial phrases. Although the nominative – or some other kind of zero marking – and default agreement marking (3rd person, singular number, neuter gender) are typical of these subjects, the study argues that they display special behavior which distinguishes them from canonical cases.

1.

Introduction

Although it is impossible to ind a universally valid deinition of the grammatical notion of subject that would apply to all languages, there are certain properties which can be regarded as prototypical, canonical subject features. Subjects have been, for example, characterized with regard to their semantic and discourse properties, and in prototypical cases, they refer to animate and agentive beings that exist autonomously and independently of the event. Moreover, subjects tend to be deinite and topical (cf. Dowty 1991; Keenan 1976; Langacker [1991] 2006: 305–317). In this paper, a semantically homogeneous group of predicates is identiied whose subjects do not behave like prototypical subjects. Instead they display some peculiar properties both within and across languages. he subject type examined consists of meteorological predicates, i.e. predicates which refer to diverse weather doi 10.1075/cal.16.08eri © 2015 John Benjamins Publishing Company

206 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

phenomena such as ‘it is snowing’, ‘it is thundering’ and ‘the wind is blowing’. Linguistically, they constitute an interesting predicate type: the situations they describe inherently lack proper participants with distinct and salient semantic roles such as agent, patient or experiencer. his makes them diferent from events expressed in intransitive and transitive clauses that can be distinguished for example on the basis of the nature of their participants. he unorthodox nature of meteorological events is, as can be expected, relected both in the semantic and in the formal characteristics of the subjects of weather predicates. his paper examines the properties of NPs functioning as subjects of meteorological predicates. We will demonstrate that they display several special characteristics which can be explained by the peculiar nature of the meteorological events. For example, what is the nature of the elements that surface as subject if proper participants, which typically surface as subjects and objects, are absent? What is the formal nature of these elements that have been ‘forced’ to surface as subjects? he goal of the paper is, in line with other papers in this volume, to shed more light on the notion of the subject by examining constructions whose subjects can reasonably be considered atypical. he organization of the paper is as follows. In Section 2, we will shortly present a typology of meteorological expressions which is relevant to the discussion in the subsequent sections. Section 3 constitutes the bulk of the paper and discusses the peculiar features of the subjects of meteorological predicates. In Section 3.1, they are examined from a semantic perspective and in Section 3.2 from a formal perspective. Finally in 3.3 their behavioral features will be touched on. Section 4 summarizes the central indings of the paper.

2. Cross-linguistic typology of meteorological expressions Eriksen, Kittilä & Kolehmainen (2010) propose a cross-linguistic typology of meteorological expressions (henceforth MEs) based on the element primarily responsible for coding the weather phenomenon in question. MEs are thus divided into three major types. In the predicate type, a predicate expresses the meteorological event, while the argument (if one is present) does not refer to the weather phenomenon itself, but has other functions (1). he argument type has the opposite characteristics: a nominal argument is primarily responsible for expressing weather, while the predicate is semantically more or less superluous (2). Finally, in the argument-predicate type the predicate and the argument are both involved in expressing the weather phenomenon (3). All types include a number of subtypes, depending on the syntactic valency and the parts of speech of the elements involved. Examples (1)–(3) illustrate the major types:

he world is raining 207

Predicate type Palestinian Arabic (1) ɨd-dunya  tɨ-shti. art-world fem.3sg-rain.impf ‘It is raining.’ Argument type Korean (2) Pi-ka o-nta. rain-nom come-plain.ind ‘It is raining.’ Argument-predicate type Turkish (3) Yağmur yağıyor. rain rain.3sg ‘It is raining.’

(Givón 2001: 119)

(Jae Jung Song, p.c.)

(Zeynep Orhan, p.c.)

In (1), the word meaning ‘world’ occurs as the subject of the clause. he predicate, in turn, is the element directly responsible for denoting the weather phenomenon in question. In (2), the subject argument refers to the weather phenomenon described, while the verb is simply a supporting element needed for creating a grammatically felicitous construction. In (3), the expression of weather is divided between predicate and argument, i.e. both refer simultaneously to the weather phenomenon in question. Each of the types examined consists of subtypes. Most of these are irrelevant to the current discussion, but one set, the predicate type, can be divided into subtypes according to how the subject is encoded, which pertains directly to the topic at hand. he following three subtypes should be mentioned in this context (the types will be further elaborated in Section 3.2): he atransitive predicate type (no explicit subject) Italian (4) Piove. rain.prs.3sg ‘It is raining.’

(personal knowledge)

he expletive predicate type (a formal/dummy element as subject) Norwegian (personal knowledge) (5) Det regn-er. it rain-prs ‘It is raining.’

208 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

he intransitive predicate type (a lexical noun as subject) Mwotlap (6) Mahē no-momyiy /ne-sew. place sta-cold /sta-warm ‘It is cold/warm.’

(François 2001: 342)

he previous examples illustrate variation between languages. It is, however, worth noting that the construction types do not only vary cross-linguistically, but that a single language may exploit several diferent construction types to express weather phenomena (for more details see Eriksen, Kittilä & Kolehmainen 2010). For example, in Russian, the argument type is used to express precipitation and the predicate type to express temperature: Russian (7) a. Idet dožd’. go.3sg.prs rain ‘It’s raining.’ b. Kholodno. cold.sg.n.pred ‘It’s cold.’

(personal knowledge)

he three major types of ME difer clearly according to the element surfacing as the subject. In the argument type (2) and the argument-predicate type (3) the weather phenomenon itself surfaces as a subject. In the predicate type (1, 5, 6), on the other hand, languages resort to other elements for subjecthood. he predicate type is thus the most fascinating type with respect to the topic of the present paper. Consequently, the focus will be on subjects of the predicate type.

3. Subjects of meteorological predicates Subjects are oten treated as difuse concepts consisting of a set of multiple properties rather than as discrete concepts (see e.g. Dowty 1991; Langacker [1991] 2006: 305–309). A study worth mentioning in this context is that by Keenan (1976) who laid the foundation for a prototypical multi-factor-treatment of subjects. Many of the typical subject characteristics proposed by him reoccur in subsequent studies, even though the details of patterning of semantic features and of encoding properties naturally vary between languages. Keenan divided subject features into three classes; semantic properties, coding properties, and behavioral and control properties (each class consisting of several features). In what follows, we will discuss the peculiar nature of subjects of meteorological predicates as found in the

he world is raining 209

predicate type outlined in Section 2 (such predicates will henceforth be referred to as MPs), in terms of a subset of these features as they pertain to MP subjects. Semantics will be discussed irst, since it constitutes the best starting point for explaining some of the unorthodox formal features displayed by MP subjects. One methodological point should be mentioned before proceeding. As the following section will deal with meteorological semantics, it necessarily follows that it will not deal with the predicate type exclusively, as the predicate type is only one out of three ways to formally encode the same set of semantic state-of-afairs. he argument type and the argument-predicate type (see Section 2) relect the same semantics and the same set of events. Consequently the following discussion will deal with MEs in general, and examples of the two latter ME types will also be used. his general discussion will nevertheless be crucial to our understanding of how MPs behave and why MP subjects are encoded the way they are. 3.1

Semantic properties

Due to the lack of distinct, salient and clearly deinable participants, weather phenomena are clearly diferent from many other events, such as ‘run’, ‘paint’ and ‘break’. For example, in the event ‘it is raining’, it is diicult to state whether the water falling from the sky is an agent that could be seen as the instigator of the denoted event, or rather a patient that is somehow afected by it. We only have the outcome of the event. We may add that weather, along with other natural phenomena, never involves sentient participants capable of inducing control over the event, which also makes the distinction between an instigating and an afected participant diicult, if not impossible, to draw. Events such as ‘the girl is running’, ‘the blacksmith painted the house’ and ‘the glass broke into hundreds of pieces’, on the other hand, can be distinguished from each other based on the semantic roles of the participants involved. Moreover, the participants in these events are salient and it is easy to make a distinction between the initiating and the afected participant. In other words, it is easy to state who did what to whom. he lack of salient participants with easily deinable semantic roles naturally has consequences for the nature of the subjects of MEs. he most straightforward consequence is, expectedly, that ME subjects cannot be accorded a semantic role in the same way that typical subjects can. As has been stated by many authors (see e.g. Keenan 1976: 321; Langacker [1991] 2006: 305–309), the typical role of a subject is that of an agent. In transitive events involving both an agent and a patient, the agent is referred to by the subject. Put another way, the participant more directly responsible for the occurrence of the denoted event is accorded primary argument status and surfaces as a subject. For weather phenomena, this

210 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

is not possible, since these events do not typically have a participant that could be held directly responsible for the given event. Prototypical subjects are typically also referential in being animate and deinite entities (Keenan 1976: 319; Comrie 1989: 189; Langacker [1991] 2006: 305–309; Dixon 1994: 8, 119). his is especially evident for transitive clauses, where patients, in turn, are usually inanimate and indeinite (see e.g. Comrie 1989: 189). Moreover, it should be noted that A, the agent subject in a transitive clause, typically represents old information, which also underlines its referential nature (see Du Bois 1987). Weather phenomena are only rarely described by transitive constructions, which stresses the non-referential and indeinite nature of their subjects. Instead, many languages describe weather phenomena (at least primarily) by intransitive constructions in which this subject characteristic is typically less strict than in transitive clauses; the sole subject argument S in an intransitive clause can basically refer to any kind of participant, since formal requirements override semantics in these cases (as in many languages all clauses must have a subject). Referents of S may furthermore be highly indeinite, since intransitive clauses are typically used for introducing new, indeinite, participants into the discourse, as shown by Du Bois (1987). In addition, weather phenomena may be claimed to be inherently indeinite entities. hey come into existence only when they occur and they thus typically represent new information. Typical examples of this include rain, snow and thunder. hey can be seen as deinite entities only when we speak of their consequences (as in ‘the rain yesterday made us wet’). An obvious exception to the indeinite nature of ME subjects is provided by clauses which express sunshine and relate to the unique, visually accessible and salient entity ‘sun’. Consider (8), where ‘sun’ appears in the deinite form: Swedish (8) Solen skiner. sun.def shine ‘he sun is shining.’

(personal knowledge)

Furthermore, referents of prototypical subjects are claimed to exist independently of the event referred to by the predicate (Keenan 1976: 313–314). Due to the lack of proper participants with an independent existence, languages have to resort to other elements for illing the subject slot for weather predicates. he most available candidates for subjecthood are usually represented by entities which are integral parts of the events described, cf. e.g.: German (9) Große Tropfen regneten ihm auf den Kopf. big drops rain.pst.3pl him onto the head ‘Big raindrops were raining onto his head.’

(Hummel 2004: 48)

he world is raining

Estonian (10) Esimene lumi sadas maha. irst snow.nom rain.pst.3sg ground.ill ‘he irst snow fell onto the ground.’

(personal knowledge)

In cases like (9) and (10), the relationship between the weather phenomenon and the entities conceptualized as its ‘participants’ is intimate. Rain(drops) and snow cannot easily be distinguished from the event itself, i.e. from snowing and raining. In other words, they are not autonomous elements and do not constitute ideal candidates for independently existing referential subjects. As Givón (1984: 89) states, “the subject and verb/predicate are […] one and the same” in constructions expressing weather and therefore “the event or state cannot be separated from an argument (‘subject’) about which the event/state is predicated”. However, since (parts of) weather phenomena are oten the only ‘participants’ present, they are accorded subject status in some languages. Finally, subjects of MEs do not constitute topics as prototypical subjects do (Keenan 1976: 318–319). he ME subjects do not typically identify what the speaker is talking about, and they are not likely to be mentioned again and tracked in the subsequent discourse. For example, in Finnish the most frequent arguments of MEs are NPs which appear in the partitive case (Kolehmainen 2010; for Finnish MEs see also Huumo & Helasvuo, this volume). heir function is to specify the nature of the precipitation in question, as with the element in boldface in (11) below. hey do not, however, introduce new referents which would form the topic of the subsequent discourse, and, according to Helasvuo (2001: 55), arguments in the partitive case in intransitive sentences are not likely to be mentioned again in the discourse. he following example – a paragraph from a newspaper article – illustrates how the referent of the highlighted partitive NP is not central to the discourse. It is a part of an ME describing the circumstances. It is not mentioned again, nor does it form the topic of the discourse (Kolehmainen 2010). Finnish (11) Jos haluaa ymmärtää, miksi Yhdysvallat valitsi Bosnian rauhanneuvottelujen tyyssijaksi lentotukikohdan Daytonin kaupungin lähellä Ohion osavaltiossa, on matkustettava paikan päälle. Sataa lunta (precipitate.3sg snow.par ‘It is snowing’). Yhdysvaltain suurinta lentotukikohtaa ympäröi piikkilangan koristama aita. Balkanin riitapukarit saavat käydä neuvottelutaistelunsa täällä rauhassa. (Finnish Text Collection) ‘If one wants to understand why the USA chose an air base near the town of Dayton in the State of Ohio as the center for the Bosnia peace negotiations, you have to travel there. It is snowing. he biggest air base in the USA is surrounded by a fence decorated by a barbwire. he Balkanian squabblers can negotiate here in peace.’

211

212 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

3.2

Coding properties

he peculiar semantic properties of meteorological events are relected in their coding. As stated earlier, we will focus in this paper on the encoding of meteorological events as MPs. We will begin by discussing the formal realization of MP subjects. his includes the complete lack of subjects, the use of expletive subjects, and the use of lexical subjects with particular focus on the fuzzy border between adverbial and core argument function in the latter type. hen morphological features connected to the subject function, i.e. case marking and agreement, will be discussed.

he lack of subjects 3.2.1 Subjects constitute privileged arguments in many languages. hey are oten obligatory constituents, and also in languages in which they can be dispensed with, they oten constitute primary arguments in the sense that their slot must be illed for a clause to have other core arguments (objects and indirect objects) (cf. e.g. Primus 1999: 52; Blume 2000: 131, 136, 144). herefore, the lack of subjects with MPs (also in languages with more or less obligatory subjects) requires an explanation. As may be expected from the discussion above, the lack of proper participants is the key to understanding the lack of subjects with MPs. Examples are found in (12)–(14), where the subject position is let unoccupied. Consider: Finnish (12) Tuulee. wind.3sg ‘It is blowing/it is windy.’ Digo (13) I-na-nya. 9-cont-rain ‘It is raining.’1 Tukang Besi (14) No-wande(=mo). 3r-rain=prf ‘It is raining.’

(personal knowledge)

(Steve Nicolle, p.c.)

(Mark Donohue, p.c.)

Bauer (2000: 100, 109) notes that this feature separates weather predicates from other impersonal predicates, which can, for example, occur as subjectless oneplace predicates with an object NP referring to the experiencer participant (e.g. physical sentience verbs and emotion verbs). 1. In the gloss, “9” stands for the noun class of mvula ‘rain’ and subject concord of noun class 9.

he world is raining

It is important to note that the absence of subjects in these constructions should not be confused with pro-drop. In the latter case an argument of the predicate is let unexpressed because it is already encoded in the person-number-gender agreement (= png) and/or is understood from the context. Atransitive weather constructions, on the other hand, are genuinely subjectless; there is no subject argument available to put into the subject slot. An illustrative example of the difference, which also emphasizes our previous argument that there are no proper participants involved in weather events, is given in (15) from Aleut. he verb is ambiguous between an ordinary intransitive (non-weather) reading with pro-drop (the irst given reading) and an atransitive weather reading (the second reading): Aleut (15) chiχta-ku-χ be.wet-prs-sg ‘It (e.g. the ground) is wet.’ or ‘It is raining.’

(Bergsland 1997: 129)

he ambiguity can be resolved by a superordinate verb of sensation. If the latter takes transitive PNG agreement, and agrees with a 3rd person object, the reading is of the intransitive non-weather type, but if it takes intransitive PNG agreement, i.e. not agreeing with any object, the reading is of the atransitive weather type, cf. (16). Aleut (16) a.

(Bergsland 1997: 129)

chiχta-ku-χ ukuχta-qa-ng be.wet-prs-sg see-rem-sg.1sg ‘I saw that it was wet.’ b. chiχta-ku-χ ukuχta-na-q be.wet-prs-sg see-rem-1sg ‘I saw that it was raining.’

In the meteorological use of chiχta-ku-χ, there is no participant available that could be said to be wet, which is manifested in the absence of object agreement in the verb ‘see’. he wetness event is the same as what we would lexicalize as raining, and such an event does not contain any inherent proper participants. hus the lack of a formally realized subject in (16b) relects the underlying lack of a referential element that could surface as a subject. 3.2.2 Expletive subjects In some languages, MPs take a subject in the form of a pronoun or deictic adverb (as in it is raining). hese subjects are known by a number of names, e.g. dummy subjects, formal subjects or pleonastic subjects, but we have chosen the term expletive subjects in this paper. he general view on these subjects is that they

213

214 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

are licensed in clauses where no lexical element is available to serve as subject, and instead a non-referential formally required element is put into the subject slot. Predicate type meteorological expressions lend themselves naturally to such subjects, as meteorological events do not involve proper participants, and a meteorological predicate will thus not provide an element for the subject position. Apart from meteorological expressions, expletive subjects are also found in presentational sentences (e.g. there is a dog in the garden) and in sentences with dislocated clausal subjects (e.g. it is hard to analyze such constructions). Based on our data, it seems that meteorological expressions with expletive subjects are largely conined to the languages spoken in Northern and Western Europe, and that they are only rarely found elsewhere, although more research is needed to establish this. As noted above, there are two main types of expletives: pronominal and adverbial expletives. Pronominal expletives are mainly pronouns that are referentially most unmarked in the language, i.e. the 3rd person neuter pronoun in English and Scandinavian languages, and the 3rd person masculine pronoun in French (due to the lack of neuter gender in French; see Achard, this volume). Examples are given in (17) and (18): Swedish (17) Det regnar 3sg.n rain.prs ‘It is raining.’

(personal knowledge)

French (18) Il pleut. he/it rain.prs.3sg ‘It is raining.’

(personal knowledge)

Adverbial expletives, in turn, are either equal to, or have developed from deictic adverbs like ‘there’ and ‘here’. For example, in the Lyngdal dialect of Norwegian, an expletive dar is used, most presumably cognate with the deictic adverb de:r in the same dialect. Norwegian, Lyngdal dialect (19) Nå reine dar. now rain.prs exp ‘Now it is raining.’

(personal knowledge)

In languages with diferent types of expletives, their use in diferent contexts is usually syntactically determined. English, for instance, has both a pronominal expletive it and an adverbial expletive there, but only the former can be used with MPs; there is restricted to presentational sentences, so *there rains oten here would

he world is raining

be ungrammatical. Nevertheless, this is not a universal restriction, as shown by the Lyngdal Norwegian data in (19), and similar cases attested in other Norwegian dialects and in Danish. It was argued above that languages usually choose their referentially most unmarked pronoun/adverb to function as expletive subjects. his does not, however, always hold, and quite intriguingly some languages employ a marked human pronoun as an MP expletive. For example, in Icelandic, Faroese and some Mainland Scandinavian dialects a pronoun meaning ‘he’ can be used as an MP expletive. Consider: Faroese (20) Hann kavar he snow.prs.3sg ‘It is snowing.’

(hráinsson et al. 2004: 287)

Even more intriguingly, in all these varieties of North Germanic, an unmarked neuter pronoun of the it-type (e.g. tað in Faroese) is used in all other expletive contexts, such as presentational sentences, showing that the marked masculine pronoun is licit only with MPs. Similarly, in the Chadic language Hausa there is an impersonal weak subject pronoun, which “[…] serves as an unspeciied subject corresponding to English ‘one/they’. Generally speaking, it is restricted to human referents […]. Second, [it] is used as the dummy subject of ‘weather’ sentences […]” (Newman 2000: 271). In Germanic, one could claim that these pronouns refer to entities involved in the weather event, which would otherwise be encoded by masculine nouns (e.g. Norwegian himmel ‘sky’ or vind ‘wind’), but which have been grammaticalized into a masculine pronoun. his explanation, however, still leaves out the impersonal pronouns of Hausa. Alternatively, a human pronoun may have been picked to mime the agent-role semantics normally found in prototypical subjects, and which may be more easily conceptualized in dynamic meteorological events than in static presentational sentences. We will not draw any irm conclusions about these MP expletives, but leave them for future research. 3.2.3 Lexical subjects To recapitulate the ME typology from Section 2, lexical subjects are hallmark features of both the argument type and the argument-predicate type. In these types, a lexical subject is the only available option, and the subjects will always be nouns referring to weather phenomena like rain, snow, hail, etc. – as shown by the Korean and Turkish examples in (2) and (3). In the predicate type, on the other hand, the predicate is the denominator of the meteorological phenomenon.

215

216 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

he subject position can thus be seen as superluous in terms of event semantics, which has led to the existence of the subjectless constructions and expletive subject constructions exempliied in 3.2.1 and 3.2.2. However, lexical subjects do exist as well, as we will show in this section. Lexical subjects of MPs generally refer to background entities in the meteorological event. hese can be nouns referring to the spatial background (like ‘place’, ‘surroundings’, ‘world’, etc.), the temporal background (‘day’ or ‘time’), or the atmospheric background (‘sky’, ‘air’ or ‘weather’). he occurrence of these elements as subjects can be considered natural, since they are readily available for weather events, and in the absence of proper participants they constitute the most ‘agent-like’ entities available. Examples are found in (21)–(23): Udihe (21) Ba: maga. place bad.weather ‘to be bad weather/storm’ Riau Indonesian (22) Hari hujan. day rain ‘It is raining.’ Kham (23) Nəm wa-ke. sky rain-pfv ‘It rained.’

(Nikolaeva & Tolskaya 2001: 510)

(David Gil, p.c.)

(Watters 2002: 234)

he nature of these subjects, i.e. that they refer to circumstances, shows that these subjects transgress the adverbial-core argument boundary. In canonical (intransitive or transitive) clauses, these NPs would typically be encoded as adverbials rather than as subjects, as in Today it is raining, but in cases such as (21)–(23), one might argue that circumstantials have rather been promoted to the subject position in order to maintain the prototypical subject-predicate structure. Langacker ([1991] 2002: 232–234; [1991] 2006: 345–348) discusses similar examples in nonmeteorological contexts. He labels the circumstantial subjects as setting subjects (see also Achard, this volume), and shows that they behave diferently from canonical participant subjects: they reject passivization. Since MPs in general cannot be passivized (cf. 3.3), passivization cannot be applied to them as a test criterion. Nevertheless, the circumstantial subjects of MPs may display other kinds of fascinating features. An interesting case in point is illustrated by the form of meteorological expressions in Samoan, illustrated in (24):

he world is raining

Samoan (Mosel & Hovdhaugen 1992: 418–419) (24) a. Sā timu Apia, ‘ae le‘i timu Satitoa. pst rain Apia but not.yet rain Satitoa ‘It rained in Apia, but not in Satitoa.’ (lit.: ‘Apia rained, but not Satitoa.’) b. Sā timu le aiai ananai. pst rain art aternoon yesterday ‘It rained yesterday aternoon.’ (lit.: ‘Yesterday aternoon rained.’) c. Sā timu Apia i le aiai ananai. pst rain Apia pr art aternoon yesterday ‘It rained in Apia yesterday aternoon.’ (lit.: ‘Apia rained on yesterday aternoon.’) d. Sā timu le aiai ananai i Apia. pst rain art aternoon yesterday pr Apia ‘It rained in Apia yesterday aternoon.’ (lit.: ‘Yesterday aternoon rained in Apia.’)

If an NP referring to either the time or the place of the event is used, it cannot be encoded as a prepositional phrase, as it would in other clauses, but must occur as an absolutive core argument, i.e. a subject. Only if both an NP referring to place and an NP referring to time are occurring simultaneously with an MP will one of them be encoded as a prepositional phrase, as there is only one subject position available. If these subjects are “promoted adverbials”, we would also be able to explain why deictic adverbials of the there-type occur as expletive subjects of MPs in certain Scandinavian varieties. In Lyngdal Norwegian (19), the expletive dar is presumably cognate with the deictic adverb de:r in the same dialect, but has the syntactic distribution of an ordinary pronominal subject. his would also indicate that expletive subjects are not necessarily as semantically devoid of function as commonly held, but that they might have an adverbial semantic function equal to the lexical subjects exempliied above. Finally, deities like Zeus or Jupiter may occur as subjects in MEs, a well-known phenomenon from classical languages: Latin (25) Juppiter pluit. Jupiter rain.3sg.prs ‘Jupiter rains.’/ ‘Jupiter sends rain.’

he explanation for cases such as (25) seems to us to be completely extra-linguistic. Within a religious world-view it would be easy to conceptualize a deity as the acting/instigating/controlling force in meteorological events, and thus this deity could be introduced as a prototypical agent and an equally prototypical subject of

217

218 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

an MP. If the notion of a deity is let out of the equation, however, the meteorological event is let wanting of proper participants, and we get the patterns illustrated elsewhere in the paper. It seems that most cultures avoid conceptualizing such an intimate link between a deity and the weather, given that such subjects do not seem to be particularly common, but the human pronominal expletives referred to in 3.2.2 might potentially have evolved from such a concept. 3.2.4 Case marking properties Case marking is unarguably one of the central features of argument marking. In many languages, for example, it assigns central grammatical functions such as subject, direct object, indirect object and adjunct/oblique to clausal constituents (similar functions may also be expressed by adpositions). he subject may, depending on the language and the deinition of a subject, appear, for example, in the nominative or the ergative case, while the object bears, for example, accusative or absolutive encoding. As shown by the typology of MPs, they do not always necessarily require a subject, and consequently such clauses would not display any data on subject case marking either. hus, for the sake of the discussion of case marking, we will here extend our scope to include not only the predicate type of MEs, but also the argument type and the argument-predicate type, as the latter two groups by necessity always include a lexical subject. Since MEs are more oten than not intransitive constructions, their subjects are more like an S than an A in the spirit of e.g. Dixon. his means that the distinction between zero and other kinds of case marking is usually not relevant to subjects of MEs. Numerous instances of MEs with zero-marked subjects have been presented in the examples examined thus far, and so no further examples are given here. Zero marking can be seen as the default marking pattern for ME subjects. In languages with ergative alignment, subjects of MEs rarely bear ergative case markings. However, some examples exist. In Northern Akhvakh, certain weather phenomena are referred to by formally transitive constructions (Denis Creissels, p.c.), cf.: Northern Akhvakh (Denis Creissels, p.c.) (26) Miλi-de gõʁwel-āri duna. sun-erg illuminate-pfv world ‘he sun is shining.’ (Lit. ‘he sun has illuminated the world.’)

he example in (26) represents the argument type, in which the entity responsible for the denoted weather phenomenon surfaces as a subject and bears ergative case marking. Another example of an ergatively coded subject in an ME is attested in Georgian, where weather verbs belong to a group of intransitive verbs called medial verbs, which take subjects in the ergative case (but only in contexts where Georgian uses the ergative). his includes verbs like ‘to rain’, ‘to snow’, ‘to thunder’, ‘to be lightening’ and ‘to blow’. Consider:

he world is raining 219

Georgian (27) kar-ma i-kr(-)ol-a wind-erg prv-blow-3sg.aor ‘he wind blew.’

(Hewitt 1996: 164)

As illustrated above, nominative marking (or some other kind of zero marking) is typical of a meteorological event. One of the evident reasons for this lies in the predominantly intransitive nature of constructions expressing MEs. An interesting example of possible variation in the case marking is provided in (28) from Finnish: Finnish (28) a.

[…] tilanne on vaikein, jos lumi situation is most-diicult if snow.nom sataa sulaan maahan. precipitate.3sg unfrozen.ill ground.ill ‘he situation is most diicult if snow falls onto unfrozen ground.’ (Finnish Text Collection) b. […] jolloin Joensuuhun satoi lumen. when Joensuu.ill precipitate.pst.3sg snow.acc ‘[…] when it snowed in Joensuu.’ (Lit. [It] fell snow in Joensuu.) (Finnish Text Collection) c. Sataa lunta. precipitate.3sg snow.par ‘It is snowing.’ (see also (11))

As shown in (28), the sole argument (S) of the MP sataa ‘precipitate/rain’ may bear nominative, accusative and partitive coding (see also Huumo & Helasvuo, this volume). his variation resembles the case marking variation attested in typical active/stative languages in which the encoding of S is luid: sometimes it is marked like A, sometimes like O. In (28), the variation pertains to one and the same verb sataa ‘rain’ with the same referent lumi ‘snow’. In (28a), ‘snow’ appears in the nominative case, and is thus case-marked as a subject. In (28b) and (28c), S appears in the accusative and the partitive case respectively, which are typical object cases in Finnish. he examples are taken from Kolehmainen (2010) who argues that the manner in which the speaker conceptualizes the role of the participant and wishes to present it governs the manner in which it is expressed. In (28a), ‘snow’ is perceived as a moving entity whose movement happens autonomously and independently of any other participant. In this respect, its role resembles some of the features of a prototypical agent (see Dowty 1991). In (28b), in turn, the role of ‘snow’ comes closer to an efected object whose referent comes into existence. It refers to the visual impression of a white snow layer which is created in the particular situation. Finally in (28c), ‘snow’ occurs in an existential clause in which it does not have the prototypical features of an agent, and the typical features

220 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

of a patient do not it it particularly well either. he clause does not describe an action initiated by the referent of the partitive NP, nor does the situation afect it. In literature, the partitive NPs in existential clauses have been claimed to display neutralizations in the no-man’s-land between the prototypical subject and the prototypical object, and the clauses which involve them are described as functioning as depictions of the circumstances in question (Vilkuna 1989: 156; Huumo 1997; Huumo 2003). his is the case in (28c) as well. 3.2.5 Agreement Case marking is closely related to the cross-referencing of arguments, and in many languages cross-referencing is an additional (or even primary) means of argument marking. For example, in Finnish and German (along with numerous other languages), only subjects are cross-referenced on the verb, while in other languages, such as Misantla Totonac and Yimas, subjects and objects are cross-referenced by distinct aixes (in Yimas this is the only mechanism of argument marking). Typical subjects may be in any person and number available in a given language, and the verb must always agree with its subject in person and/or number (if person agreement is relevant in the given language). MPs, as repeatedly pointed out, do not refer to events which include distinct participants which could function as subjects, and are instead oten encoded without subjects (as shown in 3.2.1) or with non-referential expletives (3.2.2). his lack of referential subjects leads, not surprisingly, to the MPs most oten taking on default agreement morphology, e.g. the 3rd person, singular number, neuter gender, etc. Counter-examples do exist, however. In Kiowa, weather predicates like ‘to be hot’, ‘to be cold’ and ‘to be dark’ obligatorily take plural inanimate marking, e.g.: Kiowa (29) Khí·dêl gyà-sál. yesterday pl-hot ‘It was hot yesterday.’

(Watkins & McKenzie 1984: 135–136)

A diferent type of agreement in weather predicates is found in Athabaskan languages. In San Carlos Apache, MEs do not have overt subjects, but require the “3sg form, used when the subject is a space, area, environment, time, or ‘things’” (de Reuse & Goode 2006: 118). Consider the examples below: San Carlos Apache (30) a. sidod hot ‘It is hot’

(de Reuse & Goode 2006: 118)

he world is raining

b. go-zdod 3sg-hot ‘he weather is hot’

In (30a), the reference is not to weather, and no preix occurs. Example (30b), in turn, denotes weather and the predicate takes a particular preix for this purpose. Rice (1990) reports similar data from Slave. he Athabaskan data show that subject agreement in weather predicates parallels the typology of how MPs encode their subjects. he general 3rd person singular agreement pattern attested in most languages can be regarded as the unmarked default option, as argued above. he third person singular form is the least marked form of cross-reference (and is zero in many languages), and it is thus natural that it applies whenever the subject is completely lacking; any other kind of agreement would be counter-intuitive. he agreement morphemes in San Carlos Apache and Slave, however, parallel the use of special lexical subjects in certain languages (see 3.2.3). In both language types an adverbial concept – circumstantial – is promoted to subject function, either realized through a lexical subject or through a set of agreement markers which are specialized to refer to “space, area, environment, time, or ‘things’”, as de Reuse & Goode describe the preix in San Carlos Apache above. he Kiowa data are not as easily explained, but it could still be that the obligatory plural inanimate marking is semantically akin to the background markers in Athabaskan languages – one could maybe interpret the Kiowa sentence in (29) as ‘hings were hot yesterday’, with “things” referring to the mass of entities constituting the world, just like “things” is also mentioned among the concepts which the San Carlos Apache preix may refer to. In languages where MPs take a lexical subject, this most oten occurs in the singular, and we should not be surprised that also in these languages agreement seems generally to be in the 3rd person singular. However, the reasons might be more profound than simple agreement with a singular subject noun. Consider the following data from Kham: Kham (31) a.

nəm wa-ke sky rain-pfv ‘It rained’ b. nəm-rə wa-ke sky-pl rain-pfv ‘It really rained’ c. *nəm-rə wa-ke-rə sky-pl rain-pfv-3pl ‘(It really rained)’

(Watters 2002: 234)

221

222 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

In Kham, MPs take the lexical subject nəm, ‘sky’. his word can appear in the plural for emphatic purposes, as in (31b), but notice that the MP remains in the singular – plural agreement would be ungrammatical, as shown in (31c). his disagreement might be yet another indicator of the divergent subject status of MP lexical subjects, i.e. our theory that they are adverbials promoted to subject function. his status may deprive them of ordinary agreement control abilities, leaving the MP to its default agreement pattern. With nothing apart from Kham’s data to corroborate such an analysis, however, we will leave it as a proposal rather than a claim, pending further research. 3.3

Behavioral and control properties

In this inal section, we will discuss some further properties of subjects of MPs, which also make them diferent from canonical subjects. In Keenan’s (1976) multi-factorial study of subjects, several behavioral features are demonstrated to characterize prototypical subjects. Most of these are not relevant to the current discussion, or alternatively, we lack the necessary cross-linguistic data to discuss them. herefore, we concentrate in the following on three properties: the control of coreferential deletion, voice and mood. As noted by Keenan, subjects are the most natural controllers of coreferential deletions. his means that in clauses such as he child met a friend and went home, the only possible reading is that the child, and not the friend, went home. In this respect, MP subjects do not difer from canonical subjects, since coreferential deletion is possible also for them, as in it is raining and snowing. he only diference from canonical subjects is found in the nature of the possible targets of deletion. For example, it in it is raining is coreferential only with a similar element of another MP, whereas in the case of typical subjects, the number of possible targets is naturally higher. his follows from diferences in referentiality between the typical pronominal subjects and the (formally identical) expletive subjects of MPs; only the former elements are referential. here are, however, other features that make MP subjects interesting also in this regard. One example is given in (32): Choctaw (32) a. óba-cha oktosha-h rain-ss snow-tns ‘It rained and snowed.’ b. óba-na oktosha-h rain-ds snow-tns ‘It rained and snowed.’

(Broadwell 2006: 266)

he world is raining 223

Choctaw has a typical switch-reference system. With typical subjects, the form of the switch-reference suixes varies according to whether the subjects of coordinated clauses are the same (encoded by the ss (same subject) suix -cha) or not (encoded by the ds (diferent subject) suix -na). he variation is thus functionally determined. With weather verbs, on the other hand, the same-subject and diferent-subject markers are in free variation, i.e. the verb marking lacks a functional motivation. he cross-referential properties of meteorological predicates are thus diferent from those of canonical subjects in Choctaw. In addition to the ability to control coreferential deletion, prototypical subjects participate in certain valency-changing operations, namely the passive and the causative. Both of these operations have consequences for the subject. Passivization is deined as a morphologically signaled process that promotes the object to subject position and demotes the original subject to a (typically freely omissible) oblique. Passivization applies at least to transitive verbs if a language has an operation that can be labeled as passive. An A in a transitive clause is thus the most natural target for subject demotion by passivization. In some languages, passivization also applies to intransitive verbs. In these languages (such as Finnish and Dutch), intransitive verbs usually difer according to passivizability; the passive may demote/omit the sole argument of unergative verbs (such as ‘run’ and ‘jump’), while the passivization of unaccusative verbs (‘die’, ‘fall’, ‘break’) is usually more restricted. Both of these features aid us in understanding the unacceptability of passivization of weather verbs. First, most verbs denoting weather are intransitive. Consequently, they cannot be passivized in languages where only transitive verbs may undergo passivization. Second, the exclusion of passivization in languages where intransitive verbs may be passivized can be explained by referring to the semantic nature of MP subjects: weather events do not include agents, which seem to be a prerequisite for passivization in many languages. For example, clauses such as *two cars are had by me are not possible, since ‘have’ does not include an agent. With weather verbs, there is no agent to demote, which excludes passivization and the demotion of the original subject to adjunct status. However, as usual, counter-examples are attested, as shown in (33): Dutch (33) a.

(Joost Zwaarts, p.c.) Ik ben nat geregend door een tropische onweersbui. I be.1sg wet rained through a tropical thunderstorm ‘I got wet because of a tropical thunderstorm’ b. Het balkon is onder gesneeuwd. the balcony is under snowed ‘he balcony is covered with snow.’

224 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

As shown in (33), Dutch allows the passivization of weather verbs under certain circumstances. However, it is important to note here that the primary function of passivization in this case is not to demote the original expletive subject, but rather to promote a thoroughly afected patient to subject, in other words to focus position. his function is understandable, since passivization is in many languages conined to clauses that involve a highly afected patient. he above Examples (33) clearly have highly afected patients, and the clauses are also resultative in nature. Resultatives (such as ‘the house has been built’ or ‘the book has been written’) are also characterized by their lack of an agent. hese facts may be said to account for the use of the passive instead of the active in this case. Causativization, in turn, can be seen as the mirror image of passivization: it adds an external causer to the denoted event. he introduced agent always occupies the subject slot, which typically demotes the original subject to object or oblique (or similar) status, for example, depending on the transitivity of the original clause. he introduced agent either directs its action directly at the afected participant (as with causativized unaccusative verbs, e.g. ‘the child broke the vase’) or it instigates the event without (necessarily) being actively involved in the event itself (as in ‘the ireighter made the bricklayer build a house’). he introduced agent is typically in control of the denoted event: It exercises control over the causee, who is made to act in the way speciied by the verb. Typically, the causee is also in control of what s/he is doing, which means that the main diference from a typical subject lies in volitionality. he lack of control is central to the exclusion of causativization with weather verbs. Weather (along with other natural phenomena) is typically completely out of our control. It is thus impossible for us to make weather phenomena happen. Moreover, weather phenomena themselves are not active participants, a fact which also contributes to the infelicity of causativized weather verbs. In other words, they are not typical targets of causativization. However, causativized weather verbs are not unheard of, as the examples below show: Finnish (34) a.

(personal knowledge) puutarhuri sade-tta-a nurmiko-n aamu-lla gardener rain-caus-3sg.prs lawn-gen morning-ade ‘he gardener waters the lawn in the morning.’ b. opettaja tuule-tta-a luoka-n teacher to.wind-caus-3sg.prs classroom-gen tunni-n jälkeen lesson-gen ater ‘he teacher airs the classroom ater the lesson.’

he examples in (34) involve typical weather verbs (‘rain’ and ‘blow’ (about wind)) that have been causativized. As such, these examples could therefore be seen as counter-examples to the generalization mentioned above. However, there are

he world is raining 225

evident diferences from canonical causativization. First, the verbs sataa ‘rain’ and tuulla ‘blow’ are normally subjectless in Finnish, which means that there is no subject to demote to causee status. he objects nurmikon and luokan in (34) are not original subjects demoted to causee status via causativization. Second, semantically the predicates in (34) are not causatives derived from non-causative verbs. In (34a), the causativized version of sataa actually means ‘to water’, not ‘to make/cause rain’. In (34b), in turn, the verb has the meaning ‘let in some fresh air by opening a window’, not ‘to make/cause the wind to blow’. In other words, we are really just dealing with morphological causativization of weather verbs in (34), not with typical causativization, which would include both morphological and syntactic traits. he verbs in (34) can also be seen as lexicalized verbs with somewhat idiosyncratic, not fully predictable meanings. One further note on causativization of weather verbs is worth mentioning here. In (34), we discussed morphological causativization of weather verbs and showed that it is possible, but that we are not then dealing with canonical instances of causativization. However, verbs can also be causativized periphrastically and lexically. It is relatively easy to ind examples of periphrastic causatives with MPs, as the examples in (35) illustrate: Finnish (35) a.

(personal knowledge) Jumala antaa sataa niin hyv-i-lle […] God let.3sg.prs rain.inf so good-pl-all ‘God lets it rain on the just…’

German b. Gott lässt es regnen über Gerechte […] God let.3sg.prs it rain.inf over just ‘God lets it rain on the just…’

he examples in (35) illustrate typical examples of (permissive) periphrastic causativization. Unlike the morphological examples in (35), we are dealing with canonical instances of causativization also semantically speaking. In these cases, God actually causes rain. What is especially interesting here is that deities are seen as capable of controlling weather phenomena, which makes the examples in (35) grammatical (just like the same concept made deities licit as MP subjects in Latin and Classical Greek, see 3.2.3). his thus underlines the signiicance of control to causative formation. Finally, the inability of weather verbs to causativize is closely related to their restricted imperative formation. Imperatives resemble causatives in one important respect directly related to subjects. In both causatives and imperatives, an external causer is making another animate entity act according to his/her will. Both operations are possible with highly animate entities and events that are in principle controllable. Imperatives (and causatives) can thus be readily formed from verbs that

226 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

denote actions involving a typical agent. On the other hand, imperatives formed from verbs lacking an agent are less frequent and semantically odder. In other words, imperatives such as go!, run! and build a house! are completely normal, while the likes of see the man!, break! and melt! sound less normal. Since weather phenomena are completely out of our control, the imperative formation of weather verbs yields a less felicitous construction than the imperatives formed from verbs denoting typical actions.

4. Conclusions Weather constitutes an extraordinary event type whose peculiarities are relected in its expression both in and across languages. he most conspicuous peculiarity of meteorological events is the lack of proper participants, which has linguistic consequences, some of which were discussed in this paper. he study focused on meteorological predicates (MPs) and their subjects, which do not behave like prototypical subjects. he MPs were oten demonstrated to lack a subject, and in the languages where MPs do have subjects they were oten shown to be non-referential, nontopical, indeinite, inanimate or non-agentive entities. hese are all properties that characterize the opposite of prototypical subjects. In addition, the selection of subjects was shown to be rather restricted. Such subjects are typically either expletive elements grammaticalized from pronouns and deictic adverbs, or lexical elements referring to the spatial (‘place’, ‘surroundings’, ‘world’), temporal (‘day’, ‘time’) or atmospheric (‘sky’, ‘air’, ‘weather’) background, i.e. elements which would otherwise have been encoded as adverbial phrases. he peculiar semantic properties are partly relected in the formal encoding. Although the nominative (or some other kind of zero marking) is typical of subjects of MPs which, in addition, oten take default agreement morphology (3rd person, singular number, neuter gender, etc.), the MP subjects may in many cases show special behavior which marks them of from canonical cases and highlights their special status. he discussion in this paper lends more support to the privileged status of subjects. As noted many times in the paper, weather phenomena lack proper participants, which means that MPs describing them need to resort to participants other than agents for subjecthood. his means that even in cases where the elements required for a typical subject-predicate pattern are missing, languages nevertheless tend to create this structure from other sources in the grammar and/or in the lexicon.

he world is raining 227

References Bauer, Brigitte (2000). Archaic syntax in Indo-European: he spread of transitivity in Latin and French. Berlin/New York: Mouton de Gruyter. DOI: 10.1515/9783110825992 Bergsland, Knut (1997). Aleut grammar. Fairbanks: Alaska Native Language Center University of Alaska. Blume, Kerstin (2000). Markierte Valenzen im Sprachvergleich: Lizenzierungs- und Linkingbedingungen. Linguistische Arbeiten 411. Tübingen: Niemeyer. Broadwell, George A. (2006). A choctaw reference grammar. Lincoln: University of Nebraska Press. Comrie, Bernard (1989). Language universals and linguistic typology: Syntax and morphology. Second revised ed. Chicago: University of Chicago Press. Dixon, R. M. W. (1994). Ergativity. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511611896 De Reuse, Willem J., & Goode, Phillip (2006). A practical grammar of the San Carlos Apache language. Munich: Lincom Europa. Dowty, David (1991). hematic proto-roles and argument selection. Language, 67, 547–619. DOI: 10.1353/lan.1991.0021 Du Bois, John W. (1987). he discourse basis of ergativity. Language, 64, 805–855. DOI: 10.2307/415719 Eriksen, Pål K., Kittilä, Seppo & Kolehmainen, Leena (2010). Linguistics of Weather: Crosslinguistic patterns of meteorological expressions. Studies in Language, 34, 565–601. DOI: 10.1075/sl.34.3.03eri Finnish Text Collection. Selection of electronic research material that contains written Finnish newspaper texts from 1990’s. Compiled by: he Research Institute for the Languages of Finland, he Department of General Linguistics at the University of Helsinki and he Foreign Languages Department of the University of Joensuu. Available through: CSC – IT Center for Science, http://www.csc.i. François, Alexandre (2001). Contraintes de structures et liberté dans l’organisation du discours. Une description du mwotlap, langue océanienne du Vanuatu. Doctoral thesis in linguistics, Université Paris-IV Sorbonne. Givón, Talmy (1984). Syntax: A functional-typological introduction. Amsterdam: John Benjamins. Givón, Talmy (2001). Syntax: An introduction (Vol. I). Amsterdam: John Benjamins. Helasvuo, Marja-Liisa (2001). Syntax in the making. he emergence of syntactic units in Finnish conversation. Studies in discourse and grammar 9. Amsterdam: John Benjamins. DOI: 10.1075/sidag.9 Hewitt, George (1996). Georgian: A learner’s grammar. London: Routledge. Hummel, Martin (2004). Zur Valenz der Witterungsverben: dt. regnen, engl. to rain, fr. pleuvoir, it. piovere, pt. chover und sp. llover. In Alberto Gil et al., (Eds.), Romanische Sprachwissenschat. Zeugnisse für Vielfalt und Proil eines Faches. Festschrit für Christian Schmitt zum 60. Geburtstag (pp. 35–58). Band II. Frankfurt a. M.: Peter Lang. Huumo, Tuomas (1997). Partitiivisubjekti ja tilajatkumot. Sananjalka, 39, 65–97. Huumo, Tuomas (2003). Incremental existence. he world according to the Finnish existential sentence. Linguistics 41(3), 461–493. Keenan, Edward L. (1976). Towards a universal deinition of ‘subject’. In Charles N. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press.

228 Pål K. Eriksen, Seppo Kittilä, and Leena Kolehmainen

Kolehmainen, Leena (2010). Sääverbien syntaksia ja semantiikkaa: Semanttiset roolit, osallistujien vaihteleva käsitteistäminen ja sääverbien vaihteleva valenssi. Virittäjä, 1/2010, 5–38. Langacker, Ronald W. [1991] (2002). Concept, image, and symbol. he cognitive basis of grammar (2nd ed.). Berlin and New York: Mouton de Gruyter. DOI: 10.1515/9783110857733 Langacker, Ronald W. [1991] (2006). Foundations of cognitive grammar (Vol. II). Descriptive application. Stanford, California: Stanford University Press. Mosel, Ulrike & Hovdhaugen, Even (1992). Samoan reference grammar. Oslo: Scandinavian University Press. Newman, Paul (2000). he hausa language: An encyclopedic reference grammar. New Haven: Yale University Press. Nikolaeva, Irina & Tolskaya, Maria (2001). A grammar of udihe. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110849035 Primus, Beatrice (1999). Cases and thematic roles: Ergative, accusative and active. Linguistische Arbeiten 393. Tübingen: Niemeyer. DOI: 10.1515/9783110912463 Rice, Keren (1990). Grammar of Slave. Berlin: Mouton de Gruyter. hráinsson, Höskuldur, Petersen, Hjalmar P., Jacobsen, Jógvan í Lon, & Hansen, Zakaris Svabo (2004). Faroese: An overview and reference grammar. Tórshavn: Føroya Fróðskaparfelag. Vilkuna, Maria (1989). Free word order in Finnish. Its syntax and discourse functions. Suomalaisen Kirjallisuuden Seuran Toimituksia 500. Helsinki: Finnish Literature Society. Watkins, Laurel J., & McKenzie, Parker (1984). A grammar of Kiowa. Lincoln: University of Nebraska Press. Watters, David E. (2002). A grammar of Kham. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511486883

section iii

Subjects in networks of constructions

he syntactic and semantic history of the Finnish genitive subject Construction networks and the rise of a grammatical category Jaakko Leino University of Helsinki

Prototypical Finnish subjects are marked with the nominative case. In many non-inite expression types, however, the subject is marked with the genitive case. he genitive subject has grammaticalized from three diferent sources. In some constructions, it developed from an earlier adverbial expressing a recipient, beneiciary or experiencer; in others, it originates in an adnominal possessive genitive; and its background is a syncretistic accusative. he present paper analyzes the properties of the diferent predecessors of the genitive subject which have made earlier adverbials and objects susceptible to a reanalysis into grammatical subjects. he paper sheds light on the nature and origins of the genitive subject, and tackles the development of abstract grammatical categories from more concrete context-speciic ones on a more general level.

1.

Introduction

Finnish has a rich system of non-inite verb forms, both in terms of inlectional morphology and syntactic expression types organized around non-inites. One peculiarity in many, though not all, Finnish non-inite verb constructions is the case marking of their subject – provided, of course, that the construction in question has a subject of its own. While the prototypical subject in Finnish is marked with the nominative case and the subject of inite clauses may in some contexts be marked with the partitive case (for references, see Huumo & Helasvuo, this volume),1 the subject in non-inite clauses is all but invariably in the genitive case. 1. However, Huumo and Helasvuo present an analysis which difers from the Fennistic tradition in that what has traditionally been called the partitive subject in existential sentences is not analyzed as a subject but, rather, as e-NP. doi 10.1075/cal.16.09lei © 2015 John Benjamins Publishing Company

232 Jaakko Leino

In what follows, I will go through the most common syntactic and semantic contexts of occurrence of the Finnish genitive subject. I will look at ive groups of constructions in some detail, and trace the genitive subject of each of these constructions to their historical predecessor. In each of these constructions, the genitive subject has developed from an earlier non-subject argument which has been marked by the genitive and reanalyzed as the subject in suitable contexts. Further, the genitive subject has diferent origins in diferent constructions. he genitive subject in the ive construction families originates from three diferent and diferently motivated sources. I will go through each one of these three “paths” leading to the genitive subject and point out contexts and factors leading to the reanalysis in each of the three paths. In order to provide some background to the discussion of the non-canonical genitive subject, however, let us irst look at the prototypical subject in Finnish. he prototypical Finnish subject is, typologically speaking, not a very peculiar one. According to Vilkuna (1996: 161), it has the following properties: – – – – – – –

Category: NP Case: nominative Agreement: verb agrees with subject Position: neutrally precedes the verb Government: the antecedent of e.g. a bare possessive suix Role: typically agent or experiencer Semantic type: typically animate

However, as pointed out above, in various non-inite expression types, we ind a subject marked with the genitive case. he verb, being in a non-inite form, does not agree with the subject, and the genitive subject may or may not govern possessive suixes and other anaphoric elements. he position of the subject relative to the verb also varies. Genitive subjects occur in several diferent constructions, and historically, it seems coincidental that all these structures have ended up with a genitive subject. In some cases, the genitive subject has developed from an earlier adverbial expressing recipient, and in others from one that expresses a beneiciary or experiencer more broadly; in some cases, it goes back to an adnominal possessive genitive; in some cases it has its background in a syncretistic accusative; and in some cases it is best explained with an analogy of other non-inite structures. Marking the agent of non-inite structures with the genitive case is by no means exceptional cross-linguistically. However, this typically involves either “normal” uses of the genitive or extensions thereof, e.g. the use of the genitive case to mark nominal modiiers being extended to mark similar modiiers of non-inite verb forms. As will be shown, the Finnish genitive subject originates in usages other than marking the subject. In fact, it originates in three unrelated

he syntactic and semantic history of the Finnish genitive subject

uses of the genitive which have converged and grammaticalized into what may be called a canonical subject of non-inites. While the genitive case is not the primary way to mark a grammatical subject in Finnish, it is indeed possible to interpret the genitive subject as canonical. As Sands & Campbell (2001: 277) point out, “he general rule is that the subject of anything other than a inite clause takes the genitive case. Subjects marked with genitive case in these subordinate non-inite verb forms, then, are considered to be canonically marked in Finnish.” I will not make a claim to the contrary. However, whether or not the genitive subject should be interpreted as canonical ultimately depends on what we decide to call “canonical” or “non-canonical”. As a subject overall, the Finnish genitive subject is, in a sense, non-canonical or at least secondary, while as a subject of a non-inite verb it is, in another sense, fully canonical. For a discussion of diferent types of subjects of inite verbs, see Huumo & Helasvuo (this volume).

2. Background: Finnish genitives and non-inites he Finnish genitive has a number of diferent though related uses. Most typically, it marks a modiier of a noun, most oten the owner of the referent of the noun or the NP. However, the genitive has several other uses as well. In her dissertation (2004: 13), Minna Jaakola distinguishes between four (or ive) main uses of the genitive case: – – – –

he N-genitive, meaning instances where the genitive is used to mark a modiier of a noun, typically a possessor he P-genitive, where the genitive case is used to mark the object of a preposition or a postposition he A-genitive, i.e. instances where the genitive marks a modiier of an adjective or an adverb he V-genitive, which is essentially the genitive subject focused on in the present paper N-genitive (1) a. Peka-n auto Pekka-gen car.nom ‘Pekka’s car’ b. afasia-n eri muodo-t aphasia-gen diferent form-pl.nom ‘diferent forms of aphasia’

233

234 Jaakko Leino

P-genitive (2) a. pöydä-n alla table-gen under ‘under the table’ b. pu-i-den takana tree-pl-gen behind ‘behind the trees’ A-genitive (3) a. vaalea-n sininen light-gen blue.nom ‘light blue’ b. kauhea-n hienosti terrible-gen well ‘terribly well’ V-genitive (4) a. satee-n sattu-e-ssa rain-gen happen-inf2-ine ‘in case of rain’ b. minu-n tietä-ä-kse-ni I-gen know-inf1-trnsl-poss.1sg ‘as far as I know’

More generally, the V-genitive refers to all uses of the genitive with verb forms, but it seldom is anything other than a subject of ininitive or participle forms. Jaakola (ibid.) points out that it is also possible to postulate a ith type, the O-genitive. he Finnish “accusative” case is, in fact, identical to the genitive in many instances, to the extent that there is no agreement as to whether there is a separate accusative case in Finnish or not (except in the case of personal pronouns).2 herefore, the genitive or genitive-like accusative used to mark the grammatical object could be labeled O-genitive: O-genitive (5) a. lue-n kirja-n read-1sg book-gen/acc ‘I read the book’

2. For a brief overview of Finnish subject and object case marking, see e.g. Huumo (2005: 114– 119) and Karlsson (1999); see also Huumo & Helasvuo (this volume). Roughly speaking, the genitive–accusative syncretism involves aspectually telic positive polarity sentences with a singular object with a bounded but not numerically quantiied referent. Complex though this may sound, such objects are arguably the most prototypical. herefore, the syncretism is quite noteworthy.

he syntactic and semantic history of the Finnish genitive subject

Jaakola leaves the “accusative” outside of her research but points out – like many before her – that the relationship of the Finnish genitive and accusative is problematic. Historically, the two are separate cases morphophonologically, but synchronically their singular forms are identical. Although this genitive–accusative syncretism does arise later in the present chapter, the question of how to interpret the relationship between the Finnish genitive and accusative cases need not be answered in this context. he Finnish non-inite verb inlection system is morphologically rather complex. Morphologically speaking, Finnish has three to ive diferent ininitives (e.g. Hakulinen et al. 2004: 487 list three, while e.g. Hakulinen & Karlsson 1979: 87 speak of “four or even ive”), each one of which has an incomplete case inlection. In addition, there are four participle categories, all of which have a full and regular case inlection. All four participles are also inlected for number, and three of them are also inlected for voice. Crucially for the purposes of this paper, each of the ininitive forms is more or less speciic to one or a few special constructions, to the extent that the whole concept of “ininitive” is deined by reference to such constructions. And the same is true of many participle forms, too, although they tend to be more productive and regular and less idiosyncratic and construction-speciic than ininitives overall. In any case, ininitives and participles are oten used in such similar manners that it is not easy to draw a neat line between the two categories.

3. Constructions with a genitive subject he fact that ininitives and, to some extent, also participles are associated with speciic constructions in Finnish is crucial to the present chapter because the different historical origins of the present-day genitive subject category are also construction speciic. In order to illustrate this, as well as the syntactic distribution of genitive subjects in Finnish, let us take a few examples of constructions in which the genitive subject commonly occurs: the necessive construction, the permissive construction, the temporal construction, constructions with the agent participle, and the referative construction. hese are the ive constructions that I shall focus on in what follows, and they are also examples of the kinds of constructions in which Finnish non-inite verb forms occur. Ininitive constructions Necessive: (6) Teidä-n pitä-ä lähte-ä or: lähte-mä-n. you.pl-gen must-3sg leave-inf1 leave-inf3-ins ‘You must leave.’

235

236 Jaakko Leino

Permissive: (7) Anta-kaa minu-n ol-la! give-imp.2pl I-gen be-inf1 ‘Leave me alone!’ (lit. ‘Let me be!’) Temporal: (8) Lähde-mme aamu-n koitta-e-ssa. leave-1pl morning-gen dawn-inf2-ine ‘We will leave when the day dawns.’ Participle constructions Agent participle: (9) Äidi-n teke-mä-t lihapulla-t mother-gen make-agptcp-pl.nom meatball-pl.nom ‘Meatballs made by mother’ Referative: (10) Kuul-i-n Karoliina-n lähte-neen. hear-pst-1sg Karoliina-gen leave-ptcp2.gen ‘I heard that Karoliina had let.’

he core of the necessive construction consists of the necessive verb (e.g. pitää or täytyä, ‘must’, ‘have to’), the genitive subject, and the ininitive and its arguments. Since the actor of the event expressed by the ininitive is also the participant to whom the necessity applies, it is not clear whether the genitive subject should be interpreted as the subject of the ininitive or of the necessive verb. he traditional analysis, started by Jahnsson (1871: 106–107) and Setälä (1880: 39) and repeated in textbooks throughout the 20th century, interprets the ininitive as the subject of the necessive verb and the genitive argument either as the subject of the ininitive or as a “dative adverbial” of the necessive verb (cf. P. Leino 1986: 123–126). More recently (e.g. Hakulinen et al. 2004: 500–501), the genitive subject has been more commonly interpreted as the subject of the whole construction. he permissive construction consists of one of four matrix verbs (antaa ‘give’ or ‘let’, käskeä ‘command’, sallia ‘allow’, or suoda ‘grant’), its nominative subject, and the ininitive clause, which may have a genitive subject. he ininitive clause is traditionally interpreted as the object of the matrix predicate. For a detailed account of the permissive construction, see J. Leino (2003, 2005a). he temporal construction in (8) is an adjunct-like structure, built around the E ininitive inessive.3 his structure can occur together with virtually any 3. he E ininitive is called the 2nd ininitive in the Fennistic tradition. I will adopt the more recent naming tradition adopted in Hakulinen et al. (2004) but use the traditional numbering in the glosses for the sake of brevity. hus, the glosses and the corresponding “modern” names are as follows: inf1 = A ininitive, inf2 = E ininitive, inf3 = MA ininitive.

he syntactic and semantic history of the Finnish genitive subject 237

predicate, and it expresses a secondary activity simultaneous to that expressed by the main predicate. here is also a related “second temporal construction” which expresses an earlier activity, built around the passive 2nd (or past) participle: (11) Lähde-mme aamu-n koite-ttua. leave-1pl morning-gen dawn-pass.ptcp2.par ‘We will leave ater the day has dawned.’

he agent participle in (9) may occur either as a modiier of a noun, as in (9), or as a predicative with the copula, as in (12). In both cases, the agent participle has a genitive subject (or an “agent,” hence the name). (12) Lihapulla-t o-vat äidi-n teke-m-i-ä. meatball-pl.nom be-3pl mother-gen make-agptcp-pl-par ‘he meatballs are made by mother’

he referative construction in (10) consists of a verb of cognition or communication, its nominative subject, and a non-inite complement clause. Like temporal constructions, there are two referative constructions with diferent time references. Example (10) is built around the 2nd or past participle, which positions the state of afairs expressed by the complement clause in the past relative to the situation referred to with the main predicate. he other option is to use the 1st or present participle, which expresses the two events as happening simultaneously: (13) Kuul-i-n Karoliina-n lähte-vän. hear-pst-1sg Karoliina-gen leave-ptcp1.gen ‘I heard Karoliina leave.’

All these constructions are common contexts of the genitive subject. he necessive construction is diferent from all the others in that the necessive verb has no separate subject and the genitive subject may be interpreted as the subject of either the necessive verb or the ininitive. In all the other constructions, the genitive subject is clearly the subject of the non-inite verb form, and such an analysis is also possible for the necessive construction. According to Jaakola’s data (2004: 230), shown in Table 1 below, the distribution of the genitive subject may be roughly summarized as follows: – – –

he agent participle is by far the most common use of the genitive subject. he necessive, temporal, and referative constructions follow, and are quite common, too. Everything else is marginal.

hese distributional features seem to difer somewhat from Finnish speakers’ everyday experience. Jaakola’s observations are based on written language data, which difer from everyday spoken language in many respects. For comparison, I

238 Jaakko Leino

went through the occurrences of the genitive subject in the Syntax Archive compiled at the University of Turku. he dialect data may be summarized as follows: – – –

Almost exactly two thirds of all genitive subjects occur in the necessive construction. he agent participle and the permissive, referative, and temporal constructions are somewhat common but much less so than the necessive construction. Everything else is marginal.

In all, these ive constructions (or, more accurately, ive groups of constructions) together cover far more than 90% of all instances of the genitive subject, both in written language and dialect data. Table 1. he distribution of the genitive subject in written language and dialect data.4 Written language (Jaakola 2004) Agent participle Necessive Temporal Referative Agent constr.4 Permissive Other (n = 406)

3.1

Spoken language (Syntax archive) 38% 23% 17% 15% 3% 2% 2% 100%

necessive agent participle permissive referative temporal agent constr. other (n = 836)

66% 9% 8% 6% 4% 1% 6% 100%

he necessive construction

As the distribution data shows, the necessive construction is the primary context of the genitive subject. It should be noted, perhaps, that the term necessive construction in fact refers to a group of interrelated constructions, but we need not go into the details. he interested reader is referred to more extensive works by Laitinen (1992), Laitinen & Vilkuna (1993), and Ikola (1974). Laitinen in particular goes through the history of this group of constructions in great detail. To simplify things somewhat, let us look at one of the most common necessive verbs, täytyä. According to Laitinen (and others before her), the verb täytyä originally means ‘becoming full’ (like the verb täyttyä does in modern-day Finnish), and it used to be interpreted as ‘reaching a state of readiness’ in some contexts. In this

4. he label “agent constructions” refers to a group of passive participle constructions and not to the agent participle.

he syntactic and semantic history of the Finnish genitive subject 239

meaning, a sentence such as (14) meant roughly ‘the crops ripen to be harvested’, and such a sentence could have a benefactive adverbial, marked with the genitive case, which indicates a person for whom the crop ripens. At some point, the meaning of the verb was extended to the more static meaning of ‘being in a state of readiness’, so that the expression in (14) could mean simply that the crops are ready for the man to harvest. Historically: (14) vilja täyty-y (miehe-n) korjat-a crops fulill/ripen-3sg (man-gen) harvest-inf1-lat ‘the crops ripen (for the man) to harvest’ täytyä Originally: ‘becoming full’ Interpreted as: ‘reaching a state of readiness’ Losing the meaning of change: ‘being in a state of readiness’ (14′) vilja täyty-y korjat-a crops-nom must-3sg harvest-inf1 ‘the crops must be harvested’ (~miehe-n täyty-y korjat-a vilja) man-gen must-3sg harvest-inf1 crops-nom ‘the man must harvest the crops’

What the verb täytyä used to express in sentences such as (14) was suicient conditions for the action expressed by the ininitive. Over time, these suicient conditions were reinterpreted (through what Laitinen calls “practical inference”) as the necessity of the action. he ininitive irst expressed a possible and purposeful action. In suitable contexts, this lead to the implication of the action actually taking place. And, as this implication strengthened and conventionalized, the ininitive was eventually interpreted as expressing a necessary action. For the purposes of this paper, the present-day genitive subject in the necessive construction goes back to a benefactive adverbial which originally pointed out someone who beneited from the suicient conditions and was eventually reinterpreted as someone towards whom the necessity was directed. 3.2

he permissive construction

As pointed out above, the permissive construction is peculiar in that it only allows four main verbs: antaa, käskeä, sallia and suoda. he four verbs mostly express some kind of granting of permission, letting, or allowing (hence the name permissive construction), but the verb käskeä ‘to command’ also commonly appears in

240 Jaakko Leino

the construction. By far the most common verb in the construction, antaa (which normally means ‘to give’ when used anywhere else than in this construction) may express either ‘letting’ or ‘having someone do something’. he construction consists of the main verb, its subject, and an ininitive clause. he ininitive clause has a subject of its own, and – as one might expect in the context of this paper – the subject is in the genitive case. Anna minu-n ol-la! give.imp.2sg I-gen be-inf1 ‘Let me be!’, i.e. ‘Leave me alone!’ b. Äiti käsk-i poja-n mother.nom command-pst.3sg boy-gen siivot-a huonee-nsa. tidy-inf1 room-poss.3sg ‘he mother told the boy to clean up his room.’ c. E-n salli-nut häne-n puhu-a juhla-ssa. not-1sg allow-ptcp1 he/she-gen talk-inf1 party-ine ‘I did not allow him/her to talk at the party.’

(15) a.

he permissive construction goes back to expressions of concrete ‘giving’, and the ininitive in this construction originally expresses the purpose of giving. Historically, a sentence like isä antoi omenoita lasten syödä has a regular expression of an act of giving in it, and the ininitive complement expresses purpose: (16) Isä anto-i omeno-i-ta las-ten syö-dä father give-imp.3sg apple-pl-par child-pl.gen eat-inf1 ‘father gave apples for the children to eat’ isä antoi omenoita lasten ‘Father gave apples to the children’ plus an ininitive (or a converb) expressing purpose: syödä ‘for the purpose of eating’.

hus, in the historical interpretation, the sentence involves two actions: the act of giving and the act of eating. he eating is proiled as the purpose for which the git is given to the recipient. In the modern-day interpretation, in contrast, there is only one action, the act of permitting. he eating is a part of that act in the sense that it is the permitted event. More concretely, the historical interpretation of this sentence requires the apples to be transferred from the father to the children, but the modern-day interpretation does not. And, grammatically, in the modern-day interpretation, the word lasten is a genitive subject of the ininitive clause lasten syödä omenoita, whereas in the historical interpretation, it is a complement of the verb give and expresses the benefactive participant, much like in the necessive construction.

he syntactic and semantic history of the Finnish genitive subject 241

Figure 1. he permissive construction, historical interpretation.

Figure 2. he permissive construction, present-day interpretation.

3.3

he agent participle

he two constructions discussed in Sections 3.1 and 3.2 exemplify cases where the genitive subject goes back to a benefactive participant. Let us take a diferent case in point next. In Finnish, there are several morphological categories which have the same phonological form -ma or -mä. hese include the MA ininitive, a deverbal nominal derivation suix (not a very productive one, however), the caritive participle,5 and the agent participle. I will not take a stance as to whether they have a common etymology or not, but they may. For a discussion, see Koivisto (2005). Be that as it may, the agent participle is a non-inite verb form which is typically used either as a nominal modiier or a predicative in copula clauses. It takes the head noun or the subject of the copula clause as its “logical object”, and it always has a genitive subject (or “agent”, hence the traditional name). 5. he caritive participle is roughly the negative counterpaprt of the 1st and 2nd (or present and past) participles, e.g. hajoamaton jäte ‘non-dissolving waste’, viimeistelemätön käsikirjoitus ‘non-inalized manuscript’.

242 Jaakko Leino

According to received wisdom (e.g. Hakulinen 1979: 584–586; Häkkinen 1993), the agent participle itself is etymologically a deverbal noun. For example in (17a), a madon syömä omena, i.e. ‘an apple eaten by a worm’ is etymologically ‘a worm’s eating, what a worm has eaten, namely an apple’, and the modern-day grammaticalized genitive subject is historically a genitive attribute of that deverbal noun. In other words, madon syömä would be something like ‘a worm’s eatingstuf ’, and isän leipoma would be something like ‘father’s bakings’). mado-n syö-mä omena worm-gen eat-agptcp.nom apple.nom ‘an apple eaten by a worm’ b. kakku on isä-n leipo-ma cake.nom be.3sg father-gen bake-agptcp.nom ‘he cake is [has been] baked by father.’

(17) a.

3.4

he referative construction

he three cases that we have looked at so far cover more than 60% of written language genitive subjects and over 80% of genitive subjects in dialect data. But the story continues. he fourth most common context for the genitive subject in both written language and dialect data is the referative construction, which is, as we saw in Section 2, not just one construction but in fact two participle constructions. In each of these constructions, a participial clause is used much like a inite complement clause, with verbs of speech and thought and perception. In the two referative constructions, the genitive subject is historically neither a benefactive complement nor a genitive attribute. Rather, it is the original object of the main predicate. hus, a sentence such as (18) would go back to an interpretation where I actually see the boy. he participle is claimed to be originally either a modiier of the genitive subject (or the original object) or its appositional paraphrase (cf. Ikola 1959: 53; Penttilä 1957: 632; Hakulinen 1979: 565–572). hus, the sentence näin pojan menevän originally meant either ‘I saw the going boy’ or ‘I saw the boy, the goer’, or something of the sort. Näe-n poja-n mene-vän. see-1sg boy-gen go-ptcp1.gen ‘I see the boy go.’ b. Minu-n väite-tään pitä-neen esitelmä-n. I-gen claim-pass keep-ptcp1.gen presentation-acc ‘I am claimed to have given the presentation.’

(18) a.

he syntactic and semantic history of the Finnish genitive subject 243

3.5

he temporal construction

he ith and last example case is the temporal construction. Again, there are not just one but two temporal constructions in Finnish: one which has an E ininitive inessive as its predicate, e.g. Kalle vihelteli tuulen ulvoessa (19a), and one with a passive 2nd participle partitive as its predicate, e.g. Opettaja pyyhki taulun oppilaiden lähdettyä (19b). hese two constructions intuitively go together insofar as the former is used for expressing simultaneous activity (or “present tense”) in relation to the main clause predicate, and the latter is used for expressing previous activity (or “past tense”) in relation to the main clause predicate. Especially in spoken language, the latter of these two constructions is pretty uncommon, however. Kalle viheltel-i tuule-n ulvo-e-ssa. Kalle.nom whistle-pst.3sg wind-gen howl-inf2-ine ‘Kalle whistled as the wind was howling.’ b. Opettaja pyyhk-i taulu-n teacher.nom wipe-pst.3sg board-acc oppila-i-den lähde-ttyä. pupil-pl-gen leave-pass.ptcp2.par ‘he teacher wiped the blackboard ater the students had let.’

(19) a.

In both of these two constructions, the genitive subject originates in a historical genitive attribute. hus, the temporal construction is similar to the agent participle with regard to the origin of the genitive subject. For instance, Hakulinen (1979: 576) explains the form kuultuansa (the passive 2nd participle of the verb kuulla ‘hear’ together with the 3rd person possessive suix) as historically ambiguous between ‘ater one had heard’ and ‘what one had heard’. Similarly, tuulen ulvoessa in (19a) could have been interpreted either as ‘while the wind howls’ or as ‘the wind’s howling’. In the latter interpretation, the present day genitive subject tuulen would have been interpreted as a genitive subject, and the present day ininitive ulvoessa would have been a deverbal noun.

4. he genitive subject: hree paths We have briely reviewed ive example cases and found three diferent diachronic origins for the synchronically rather uniform category “genitive subject.” First, there is the dative genitive origin, which we found in the necessive and permissive constructions and which is by far the most usual in terms of present-day frequencies. Second, there is the genitive attribute origin, which we found in the agent participle and in the temporal constructions. And third, there is the “previous object” point of origin, which is made possible by the syncretism of the genitive

244 Jaakko Leino

and the accusative cases in Finnish (to the extent that it is not at all obvious that Finnish even has an accusative in the irst place). We saw the genitive–accusative syncretism to be the origin of the genitive subject of the referative construction and, in fact, a certain part of the permissive construction actually goes back to an object control structure, so that this path is actually also relevant for the permissive construction if we were to look at it in more detail (cf. J. Leino 2003: 298–304). 4.1

Path 1: he “dative genitive”

Instead of looking at any speciic construction in more detail, let us take a closer look at the three paths leading to the genitive subject, starting with the “dative genitive” (or benefactive complement) path. As background, the relevant kind of dative-like use of the genitive is no longer productive in Finnish: the allative case has assumed the place that allegedly used to belong to the genitive. But apparently the genitive has been used to express recipient, benefactive, and experiencer arguments in a number of constructions at some point in the history of the Finnic languages, and some traces remain of such usage in old written Finnish (for a discussion and somewhat diferent accounts on the development of the dative genitive, see e.g. Korhonen 1991; Huumo 1995; Inaba 2000; P. Leino 2001). As we saw, in the permissive construction, the dative genitive is associated with the recipient role of the giving frame, and in the necessive construction, the historical traces lead back to an obligation created by “practical necessity.” Given what we know of (our conception of) giving, possession, recipients, and obligation, we may deduce that, incidentally, both kinds of dative genitives require animate and, in fact, human referents. Originally, in both of these constructions, the object of the matrix verb has been interpreted as expressing the patient (or “logical object”) of the ininitive as well, and in both cases the object has been reanalyzed as the object of the ininitive when the ininitive construction with a subject of its own has emerged. If we compare the dative genitive with Vilkuna’s (1996: 161) list of prototypical subject properties in Finnish (cf. Section 1), we ind that a dative genitive complement is also an NP (for a thorough account of the subject in Finnish, see Huumo & Helasvuo, this volume). Since it is a dative genitive, it is not in the nominative case like the prototypical subject. Since verbs only agree with nominative subjects in Finnish, there is no agreement in the case of genitive subjects nor dative genitives.6

6. However, in certain cases a genitive attribute triggers a possessive suix in the head noun. his phenomenon applies to some extent to ininitives and participles and their genitive subjects

he syntactic and semantic history of the Finnish genitive subject 245

Table 2. Dative genitive – prototypical subject properties. ✓ X X (?) ? X ? ✓

Category: NP Case: nominative Agreement: verb agrees with subject Position: neutrally precedes the verb Government: the antecedent of e.g. a bare possessive suix Role: typically agent or experiencer Semantic type: typically animate

As for position, it is not clear that the genitive subject neutrally precedes the verb. In the permissive construction it does typically precede the ininitive, but in the necessive construction, it typically precedes the inite verb – and neither is it totally clear whether it should be interpreted as the subject of the ininitive or the necessive verb. However, as far as the historical dative genitive goes, it has typically preceded the ininitive in sentences expressing ‘giving’ and instantiating the pre-permissive-construction (though oten not immediately). For sentences expressing‘ripening’ and ‘practical necessity’ or ‘obligation’ and instantiating the pre-necessive-construction, we can only speculate, but the dative genitive probably has also preceded the ininitive more oten than not, given what we know of corresponding constructions in modern-day Finnish. Dative genitives are typically not possible antecedents of a bare possessive sufix – or at least present-day allative arguments are not, and we may assume that the same is true of dative genitives. And as for semantic role, the dative genitives are mostly benefactives, but also experiencers to some extent. At any rate, they are not agents like the prototypical subject. On the other hand, they have been interpreted as secondarily expressing the actor of the ininitive in the relevant expressions, and this secondary role seems crucial here. And, as we already saw, they are animate and even human. All in all, dative genitives have at least something in common with prototypical subjects, though not particularly much. 4.2

Path 2: he genitive attribute

As we saw in Section 3, the genitive subject of the agent participle, as well as that of the temporal construction, originates from an earlier genitive attribute of a deverbal noun which has grammaticalized into a non-inite verb form. Overall, the genitive attributes of deverbal nouns tend to be ambiguous, both in Finnish and as well. hus, there are cases where the genitive subject triggers a coreferential possessive suix on the ininitive or participle in a manner highly reminiscent of person agreement. For a discussion of this phenomenon, see J. Leino (2005b: 251–254).

246 Jaakko Leino

in several other languages. hey oten have two possible interpretations, which have been dubbed “subjective” and “objective” in the Finnish grammar tradition, i.e. a genitive attribute may be interpreted as corresponding to either the subject or the object of the base verb. hus, in a sentence such as (20), we may interpret the word koiran either as the eater or the thing being eaten: (20) Koira-n syöminen on nopea-a. dog-gen eating.nom be.3sg quick-par ‘Eating a dog is quick.’ / ‘A dog’s eating is quick.’

Genitive attributes typically express either possession or ownership (although they are not limited to this, of course), and this provides an obvious link to animacy; the referent of a genitive attribute is typically someone capable of possessing: necessarily animate and typically human. Even though this does not, strictly speaking, have to be the case with deverbal nouns, it typically is with the genitive attribute category overall. As compared to prototypical subjects, a genitive attribute is also an NP, but not in the nominative but in the genitive case. As was the case with dative genitives, there is no agreement, but genitive attributes especially tend to trigger possessive suixes (cf. the footnote on p. 428). Genitive attributes tend to precede their head noun in Finnish, so they also precede the deverbal noun. hey also trigger possessive suixes and act as their antecedents. heir role is hardly agentive or experiencer-like overall, except in the case of deverbal nouns when it is, in fact, typical. Finally, as pointed out, they are typically animate. hus, all in all, genitive attributes make surprisingly good candidates for subjects. Table 3. Genitive attribute – prototypical subject properties. ✓ X ? ✓ ✓ ? ✓

Category: NP Case: nominative Agreement: verb agrees with subject Position: neutrally precedes the verb Government: the antecedent of e.g. a bare possessive suix Role: typically agent or experiencer Semantic type: typically animate

4.3

Path 3: Genitive–accusative syncretism

he third historical path leading to the genitive subject involves a genitive–accusative syncretism. In these instances, what is a genitive subject of a non-inite verb today, used to be an accusative object of its matrix verb in previous times. It may seem odd

he syntactic and semantic history of the Finnish genitive subject 247

that an object becomes a subject, but one thing that makes this reanalysis easier in the particular context is the group of matrix verbs involved. he relevant construction discussed in Section 3.4 was the referative construction. he inite verbs that occur in that construction are verba sentiendi et dicendi, which are only marginally transitive. herefore, they do not provide a particularly strong motivation for accusative NP objects. Rather, what these verbs seek are clausal complements – and, indeed, once the previous object is reanalyzed as the subject of the non-inite verb form, the resulting structure is precisely a non-inite complement clause. his reanalysis has, therefore, been well motivated. A second example case of objects becoming subjects is something that was not discussed in any detail in Section 3. As was mentioned in passing, one part of the permissive construction discussed in 3.2 goes back to a previous object control structure in which the predecessor of the present-day genitive subject is the object, rather than the recipient argument, of the matrix verb. In other words, what we are dealing with are sentences such as I commanded you to do something. Such sentences involve a transitivity chain, or a causal chain, which goes from the subject of the inite verb towards the object of the ininitive. And since the actor of the ininitive is in the middle of this chain, it has properties of both a typical subject and a typical object. However, since we are dealing with ‘command’ type of predicates, the relevant referents are necessarily animate and typically human again. When we compare syncretistic objects with prototypical subjects, we need to take into account what kind of objects are being examined. In any case, we can conidently say that syncretistic objects are NPs but not in the nominative case and that there is no agreement between them and the non-inite verb. With regard to position, an informative analysis would require a reconstruction of the basic word order of the predecessors of the referative and the permissive constructions. he permissive construction seems fairly easy: object control structures in Finnish and other relevant languages tend to be ordered so that the object precedes the ininitive that it controls. he referative construction is nontrivial, however; it seems to involve, among other things, a major change in Finnic basic word order (from SOV to SVO), and reconstructing the order of the predecessor of the referative construction would involve so many variables that it would inevitably be reduced to mere speculation. Within the relevant non-inite clause, this kind of accusative object tends to be a possible antecedent of a possessive suix. And as for the role and semantic type, I have already pointed out that in the proto-permissive construction, the object was a part of the causal chain and typically animate. In the proto-referative construction, the object could, at least in principle, be anything that we conceive of or talk or think about.

248 Jaakko Leino

Table 4. Accusative object – prototypical subject properties. ✓ X X ? ? ? ?

Category: NP Case: nominative Agreement: verb agrees with subject Position: neutrally precedes the verb Government: the antecedent of e.g. a bare possessive suix Role: typically agent or experiencer Semantic type: typically animate

5. Conclusion: he emergence of the genitive subject category he Finnish genitive subject is, as we have seen, the result of three originally separate paths or reanalyses. Each has a diferent point of origin with diferent subject properties and contextual factors motivating the reanalysis. What, then, is the genitive subject itself like? What has emerged through these three paths? When we look at the properties of a typical modern-day genitive subject, we have to acknowledge that many but not all of these properties come from historical constructions. To a certain degree, the present-day genitive subject relects the properties of its predecessors: the benefactive or “dative” genitive, the genitive-like accusative object, and the genitive attribute. Yet, not all of its properties come from these three predecessors; the genitive subject acquires a signiicant part of its properties – in many cases notably the semantic role – from the valency of the matrix predicate. Overall, the end result of the historical development, i.e. of the genitive subject, is essentially identical to the nominative subject with two exceptions: 1. By deinition, the genitive subject is not in the nominative case. 2. here is no subject–verb agreement (except for the possessive suix “agreement” briely mentioned above). his is to be expected: the genitive subject is used with verb forms which do not have person-marking morphology and thus cannot show agreement. hese two exceptions aside, the genitive subject is a quite ordinary and typical subject, however. From a somewhat diferent perspective, we might also say that it has become a conventionalized category and that it has acquired enough properties typical for a subject in order to be called a subject. he development of the Finnish genitive subject is a complex case of intertwined reanalyses. As such, it is a diachronic manifestation of the usage-based approach to language (cf. e.g. Langacker 1988; Barlow & Kemmer 2000): in order

he syntactic and semantic history of the Finnish genitive subject 249

Table 5. Genitive subject – prototypical subject properties. ✓ X ? ✓ ✓ ✓ ✓

Category: NP Case: nominative Agreement: verb agrees with subject Position: neutrally precedes the verb Government: the antecedent of e.g. a bare possessive suix Role: typically agent or experiencer Semantic type: typically animate

for a construction to be reanalyzed, it has to be acquired by someone, through its usage, but in a manner which is diferent from the analysis intended by the person(s) that produced that usage. Crucially, what the history of the genitive subject exempliies is the emergence of a category. Apparently, there used to be a time when no such category as the genitive subject existed in early Finnish or proto-Finnic, but in modern Finnish, there clearly is such a category. And, as we have seen, that category has emerged through the reanalysis of quite a few diferent constructions which have included originally diferent kinds of predecessors to the genitive subject. In the light of the existing evidence, we can merely speculate about the manner in which these diferent sources have collapsed together and about what it means for a language to have such a category. Has there been some kind of a “superschema” that arose at some point and brought the three paths together into a uniform “genitive subject”? Have the three paths afected one another during the process? If they have, how have they afected one another? And how should we conceive of interdependencies between diferent constructions in the irst place? What should we think about the internal organization of grammar? What do we mean by “a structured inventory of linguistic signs,” as the standard epithet goes? How is it structured? hese are intriguing questions which we, unfortunately, cannot address within the limits of the present chapter. Yet, the discussion above does seem convincing to the efect that when we look at the emergence of grammatical categories, we need to take both formal and semantic factors seriously; even such basic categories as the subject are, really, in essence clusters of both morphosyntactic and semantic (and conceivably also pragmatic) properties. A proper understanding of such seemingly unproblematic categories requires examining their history and borderlines. herefore, an understanding of what non-canonical subjects are like and how they came about will clarify our understanding of what exactly we mean by the notion of subject.

250 Jaakko Leino

References Barlow, Michael & Kemmer, Suzanne (Eds.). (2000). Usage-based models of language. Stanford: CSLI Publications. Hakulinen, Auli, & Karlsson, Fred (1979). Nykysuomen lauseoppia. Helsinki: Finnish Literature Society. Hakulinen, Auli, Vilkuna, Maria, Korhonen, Riitta, Koivisto, Vesa, Heinonen, Tarja Riitta, & Alho, Irja (2004). Iso suomen kielioppi. Helsinki: Finnish Literature Society. Hakulinen, Lauri (1979). Suomen kielen rakenne ja kehitys. Fourth, supplemented and corrected edition. Helsinki: Otava. Häkkinen, Kaisa (1993). Ongelmoita oppimia: Suomen agenttipartisiipin historiaa ja nykyisyyttä. Festschrit für Raija Bartens. Helsinki: Suomalais-Ugrilainen Seura. 129–148. Huumo, Tuomas (1995). Paikallissijan kieliopillistuminen datiivi-genetiivin funktioon: uralilaisen n-sijan ja itämerensuomalaisen adessiivin kehityksen vertailua. Sananjalka, 37, 55–79. Huumo, Tuomas (2005). Subjectivity and the challenge of aspect marking: he riddle of the Finnish Quasi-resultative construction. Cognitive Linguistics, 16(1), 113–144. DOI: 10.1515/cogl.2005.16.1.113 Karlsson, Fred (1999). Finnish: An essential grammar. London: Routledge. Ikola, Osmo (1959). Eräistä suomen syntaktisista siirtymistä. Sananjalka, 1, 39–60. Ikola, Osmo (1974). Lauseenvastikeoppia. Tietolipas 76. Helsinki: Finnish Literature Society. Inaba, Nobufumi (2000). Genetiivin ja partitiivin datiivinen käyttö vanhassa kirjasuomessa: sijanmerkinnän ja sijajärjestelmän suhteesta. Sananjalka, 42, 47–85. Jaakola, Minna (2004). Suomen genetiivi. Helsinki: Finnish Literature Society. Jahnsson, A. W. (1871). Finska Språkets Satslära. För läroverkens behof. Helsingfors: Utgifvarens förlag. Koivisto, Vesa (2005). Monikasvoinen mA ja suomen agenttipartisiipin tausta. In Ilona Herlin & Lari Kotilainen (Eds.), Elävä kielioppi: suomen ininiittisten rakenteiden dynamiikkaa (pp. 146–172). Helsinki: Finnish Literature Society. Korhonen, Mikko (1991). Remarks on the structure and history of the Uralic case system. Journal de la Société Finno-Ougrienne, 83, 163–180. Laitinen, Lea (1992). Välttämättömyys ja persoona. Suomen murteiden nesessiivisten rakenteiden semantiikkaa ja kielioppia. Helsinki: Finnish Literature Society. Laitinen, Lea, & Vilkuna, Maria (1993). Case-marking in necessive constructions and split intransitivity. In Anders Holmberg & Urpo Nikanne (Eds.), Case and other functional categories in Finnish syntax (pp. 23–48). Berlin: Mouton de Gruyter. Langacker, Ronald W. (1988). A Usage-Based Model. In Brygida Rudzka-Ostyn (Ed.), Topics in cognitive linguistics. Current issues in Linguistic heory 50 (pp. 127–161). Amsterdam: John Benjamins. DOI: 10.1075/cilt.50.06lan Leino, Jaakko (2003). Antaa sen muuttua. Suomen kielen permissiivirakenne ja sen kehitys. Helsinki: Finnish Literature Society. Leino, Jaakko (2005a). Frames, proiles and constructions. Two collaborating CGs meet the Finnish permissive construction. In Jan-Ola Östman & Mirjam Fried (Eds.), Construction grammars. Cognitive grounding and theoretical extensions. Constructional approaches to language (Vol. 3, pp. 89–120). Amsterdam: John Benjamins. DOI: 10.1075/cal.3.05lei

he syntactic and semantic history of the Finnish genitive subject

Leino, Jaakko (2005b). Passiivin ininitiivimuodot suomen murteissa. In Ilona Herlin & Lari Kotilainen (Eds.), Elävä kielioppi: suomen ininiittisten rakenteiden dynamiikkaa (pp. 231– 257). Helsinki: Finnish Literature Society. Leino, Pentti (1986). Ininitiivin asema lauseessa. In Pentti Leino & Jyrki Kalliokoski (Eds.), Kieli 1 (pp. 103–165). Helsinki: University of Helsinki. Leino, Pentti (2001). Henkilöviitteinen allatiivi. In Pentti Leino, Ilona Herlin, Suvi Honkanen, Lari Kotilainen, Jaakko Leino, & Maija Vilkkumaa (Eds.), Roolit ja rakenteet. Henkilöviitteinen allatiivi Biblian verbikonstruktioissa (pp. 456–513). Helsinki: Finnish Literature Society. Penttilä, Aarni (1957). Suomen kielioppi. Helsinki: WSOY. Sands, Kristina, & Campbell, Lyle (2001). Non-canonical subjects and objects in Finnish. In Alexandra Y. Aikhenvald, R. M. W. Dixon & Masayuki Onishi (Eds.), Non-canonical marking of subjects and objects (pp. 251–305). Amsterdam: John Benjamins. DOI: 10.1075/tsl.46 Setälä, Emil Nestor (1880). Suomen kielen lause-oppi. Oppikirjan koe. K. E. Holm, Helsinki. Vilkuna, Maria (1996). Suomen lauseopin perusteet. Helsinki: Finnish Literature Society.

251

From canon and monolith to clusters A constructionist model of subjecthood in Russian Aki-Juhani Kyröläinen University of Turku

he status of oblique subjects is a contested topic. Subjecthood is typically deined relative to syntactic tests in formal approaches and diferent tests have been proposed in the literature. However, they have not been applied systematically to cover a wide range of constructions such as oblique subjects in Russian. his study presents a clustered model of subjecthood building on ten construction types and twenty features of subjecthood in Russian based on construction grammar and Keenan’s (1976) study. he structure of the model indicates that Russian has two clustered subject constructions: the nominative and the dative. hese types motivate the deviations displayed by the other constructions. Hence, subjecthood is deined as a fully motivated network structure.

1.

Introduction

he status of dative subjects, or oblique subjects in general, is a widely debated topic and has received enormous attention in formal approaches to the Slavic languages, speciically Russian. Although formal approaches employ rigorous parameters to establish categories, there is no agreement among these various proposals on the status of possible dative subjects. Russian is an excellent domain for testing the notion of dative subjects by examining a substantial number of impersonal construction types, which lack the traditional deinition of subjecthood in terms of subject-verb agreement found in nominative-accusative alignment. he controversial status of dative subjects is perhaps due to several crucial factors which lead to observable discrepancies in the description. he irst such factor involves the precise target of the proposed syntactic tests. he tests are assumed to target such categories as subject, indirect object, and possibly a dative-subject, and the deinition of these categories precedes the syntactic tests. Secondly, the tests are typically not applied with equal coverage to their possible targets. hirdly, not all tests are considered to be equal in teasing apart subjects doi 10.1075/cal.16.10kyr © 2015 John Benjamins Publishing Company

254 Aki-Juhani Kyröläinen

from non-subjects; no criteria, however, have been proposed to determine the alleged inequality between diferent tests. Finally, it is possible that not all datives are equal. (Barðdal 2006; Barðdal & Eythórsson 2003; Crot 2001; Eythórsson & Barðdal 2005). hese four factors can be considered to contribute, at least in part, to the status of subjecthood in Russian. In this study, a cluster-based approach to subjecthood is posited, allowing us to simultaneously take into consideration a number of factors which are not commonly discussed jointly in previous studies on subjecthood: (1) (dis)similarity among diferent constructions; (2) the evaluation of possible subject candidates; (3) the status of the proposed tests; (4) a possible hierarchy among diferent features; and lastly, (5) a network model of subjecthood in Russian based on the inheritance of features. he adopted approach utilizes a multivariate statistical method.his kind of feature-based model of subjecthood is absent from the Russian tradition (cf. Leinonen 1985: 14; for discussion, see Section 2). I consider ten constructions, along with twenty features forming the basis of the clusters. he construction types are discussed in Section 4 and the proposed set of features is outlined in Sections 5–7. he proposed network structure of the construction types is given in Section 8. Finally, a hierarchy of subjecthood is established in Section 9 building on Keenan’s (1976) clustered concept of subjecthood and on Dowty’s (1991) principle of Proto-Role.1

2. Outlook on subjecthood in Russian he argument supporting the status of oblique subjecthood is complicated by a lack of commonly accepted criteria for validating the status of certain arguments as subjects. According to the traditional deinition, subject status is deined in terms of case-marking, i.e. the nominative case, and subject-verb agreement in number, person and gender. In the Russian linguistic tradition subjecthood is commonly viewed as a two-tiered system, consisting of a semantic subject, sub’ekt, and a grammatical subject, podležaščee. his system allows the incorporation of oblique arguments of various impersonal constructions as pertaining to semantic subjecthood. (Leinonen 1985; Švedova (ed.) 1982b: 94–95; Zolotova 2000). Rigorous deinitions of these two categories, however, are lacking. hus most theories on argument structure posit additional syntactic tests to determine 1. I am grateful to two anonymous reviewers for their comments and suggestions and to Mikhail Voronov for discussions on grammaticality and acceptability. I would also like to thank Marja-Liisa Helasvuo, Tuomas Huumo and Riitta Pyykkö for commenting on earlier versions of this article.

From canon and monolith to clusters

subjecthood (cf. Barðdal 2006). Zimmerling (2009: 254–255, 264) posits a tiered system for subjecthood in Russian: strong and weak subjects. Zimmerling conines the tiered subject structure to syntactic tests, stipulating that a strong subject is capable of displaying control-and-binding features, while in the case of a weak subject these parameters vary depending on the precise syntactic type in question. he dative in the so-called dative-ininitive construction, consisting of a dative argument and an ininitive, is a strong subject, while the remaining impersonal types are weak. On the other hand, Moore and Perlmutter (2000) acknowledge that the dative-subject in the dative-ininitive construction occupies a subject position (see also Perlmutter & Moore 2002). However, they maintain that the other impersonal types are I(nversion)-nominals; in other words, they occupy the subject position in the underlying structure but at the surface structure functioning as indirect objects (cf. Greenberg & Franks 1991). In contrast, Sigurðsson (2002) considers eleven features and denies the status of oblique subjects altogether in Russian, by positing two underlying zero structures which then can be used to derive the grammaticality of the syntactic tests. his brief survey shows that even if we ignore theory-internal discussion, any comparison of diferent approaches is diicult to sustain. hus the only factor that can be compared among them is the outcome of the syntactic tests. Moore and Perlmutter (2000) posit three explicit criteria for subjecthood in Russian, i.e. control, raising and agreement; they exclude linear order and the ability to antecede relexives. In sum, the studies apply diferent constructions, yielding conlicting results. In addition, the same tests are not applied across diferent construction types, leading to inconsistencies between possible subject candidates.

3. Constructions, syntactic roles and clustered structures The notion of constructions figures prominently in cognitively oriented approaches. Constructions are understood as form-meaning pairings, and are also considered to be the primitive units of language (Crot 2001; Goldberg 2006; Kay & Fillmore 1999; Langacker 2009b). Goldberg deines constructions as formmeaning pairings which are not strictly predictable from their components or from the other constructions in a language. Predictable patterns also constitute constructions, as long as they appear with suicient frequency. (Goldberg 2006: 5). his view posits that all linguistic units, from morphemes to argument patterns, are constructions; the only diferences between these units lie in their schematicity and complexity (Crot & Cruse 2004: 555; Goldberg 2006: 5). Radical Construction Grammar goes even further, denying the status of grammatical relations. Traditional concepts, such as subject, object and indirect object, are derived

255

256 Aki-Juhani Kyröläinen

from constructions as a whole. In this vein, syntactic categories are viewed as roles constituting a part-whole relationship; for example subject is deined in relation to the construction as a whole and not to the predicate (part-part relation). Traditionally, subjecthood is determined based on syntactic tests. However, an analysis based on features is faced with at least two issues: the nature of the features employed and their number (Crot & Cruse 2004: 10–11, 88–89). In most formalist approaches, the syntactic tests, i.e. features, are used to determine the status of subjecthood. he category of subjecthood is established and deined in terms of n-features, which are by deinition necessary and suicient. hus the category of subjecthood is deined in relation to itself. Although the tests are independent of the status of subjecthood, the very meaningfulness of their implementation is based on the pre-established category which they are supposed to deine; otherwise they would fail to satisfy the requirement of necessary and suicient features. his is also indirectly relected by the labels chosen for the test, such as “raising to subject” and “raising to object”. his circularity can be avoided by adopting the position of Radical Construction Grammar with regard to subjecthood. In a construction grammar approach, the various syntactic tests proposed can be understood as constructions, simply displaying patterns of combinability and usage, and not as relections of any particular syntactic status. he notion of subject, however, is here let undeined: what kind of nominal can occur in the subject slot in a given construction? More precisely, what kind of nominal qualiies as a subject in a given construction type? (cf. Bhaskararao & Subbarao 2004: x) Here I combine a model based on Keenan’s (1976) concept of the clustered structure of subjecthood with the proto-role principle formulated by Dowty (1991). he insight of Keenan’s proposal lies in the fact that subjecthood does not originate in a single monolithic linguistic dimension. Rather, it is a multi-faceted concept, that of clustered structure. Keenan’s proposal consists of +30 features, grouped into three kinds of properties: coding properties, behavior properties, and semantic properties (Keenan 1976: 311, 324). In the spirit of this tripartite structure, the following labels are proposed for the clustered structure of subjecthood: Argument properties, Indexicality and Schematicity. he category of Argument properties contains features which capture the structure of a given slot in a construction. he notion of Indexicality follows Crot’s (2001: 247) statement that agreement morphemes index discourse referents. hus Indexicality covers constructions whose basic function is to track discourse referents. In formal approaches these are typically labeled as control and binding operations. he last category is intended to mirror Barðdal’s (2006: 94) observation that Icelandic and German oblique subject constructions difer in schematicity. Together the labels are used to capture twenty features relevant to subjecthood in Russian. hey are based on previous studies of subjecthood

From canon and monolith to clusters 257

in Russian in general (Barðdal 2006; Keenan 1976; Moore & Perlmutter 2000; Perlmutter & Moore 2002; Zimmerling 2009). he proposed features are deined and exempliied in Sections 5–7. Dowty proposes a prototype-based structure for subject and object selection based on ive Proto-Agent and ive Proto-Patient features. he proto-role account builds on the principle that not all features are necessarily present in any given type. hus subject/object selection is based on the highest number of features present in a given argument slot. (Dowty 1991: 572–579). his principle is appropriated in this study for possible subject selection. he absence of a particular feature does not necessarily entail exclusion from possible subjecthood status. his makes it possible to account naturally for the fact that not all constructions are equal. Constructions have a prototype structure, which does not pertain to minimal pairs. Lastly, another advantage follows from Dowty’s proto-role account, namely that certain features take precedence over others. Pairings implicitly impose structure on the semantic relation they are set to deine; a Proto-Agent, for example, causes an event, whereas a Proto-Patient undergoes a change of state (Dowty 1991: 574; Koenig et al. 2003: 83). While the feature set proposed in this study is not a priori ordered, nor does it come in a paired form, I demonstrate in the inal section that an ordering emerges naturally from the interaction of constructions and features.

4. Constructions deined In this paper, the discussion of the status of subjecthood is limited to ten constructions. he basic principle for selecting these particular constructions lies in their overlapping semantic structures and their generality; in other words, they are typically used to establish the basic argument construction types in Russian. I adopt the notation in Divjak and Janda (2008), in order to maintain compatibility between these two approaches to subjecthood in Russian. Janda (2008) posits the transitive construction as the basic predicate construction type in Russian, following Hopper’s and hompson’s (1980) semantic deinition of transitivity. his typicality is achieved through comparison with the canonical event model. Langacker (1991: 283–284) deines the model as follows: “a person who volitionally initiates physical activity resulting, through physical contact, in the transfer of energy to an external object”. his produces an inherently asymmetrical relation, as relected in the transitive construction in Russian, yielding a prototypical n(ominative) v(erb) a(ccusative) pattern. he semantic deinition of transitivity allows the establishment of a smooth continuum from prototypical transitive construction to intransitive construction, the prototypical

258 Aki-Juhani Kyröläinen

form of which is n(ominative) v(erb). he basic transitive construction can be extended by adding a dative argument, yielding a n(ominative) v(erb) a(ccusative) d(ative) pattern corresponding to a ditransitive construction. hese three constructions are the typical personal construction types in Russian. he modal constructions show a divergent alignment pattern in Russian, displaying both personal and impersonal types. Additionally, the predicate slots of these constructions are deviant compared to the typical personal types, consisting of adjectival, adverbial and defective verbal predicates (cf. Leino on Finnish ininitival constructions in this volume). he personal types take two forms: n(ominative) adj(ective) inf(initive) and d(ative) adj(ective) n(ominative). he predicate dolžen ‘must’ exempliies the former type, nužen ‘necessary’ the latter. he impersonal types are d(ative) v(erb) inf(initive) and d(ative) adv(erb) inf(initive). he verb prixoditsja ‘have to’ proiles the irst type and the adverb nado ‘need’ the latter. he semantic content of these modal constructions is not important for the status of possible subjecthood; for a recent discussion on Russian modal semantics cf. Divjak (2009) and Fortuin (2007). he d v inf construction is labelled as d + v by Janda and Divjak (in this volume). he last group of constructions considered in this study consists of impersonal constructions. he d(ative) inf(initive) construction may be considered as an alternation of the canonical event model grounded through the transitive construction in Russian. Both transitive and intransitive verbs can appear in the predicate slot. hus, the minimal construction consists of a dative argument and an ininitive. he construction is used to express various shades of modality depending on the context and the exact proile of the construction (cf. Fortuin 2005; Fortuin 2007; Halizeva 1969). Greenberg and Franks (1991: 91) argue that the dative cannot be overtly expressed under certain conditions, illustrated in (1). Consistent with their analysis, Zimmerling (2009) also considers similar patterns to pertaining to the d inf type by positing a silent pro-drop (cf. Moore & Perlmutter 2000: 387). Contrary to this mode of analysis, it is claimed in this study that these patterns are not instances of the d inf construction. he reason for this sudden ungrammaticality is already present in the manner of formatting the example. hese patterns are analyzed as instances of another impersonal type, namely d(ative) adv(erb) in this study. his type is used to proile a number of state expression such as physical and mental among others. his type can be conjoined with an ininitival complement. (1) Mne važn-o [(*vam) žit’, odn-omu]. 1sg.dat important-adv (*2pl.dat) live-inf alone-dat ‘It is important for me *for you to live alone.’ (Greenberg & Franks 1991: 91)

From canon and monolith to clusters 259

he last type is another state expression in Russian. he form consists of a dative argument, a modiier, and a verbal predicate with the relexive marker (-sja) as in (2). (Cf. Kyröläinen 2008 and references therein). (2) Как vam rabota-et-sja s prem’er-om? How 2pl.dat work-3sg.prs-rm pr Prime.Minister-ins ‘How is the work going with the Prime Minister?’ (RNC, Vladislav Starkov. Sovetnik prezidenta. Argumenty i fakty, Moskva 2001.04.04)

his type is marked with the abbreviation d v[sja] in this study. In the Russian tradition, the construction is also considered to express modality (cf. Geniušienė 1987: 273–274). In sum, this section introduced the construction types investigated in this study. In the subsequent sections (5–7), the properties of these construction types are considered in terms of subjecthood in Russian.

5. Features of Argument-properties 5.1

Nominative case and subjecthood

Traditionally, the most prominent feature associated with subjecthood in languages following the nominative-accusative alignment pattern is the nominative case, and this is the feature acknowledged by the Russian Grammar (Švedova (ed.) 1982a: 480–481). Case marking generally does not involve reference tracking within or across clauses, but may be considered to be conined to referentiality and to the roles of participants in the speech event (Silverstein 1976). hus case marking emerges for less prototypical referents in terms of the canonical event model. Nonetheless, the distribution of the nominative case among the constructions simply displays the traditional division of personal versus impersonal construction types in Russian. he nominative case marking is found in ive constructions: n v a, n v, n v a d, n adj inf and d adj n. he remaining constructions, according to the Russian linguistic tradition, are thus impersonal. 5.2

Animacy and subjecthood

he second feature of Argument properties is animacy, which igures prominently throughout linguistic theories of syntactic relations. Animacy can be regarded as a conceptual feature connecting syntactic and semantic roles: certain semantic roles are more oten linked to the status of subjecthood, while at the same

260 Aki-Juhani Kyröläinen

time certain syntactic roles are more closely associated with animacy. Bock et al. demonstrate that animacy is a determining factor in subject/object selection in English: animate subject arguments occur more oten with verbs that allow both animate and inanimate subjects, contrary to object selection. hey also show, on the basis of a priming test with the passive construction, that the status of a subject is best understood as a direct mapping between arguments rather than an underlying structure. (Bock, Loebell & Morey 1992: 154–159, 162). Crucially, this can be taken as indirect evidence in favor of the syntactic role position.2 he signiicance of the feature ‘animacy’ stems from the canonical event model. hus animacy igures prominently in the deinition of the n v a construction. Analogously, the n v a d construction is deined in terms of a transfer from one animate being to another. (Crot 2001: 159–164; Langacker 1991: 306–307). However, the semantic content encoded by the N V construction varies greatly, from controlled activities (begat’ ‘run’) to temporal states (dlit’sja ‘persist’), with various intermediate constructions such as bodily actions (povoračivat’sja ‘turn oneself ’) and uncontrolled activities (spat’ ‘sleep’). he modal constructions do not display a uniied case-marking pattern in encoding the entity undergoing the proiled modal force. he n adj inf construction proiles the entity undergoing the modal force in the nominative case, while in the other three constructions the entity is marked with the dative case. he structure of the d adj n construction can be seen to approximate a pattern typically associated with mental verbs. he entity undergoing the modal force, the modal experiencer, directs his/her attention to the secondary entity in the proiled relation, which can be seen as the stimulus. (On mental verbs cf. Crot 1998: 31, 2001: 155–156; Dowty 1991: 579–580, 586–587). he semantics of d adv and d v[sja] constructions are strongly connected to the experiencing of various mental states. he d v[sja] construction, however, can be used with inanimate NPs through metaphorization, as in (3). (3) Čem dal’-še gosudarstv-o ot rynk-a What.ins far-cmpr state-nom pr market-gen tem leg-če živ-et-sja biznes-u that.ins easy-cmpr live-3sg.prs-rm business-dat ‘he farther away the state is from the market, the better it is for the business.’ (RNC, Marina Borisova. Glavnaja problema Rossii – nepredskazuemost’ vlasti. Nezavisimaja gazeta 2003.05.27) 2. he rationale behind the priming test is that if the subject slot is indeed a byproduct of an underlying structure, it should show a priming efect with the underlying object-arguments, i.e. a priming efect with the surface objects of the active and the surface subject of the passive. According to the direct-mapping hypothesis, the priming efect should follow the animacy of the primes and not of the underlying structure.

From canon and monolith to clusters

(4) Universitet-u – ras-ti i procveta-t’! university-dat – grow-inf and thrive-inf ‘he university has to grow and thrive.’ (invocation) (Zimmerling 2009: 259)

he d inf construction can also be used with inanimate NPs, as shown in (4). Nonetheless, the ability to contain an inanimate NP in the dative slot does not constitute an ainity to inanimacy in usage-based models. his is a direct contrast to the notion of speciication in formal approaches; Zimmerling (2009: 259), for example, claims that the d inf is not speciied as [+animate]. hus none of the subject candidates show a clear discrepancy in animacy. 5.3

Linear order and subjecthood

Another factor oten considered in determining subjecthood is linear order in the clause, which is taken to relect the conceptual structure, following the principle of iconicity. A fronted element is considered to possess a higher level of saliency. Russian does not have rigid linear word-order rules; rather, it is best characterized in terms of information structure (Švedova (ed.) 1982b: 91–92; cf. also Helasvuo and Huumo in this volume). Consequently, the null hypothesis for the linear order of the possible subject candidates is thus the pre-predicative position in Russian.3 here are hardly any quantitative studies of linear order in Russian including a comprehensive set of diferent construction types. Sirotinina’s (2006 [1965]) study of linear order in Russian is still perhaps the most exhaustive in terms of variation, number of diferent construction types, and quantitative data. he salience of the irst position becomes apparent with nonsense sentences. Additionally, the shape of the lexical item inluences selection in terms of subjecthood. Sirotinina reports an experiment in which 48 participants were asked to identify the subject in the following nonsense sentences: (1) Kuzdru bokra budlanula, (2) Kuzdra bokra budlanula, (3) Vogra budlanula tydra, (4) Budlanula tydra vogra. he shape of the nonsense words was found to be one factor inluencing subject choice, along with linear order. he shape of the suix -u can be interpreted as the accusative case, i.e. the direct object, whereas the shape of the suix -а is interpreted as the feminine nominative. Lastly, the shape -la points to a feminine past tense form and might thus be considered to function as a predicate. Not surprisingly, the “word” bokra in the irst example was chosen as the subject by 43 participants. For the second and the third example, the fronted “word” was chosen as the subject by 42 participants 3. he label ‘pre-predicative’, rather than ‘preverbal’, is used in this study to cover the modal constructions.

261

262 Aki-Juhani Kyröläinen

in both cases. he last example does not contain any clues in terms of shape or linear order. hus the results are as expected: 29 participants chose tydra and 16 vogra. (Sirotinina 2006 [1965]: 95–96). With regard to linear order, possible subject candidates in the declarative main clause should prototypically occur in the pre-predicative position. (Keenan 1976: 319). It has been claimed that it is this feature that separates the n v a d and the d adv constructions from the remaining types. his claim is further supported by Sirotinina’s (2006 [1965]: 112, 118–121) indings on impersonal, modal and copular constructions in written Russian. Her data consist of 47 lexical items. he results, however, are not divided according to clause type. he observed regularities in linear order are nevertheless suiciently constant to postulate that in modal constructions the dative appears in the pre-predicative position regardless of the clause type: dolžen ‘must’ (99.8%), možno ‘it is possible’ (99.4%), nado ‘ought’ (99.4%), nužno ‘should’ (99.4%) and nel’za ‘it is impossible’ (88.8%). he most interesting part of Sirotinina’s indings, however, is that concerning the d adv construction type. his construction type seems to favor the post-predicative position over the fronted one, as clearly demonstrated by such predicates as stydno ‘shame’ (20.0%) and žal’ ‘pity’ (14%). At the same time there is lexical variation in preferred linear order: for example stranno ‘strange, odd’ (53.4%) and trudno ‘diicult’ (52.0%) indicating that these predicates can appear in either position. However, the majority of lexical items in Sirotinina’s data which can be grouped under this construction type favor the post-predicative position. Her indings seem to indicate that the d adv construction does not have a clear preference and that linear order is closely tied to the information structure of the discourse. I follow the standard assumption that the dative argument in the n v a d construction favors the post-predicative position. Quantitative studies for the remaining constructions are, at least to my knowledge, missing; I assume, following the null hypothesis, that they follow the standard pre-predicative position. 5.4

Autonomy and subjecthood

Autonomy is another concept oten evoked to deine subjecthood. However, the concept of autonomy in Keenan’s account is a not a single feature; rather, it is perhaps best understood as a clustered concept, in the sense that he considers autonomy to consist of multiple features, including independent existence, indispensability, autonomous reference and various deletion operations. (Keenan 1976: 312–320). Autonomy is likewise a feature among Dowty’s proto-role selection: the Proto-Agent exists independently of the event, whereas the Proto-Patient is dependent on it or does not exist at all (cf. Primus 1999: 266–268). his feature

From canon and monolith to clusters 263

is oten evoked, but no exact criteria are ofered to establish the independence of a given element. Gerritsen (1990: 34) views the autonomy of a possible subjectreferent as a testable feature. He proposes that autonomy can be teased apart in Russian by means of so-called semi-predicates sam ‘alone’. Typically, the semipredicates index the instigator of the event. he semi-predicates are typically used as a test of subjecthood in terms of agreement. his would lead to deining autonomy at least in part in relation to agreement. In this study, semi-predicates are used as a separate property discussed in Section 6.2. In Cognitive Grammar, the concept of autonomy is typically associated with transitivity. he subject of the transitive construction is the instigator of the proiled event and in this sense independent of the causal chain, while the object is dependent on it, thus yielding an asymmetric alignment: autonomous versus dependent (A/D) (Langacker 1991: 282–283). Independence is clearly indicated by the presence of verbs pertaining to the domain of creation (sozdavat’ ‘create’) or destruction (razrušat’ ‘destroy’). he manifestation of the object of creation is dependent on the force which brings the event about. I will argue that all the possible subject candidate slots in the constructions discussed here pertain to autonomy, except for the dative arguments in the n v a d and d adj n constructions. he n v a d construction already has a slot reserved for the independent argument, namely the nominative. Compared to other constructions the d adj n construction is problematic; in fact, due to its semantic structure it resembles a chicken-and-egg problem. Example (5) illustrates this construction. Additionally, the subject (vrač ‘doctor’) loses its referential value and is construed as a type speciication, in the Langackerian sense (Langacker 1991: 68, 2009b: 119– 120), rather than a fully individualized entity. In contrast, the referent of the dative argument is fully referential, as demonstrated by its ability to be co-indexed with the omitted nominative argument of the imperative construction. (5) Nu, vrači namj i nuž-en, nu Well doctori.nom.m 1plj.dat also need-nom.m well __*i/j nes-i-te её sjuda. carry-imp-2pl 3sg.acc here ø*i/j ‘We need a doctor at this very moment, well, bring her here.’ (RNC, Ordinamenti (2004). Èkran i scena 2004.05.06)

he dative argument may be freely omitted as long as it is grounded in the speech event. In contrast, the nominative argument is close to an obligatory argument. Hence, it can be considered autonomous.

264 Aki-Juhani Kyröläinen

5.5

Subject-predicate Inversion construction and subjecthood

he subject-predicate inversion construction, more simply ‘inversion’, refers to a change in the prototypical linear order of arguments when an inverted constituent, typically a temporal or spatial one, is foregrounded and appears in the fronted position. his feature is typically used as a test for subjecthood, especially in languages with ixed linear order structure, albeit rarely in Russian (Barðdal 2006; Eythórsson & Barðdal 2005; cf. Paducheva 2010). he test is supposed to be able to identify the subject of a construction, assuming that the default order follows the linear pattern of subject–predicate. By introducing a fronted element, the prototypical linear order of the arguments is changed and the subject appears in the post-predicative position, yielding an inversion construction. Example (6) illustrates a prototypical intransitive construction with a locative adjunct. Fronting of the locative adjunct typically triggers the subject-predicate inversion construction, exempliied in (7). Paducheva shows that the inversion construction in Russian may also be accompanied by a number of changes in intonation, especially with the location and type of the phrasal stress. In (6), the locative adjunct has a falling stress, indicated by the backslash, while in the fronted position it acquires the rising stress as indicated by the slash. (Paducheva 2010; cf. Švedova (ed.) 1982b: 92, 201–202). he rising stress pattern is also present in complex inversion structures as in Example (8). (6) Moj brat / | živ-et v Kazan-i \. 1sg.poss.nom brother.nom live-3sg.prs pr kazan-prep ‘My brother lives in Kazan.’ (Paducheva 2010: 117) (7) V Kazan-i / | živ-et moj brat \. pr kazan-prep live-3sg.prs 1sg.poss.nom brother.nom. ‘Kazan is where my brother lives.’ (Paducheva 2010: 117) (8) Teper’ pisa-t’ za nego roman / | prixod-it-sja Now write-inf pr 3sg.acc novel.acc must-3sg.prs-rm mne \. 1sg.dat ‘Now I must write a novel for him.’ (RNC, Aleksandr Pjatigorskij. Vspomniš’ strannogo čeloveka 1997)

In terms of stress shit the inversion construction is most typically attested in the n v construction, and is less likely to occur with the n v a construction (Paducheva 2010: 115–116). Paducheva demonstrates that the shit in stress is also a property of the inverted constituent. Moreover, the possible absence of the shit in the transitive construction raises the question whether the shit in stress is a property of

From canon and monolith to clusters 265

the inversion construction or a vestige of something else, especially when the shit is not applicable to every type of inverted constituent. As noted by Paducheva, the inversion element does not always acquire a rising stress. hus this shit is not considered as a feature of the inversion construction in this study. Sirotinina’s study of linear order ofers an alternative perspective on the inversion construction in Russian. he almost totally ixed linear order with the n adj inf and d adj inf types show that the inversion construction is not prototypically associated with them. Secondly, the dative argument in the n v a d construction prototypically occurs in the preverbal position and is not afected by the inversion construction. Rather, fronting the dative argument can itself trigger inversion. Based on this, these three construction types do not typically undergo inversion in Russian.

6. Features of indexicality 6.1

Agreement and subjecthood

Agreement is taken as one of the fundamental criteria to establish the status of subjecthood. According to the canonical deinition, only the subject argument triggers agreement with the predicate. In Russian, the canonical agreement pattern is attested in person, number and gender. As the division between personal and impersonal construction types is determined by agreement, the classiication of the constructions is relatively straightforward. he subject candidate triggers agreement with the predicate in the following constructions: n v a, n v and the two modal constructions (n adj inf and d adj n). he d inf construction is a special type among these constructions. For obvious reasons the ininitival form itself does not show indexing properties, but it can be conjoined with the passive past participle construction. his augmented proile of the construction displays agreement, explicated in (9). he past passive participle agrees with the dative argument in the d inf construction both in case and gender. (9) Toj rukopis-i ne by-t’ opublikova-nn-oj […]. hat.dat.f manuscript-dat.f neg be-inf publish-ppp-ins.f ‘It is not in the cards to publish that manuscript.’ (Moore & Perlmutter 2000: 389)

he remaining dative arguments do not trigger agreement with the predicate.

266 Aki-Juhani Kyröläinen

6.2

Semi-predicates and subjecthood

Another form of agreement is demonstrated in Russian by the so-called semipredicates sam ‘alone’ and odin ‘one’. he semi-predicates have a full paradigm, consisting of six cases, both singular and plural form, and agreement in case with their antecedent. hey are commonly used as a test for subjecthood in formal approaches (Moore & Perlmutter 2000; Zimmerling 2009). Examples (10), (11) and (12) demonstrate the use of semi-predicates with the impersonal construction types. (10) Tem bolee čto emui sam-omui ne rabota-l-o-s’. hat.ins more that 3sgi.dat alonei-dat neg work-pst-n-rm ‘All the more reason that he was unable to work alone.’ (RNC, G. JA. Baklanov. Žizn’, podarennaja dvaždy 1999) (11) Vас-еi by-l-o trudn-o reši-t’ èt-u namei-dat be-pst-n diicult-adv solve-inf this-acc zadač-u sam-omui/ odn-omui. assignment-acc himselfi-dat alonei-dat ‘It was diicult for Vasa to solve this assignment by himself/alone.’ (Zimmerling 2009: 257)4 (12) Vot i prixod-it-sja mnei sam-omui exa-t’. Well and must-3sg.prs-rm 1sgi.dat alonei-dat travel-inf ‘Well, I must travel alone.’ (RNC, Aleksandr Rekemčuk. Mal’čiki 1970)

However, it is unanimously accepted that in impersonal constructions the semipredicates can be conjoined with the dative argument (Greenberg & Franks 1991; Moore & Perlmutter 2000; Perlmutter & Moore 2002; Zimmerling 2009). (13) Saš-ai da-l emuj sam-omu*i/j/ namei-nom give-pst.m 3sgj.dat alone*i/j-dat/ odn-omu*i/j bilet-y. one*i/j-dat ticket-acc.pl ‘Saša gave him alone/only him tickets. (…the tickets to him alone/only to him)’

In (13), the semi-predicates are conjoined with the dative argument in the n v a d type. he semi-predicate odin ‘one’ adds a strong distributive reading but it is still acceptable. hus the semi-predicates do not display any discrepancy between the nominative and the various dative arguments examined in this study. 4. Zimmerling analyzes Example (11) as an instance of the d inf construction. he ininitive rešit’ ‘solve’ has a silent pro-drop. In the present study these are analyzed as instances of the d adv construction with the ininitive as complement.

From canon and monolith to clusters 267

6.3

Instantiation and subjecthood

he concept of instantiation is used to analyze usage patterns with various omitted arguments based on Crot (2001), Fillmore (1985) and Kay and Fillmore (1999). I follow Crot’s argumentation that instantiation is always a property of a construction and constitutes a symbolic relation between syntactic and semantic elements. Crot posits four types of referent encodings, but only three of them are relevant to the status of subjecthood in Russian. Nonnull instantiation is the baseline category; it requires that the referent must be accessed in accordance with the information provided in the grounded speech event, and is overtly expressed. he second category is indeinite null instantiation: this is an instantiation of the referent without any overt expression, with indeinite status of the referent. he third type is deinite null instantiation, the opposite of indeinite. he referent is grounded in the speech event and must be accessed. (Crot 2001: 276–277). Fortuin (2006: 321–322) observes that the modal meaning of the d inf construction is not tied to the presence of the dative argument. he dative can be overtly proiled, exemplifying a nonnull instantiation, as in (14). In the absence of the dative, as in (15), the modal meaning of the construction is still intact, the diference thus lying in the instantiation type, i.e. the indeinite null. (14) Temne-et, – dal’-še mne ne proj-ti. Darken-3sg.prs – far-cmpr 1sg.dat neg go-inf ‘It is getting dark, I cannot go farther.’

(Fortuin 2006: 322)

(15) Temne-et – dal’-še __ ne proj-ti. Darken-3sg.prs – far-cmpr ø neg go-inf ‘It is getting dark, one should not go farther.’

(Fortuin 2006: 322)

he possibility of deleting an argument in a construction is oten considered to be a feature of subjecthood (Barðdal 2006: 48). However, Russian is not a prototypical pro-drop language compared to such languages as Spanish and Italian.5 Furthermore, the Russian linguistic tradition attributes the possibility of omitting the subject candidate to variation in genre (cf. Zdorenko 2010 and references therein). Example (16) demonstrates an instance of indeinite null across clause boundaries with the d v[sja] and d adv inf constructions.

5. Variation is noticeable even the between Slavic languages. Based on Seo’s (2001) corpus study, Russian is the most conservative among the Slavic languages in the sample, with 22% omitted subjects; the igures for the other Slavic languages varied between 80%–90%.

268 Aki-Juhani Kyröläinen

(16) V molod-ye god-y, kogda __i na pr young-acc.pl year-acc.pl when øi pr mest-e ne sid-it-sja i __i objazatel’n-o place-prep neg sit-3sg.prs-rm and øi necessary-adv nado pokinu-t’ svoji dom, […]. need abandon-inf possi.acc home.acc ‘In youth, when one is unable to sit still and necessarily needs to abandon one’s own home.’ (Asar Èppel’. Na trave dvora 1992. Sel’skaja nov’. 01.01.2010)

However, Perlmutter and Moore argue that the d inf construction cannot appear in a pro-drop. his is considered to be a necessary and suicient feature to separate d inf from other dative arguments (Perlmutter & Moore 2002: 632–633). Examples (17) and (18) show the basis of their argumentation. In the constructionist interpretation, the crux of ungrammaticality is claimed to be in the instantiation type of the null. According to their description, Example (18) can be interpreted only in terms of an unspeciied human referent but not as a deinite null, which would correspond to the pro-drop. (Cf. Perlmutter & Moore 2002: 632–633). (17) Mne ne sda-t’ èkzamen. 1sg.dat neg pass-inf exam.acc ‘It is not (in the cards) for me to pass the exam.’ (Perlmutter & Moore 2002: 632) (18) *__ ne sda-t’ èkzamen. ø neg pass-inf exam.acc

(Perlmutter & Moore 2002: 632)

Certainly, it is possible to envision a fragment of discourse where Example (18) can have a deinite reading. However, wrangling over on the placement of the asterisk is not needed. he d inf construction can be used in fronted position in a clause acquiring a speciic discourse function, namely a discourse-anchoring element which grounds the whole speech event (cf. Langacker 2009b: 250–251). Example (19) illustrates the usage of the d inf in this function; moreover, the d inf construction can only have a deinite null instantiation. (19) __i Spa-t’ __i/ ne spa-l-o-s’, i Žen-ei øi Sleep-inf øi/ neg sleep-pst-n-rm, and namei-dat zaxote-l-o-s’ bulk-i s varen’-em. crave-pst-n-rm roll-gen pr jam-ins. ‘Sleep, sleep did not come and Ženja began to crave a roll with jam.’ (RNC, Arkadij Gajdar. Timur i ego komanda 1940)

In (19), the overtly proiled dative argument is indexed with the omitted arguments in the d inf and d v[sja] constructions. However, when the d inf construction is used in the fronted anchoring position, the dative argument cannot be overtly

From canon and monolith to clusters 269

expressed without involving a high degree of semantic anomaly. Furthermore, this usage pattern comes with its own restriction, which might be considered as a form of reduplication. he fronted anchoring function requires that the following predicate is a duplicate of the ininitive. Instantiation is thus typically associated with the various impersonal dative arguments. In contrast, the dative in the n v a d is the exceptional case. he meaning of the construction is proile-determining and requires the dative argument to be overtly encoded. 6.4

Coordination construction and subjecthood

he second feature of Indexicality is coordination, or – as it is labeled in formal approaches –conjunction reduction. Coordination is a typical feature of subjecthood and in formal approaches it is linked to the notion of control. he tests are used to weigh the argument’s ability to control the omitted subject in the second conjunct. (Eythórsson & Barðdal 2005: 842; Sigurðsson 2002: 693). hus coordination construction is a speciic type of instantiation; a form of complex sentence whose basic semantic function is to link two events. he precise linking function can vary in range, for example from causal to sequential, and the coordination construction is known to elude exact deinition. However, Wierzbicka reasons that not everything is naturally joinable. She further argues that only two events which the speaker is able to perceive as a singular event can be proiled as a coordination construction. (Wierzbicka 1980: 246–254). hus the motivation behind the coordination construction as a possible test for subjecthood is established. If the conceptual basis of coordination lies in the ability to perceive two events as conjoined in a single gestalt whole, then the primary entities in the events thus perceived must of necessity possess a similar identity if either one of them is omitted. In Russian, the coordination construction is typically proiled with the conjunction i ‘and’ (where overtly expressed). It is commonly accepted that various (impersonal) dative arguments can be used in coordination. Additionally, dative arguments can be used to coordinate across diferent construction types, as exempliied in (20) by d v[sja] and d v inf. (20) No __i ži-t’ s ètim zabolevani-em But øi live-inf pr that.ins disease-ins trudn-o, __i prixod-it-sja vo mnog-om diicult-adv, øi must-3sg-rm pr many.prep sebjai/j ograničiva-t’. selfi/j.acc limit-inf ‘To live with this disease is diicult; one has to limit oneself in many things.’

270 Aki-Juhani Kyröläinen

In (21), the dative argument of the d v[sja] is the only overtly encoded subject candidate. If the dative argument were unable to be combined with the omitted referents of the n v constructions, a stark mismatch in referentiality would occur. (21) Ne spa-l-o-s’ v ètu noč-’ knjaz-jui, neg sleep-pst-n-rm pr this.acc night-acc princei-dat __i xodi-l iz ugl-a v ugol: øi walk-pst.m pr corner-gen pr corner.acc __i obdumyva-l put-’ na Kiev. øi ponder-pst.m journey-acc pr Kiev.acc ‘he prince could not sleep that night and paced up and down, pondering the journey to Kiev.’ (RNC, Vitalij Gubarev. Predan’e stariny glubokoj 1960–1980)

Example (22) illustrates the usage pattern of a coordinated complex event structure consisting of the n v a d and the n v a types. he dative argument of the n v a d cannot be indexed, with deinite null instantiation in the second conjunct. (22) Saš-ai da-l Maš-ej bilet-y i __i/*j namei-nom give-pst.m namej-dat ticket-acc.pl and øi/*j kupi-l tože konfetk-i. buy-pst.m also candy-acc.pl ‘Saša gave Maša the tickets and bought some candy too.’

In terms of coordination, the dative in the n v a d clearly deviates from other dative arguments. 6.5

Possessive relexive pronoun and subjecthood

he possessive relexive pronoun svoj ‘own’ has a complete paradigm: six cases, three genders, and singular and plural form. It functions as the modiier of an NP. Example (23) displays its typical usage with a nominative antecedent. (23) Stiven Spilberg ob’javi-l svoj name.nom name.nom announce-pst.m poss.acc sledujušč-ij proekt […]. next-acc project.acc ‘Steven Spielberg announced his next project.’ (RNC, Rendez-vous. Èkran i scena 2004.05.06)

From canon and monolith to clusters

Although the possibility of indexing the relexive possessive pronoun is a typical criterion for subjecthood, it is generally argued that it does not ofer a clear-cut criterion for classiication in Russian (Greenberg & Franks 1991: 77; Moore & Perlmutter 2000). In (24), the argument in the instrumental case is the antecedent of the possessive relexive pronoun. (24) Meždu nimii ustanovi-l-i-s’ svoii pr 3pli.ins settle-pst-pl-rm possi.nom.pl osob-ye otnošeni-ja. special-nom.pl relationship-nom.pl ‘heir own special relationships formed between them.’ (Timberlake 1980)

At the same time, the ability to distinguish between subject and oblique argument is apparent. Example (25) illustrates a potential mismatch in indexing in the n adj inf construction. However, the nominative argument is the only possible antecedent for the possessive relexive pronoun. (25) Jai dolž-en vrem-ja ot vremen-i čestn-o 1sgi.nom obligate-m time-acc pr time-gen honest-adv otčityva-t’-sja pered čitatel-emj o svoeji/*j rabot-e. report-inf-rm pr readerj-ins pr possi/*j.prep work-prep ‘From time to time, I am obligated to honestly report about my work to the reader.’ (Rossijskaja gazeta 31.12.2009, 255) (26) Saš-ai kupi-l emuj svoju?i|j mašin-u. namei-nom buy-pst.m 3sgj.dat poss?i/j.acc car-acc ‘Saša bought him his car.’

Additionally, the dative argument in the n v a d construction cans also antecede the possessive relexive pronoun, as in (26). his points to the conclusion that all the arguments in the constructions considered in this study are able to index the possessive relexive pronoun. 6.6

Short-distance relexive pronoun and subjecthood

Russian has a relexive pronoun sebja ‘oneself ’, which has defective paradigm consisting of only ive cases. It never occurs in the nominative case. It is commonly argued that the Russian relexive pronoun is capable to index both subjects and non-subjects (Greenberg & Franks 1991: 78). In (27) the dative argument, although omitted, in the d inf construction is indexed with the relexive pronoun. he dative in the d v[sja] construction can also be indexed with the relexive pronoun, exempliied in (28), although the structure of the construction is limited to ixed relexive expressions.

271

272 Aki-Juhani Kyröläinen

(27) Kak že __i pomo-č’ sebei i blizk-im How part øi help-inf selfi.dat and close-dat.pl ži-t, s ètim zabolevani-em? live-inf pr this.ins disease-ins ‘How to help oneself and one’s loved ones to live with this disease?’ (28) Boris-u ne rabota-et-sja u sebja doma. name-dat neg work-3sg.prs-rm pr self.gen at.home ‘Boris does not feel like working at his own house.’ (Moore & Perlmutter 2000: 378)

In sum, the short-distance relexive pronoun does not distinguish subject candidates in Russian. 6.7

Referential mismatches, long-distance relexive pronoun and subjecthood

Ackerman and Moore (2009) claim that in Russian the long-distance relation is the only test for subjecthood in terms of relexivity. Before turning to the complexities involved in the indexing capabilities of the arguments, I give counter-evidence against Ackerman and Moore’s claim as to the long-distance relation. Example (29) illustrates the use of the d adj n construction with the long-distance relexive pronoun. he relexive pronoun sebja ‘self ’ is located in the purpose clause and is indexed with the dative in the main clause. he purpose construction is discussed in Section 6.8. Moreover, Example (29) can be manipulated to demonstrate the indexing capability of the dative argument in other impersonal types, exempliied in (30). (29) Mnei ne nuž-en dopingj, čtoby 1sg.dat neg need-nom.m dopingj.nom, that __i sebjai/*j vzbodri-t’. øi self i/*j.acc make.better-inf ‘I do not need doping to make myself /*itself better.’ (Molodoj leninec. Penza 05.01.2010, 2) (30) Mnei ne stydn-o upotreblja-t’ dopingj, 1sg.i.dat neg shame-adv use-inf dopingj.acc, čtoby __i seb.jai/*j vzbodri-t’. that øi self i/*j.acc make.better-inf ‘I am not ashamed of using drugs to make myself /*itself better.’

he d v[sja] construction can also be conjoined with a long-distance relexive. Example (31) demonstrates a complex indexing relation across clause boundaries. he relexive pronoun can only be indexed with deinite null instantiations: null in the n v and in the d v[sja] construction, cf. Section 6.3.

From canon and monolith to clusters 273

(31) I kak raz vo vtorom tajme! Da ešče na pole sopernika! […] I __i dobyva-et očen’-očen’ nužn-uju nič’-ju. And øi gain-3sg.prs very-very need-acc draw-acc Nužn-uju ne tol’ko sebei/*j, no i, bezuslovn-o, Need-acc neg only selfi/*jdat but also certain-adv sovpadeni-e – toj komand-ej, coincidence-nom – that.dat teamj-dat protiv kotor-oj sovsem ne __i bega-lo-s’ pr which-gen quite neg øi run-pst-n-rm nedel-ej ranee. week-ins before ‘All of a sudden in the second half! On the ield of the competitor! (We) gained a very much needed draw. Not only needed for ourselves, but certainly, a coincidence, for the team against which our game did not go so well the week before.’

his leaves the indexing ability of the n v a d construction open. In general, there is tension – more precisely competition – between a dative and a nominative argument in terms of indexing potentiality. Czeczulin reports interesting empirical studies (sentence and picture selection) on the relexive pronouns in Russian. Although her study is focused on second language acquisition, the results with the native participants (N = 10) are noteworthy. he stimuli in the experiments are based on theoretically ambiguous coreferentiality, following Government and Binding heory (Chomsky 1981). Example (32) illustrates the issue. (32) Babušk-ai vele-l-a vnučk-ei nali-t’ Grandmotheri-nom order-pst-f granddaughterj-dat pour-inf sebei/j ča-ju. self i/j.dat tea-gen ‘he grandmother told the granddaughter to pour herself tea.’

From a usage-based perspective, Example (32) contains hardly any ambiguity: the most natural reading is the short-distance relation. his is in accordance with the results reported by Czezculin, contrary to the theoretical assumptions of the study (Czeczulin 2007: 124). Another interesting result is the discrepancy between the possessive and non-possessive relexive pronouns, suggesting that they should be treated diferently, contrary to the theoretical background of Government and Binding (cf. Czeczulin 2007: 120–121). Another issue which arises from this is the concept of distance. In formal approaches the dependency relation between the antecedent and the relexive pronoun is considered to be an autonomous operation, unafected by speciic lexical items or by various clause structures.

274 Aki-Juhani Kyröläinen

Russian, at least, suggests another interpretation. When the dative argument is fronted for purposes of emphasis, the reading changes as in (33). he most natural reading is that the nominative argument becomes the antecedent. Moreover, Example (33) demonstrates that the concept of distance is connected to minimal construction frames; in other words, the dative argument cannot index the relexive pronoun even though it is a short-distance dependency relation. (33) Vnučk-ei babušk-aj vel-e-l-a nali-t’ Granddaughteri-dat grandmotheri-nom order-pst-f pour-inf sebei/j ča-ju! selfi/j.dat tea-gen ‘he grandmother instructed/told the granddaughter to pour her tea!’

hus the data seem to indicate a discrepancy in indexing potential which is connected to speciic construction type, lexical verb and clause structure. he dative in the n v a d cannot be typically indexed with the relexive pronoun across constructions and clause boundaries. 6.8

Purpose construction and subjecthood

Crot argues, citing Haspelmath (1989), that the most typical sources for ininitival complements are purpose clauses. Secondly, purpose clauses are closely connected to coordination construction, in the sense that both constructions imply a consecutive ordering of the linked events (Crot 2001: 352). hese semantic dimensions are clearly displayed by the purpose construction in Russian. he purpose construction is proiled with the conjunction čtoby ‘in order to’. Because of the close connection to coordination construction, it can be used as a test for subjecthood (cf. Moore & Perlmutter 2000: 397). When the main and subordinate clause have a coreferential subject the subject is not overtly expressed in the purpose clause, yielding a typical pattern čtoby null inf (cf. Rubinstein 1986: 368). Example (34) illustrates this with the d inf construction. (34) Čtoby __i ne ošibi-t’-sja v hat øi neg mistake-inf-rm pr vybor-e cvet-a krask-i dlja volos, selection-prep colour-gen dye-gen pr hair.gen.pl, __i nužn-o učityva-t’ ottenok svoeji kož-i. øi, need-adv consider-inf hue.acc possi.gen skin-gen ‘To avoid making a mistake in choosing the colour of a dye for hair, one has to consider the hue of one’s skin.’ (RNC, Sovety parikmaxera. Daša 10, 2004)

From canon and monolith to clusters

Moore and Perlmutter (2000: 398) demonstrate that the d v[sja] construction cannot appear in the purpose construction. Example (35) illustrates the usage pattern with čtoby ‘in order to’ and d v[sja]. However, this is not a purpose construction. (35) Jai by xote-l, čtoby __1 rabota-l-o-s’ 1sg.nom cond want-pst.m that øi work-pst-n-rm by tak že družn-o, […]. cond part part vigorous-adv, ‘I would want also the work to be coming along vigorously.’ (RNC, Arkadij Gajdar. V dni poraženij i pobed 1922–1925)

he predicates in the remaining impersonal constructions lack the semantic content to be conjoined with purpose constructions proiling various aspects of either modality or states. his is clearly demonstrated by the d v[sja] construction. Russian has two interrogatives: začem ‘for what purpose’ and počemu ‘why’. he former can be perceived as the interrogative counterpart of the purpose construction. Another contributing factor is the semantic structure of purpose, which presupposes directionality and conscious action. (Arutjunova 2003). his is illustrated in (36): (36) Počemu /* začem tak pečal’n-o živ-et-sja Why/for.what.purpose so sad-adv live-3sg.prs-rm pod našim ser-ym neb-om? pr 1pl.poss.ins gray-ins sky-ins ‘Why is life so sad under our gray sky?’ (RNC, Nevol’nik dobryx del 2003. Izvestija 2003.01.31)

he interrogative začem ‘for what purpose’ is downright ungrammatical. hus the ability to be conjoined with the purpose construction is limited to the n v a, n v and d inf constructions. 6.9

Gerundive construction and subjecthood

he gerundive construction is another feature typically used to test the status of subjecthood of a particular argument in Russian. In terms of instantiation, the proile of the gerundive construction has a null, which cannot be overtly encoded. hus the test for subjecthood is again an indexing feature, deined in formal approaches as control. he gerundive construction in Russian occurs in both the present and the past tense; the ending of the verbal form is dependent on both the stem and the tense. In a constructionist perspective, the gerundive construction can occur in the following format vger null. Similarly to the relexive pronouns, linear order is not relevant to indexing.

275

276 Aki-Juhani Kyröläinen

Zimmerling (2009: 255–256 and references therein) demonstrates that various impersonal constructions can be conjoined with the gerundive, as in (37) (d adv) and (38) (d v inf). (37) Mnei stydn-o __i gljad-ja na 1sgi.dat shame-adv øi look-prs.ger pr igr-u svoeji komand-y. game-acc possi.gen team-gen ‘I am ashamed when I look at the game of my team.’ (Zimmerling 2009: 255) (38) […] i imi prixod-it-sja sam-imi and 3pli.dat have.to-3sg.prs-rm alonei-dat iska-t’, __i prokladyva-ja svoji put-’ seek-inf øi lay-prs.ger possi.acc path-acc v literatur-u. pr literature-acc ‘And they have to seek alone in creating their path to literature.’ (RNC, Karelija (Petrozavodsk); 31.12.2009; 147)

he gerundive construction is conjoinable with the various dative arguments. However, the n v a d type is unable, at least, prototypically, to index the null instantiation of the gerundive construction.

7. 7.1

Features of schematicity Paradigmatic features and subjecthood

he feature paradigm is posited to take into consideration the gestalt whole of constructions, not just a single slot as in the constructionist notion of linguistic signs.6 At the same time, the standard tests proposed in the literature do not strictly target the argument slot either but a range of potential compatibility with other constructions, for example the purpose and gerundive constructions. Moreover, the paradigm feature is proposed to evaluate the schematicity of the predicate in a construction, i.e. whether a given predicate slot in a construction displays the full range of categories available in a language for predicates or whether it involves restrictions. In this study the range of paradigmatic features is limited to indexing, aspect and productivity.

6. Inclusion of this feature was inspired by Moore’s and Perlmutter’s (2002) study on the compatibility of various clause level compatibility limitations with the impersonal constructions.

From canon and monolith to clusters 277

he indexing feature describes the predicate slot in terms of number and person for verbal predicates and number and gender for adjectival ones. he feature is fairly straightforward in Russian, dividing the constructions into three groups: personal verbal predicates (n v a, n v, and n v a d), personal adjectival predicates, displaying indexing in both number and gender (n adj inf and d adj n), and a residual group of constructions which follow the default indexing pattern. In the Slavic languages, the aspectual system is highly grammaticalized. he aspectual distinction between perfective and imperfective has obligatory morphological marking in all verbal categories. Prototypically, verbs have both aspects available. hus the verbal system in Russian is typically described in terms of aspectual pairs. However, the paradigmatic feature aspect is intended to depict a potential restriction in the slot of a construction, not a particular lexical item. he personal verbal predicate constructions (n v a, n v, and n v a d) do not have aspectual restrictions. Aspectual choice is not based on the semantics of the construction. Secondly, the same holds for the d inf construction, allowing both forms to appear in the predicate slot (Fortuin 2007; Perlmutter & Moore 2002). hirdly, the d v inf construction is lexically limited. he slot, however, can contain both perfective and imperfective verbs: for example such imperfective verbs as nadležit ‘necessary’, sleduet ‘must’ and prixoditsja ‘have to’, or perfective ones such as ponadobitsja ‘need to’. hirdly, the d v[sja] construction is generally considered to be tied to imperfective semantics (Kyröläinen 2008 and references therein). Lastly, adjectival constructions are ruled out by default. hey deviate from the canonical event model, which is relected in the type of the predicate slot. In these constructions, the aspectual distinction is achieved through nonpredicative means. he last feature to be considered under the label paradigm is productivity. Divjak and Janda (2008: 169), citing Daum and Schenk (1992), consider that the “stock” of verbal predicates in Russian consists of at least twenty thousand items. he concept of productivity escapes precise deinition, but in this study one aspect of productivity is touched upon in terms of coverage (cf. Barðdal 2008 for discussion on productivity in general). If a construction is productive, it will cover an array of lexical items and will be readily available for lexical innovation (cf. for this aspect of productivity Goldberg 2006; Osherson, Wilkie, Smith, Lopez & Shair 1990).7 In this vein, the aspect of productivity applied in this study pertains to the notion of central vs. peripheral membership status, discussed by Divjak and Janda (2008). he productivity of the n v a and n v constructions is indisputable. he n v a d construction is certainly less productive compared to the other two. Nonetheless, it is applicable to new lexical items. he modal constructions n adj inf and d adj n cover only a handful of lexical items and are basically lexical 7. Another term commonly used for this phenomenon in cognitive linguistics is ‘entrenchment’. For discussion of this concept I refer to Crot (2001) and Langacker (2009a).

278 Aki-Juhani Kyröläinen

constructions. he d v inf type, according to Divjak and Janda (2008: 169), is supported in Russian by only 37 verbs, and can thus hardly be considered a productive type. he d inf is generally accepted as productive, which is also indirectly supported by the features proposed in this study; in other words, the range of features available is closer to those of the n va and the n v types; cf. Table 1 in Section 8. hus it can be seen as a noninite counter-part of the n v a and a v constructions. he d v[sja] is also generally accepted as a productive type, albeit an infrequent one; for the productivity of this type in terms of coverage see Kyröläinen (2008). Finally, the d adv and d adv inf can be regarded as having a peripheral status in the overall constructional network of predication types in Russian. Zimmerling (2009: 256 and references therein) does not diferentiate between these two types, but inds that together they cover approximately 140–160 adverbial predicates. At the same time, the d adv construction is generally considered to be productive. Leinonen regards the ability to cover various domains of states, such as emotional and physical ones, as well as expressions of perceptional experience, as an indication of productivity. he class of adjectives which cannot be rendered in this construction type is prototypically used to proile the manner of an action: for example medlenno ‘slowly’ and pospešno ‘hastily’. (Leinonen 1985: 88–89). hus the distinction between the d adv and the d adv inf types is further supported and should be taken into consideration in describing various dative experiencer patterns in Russian. 7.2

Passive construction and subjecthood

he passive construction appears occasionally in the literature on subjecthood. he discussion is limited to the periphrastic passive construction with the passive past participle (ppp), which has a distinctive form in Russian. he prototypical ppp construction has the following characteristics: a nominative argument illing a nominal predicate slot, morphologically marked (depending on the verb stem) with the endings -t- and -n-, and aspectually limited to the perfective. Finally, the proile of the passive construction can contain an optional instrumental argument, yielding a form Nominative v[ppp] (Instrumental). he nominative argument and the ppp agree in gender and number. From a semantic perspective the ppp proiles an end result state, indicating a diference in construal from the canonical event model (Langacker 1991: 335–336). In terms of the conceptualization of event types, its counter-part is the n v a pattern. Hence the ppp is traditionally used as a test for direct objecthood (Crot 2001: 34–36, 41–44). Verbs occurring in the a v construction do not typically appear in the ppp construction.

From canon and monolith to clusters 279

Barðdal (2006: 50–51) demonstrates that certain subject-like dative obliques in German and Icelandic can appear in the passive construction, and keep their case form. Similarly, Moore and Perlmutter (2000: 392–393) show that the d inf construction can be conjoined, albeit under very restricted conditions, with the passive construction. his is illustrated in (39). he instrumental case of the ppp is due to the ininitival predicate byt’ ‘be’, which obligatorily assigns the case. he ppp nevertheless agrees in number and gender, which is a prototypical feature of adjectival predicates. Additionally, the dative argument in n v a d construction cannot trigger agreement in (39). (39) Fil’m-am Kislovsk-ogo ne by-t’ Film-dat.pl.m name-gen neg be-inf dostup-n-ymi publik-e access-ppp-ins.pl.m public-dat.f ‘It is not (in the cards) for Kieslowski’s ilms to accessible to the public.’ (Moore & Perlmutter 2000: 393)

he remaining construction types in study cannot appear in the passive construction. his restriction is both semantically and structurally motivated. From the semantic perspective, the constructions proile various state-like event types which do not have a resultative end state. Structurally, the predicate slots in the constructions are restricted, as was demonstrated in Section 7.1. 7.3

Raising construction and subjecthood

Raising to subject position is another feature employed in formal approaches to subjecthood. Accordingly, by inserting the raising predicate the subject of the ininitive is “raised” to the subject position. In constructionist terms, there is no need to posit any movement. If a construction is compatible with the raising construction, the subject slots of the constructions are uniied. hus raising constructions pertain to complex event structures and indexicality. Langancker ofers an elaborated account of various raising constructions based on metonymy, i.e. the ability to perceive an entity as a conceptual reference point for another. (Cf. Langacker 1995: 27–30). he raising construction thus pertains to indexing referents across constructions. Russian has a small number of raising predicates, such as načinat’ ‘begin’, stat’ ‘become’ and okazat’sja ‘seem, appear’. he latter verb cannot take an ininitival complement in Russian. Instead, it combines with a nominal argument in the instrumental case. hus it can be only used as a raising predicate for the constructions with an adjectival predicate slot.

280 Aki-Juhani Kyröläinen

Moore and Perlmutter argue that the d inf construction can be conjoined with a raising construction whereas other datives cannot. Example (40) illustrates a typical raising construction with an intransitive (n v) construction. he d inf is compatible with a raising predicate, as shown in (41). Moreover, the subject of the raising predicate is in the dative case, inherited from the d inf construction. his is further supported by the indexing with the semi-predicate odin ‘one’. (40) Onii načina-jut rabota-t’ odn-ii. 3pli.nom begin-3pl.prs work-inf alonei-nom.pl ‘hey are beginning to work alone.’ (Moore & Perlmutter 2000: 399) (41) Imi ne nača-t’ rabota-t’ odn-imi. 3pli.dat neg begin-inf work-inf alonei-dat.pl ‘It is not (in the cards) for them to begin to work alone.’ (Moore & Perlmutter 2000: 399)

In contrast, Moore and Perlmutter (2000: 399–400) argue that the remaining impersonal construction types cannot appear in the raising construction. However, under certain conditions, the d adv construction can appear in the raising construction, as in (42). he same inherited dative argument is in the subject position of the raising predicate. Omitting the verb stanovit’sja ‘become’ yields an ungrammatical construction.8 Additionally, the d v[sja] construction can also appear in the raising construction, as in (43). (42) Mne nača-l-o stanovi-t’-sja tošn-o. 1sg.dat begin-pst-n become-inf-rm nauseous-adv ‘I was feeling sick.’ (Zimmerling 2009: 258) (43) Dolg-o side-l smotritel-’ nepodvižn-o, Long-adv sit-pst.m inspector-nom still-adv mne sta-l-o drema-t’-sja. 1sg.dat become-pst.n doze-inf-rm ‘For a long while the inspector sat still; I dozed.’ (RNC, I. A. Gončarov. Fregat, Pallada 1855)

However, none of the modal constructions can appear in the raising construction. Modality is neither a state one can enter nor an activity one can begin. he semantic

8. Perlmutter and Moore have revised their position in recognizing these patterns. hey argue, however, that these are instances of staged raising predications and are explained by an underlying silent explicative subject in the nominative case (Perlmutter & Moore 2002: 637–639). According to the position advocated in this study, the restrictions are explained in semantic and structural terms. hus the surface structure is suicient to motivate these conjoined complex constructions. Moreover, Example (43) cannot be analyzed as a staged raising predication.

From canon and monolith to clusters

motivation also holds for the n v a d construction. Recipiency is neither a state nor an activity a referent can commence upon oneself. 7.4

Imperative construction and subjecthood

he last feature considered in this study is the imperative. In terms of constructions, the imperative has distinct morphological form in Russian depending on the stem of the verb. he imperative is formed for the second person singular by the endings -j, -i and -’, for the second person plural by -jte, -ите and -’te. In general, the imperative construction is another test of compatibility and indexicality. When a construction is conjoined with the imperative, the argument occupying the subject slot is typically omitted. his omission of the subject in the imperative construction is typically attributed to its function: the subject argument is indexed to the addressee in the speech event. Dryer (2005) notes that typologically many languages which utilize pronouns in the subject position in a declarative clause do not overtly encode them in the imperative construction. Example (44) illustrates a typical imperative construction. However, the test is not without exceptions in Russian, as discussed by Fortuin (2010) and exempliied in (45). (44) Sid-i-te tam i rabota-j-te! sit-imp-2pl there and work-imp-2pl ‘sit there and work!’ (45) Net, ty skaž-i! neg, 2sg.nom say-imp.2sg ‘No, you tell me!’ (Fortuin 2010: 432; www.aldebaran.ru/det/vasina/vasina6/?13)

According to Barðdal (2006: 54 and references therein), the imperative test is not used in Icelandic for subject-like oblique arguments simply because many of the predicates do not have the semantic content to be conjoined with the imperative construction. he same holds true for Russian. n v a and n v are the only constructions which can be conjoined with the imperative.

8. Towards a clustered structure of subjecthood in Russian Table 1 summarizes the main points made in the previous sections. he digit 1 denotes the prototypical presence of a feature, zero its prototypical absence. he results indirectly support the notion of a canonical event type. n v a is the only construction which displays the full range of the features, albeit the only diference between n v a and n v occurs in the passive construction.

281

282 Aki-Juhani Kyröläinen

Before ofering a potential radial network of these constructions, certain methodological considerations need to be taken into account. he radial network approach assumes that distributional (dis)similarities between constructions are associated with diferences in membership status. In this vein, building a network is nontrivial. In this study, a neighbor-joining algorithm is used to build the radial network (Paradis, Claude & Strimmer 2004). he algorithm belongs to the family of various clustering methods which are widely used in diferent taxonomic applications (Kaufman & Rousseeuw 2005 [1990]). he basic principle in cluster methods is to detect (dis)similar objects in a given set; thus it is analogous to categorization. he package ofers a wide range of visualization options. he visual representation of the binary data is easily seen in the clusters, as shown in Figure 1.9 Table 1. Constructions and prototypical presence/absence of features. Construction

Argument-properties Nominative

Animacy

Autonomy

Linear order

Inversion

nva nv nvad n adj inf d adj n d v inf d adv inf d inf d v[sja] d adv

1 1 1 1 1 0 0 0 0 0

1 1 1 1 1 1 1 1 1 1

1 1 0 1 0 1 1 1 1 1

1 1 0 1 1 1 1 1 1 0

1 1 0 0 1 1 0 1 1 1

Construction

Indexicality Agreement

Semi-pred

Instantiation

Coordination Poss ref

nva nv nvad n adj inf d adj n d v inf d adv inf d inf d v[sja] d adv

1 1 0 1 1 0 0 1 0 0

1 1 1 1 1 1 1 1 1 1

1 1 0 1 1 1 1 1 1 1

1 1 0 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1 1

9. R version 2.11.1 (64bit) and the package Ape version 2.5.3 were used in this study.

From canon and monolith to clusters 283

Construction

Indexicality

nva nv nvad n adj inf d adj n d v inf d adv inf d inf d v[sja] d adv Construction nva nv nvad n adj inf d adj n d v inf d adv inf d inf d v[sja] d adv

Short ref

Long ref

Purpose

Gerund

1 1 1 1 1 1 1 1 1 1

1 1 0 1 1 1 1 1 1 1

1 1 0 0 0 0 0 1 0 0

1 1 0 1 1 1 1 1 1 1

Schematicity Paradigm I Paradigm A Paradigm P Passive

Raising

Imperative

1 1 1 1 1 0 0 0 0 0

1 1 0 0 0 0 0 1 1 1

1 1 0 0 0 0 0 0 0 0

1 1 1 0 0 1 0 1 0 0

1 1 1 0 0 0 0 1 1 1

1 0 0 0 0 0 0 1 0 0

he last point to be made before looking at the results involves the binary data shown in Table 1. From the perspective of Russian, a number of features do not add any contrast between constructions. In this sense, they are redundant to the actual clustering; nonetheless, they have been kept in the cluster structure. he main reason for this is the clustering of features. his point will be considered later in this section. he clustering solution is shown in Figure 1 with the unrooted tree plotting function (cf. to the diagram presented by Janda and Divjak in this volume). he radial structure seems to indicate that Russian has three major clustered concepts. he canonical notion of subjecthood arises naturally from the data. Interestingly, the almost unanimous view that the dative in the d inf construction occupies a subject position is supported by the tree structure. he right branch contains four construction types, with two subbranches: one consisting of d adv and d v[sja], the other of d adv inf and d v inf. In the lower branch, the

284 Aki-Juhani Kyröläinen

NVA NV D INF

D ADV D V[SJA]

D ADV INF D V INF NVAD

N ADJ INF D ADJ N

Figure 1. Radial structure of subjecthood in Russian.

d adj n and n adj inf types form a small separate clustered structure of their own. Although the n v a d construction is linked to this clustered structure, it is clearly separated. his is illustrated by the distance in the tree structure. Figure  1 shows the clustered structure with all the proposed features. As pointed out earlier, the number of chosen features is always somewhat arbitrary, and the possibility of adding another feature, which might perform even better, is always present. he neighbor-joining algorithm has a built-in method for evaluating and visualizing this uncertainty. If the feature set is still able to establish category boundaries with a skewed input, the uncertainty is at least partly reduced. he basic idea behind the method chosen is to allow features to be present multiple times in a given sample. By exploiting sampling with replacement, we grow new trees (n = 200) from the original data. Allowing the features to appear multiple times will efectively skew the original distribution in the data. If the clustered structure is still present across these skewed samples, we have stronger support for the cluster solution. he neighbor-joining algorithm allows visual representation of this by comparing the proportions of the sampled trees that support the groupings in the original tree (Baayen 2008: 145–146), visualized in Figure 2. he support is represented visually in Figure 2 by the thermometers ranging from 0 to 1: the higher the proportions in the thermometers, the better the support for the grouping (dark shade). he crux of subjecthood in Russian arises naturally from the samples. he irst split in the tree has very little support, indicating that the actual structure of the tree can take any shape. If we had a uniied category based on similarity, the expectation would be inversed: the core would have high support and variation might be observed in the outer branches. In contrast, the outer branches are almost fully supported, except for d adv inf and d v inf. he irst split in this branch also has the weakest support compared to the other

From canon and monolith to clusters 285

NVA NV D ADV

D INF

D V[SJA]

D ADV INF D V INF NVAD

N ADJ INF D ADJ N

Figure 2. Support of original clustered structure based on 200 resamplings.

branches. he strongest support is found for the n v a d construction. he positions of the canonical subject constructions, n v a and n v, are also well supported throughout the samples. he last question regarding clustering-based subjecthood that needs to be evaluated concerns the features themselves. he set was divided a priori into three groups: Argument-slot, Indexicality and Schematicity. Assuming that subjecthood in a given language arises based on usage, the constructions can be taken as a starting point and used to build up the clustered structure of the features, as shown in Figure 3. he clustering suggests that the data contain two major categories with subbranches. he algorithm clusters Agreement, Paradigm  I(ndex) and Nominative together, the canonical instances afecting agreement. he right branch consists of those features which are most likely to be afected by schematicity, the ability to be conjoined with other construction types in a language. he let branch, with two subtypes, is a mix of Indexicality and Argument-slot properties. he lower subbranch groups features which in Russian are closely connected with animacy. Generally, the relexive pronouns have a strong association with animacy and repel inanimate antecedents. he same holds for semi-predicates. hus this centering around animacy arises naturally from the constructions, without being explicitly speciied. he upper subbranch mostly includes features which are tied to the properties of a speciic slot in a construction and instantiation. Interestingly, the long-distance relexive is found on this branch. As the examples suggested, long distance relexive pronoun is afected in Russian by linear order and clause structures, both of which are located on this branch.

286 Aki-Juhani Kyröläinen

Raising Linear order Autonomy Inversion Instantiation Coordination

Paradigm P Agreement

Gerund Long ref

Paradigm I Nominative

Paradigm A Imperative Short ref Animacy

Poss ref

Passive Purpose

Semi pred

Figure 3. Clustered structure of features based on constructions.

he last branching based on the constructions is displayed by the raising construction and paradigmatic feature productivity. heir position seems to indicate that they are not strongly related to subjecthood compared to the other branches in the tree structure. Lastly, the Inversion construction seems to be an isolated instance. In this sense, the tree structure relects the fact that Inversion is rarely considered to be a feature of subjecthood in Russian. hus the tree structure suggests that Russian has four clustered feature sets for subjecthood. he Primary cluster contains canonical deinitions of subjecthood – Agreement, Nominative case and Paradigm I(ndex), i.e. whether the predicate slot of the construction displays indexing. he lower right-hand branch groups features pertaining to Conjoinability, including aspect. he third clustered structure is related to Instantiation. he last set displays a close proximity to Animacy.

9. Entailments and conclusions he dative case does not form a single uniied category which could be simply classiied as a syntactic relation under the label of indirect object. he same observation also stands for the nominative. he distinction between the d adj n and n adj inf constructions is supported even through the resampling method. he canonical event model seems to motivate deviances in both of these cases. Abstracting across the constructions, the clustered structure indicates that contemporary Russian has two clustered subject construction, n v a and d inf, the basic structure of which is partly inherited by their related constructions based on the proposed set of features, given in Table 2.

From canon and monolith to clusters 287

Table 2. Prototype-based structure of subjecthood in terms of clustered feature sets. Primary cluster

Conjoinability Instantiation Animacy cluster cluster cluster

Subjecthood

n_v_a

3

4

6

4

17

n_v

3

3

6

4

16

n_v_a_d

2

1

0

4

7

n_adj_inf

3

0

6

4

13

d_adj_n

3

0

4

4

11

d_inf

1

3

6

4

14

d_v_inf

0

1

6

4

11

d_adv_inf

0

0

6

4

10

d_v[sja]

0

0

6

4

10

d_adv

0

0

5

4

9

Table 2 shows the structure of the constructions in terms of clustered features, excluding those of Inversion, Raising and Paradigm P(roductivity). Extending Dowty’s principle of the proto-roles, the features are collapsed under their labels, yielding a ranking in terms of subjecthood. Based on this reappropriation of Dowty’s proto-roles, we can state that Russian has two basic subject constructions functioning as the baseline categories, given in bold. hey possess features from all possible cluster types, with no gaps. he Subjecthood column demonstrates that the basic subject constructions are truly the baseline categories. he remaining constructions deviate from these two types, forming a continuum in terms of subjecthood. he n v a d construction shows a clear deviation compared to the all others. his same result was already present in the clustering, which demonstrated the (dis)similarity between the various constructions. he clustered structure of the features allows setting strong predictive entailment conditions. he Animacy cluster is the weakest set, followed by the Instantiation cluster, the Conjoinability and inally the Primary Cluster. his makes it possible to posit a language-speciic pathway to subjecthood. Moreover, we can posit a constructionist hypothesis in terms of an ordering of features for subjecthood: Animacy Cluster > Instantiation Cluster > Conjoinability Cluster > Primary Cluster. Whether this pathway holds diachronically is an empirical question. However, the Primary Cluster takes precedence over the others. his is clearly illustrated by the d adj n construction. he nominative argument lacks most of the features associated with subjecthood compared to the dative argument, but the ability to trigger features from the Primary Cluster deines its status within the construction.

288 Aki-Juhani Kyröläinen

Lastly, a second entailment can be stated based on the clustered structure. If there is competition between a predicate construction and an argument construction, the entailment pathway may be used to motivate the seemingly anomalous behavior of certain predicates. his is perhaps most prominent in the n v a d construction, which prototypically lacks the features from the Instantiation Cluster. he mismatch in the structure of this construction is prominently demonstrated by such verbs as dat’ ‘give’ and nalit’ ‘pour’. he former instantiates the n v a d construction and easily allows the omission of the dative argument. In contrast, the same semantic description does not pertain to the verb nalit’ ‘pour’ which prototypically requires overt encoding of the dative argument; otherwise the semantics of the verb simply proile a transitive construction. Following the entailment pathway, these verbs will thus have diferent properties. his allows us to posit a predication hypothesis: strong predicates are able to acquire features from adjacent clusters compared to the argument construction. his should also hold in the reversed situation, i.e. with weak predicates; they can only possess features from the preceding cluster(s) in the pathway of the argument construction. To sum up: the study ofers strong evidence for a clustered structure of subjecthood, which can be used to extend both a (dis)similarity between constructions and an ordering of features based on the concept of proto-role. his notion of subjecthood makes it possible to illustrate the hierarchical constellation of features, yielding strong entailments in terms of both argument and predicate constructions. he constellation arises naturally from the interaction of features and constructions, without positing any a priori deined possible constraints.

References Ackerman, Farrell & Moore, John (2009). Proto-properties and obliqueness. Paper given at the conference Case in and across Languages. Helsinki 27–29.8.2009. Arutjunova, N. D. (2003). Jazyk celi. In N. D. Arutjunova (Ed.), Logičeskij analiz jazyka. Izbrannoe 1988–1995 (pp. 386–396). Moskva: INDRIK. Baayen, Harald R. (2008). Analyzing linguistic data: A practical introduction to statistics using R. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511801686 Barðdal, Jóhanna (2006). Construction-speciic properties of syntactic subjects in Icelandic and German. Cognitive Linguistics, 17(1), 39–106. DOI: 10.1515/COG.2006.002 Barðdal, Jóhanna (2008). Productivity: Evidence from case and argument structure in Icelandic. Amsterdam: John Benjamins. DOI: 10.1075/cal.8 Barðdal, Jóhanna, & Eythórsson, hórhallur (2003). Icelandic vs. German: Oblique subjects, agreement and expletives. Chicago Linguistic Society, 39(1), 755–773. Bhaskararao, Peri, & Subbarao, Venkata Karumuri (Eds.). (2004). Non-nominative subject (Vol. 1). Amsterdam: John Benjamins.

From canon and monolith to clusters 289

Bock, Kathryn, Loebell, Helga, & Morey, Randal (1992). From conceptual roles to structural relations: Bridging the syntactic clet. Psychological Review, 99(1), 150–171. DOI: 10.1037/0033-295X.99.1.150 Chomsky, Noam (1981). Lectures on goverment and binding. Dordrecht: Foris. Crot, William (1998). Event structure in argument linking. In M. B. W. Geuder (Ed.), he projection of arguments (pp. 21–63). Stanford: CSLI Publications. Crot, William (2001). Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780198299554.001.0001 Crot, William, & Cruse, Alan D. (2004). Cognitive linguistics. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511803864 Czeczulin, Annalisa (2007). he Russian relexive in second-language acquisition: Binding preferences and L1 transfer. Doctoral Dissertation, University of Pittsburgh. Daum, Edmund, & Schenk, Werner (1992). Die Russische Verben. Leipzig: Langenscheidt. Divjak, Dagmar (2009). Mapping between domains. he aspect-modality interaction in Russian. Russian Linguistics, 33(3), 249–269. DOI: 10.1007/s11185-009-9045-8 Divjak, Dagmar, & Janda, Laura A. (2008). Ways of attenuating agency in Russian. Transactions of the Philological Society, 106(2), 138–179. DOI: 10.1111/j.1467-968X.2008.00207.x Dowty, David (1991). hematic proto-roles and argument selection. Language, 67(3), 547–619. DOI: 10.1353/lan.1991.0021 Dryer, Matthew (2005). Expression of pronominal subjects. In M. Haspelmath, M. Dryer, D. Gil, & B. Comrie (Eds.), he world atlas of language structure (pp. 410–413). Oxford: Oxford University Press. Eythórsson, hórhallur, & Barðdal, Jóhanna (2005). Oblique subjects: A common Germanic inheritance. Language, 81(4), 824–881. DOI: 10.1353/lan.2005.0173 Fillmore, Charles J. (1985). Pragmatically controlled zero anaphora. Berkeley Linguisitc Society, 12, 163–182. Fortuin, Egbert (2005). From necessity to possibility: he modal sprectum of the dativeininitive construction in Russian. In B. Hansen & P. Karlík (Eds.), Modality in Slavonic languages: New perspectives (pp. 39–60). München: Sagner. Fortuin, Egbert (2006). On the use of dative subjects in the construction of anteriority in Russian. Russian Linguistics, 30(3), 321–357. DOI: 10.1007/s11185-006-0706-6 Fortuin, Egbert (2007). Modality and aspect: Interaction of constructional meaning and aspectual meaning in the dative-ininitive construction in Russian. Russian Linguistics, 31(3), 201–230. DOI: 10.1007/s11185-007-9015-y Fortuin, Egbert (2010). Explicit second-person subjects in Russian imperatives: Semantics, word order, and a comparison with English. Linguistics, 48(2), 431–486. DOI: 10.1515/ling.2010.014 Geniušienė, Emma (1987). he typology of relexives. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110859119 Gerritsen, Nelleke (1990). Russian relexive verbs. In search of unity in diversity. Amsterdam: Rodopi. Goldberg, Adele E. (2006). Constructions at work. he nature of generalization in language. Oxford: Oxford University Press. Greenberg, Gerald R., & Franks, Steven (1991). A parametric approach to dative subjects and the second dative in Slavic. Slavic and East European Journal, 35(1), 71–97. DOI: 10.2307/309034

290 Aki-Juhani Kyröläinen

Halizeva, V. S. (1969). Semantičeskij analiz sojuzov predšestvovanija. Russkij jazyk za rubežom, 2, 77–82. Haspelmath, Martin (1989). From purpose to ininitive: A universal path of grammaticalization. Folia Linguistica Historica, 10(1–2), 287–310. Hopper, Paul J., & hompson, Sandra A. (1980). Transitivity in grammar and discourse. Language, 56(2), 251–299. DOI: 10.1353/lan.1980.0017 Janda, Laura A. (2008). Transitivity in Russian from a cognitive perspective. In G. Kustova (Ed.), Dinamičeskie modeli: Slovo. Predloženie. Tekst. Sbornik statej v čest’ E. V. Padučevoj (pp. 970–988). Moskva: Jazyki slavjanskoj kul’tury. Kaufman, Leonard, & Rousseeuw, Peter J. (2005) [1990]. Finding groups in data: An introduction to cluster analysis. New Jersey: John Wiley & sons Inc. Kay, Paul, & Fillmore, Charles J. (1999). Grammatical constructions and linguistic generalizations: he What’s X doing Y? construction. Language, 75(1), 1–33. DOI: 10.2307/417472 Keenan, Edward L. (1976). Towards a universal deinition of “subject”. In C. N. Li (Ed.), Subject and topic (pp. 303–333). New York: Academic Press. Koenig, Jean-Pierre, Mauer, Gail, & Bienvenue, Breton (2003). Arguments for adjuncts. Cognition, 89(2), 67–103. DOI: 10.1016/S0010-0277(03)00082-9 Kyröläinen, Aki-Juhani (2008). Low-frequency constructions and salience: A case study on Russian verbs of motion of dative impersonal construction type. In A. Mustajoki, M. V.  Kopotev, L. A. Birjulin, & E. J. Protasova (Eds.), Instrumentarij rusistiki: korpusnye podhody (pp. 176– 197). Helsinki: Department of Slavonic and Baltic Languages and Literatures. Langacker, Ronald W. (1991). Foundations of cognitive grammar. Vol. 2. Descriptive application. Stanford: Stanford University Press. Langacker, Ronald W. (1995). Raising and transparency. Language, 71(1), 1–62. DOI: 10.2307/415962 Langacker, Ronald W. (2009a). A dynamic view of usage and language acquisition. Cognitive Linguistics, 20(3), 627–640. Langacker, Ronald W. (2009b). Investigations in cognitive grammar. Berlin and New York: Mouton de Gruyter. DOI: 10.1515/9783110214369 Leinonen, Marja (1985). Impersonal sentences in Finnish and Russian: Syntactic and semantic properties. Helsinki: Slavica Helsingiensia. Moore, John, & Perlmutter, David M. (2000). What does it take to be a dative subject? Natural Language & Linguistic heory, 18(2), 373–416. DOI: 10.1023/A:1006451714195 Osherson, Daniel N., Wilkie, Ormond, Smith, Edward E., Lopez, Alejandro, & Shair, Eldar (1990). Category-based induction. Psychological Review, 97(2), 185–200. DOI: 10.1037/0033-295X.97.2.185 Paducheva, Elena (2010). Subject-predicate inversion and its cognitive sources. Russian Linguistics, 34(2), 113–121. DOI: 10.1007/s11185-010-9054-7 Paradis, Emmanuel, Claude, Julien, & Strimmer, Korbinian (2004). APE: Analyses of phylogenetics and evolution in R language. Biostatistics, 20(2), 289–290. Perlmutter, David, & Moore, John (2002). Language-internal explanation: he distribution of Russian impersonals. Language, 78(4), 619–649. Primus, Beatrice (1999). Cases and thematic roles. Tübingen: Niemeyer. DOI: 10.1515/9783110912463 Rubinstein, George (1986). Subjective dative in Russian ininitival clauses of purpose. he Slavic and East European Journal, 30(3), 367–379. DOI: 10.2307/307889

From canon and monolith to clusters

Seo, Seunghyun (2001). he frequency of null subject in Russian, Polish, Czech, Bulgarian and Serbo-Croatian: An analysis according to morphosyntactic environments. Doctoral Dissertation, Indiana University at Bloomington. Sigurðsson, Halldór Ármann (2002). To be an oblique subject: Russian vs. Icelandic. Natural Language & Linguistic heory, 20(4), 691–724. DOI: 10.1023/A:1020445016498 Silverstein, Michael (1976). Hierarchy of features and ergativity. In R. M. W. Dixon, (Ed.), Grammatical categories in Australian languages (pp. 112–171). Canberra: Australian Institute of Aboriginal Studies. Sirotinina, O. B. (2006) [1965]. Porjadok slov v russkom jazyke. Moskva: KomKniga. Švedova, N. Ju. (Ed.). (1982a). Russkaja grammatika. Tom 1. Moskva: Nauka. Švedova, N. Ju. (Ed.). (1982b). Russkaja grammatika. Tom 2. Moskva: Nauka. Timberlake, Alan (1980). Oblique control of Russian relexivization. In C. V. Chvany, & R. D. Brecht (Eds.), Morphosyntax in Slavic (pp. 235–259). Columbus Ohio: Slavica Publishers. Wierzbicka, Anna (1980). Lingua mentalis: he semantics of natural language. New York: Academic Press. Zdorenko, Tatiana (2010). Subject omission in Russian: A study of the Russian national corpus. In S. T. Gries, S. Wulf, & M. Davies (Eds.), Corpus-linguistic applications: Current studies, new directions (pp. 119–133). Amsterdam, Netherlands: Rodopi. Zimmerling, Anton (2009). Dative subjects and semi-expletive pronouns in Russian. In G.  Zybatow, U. Junghanns, D. Lenertová, & P. Biskup, (Eds.), Studies in formal Slavic phonology, morphology, syntax, semantics and information structure. Proceedings of FDS 7, Leipzig 2007 (pp. 253–265). Frankfurt am Main: Peter Lang. Zolotova, G. A. (2000). Ponjatie ličnosti/bezličnosti i ego intepretacii. Russian Linguistics, 24(2), 103–115. DOI: 10.1023/A:1007054326390

291

he role of non-canonical subjects in the overall grammar of a language A case study of Russian Laura Janda and Dagmar Divjak University of Tromsø / University of Sheield

We ofer a model of Russian core syntax in terms of a radial category network of constructions. he prototype corresponds to Langacker’s “canonical event model”, namely a prototypical transitive event, and more peripheral constructions are related to it via metaphor and metonymy. From this perspective we focus on non-canonical subjects marked in the dative case, highlighting the complex interaction of lexical items (verbs) with constructions, and building on our previous work on case and ininitives. We hypothesize that a speaker’s perception of cause may be inluenced by the use of non-canonical subjects (in Russian) rather than canonical subjects (in English) and present the results of an experiment. We are unable to prove any conclusive efect, but show the importance and need for further testing.

1.

Constructing an entire grammar: heoretical background

An important premise of cognitive linguistics is that it is possible to explain linguistic phenomena in terms of general cognitive strategies (Langacker 1987: 12–13). hese strategies include categorization in terms of radial categories, the abstraction of embodied experience as image schemas, metaphor, and metonymy (Feldman 2006). Turner (1996: 9) demonstrates how embodied experience makes it possible for us to recognize “small stories as involving objects and events” and to “form conceptual categories of objects and events”. hese “small stories” are primarily “events in space: he wind blows clouds through the sky, a child throws a rock, a mother pours milk into a glass, a whale swims through the water” (Turner 1996: 13). Turner & Lakof (1989, cf. also Lakof 1993) show how the events are actions1 metaphor 1. In Lakof and Turner’s model, an event is merely something that happens, whereas an action necessarily has an agent that brings it about. doi 10.1075/cal.16.11jan © 2015 John Benjamins Publishing Company

294 Laura Janda and Dagmar Divjak

extends and abstracts the concepts of actors, movements, and objects to produce the building-blocks of constructions: subjects, objects, verbs. he failure of a boulder to move, for example, is an event that does not have an agent. However, events are actions “imputes agency to something causally connected to the event” (Lakof & Turner 1989: 37), thus metaphorically creating agents in events that are otherwise agentless, as in he boulder resisted all of our eforts to move it where the boulder becomes the agent of the otherwise agentless event. As a result of events are actions, the repertoire of potential subjects, both canonical and non-canonical, is extended. he events are actions metaphor further organizes these elements in structures that yield transitive and intransitive and other types of constructions. Turner (1996: 145) directly identiies the constructions that result from projecting the structure of “small stories” onto grammar with the construction grammar proposed by Goldberg (1995). Turner (ibid.) furthermore claims that “[r]udimentary grammar is a repertoire of related grammatical constructions” and that “[t]he backbone of any language consists of grammatical constructions that arise by projection from basic abstract stories”. In other words, Turner asserts that the grammar of a language consists of a network of related grammatical constructions. Although Goldberg in places echoes this view,2 she also admits that linguists working on construction grammar “have focused on idiomatic constructions” (Goldberg 2006: 14) rather than tackling the larger task of describing an entire grammar in terms of constructions. he only network of constructions Goldberg (2006: 166–182) presents herself consists of Subject-Auxiliary Inversion constructions in English, which form a small and peripheral subset of English grammar. To summarize: construction grammar is predicated on the assumption that the constructions in a grammar are related to each other, forming a network akin to a radial category, but scholars have focused on speciic idiomatic constructions or minor groups of constructions. No one has presented the grammar of a language in terms of a network of constructions. Part of the reason for this gap may be that so much of the work on construction grammar has focused on English, where the grammar is dependent on the ordering of elements, rather than being a more straightforward matter of elements and their combinations. It may be that a grammatical system based on word order is inherently harder to handle in the framework of construction grammar, but this issue goes beyond the scope of our article, which focuses on Russian rather than English. A highly inlected language, such as Russian, clearly marks elements in terms of their roles in a construction, relegating word order to pragmatic functions. he description of a grammar as a network of constructions may thus be easier to realize for a language like Russian 2. Cf. Goldberg (2006: 18) “the network of constructions captures our grammatical knowledge of language in toto” and (2006: 227) “our knowledge of linguistic constructions, like our knowledge generally, forms an integrated and motivated network”.

he role of non-canonical subjects in the overall grammar of a language 295

with transparent morphological patterns for word classes and inlectional endings that specify case for noun phrases and agreement for verbs. A language like Russian makes it possible to organize constructions according to the elements that they contain, without the added parameter of order and the ambiguities word order must solve in the face of a limited morphological system in a language like English. English is of course typologically unusual in this respect, as most languages have more inlection than English does. hus a description of the grammar of Russian as a network of constructions is also probably more representative of how this potential of construction grammar can be realized. Crot’s typological model (2001) is non-reductionist and, unlike Goldberg’s model, not idiom-biased, not restricting the notion “construction” to those with a readily identiiable, tangible meaning, e.g. the let alone construction. For Radical Construction Grammar, the “primitive construct” is the construction, a complex entity containing categories and relations that are deined by the constructions they appear in; they are not theoretical primitives. In other words, constructions are not derived from their parts, but instead the parts are derived from the constructions. Hence, the parts of a construction do not have an independent existence outside of the whole construction. Crot suggests that linguists abandon the assumption that syntactic structures are made up of primitive categories and relations. He explicitly claims that the grammatical description of any language should exclusively consist of an analysis of the constructions in that language and the network of relationships among them. hese constructions or grammatical patterns of a language are aptly described by means of the distributional method (Crot 1999: 69–74). In the distributional method, constructions deine categories, albeit not in the strict sense. What matters is “the semantic interpretation of a word in a particular grammatical construction. By examining the meanings of verbs in constructions, we can establish semantic classes of events and conceptual meanings of constructions” (Crot 1998: 91). he representation of Russian (and potentially any other) grammar as a radial category of constructions yields added beneits. Speakers’ knowledge of a language may well be organized by their knowledge of relations between (the meanings of) constructions (van den Eynde 1995: 116f; Goldberg 1995: 67; Crot 2001: 25–29). Human beings have the capacity to construe a situation in alternate ways; hence, knowledge of one construction involves simultaneous knowledge of a complex network of constructions. Networks of constructions can be used to distinguish between the forms of a lexeme and to categorize verbs into (sub)classes, as Apresjan (1967) has shown for Russian, Eggermont and Melis (1992) have shown for French, Levin (1993) has shown for English, and Schøsler and Van Durme (1996) have shown for Danish. Within cognitive linguistics and especially within Construction Grammar approaches, both constructions and lexemes and, in fact, all linguistic units are considered form-meaning pairings that form larger sets or networks.

296 Laura Janda and Dagmar Divjak

If both grammatical constructions and lexical elements are meaningful units, their meanings need to be compatible in order to yield felicitous combinations. Comparing the sets of constructions verbs can occur in then tells us something about the meaning components those verbs share. A network account also reveals a center-periphery structure based on a prototype that corresponds to Langacker’s (1991: 285–286) “canonical event model”, the prototypical transitive event. From this structure it is possible to discover which constructions are more peripheral in the system, and these include impersonal constructions with non-canonical subjects. It is also possible to show how neighboring constructions in the network are related to each other and how they inluence each other semantically. In many places we see that the transition between neighboring constructions is continuous rather than discrete. All of these points are illustrated in detail in Section 3.1.

2. Non-canonical subjects in the dative case in Russian For the purposes of this article we deine grammatical subjects as those subjects that are expressed by a nominative noun phrase that triggers verbal agreement, and this type of subject is most typically an agent. Under this deinition, many sentences in Russian do not have a grammatical subject, and most of these sentences will use the neuter singular verb morphology as their “default” solution. However, many of the sentences that lack a grammatical subject do have a non-canonical subject. his is possible when subject functions are realized by either something that is not a noun phrase (we will see examples of ininitives and clauses that ill this role below), or is a noun phrase marked by another case, such as the dative or accusative. his latter type of non-canonical subject is typically an experiencer and the potential agent of a further action. Construction grammar presents an unrealized potential: despite claims that it is in principle possible to describe the entire grammar of a language in terms of a network of basic constructions, no one has ever presented such a network for the grammar of a language. We take on this challenge and in Section 3.1 we present the network of basic grammatical constructions in Russian. his network includes a peripheral subset of impersonal constructions in which the dative case marks noncanonical subjects. he focus of Section 3.2 is a Russian construction in this subset that is composed of a dative noun phrase (tagged “d”), a inite impersonal verb, and an ininitive form (“inf”). Examples (1a) and (2a) share this dative impersonal construction (underlined):3 3. Examples 1 and 2 also contain glosses for various cases: n = nominative, a = accusative, g = genitive, l = locative.

he role of non-canonical subjects in the overall grammar of a language 297

Nam nadoelo tratit’ vremja i sily, Us-d bored waste-inf time and energy-a čtoby dokazyvat’ očevidnye… vešči. in-order prove-inf obvious things-a ‘We got bored with wasting time and energy to prove obvious things.’ b. Nam nadoela èta situacija/utrata vremeni i sily. Us-d bored this situation/loss-n time-g and energy-g ‘We got bored with this situation/the loss of time and energy.’

(1) a.

Xorošo by zadumat’sja ob ètom do togo, Good would think-inf about this-l before that-g kak vam dejstvitel’no ponadobitsja pokazat’sja vraču. how you-d really needs show-self-inf doctor-f ‘It would be good to think about this before you really need to go and see a doctor.’ b. *Vam ponadobitsja vrač. You-d needs doctor-n ‘*To you will be needed doctor.’

(2) a.

hough the structures in (1a) and (2a) look the same, they do not behave in entirely the same way: the inite verbs they contain have distinct properties. Examples like these difer in the degree of integration of inite and ininitive verbs, the status of the ininitive (whether or not it functions as a grammatical subject), and the status of the dative (whether or not it functions as a semantic subject). hese diferences correspond to the existence vs. non-existence of related constructions, as in (1b), where we see that the verb nadoest’ ‘bore’ can have a nominative subject, whereas ponadobit’sja ‘need’ cannot. In other words, relationships to other constructions are crucial in determining how diferent verbs behave in this construction. An elicitation test showed that speakers agree on which inite verbs behave like nadoest’ ‘bore’ and which behave like ponadobit’sja’ ‘need’, making it possible to establish which verbs belong to each type. Section 4 contrasts the Russian dative impersonal constructions examined in Section 3.2 with their translation equivalents in English, which consist of personal constructions. A questionnaire was used to probe whether there is a diference between Russian and English perception of the causes motivating situations that are described using impersonal constructions in Russian, but personal constructions in English. We did not ind any signiicant diference between Russian and English respondents. It may be the case that although the grammatical constructions in the two languages are diferent, the “thinking for speaking” (Slobin 1996) involved is not diferent enough to be measured, at least not by this kind of test.

298 Laura Janda and Dagmar Divjak

3. he place of non-canonical subjects in the grammar of Russian Grammatical case plays a pivotal role in Russian syntax. A brief characterization of each of the six grammatical cases is presented here, focusing only on relevant uses; for a fuller description see Janda & Clancy (2002). he nominative case typically marks grammatical subjects, the accusative marks the direct object, the dative marks the indirect object, and the instrumental marks instruments. However, some verbs govern the dative or instrumental, marking their objects with those cases instead of the accusative. Additionally, the accusative, dative, and instrumental can be used with various prepositions and in adverbials such as time expressions. he genitive case is primarily used adnominally and with prepositions, but there are verbs that govern this case as well. he locative case is used only with prepositions. 3.1

he network of basic constructions in Russian

We present a major portion of Russian grammar as a network of grammatical constructions depicted in Figure 1, which we invite the reader to compare with the diagrams presented by Kyröläinen in the previous chapter in this volume.

V+A+I N+V+A+I

N+V+I

V+A

N+V+A

D+V+A

N+V+A+D

D+V

N+V+G N+V+D N+V+PP N+V

Figure 1. he basic grammatical constructions of Russian as a radial category.

he network in Figure 1 is mostly limited to active indicative inite clauses, though in places it overlaps with passive and non-inite constructions as well as with constructions in other moods. Still, it represents the basic constructions that constitute a core portion of Russian grammar, centered around the basic transitive

he role of non-canonical subjects in the overall grammar of a language 299

construction (n + v + a in the igure), as illustrated in Example (3) (all other constructions in this network are illustrated with examples below). (3) Ženščina šila kostjum. Woman-n sewed suit-a ‘he woman sewed a suit.’

he elements in the constructions are verbs (v), noun phrases marked in the nominative (n), accusative (a), dative (d), instrumental (i), or genitive (g) case, or prepositional phrases (pp). hough the elements are presented in linear order (such as n + v + a) in Figure 1, the actual order is variable; only the combination of elements is crucial. We see two kinds of relationships in Figure 1: ones that we term “metonymical”, and ones that we term “metaphorical”. he solid lines show relationships where there is one more/less element in the neighboring construction. For example, a solid line connects the n + v + a construction with the neighboring n + v + a + d construction, since the latter construction has one more element (“d”, a dative-marked noun phrase) than the former one. he solid lines thus show extensions via metonymy, since some constructions are parts of other, larger constructions. We consider this relationship to be metonymical because the patterns that are afected by it show a part-whole relationship, although alternatively one could also interpret this as a relationship of schematicity in which the “extra” element in the larger construction is an optional part that may or may not be added to the simpler construction. he dotted lines indicate relationships where one element appears instead of another one. For example, a dotted line connects the n + v + a construction with the neighboring n + v + i construction because the two difer in the presence of an accusative noun phrase in the former as opposed to an instrumental noun phrase in the latter. he dotted lines thus show extensions via metaphor, where a structure with one element is mapped to a structure with a diferent element. Alternatively this relationship can also be interpreted as an alternation between two diferent cases.4 he n + v + a + d construction and the n + v + i construction are illustrated in Examples (4) and (5).

4. Technically it would be possible to put a dotted line between n + v + a + i and n + v + a + d. However, whereas it is possible to ind examples that show a semantic relationship between all of the relationships that have dotted lines in Figure 1, we have not been able to discover relevant examples for n + v + a + i and n + v + a + d. In other words, for all of the dotted lines in Figure 1, we can show examples where closely synonymous meanings are involved in the linked constructions (cf. Examples (15)–(17) below), but this is not the case for n + v + a + i and n + v + a + d.

300 Laura Janda and Dagmar Divjak

(4) Ženščina šila mužu kostjum. Woman-n sewed husband-d suit-a ‘he woman sewed a suit for her husband.’ (5) Ženščina xlopnula dver’ju. Woman-n slammed door-i ‘he woman slammed the door.’

he remaining constructions depicted in Figure 1 are illustrated in Examples (6) through (14), beginning with the v + a construction and proceeding clockwise around the periphery of the igure. (6) V + A:

Ženščinu tošnilo. Woman-a nauseated ‘he woman felt nauseated.’

(7) D + V + A:

Gde nam iskat’ pravdu? Where us-d seek-inf truth-a ‘Where can we look for the truth?’

(8) D + V:

Ženščine nadoelo rabotat’. Woman-d bored work-inf ‘he woman got bored of working.’

(9) N + V + D:

Ženščina zaplatila advokatu. Woman-n paid lawyer-d ‘he woman paid the lawyer.’

(10) N + V:

Ženščina spala. Woman-n slept ‘he woman slept.’

(11) N + V + PP:

Ženščina rabotala v kabinete. Woman-n worked in oice-l ‘he woman worked in the oice.’

(12) N + V + G:

Ženščina izbegala neprijatnostej. Woman-n avoided unpleasant-things-g ‘he woman avoided unpleasant things.’

(13) N + V + A + I: Ženščina rezala rybu nožom. Woman-n cut ish-a knife-i ‘he woman cut the ish with a knife.’ (14) V + A + I:

Ženščinu ubilo tokom. Woman-a killed electric-shock-i ‘he woman was killed by an electric shock.’

he role of non-canonical subjects in the overall grammar of a language

Figure 1 presents the core constructions of Russian, illustrated in Examples (3) through (14), as a radial category, centered around the n + v+a construction, which we argue is prototypical. Note that the v + a + i construction is quite parallel to the comitative, elative, and läbi ‘through’ constructions typical of non-volitional agents and inanimate actors as described by Lindström in her chapter in this volume. here are at least two reasons to consider n + v + a construction the prototype for this network. One reason is semantic, in that this construction instantiates Langacker’s (1991: 285–286) “canonical event model”, the most salient type of event. he other reason has to do with the structure of the network. A prototype is usually directly connected to more subcategories than any other subcategory (Lakof 1987; Geeraerts 1995; Crot & Cruse 2004; and Lewandowska-Tomaszczyk 2007). As we see in Figure 1, the n + v + a construction has more direct connections to the remaining constructions than any other construction: there are four links motivated by relations of addition/removal of participants (solid lines = metonymic), and ive links motivated by substitution of participants (dotted lines = metaphorical), giving a total of eight links. he use of a thicker frame around n + v + a symbolizes the prototypical status of this construction in the igure. he primary focus of this article is the d + v construction in (8). his construction is labelled “d v inf” by Kyröläinen (in the previous chapter in this volume), and he inds the dative noun phrase in this construction to be very similar to a nominative subject. his construction is inluenced by its mapping (metaphorical) relationship to v + a and its metonymic relationship to n + v + d. However, before detailing these two relationships, it is necessary to understand the overall nature of the relationships between constructions in the radial network presented in Figure 1. Although the network is drawn in terms of discrete nodes and the diferences between the core constructions involve discrete syntactic changes (addition/ removal or change of a participant), the relationships that connect the constructions have a continuous semantic nature. his is true of both the metaphorical and the metonymic relations in the radial category, as we have demonstrated in detail in Divjak & Janda (2008) and Janda (2008). hose arguments are summarized here. Concerning the metaphorical transitions between n + v + a and constructions with the same number of participants, but the genitive, dative, or instrumental case instead of the accusative, we ind that oten the same verb or a verb with a near-synonymous meaning can be found in parallel constructions. Examples (15a–b) illustrate the use of the same verb in both n + v + a and n + v + g constructions, and Examples (16a–b) and (17a–b) illustrate the use of near-synonyms in the n + v + d and n + v + i constructions. In all three instances, only subtle diferences in construal are involved in the choice between the “a” and “b” examples.

301

302 Laura Janda and Dagmar Divjak

N + V + A: Ženščina bojalas’ smert’. Woman-n feared death-a ‘he woman feared death.’ b. N + V + G: Ženščina bojalas’ smerti. Woman-n feared death-g ‘he woman feared death.’

(15) a.

N + V + A: Ženščina soprovoždala muža. Woman-n accompanied husband-a ‘he woman accompanied her husband.’ b. N + V + D: Ženščina soputstvovala mužu. Woman-n accompanied husband-d ‘he woman accompanied her husband.’

(16) a.

N + V + A: Ženščina brosala kamni. Woman-n threw stones-a ‘he woman was throwing stones.’ b. N + V + I: Ženščina brosalas’ kamnjami. Woman-n threw stones-i ‘he woman was throwing stones.’

(17) a.

‘Death’ in (15a) can refer to something more speciic, deinite, than in (15b), where it is potentially more general. (16a) is a more neutral statement in relation to (16b) which can imply that the woman submits herself to follow the path of her husband. (17a) is a statement about what happened to some stones, whereas in (17b) the stones are conceptualized more as an instrument for carrying out an act of throwing. Of course there are also many verbs that show more disparate semantics, and the core meaning of each case is quite distinct. Overall, the accusative is the most neutral way to mark a patient in a transitive construction. he genitive is largely restricted to use with verbs describing approach or withdrawal; in (15b) bojat’sja ‘fear’ denotes an emotional withdrawal (Janda & Clancy 2002: 118). he dative appears with verbs denoting giving, communication, beneit, harm, submission, and matching; in (16b) soputstvovat’ ‘accompany’ is an example of a verb denoting a matching action (Janda & Clancy 2002: 102). Some actions require the presence of an instrument under the agent’s control, and brosat’sja ‘throw’ requires that something be thrown (Janda & Clancy 2002: 26). However, as Examples (15)–(17) show, there are areas of semantic contiguity in this sub-network of Russian syntax. All of these examples could be multiplied, and together they show that the metaphorical relationships between the n + v + a construction and its neighbors are continuous, presenting language users with some nearly synonymous choices, despite discrete syntactic diferences.

he role of non-canonical subjects in the overall grammar of a language 303

A similar continuum is established in the metonymic relationship between the n + v + a construction and the n + v construction by the fact that many transitive verbs can be used without overtly stating a direct object. In other words, transitive verbs oten can be used in both constructions, as we see in Examples (18a–b). N + V + A: Ženščina šila kostjum. Woman-n sewed suit-a ‘he woman sewed a suit.’ b. N + V: Ženščina šila. Woman-n sewed ‘he woman sewed.’

(18) a.

he overt expression of the object is oten optional, creating a continuum between the two constructions, and parallel examples could also be cited for the relation of the n + v construction to the other constructions it has metonymic links to, namely n + v + pp, n + v + g, n + v + i, and n + v + d. his article focuses on the d + v construction (see Example (8) above), which, like all other constructions in the network, is inluenced by its relationships to neighboring constructions. Here the most crucial links are the metaphorical one to the v + a construction and the metonymic one to the n + v + d construction, paralleled by the link between v + a and n + v + a. Table 1 presents examples illustrating the parallels among these four constructions. Table 1. he v + d construction and its neighborhood. n+v+d

v+d

v+a

n+v+a

v can have n subject

Ženščine nadoel mužčina [Woman-d bored man-n] ‘he man bored the woman’

Ženščine nadoelo (rabotat’) [Woman-d bored (work-inf)] ‘he woman got bored (of working)’

Ženščinu trjaslo [Woman-a shook] ‘he woman got shaken up’

Ženščinu trjas mužčina [Woman-a shook man-n] ‘he man shook the woman’

v cannot have n subject

XXX

Ženščine prišlos’ rabotat’ [Woman-d forced work] ‘he woman had to work’

Ženščinu tošnilo [Woman-a nauseated] ‘he woman felt nauseated’

XXX

he verbs in Table 1 describe situations of discomfort or compulsion, although verbs that describe good fortune and enjoyment can also be used to illustrate the relationships among these constructions. he irst thing to notice in Table 1 is the mid portion of the table containing the v + d and v + a constructions and the two

304 Laura Janda and Dagmar Divjak

examples that illustrate each construction. hese are both impersonal constructions and they have contiguous, if not directly overlapping, semantics in that they involve situations that are beyond the control of the referent of the noun phrase (the woman), whether it is marked in the dative or the accusative. here is thus a continuum between these two constructions similar to the continua illustrated in Examples (15)–(17) above. Both the v + d and the v + a construction use the default neuter singular morphology on the verb (with -o desinence in the past tense) since they have no grammatical subject to agree with. he second thing to notice in Table 1 is the presence of the ininitive verb form rabotat’ ‘work’ in the two examples of the v + d construction. he ininitive expresses the action that the dative-marked noun phrase can be the subject of. In the construction with nadoelo ‘bored’, the ininitive is not obligatory, since it is possible to say just Ženščine nadoelo ‘he woman got bored’, making this construction parallel to the v + a example Ženščinu trjaslo ‘he woman got shaken up’, except for the diference in case. It is oten possible to have an ininitive verb form in the v + d construction, but not in the v + a construction; the experiment described in Section 4 focuses speciically on the v + d construction with an ininitive. We turn now to the let-most column in Table 1, labeled “n + v + d”, and compare it with the column under “v + d”. In addition to the optional presence of an ininitive in the v + d construction, we note that some verbs in this construction can also appear in the n + v + d construction with a nominative subject, whereas other verbs cannot appear in this construction. his distinction is represented by the two rows in Table  1: the irst row, labeled “v can have n subject” presents examples showing that the same verb, here nadoest’ ‘bore’, can appear in both constructions, and such verbs serve to metonymically link the two constructions since these verbs can appear in both the larger, full construction (n + v + d) and the reduced, partial construction (v + d). Note also that whereas the form nadoelo ‘bored’ in the v + d construction shows the neuter singular -o, the form nadoel ‘bored’ in the n + v + d construction is masculine singular, agreeing with the subject mužčina ‘man’. If we compare the rightmost two columns in Table 1, we ind that the same metonymic link is observed also between the v + a and n + v + a constructions, where the verb trjasti ‘shake’ appears in both constructions, with the neuter singular form trjaslo ‘shook’ in the impersonal v + a construction, but the masculine singular trjas ‘shook’ in the personal n + v + a construction. In the bottom row of Table 1, labeled “v cannot have n subject”, we present examples of verbs that can only appear in impersonal constructions. hese verbs cannot open a nominative slot. For the v + d construction, prišlos’ ‘was forced to’ is such a verb and happens to require the presence of an ininitive, although some other verbs of this type do not require an ininitive (such as povezlo ‘got lucky’).

he role of non-canonical subjects in the overall grammar of a language 305

he v + a construction also has some verbs like tošnit’ ‘feel nauseated’ and znobit’ ‘feel cold’ that function exclusively in impersonal constructions. We argue that the dative participant in the v + d construction is a non-canonical subject because (a) it is always a human experiencer and (b) in the presence of an ininitive, the dative participant identiies the agent of the action described by the ininitive. Returning to the v + d examples in Table 1, in both examples the woman is the experiencer of a situation of annoyance or compulsion and the woman is also the agent of the action named by rabotat’ ‘work’, since both examples assume that it is possible to say Ženščina rabotaet ‘he woman works’. he remainder of this article focuses on verbs in the v + d construction with an accompanying ininitive form. his includes both verbs that can open a nominative slot and verbs that cannot do so. Both types of verbs are investigated in the experiment described in Section 4. 3.2

Two kinds of non-canonical subject status

his section presents a close examination of the v + d construction with an ininitive form. In these constructions, as demonstrated above, the dative participant is an experiencer, and thus a kind of non-canonical subject. We developed a test that was run on all verbs that can appear in this construction (Divjak & Janda 2008). he test makes it possible to distinguish which verbs can open a nominative slot and which cannot. hanks to this test we were able to identify the two groups of verbs and further analyze their behavior. We show that if a verb can open a nominative slot, the ininitive occupies that slot in this construction, serving also as a kind of non-canonical subject. Verbs that cannot open a nominative slot have no grammatical subject at all, only the dative experiencer. We argue that there is also diference in the role of the dative participant: if the verb can open a nominative slot, the dative is a true experiencer, but if it cannot open a nominative, the dative is a more agentive experiencer. In Examples (19a–c) we see that the verb nadoest’ ‘bore’ can have three kinds of subjects. In (19a) the noun gosslužba ‘civil service’ is the grammatical subject, appearing in the nominative case. hus, as we saw above in Table 1, this verb can open a nominative slot. But this slot can also be illed with two other items, neither of which are noun phrases. In (19b) the ininitive form tratit’ ‘waste’ ills this slot, since it is wasting that causes or is the agent of boredom. his example illustrates the v + d construction with an ininitive form that we focus on here. (19c) shows that there is a third option, namely that an entire clause can ill this slot.5 5. Examples (19)–(22) are cited from the Russian National Corpus, together with the listing of the author, source, and date.

306 Laura Janda and Dagmar Divjak

(19) a.

Vy motivirovali svoj uxod tem, čto You-n motivated own departure-a that-i that gosslužba nadoela. civil service-n bored ‘You motivated your departure by (saying) that civil service had become boring.’ [Svetlana Oitova. Glavnaja zadača – oslablenie gosudartsva // “Nezavisimaja gazeta”, 2003. 04.09]

b. Nadoelo tratit’ vremja i sily, Bored waste-inf time-a and energy-a čtoby dokazyvat’ očevidnye … vešči. in-order prove-inf obvious things-a ‘Wasting time and energy to prove obvious things has gotten boring.’ [Vtoroj s”jezd // “Specnaz Rossii”, 2003.05.15] c.

Generalu nadoelo, čto gazety pišut General-d bored that newspapers-n write pro nego vsjakie domysly. about him-a various conjectures-a ‘he general got bored with the newspapers writing all kinds of conjectures about him.’ [Petr Akopov. Aušev dal kljatvu ne byt’ prezidentom // “Izvestija”, 2002.01.24]

We follow Browne (1987: 166) in recognizing the ininitive form as a noncanonical subject: “[Subjects are] those noun phrases with which the verb agrees in person and number (in gender too, for some verb forms). hen we observe that an ininitive construction or a subordinate clause can play the same role as a noun phrase and is mutually exclusive with it; therefore we extend the term “subject” to these ininitive constructions or clauses, and mention in our description the special verb-agreement which they are associated with (3rd singular neuter)”. By contrast, there are also verbs that appear in the v + d construction with an ininitive, but cannot open a nominative slot, such as those in Examples (20)– (21). Here we have no grammatical subject at all, only the dative. In Example (20), the dative-marked pronoun Vam ‘you’ identiies the agentive experiencer since the inite verb ponadobitsja ‘needs’ has no grammatical subject. his verb can open a nominative slot in its here intended sense but the ininitive pokazat’sja ‘show oneself ’ cannot serve this role. he verb nadležit ‘requires’ in (21) also has the dative agentive experiencer Vam ‘you’, but does not open up a nominative slot hence the accompanying ininitive vstretit’ ‘meet’ cannot serve as a subject.

he role of non-canonical subjects in the overall grammar of a language 307

(20) Xorošo by zadumat’sja ob ètom do togo, Good would think-inf about this-l before that-g kak vam dejstvitel’no ponadobitsja pokazat’sja vraču. how you-d really needs show-self-inf doctor-d ‘It would be good to think about this before you really need to go and see a doctor.’ (21) Vam nadležit vstretit’ ètu nuždu, You-d requires meet-inf that need-a nakormit’ ètix golodnyx ljudej! feed-inf these hungry people-a ‘You have to meet that need, feed these hungry people!’ [Antonij (Blum), mitropolit Surožskij. Čudo pjati xlebov i dvux rybok. (1980)]

here are eighty-one verbs in Russian that can appear in the v + d construction with an ininitive. However, there is no a priori way to know which of these verbs have an ininitive non-canonical subject and which ones do not. We designed an elicitation test for argument structure based on substitution with pronouns and ininitive phrases in order to distinguish the two types of verbs (Divjak & Janda 2008). Native speakers were presented with Russian phrases containing the structure inite verb + ininitive, such as His parents decided to buy him a Mercedes. hey were then asked to form or rate phrases with the same inite and ininitive verbs with this structure: What did his parents decide?; To buy him a Mercedes; To buy him a Mercedes, that’s what they decided. hese phrases show whether it is possible, as in English, to substitute the ininitive form with a pronoun such as what or that. If this is possible, then the verb is of the type that opens a nominative slot and the ininitive occupies that slot. If this is not possible, the verb has no grammatical subject. he experiment was set up as a small number design in which we worked with ive native speakers of Russian, between the ages of twenty-ive and ity. hey judged the constructional possibilities of these verbs on a three-point scale. Several measures were taken to minimize possible problems with this set-up. Native speakers were asked both to rate ready-made sentences and to form sentences using particular constructional devices; these sentences were on a later occasion presented to the participant who had constructed them as well as to other participants. To guard against lexical efects, the tests were carried out using pronouns and other pro-forms, which ensures that the mutual efect of lexical items in a construction is minimized. Moreover, to check for repetition efects in judgments of grammaticality ten control judgments were collected for every verb from other native speakers. In this experiment, the trigger questions were mixed with other, non-related questions about Russian syntax and semantics. he consistency in the answers was almost without exceptions.

308 Laura Janda and Dagmar Divjak

Here are what the results looked like, starting irst with ininitives that are used as a grammatical subject with a verb such as nadoest’ ‘bore’. In Examples (22a–c), the ininitive event, i.e. čitat’ ‘read’, can be substituted with the pronoun čto ‘what’. As we see in (22c), one can also use the ininitive čitat’ ‘read’ to answer the question posed in (22b); an alternative is a noun such as čtenie ‘reading’. Rebenku nadoelo čitat’. Child-d bored read-inf ‘Reading bored the child.’ b. Čto rebenku nadoelo? What-n child-d bored ‘What bored the child?’ c. Čitat’. OR Čtenie. Read-inf OR Reading-n ‘Reading.’

(22) a.

With this type of inite verb there is a nominative slot, and the ininitive čitat’ ‘read’ its into the nominative slot. his makes the ininitive a subject, although not a prototypical subject. he dative slot, occupied by a person, is then free to take on its typical function, that of an experiencer. Reiication (Langacker 1987) explains these facts: an ininitive event is reduced to and treated like any other “thing” (cf. Smith 1994) that can be the subject of the inite verb event and bring it about. his substitution is possible for forty-four of the eighty-one inite verbs that are known to occur in the v + d construction with an ininitive. For the remaining thirty-seven inite verbs, the ininitive cannot ill the subject slot. In Example (23), we see that only a noun such as lekarstvo ‘medicine’ can appear in that slot. Bol’nomu ponadobitsja pokazat’sja vraču. Patient-d needs show-self-inf doctor-d ‘he patient will need to go and see a doctor.’ b. Čto bol’nomu ponadobitsja? What-n patient-d needs ‘What will the patient need?’ c. *Pokazat’sja vraču. VS Lekarstvo. Show-self-inf doctor-d VS Medicine-n ‘*To go and see a doctor.’ ‘Medicine.’

(23) a.

Some of the inite verbs that cannot open a nominative slot and do not tolerate the ininitive in that slot are morphologically defective, having forms only in the third person (neuter) singular. An example is given in (24), where we see that even the question with the pronoun (24b) is infelicitous, though this can be corrected by adding a pro-verb such as sdelat’ ‘do’ (24c):

he role of non-canonical subjects in the overall grammar of a language 309

Vam nadležit javit’sja v ukazannyj srok. You-d requires appear-inf in appointed time-a ‘You are required to appear at the appointed time.’ b. *Čto vam nadležit? What-n you-d requires ‘*What are you required?’ c. Čto vam nadležit sdelat’? What-n you-d requires do-inf ‘What are you required to do?’

(24) a.

he ininitive phrases pokazat’sja vraču ‘go and see a doctor’ and javit’sja v ukazannyj srok ‘appear at the appointed time’ in (23) and (24) do not it in the nominative slot occupied by čto ‘what’. he inite verb does not function as construction kernel (semantically the main verb rather than one that serves an auxiliary function) (cf. Butler 1967). No reiication is possible here: the ininitive events go and see a doctor or appear at the appointed time cannot be reduced to, or treated like any other “thing” that can be the subject of the inite verb event need or require. In other words, the ininitive event does not initiate the inite verb event. With these inite verbs, the ininitive is stronger than usual, and the inite verb needs the ininitive to carry the load of the construction. Here the role of the inite verb is similar to that of an auxiliary. he inite verb event modiies the ininite verb event and together the inite verb and the ininitive indicate a complex event. his status is claimed for modal verbs in general and its the verbs that display this pattern as well. Among such inite verbs there are two major semantic groups: a group of non-implicative verbs that express modal-like concepts, and a group of implicative verbs that stress the result obtained. Some examples of these inite verbs are presented in (25): (25) a.

Modality verbs: Volition: xočetsja ‘feels like’, ne terpitsja ‘is impatient’, xvatit ‘is suicient’ Suitability: goditsja ‘is suited for’, nadležit ‘requires’, polagaetsja ‘is supposed to’ Necessity: trebuetsja ‘is needed’, predstoit ‘is in store for’, ostaetsja ‘is necessary’ b. Result verbs: Success: udastsja ‘succeeds’ Success + Associated (mis)fortune: povezet ‘gets lucky’, posčastlivitsja ‘is lucky’ Success + Cause: dovedetsja ‘gets the chance to’, slučitsja ‘happens’

When these inite verbs appear in the v + d construction with an ininitive, the ininitive deines the construction’s kernel. he event expressed by the ininitive is the only one that can combine with a nominative (canonical) subject, as in Example (26). We label this subject a “True Agent”:

310 Laura Janda and Dagmar Divjak

(26) Bol’noj pokazalsja vraču. Patient-n showed-self doctor-d ‘he patient went to see a doctor.’

he True Agent in (26) uses the same noun phrase as the one that appears in the dative case as a grammatical subject in (23a). In (23a) the inite verb cannot open a nominative slot, the ininitive cannot serve as a subject, and indeed there is no subject. he person named in the dative noun phrase is at once the experiencer of the inite verb + ininitive complex and the agent of the action named in the ininitive. We call this participant an “Agentive Experiencer”. he dative noun phrase in Example (22) is combined with a inite verb that does open a nominative slot. Here the ininitive does serve as a non-canonical subject, since it can be replaced with a nominative noun phrase. We call the dative noun phrase here a “True Experiencer”. Overall, the three types of participants can be viewed as lying at diferent points along a continuum of agentivity, where the True Agent has the highest agentivity, the True Experiencer has the lowest agentivity, and the Agentive Experiencer is between the two extremes, as visualized in (27): (27) Scale of Agentivity True Agent > Agentive Experiencer > True Experiencer

his section focused on the impersonal verbs that appear in the v + d construction with an ininitive. Native speakers concur that forty-four inite verbs in this construction open a nominative slot, while thirty-seven do not. In combination with the inite verbs that open a nominative slot, the ininitive ills that slot and the dative participant is a True Experiencer. In combination with the inite verbs that do not open a nominative slot, the ininitive forms a complex event, and the inite verb behaves like an auxiliary. With these auxiliary-like inite verbs, the dative participant straddles two roles as a non-canonical subject, since it is at once the experiencer of the situation described by the inite verb and the initiator of the event named by the ininitive. his dative participant is an Agentive Experiencer. he next section presents an experiment comparing the inferences made by Russian speakers in response to sentences containing Agentive Experiencers and True Experiencers with the inferences made by English speakers who are confronted with True Agents in the translation equivalents.

he role of non-canonical subjects in the overall grammar of a language

4. A discourse-cohesion experiment6 Russian and English express agency diferently. Russian has no modal verbs (except for moč’ ‘be able’), but abounds in impersonal constructions, e.g. mne xolodno/48 let [Me-d cold/48 years] ‘I’m cold/48 yrs old’. English by contrast has many modal verbs and favors personal subject-headed constructions. Oten, Russian impersonal v + d constructions correspond to English personal constructions containing a inite verb that agrees with a (nominative) subject, as in Mne xočetsja spat’ [Me-d wants sleep-inf] ‘I feel like sleeping’. For the expression of experiences such as necessity, obligation, volition, suitability, and success Russian avoids employing personal constructions. English, on the contrary, lacks true impersonal constructions altogether. In other words, although the (nominative) subject typically ills the role of agent in both English and Russian, in English constructions expressing the above-mentioned experiences a grammatical subject is present, while Russian uses a non-canonical subject, namely an experiencer participant. Does this grammatical diference inluence the way speakers of Russian and English think about a situation? In particular, given an event that is an experience, do we see diferent results of the events are actions metaphor, identifying different items that can be attributed agent-like status? One might expect that speakers of English will be more likely to assign agency to the person that serves as the subject of the verb, whereas Russian speakers would be less likely to identify the dative-marked experiencers in equivalent phrases as the agents. Perhaps Russians will be more likely to choose a direct object or a circumstance as the cause? We designed an experiment to address these questions. We used a questionnaire to probe whether speakers of Russian and English perceive a diference in discourse cohesion between causes motivating situations that are described using impersonal constructions in Russian, but personal constructions in English. Native speakers of the two languages were given trigger sentences that contained an impersonal v + d construction in Russian, but the translation equivalent in a personal construction in English. he participants were shown a trigger sentence and asked to decide whether a second sentence would be a good way to continue the narrative in the trigger sentence.

6. We express our thanks to: Dr. Stef Grondelaers (K.U. Leuven, Belgium), Dr. Christina Hellman (SU, Sweden) and Prof. Stefan Gries (UCSB, USA) for discussing the experimental set-up; Prof. Masja Koptjevskaja (SU, Sweden) and Eleonora Magomedova (UNC, USA) for scrutinizing the experimental items; our 72 participants for illing out the questionnaires; Dr. Chris Wiesen (UNC, USA) for statistical analysis.

311

312 Laura Janda and Dagmar Divjak

Example (28) presents a token set from the experiment. (28a) is a trigger sentence expressed with an impersonal construction in Russian, but a personal construction in English (here only the English is presented). (28b–d) ofer three alternative continuations for the narrative: (28) a. b. c. d.

I managed to read War and Peace I set all projects aside and only read for a whole week. he book was less boring than I had assumed. It rained all summer, which kept me at home.

(28b–d) make diferent assertions about the cause for success stated in (28a). (28b) asserts that I succeeded because of something that I did. his alternative assigns agency to the referent of the pronoun I, which is also the nominative subject of the inite verb in (28a) in English, but in the Russian equivalent, this referent appears as a dative experiencer. (28c) gives a diferent interpretation, namely that the object, the book, is what made success possible. (28d) identiies an external factor, namely the weather, as the cause. Given that the English trigger sentence presents the person experiencing success as an agent, one might expect English speakers to prefer (28b) with the same referent in the same canonical subject position. Russian marks the person who succeeds with the dative case in an impersonal construction; the success is grammatically presented as something that happens to that person, not something that they have brought about. Given this fact, one might expect Russian speakers to prefer (28c) or (28d), which assign the cause for success to the object or an external circumstance. his experiment was implemented by means of a questionnaire containing pairs of sentences from token sets similar to Example (28), where the inite verb in the trigger sentence was an impersonal verb in a v + d construction in Russian, but a personal verb in English. he second sentence corresponded to (28b), (c), or (d). In Russian the questionnaire contained six verbs that do not open a nominative slot, where the dative is an Agentive Experiencer, plus six verbs that do open a nominative slot, where the dative is a True Experiencer. he twelve Russian verbs used in the questionnaire and their English equivalents are presented in (29):7

7. Note that many of these verbs are the same as those cited in (25). However in (25) all the verbs are cited in their third person present forms, whereas in (29) some of the same verbs are cited in their neuter singular past forms. hese are the only two forms that are relevant since these verbs lack a full paradigm in the v + d construction. he forms cited in (29) relect those used in the questionnaire, where two verbs appeared in the present tense form (since that is the form they are mostly found in), as relected in the English translations.

he role of non-canonical subjects in the overall grammar of a language

(29) a.

Russian verbs that do not open a nominative slot: ostalos’ ‘had to’, povezlo ‘got lucky’, polagalos’ ‘was supposed to’, prišlos’ ‘had to’, xotelos’ ‘felt like’, udalos’ ‘managed to’ b. Russian verbs that do open a nominative slot: grozilo ‘was in danger of ’, idet ‘suits, looks good (on)’, l’stilo ‘was lattered to’, nravilos’ ‘pleased, enjoyed’, oprotivelo ‘got sick of ’, ne svetit ‘is not fated to’

For each trigger sentence as in (28a), respondents were asked to rate the appropriateness of a continuation sentence, as in (28b–d). We hypothesized that Russian respondents would be most likely to rate the Object or Circumstance continuation sentences higher than the Subject continuations, since the trigger sentences all presented the events in impersonal constructions. English respondents should show a higher preference for Subject continuations, since they parallel the structure of the personal constructions in trigger sentences. We distributed 36 questionnaires per language; the English ones went to college-age, non-linguists, non-Slavists, who responded in a classroom setting. he Russian ones were sent to native speakers of Russian who responded via email, and who represented a broader range of ages and backgrounds. Each questionnaire contained six benchmark sentence pairs, twelve iller sentence pairs, and six trigger sentence pairs. Only the trigger sentence pairs contained the relevant verbs, and only those items were considered in the statistical analysis. he benchmark sentence pairs (three at the beginning of the questionnaire and three at the end) were provided for training purposes and to make sure that the participants understood the task (participant reliability). he purpose of the iller sentence pairs was to prevent participants from guessing what we were testing. he trigger sentence pairs contained the items of interest, i.e. the independent variables. here were two types of independent variables, corresponding to the experiencer and the instigator. he experiencer had two levels: True Experiencer vs. Agentive Experiencer. he instigator had three levels, corresponding to the three possible continuation sentences: Subject (like 28b), Object (like 28c), and Circumstance (like 28d). For each of the twelve verbs we created three diferent token sets to minimize the inluence of lexical efects. Each participant judged each factor level combination only once and judged only one variable level combination from each token set. In other words, each participant saw only one possible continuation (Subject, Object, or Circumstance) for each trigger sentence. Both illers and triggers were presented in randomized order in every questionnaire to avoid order efects, making sure that no two triggers followed each other and that no questionnaire started or ended with a trigger item. Each questionnaire thus contained a series of pairs of sentences. No participant saw the same trigger sentence, or even the same trigger verb, more than once.

313

314 Laura Janda and Dagmar Divjak

he dependent variable was discourse coherence. In order to measure this parameter, each participant was asked to judge how well the two sentences in each pair it together. his judgement was measured on a ive-point Likert scale, where +2 meant that the sentences it very well together, +1 meant that they it well together, 0 meant there was a neutral relationship between sentences in a pair, −1 meant that the sentences did not it well together, and −2 meant that the sentences it together very poorly. As a result we collected 36 judgments for each variable level combination in each language. he data were analyzed using both the Means Model (which models means) and the Multinomial Model (which models proportions) and no statistically signiicant contrasts were found. his means that our data does not contain any evidence to support the hypothesis that speakers of Russian and English have diferent assumptions as to who or what is responsible for the need, opportunity, requirement, etc. to do something. It appears that speakers’ expectations are not guided by the structure of their language, but by the situation itself. Although we failed to ind a statistical diference in the way that Russian vs. English speakers rate the coherence of sentences from token sets like (28), this does not necessarily mean that there is no diference in the “thinking for speaking” that Russian and English speakers engage in. he test we used did not reveal any such diferences, but we must remain agnostic about whether such diferences do exist. Perhaps if these diferences exist, they could be revealed by diferent means. For example, it might be possible to design a diferent kind of psycholinguistic experiment with a diferent task, for example a (speeded) forced-choice task, or an online task using a very sensitive instrument such as an eye-tracker. Another alternative might involve an analysis of corpus data, perhaps from parallel Russian/ English corpora.

5. Conclusion his article has taken on an implicit challenge within construction grammar by showing that the core constructions of a language can be modelled as a radial category. he radial category of Russian core syntax reveals a network of constructions that are linked to each other via metonymic (addition/removal of participants) and metaphorical (mapping between diferent participants) relationships. Neighboring constructions show parallels and involve contiguous, partially overlapping semantics. For example the impersonal v + d construction behaves in ways somewhat parallel to the impersonal v + a construction, and both are inluenced by their metonymic relationships to personal constructions containing a canonical

he role of non-canonical subjects in the overall grammar of a language

nominative subject, namely the n + v + d and the n + v + a constructions. he v + d construction and the v + a construction are similar in that they admit two kinds of verbs: verbs that can appear in personal constructions (with a nominative subject), and also verbs that can only appear in impersonal constructions. For the v + d construction, this means that an accompanying ininitive can either serve as a type of non-canonical subject, illing the slot of the nominative, or, if the verb does not open a nominative slot, the inite verb is more auxiliary-like and there is no subject position available at all. In the irst case we claim that the dative participant is a True Experiencer, whereas in the second case we claim that the dative participant is an Agentive Experiencer. An elicitation experiment found that native speakers of Russian can consistently diferentiate between inite verbs that open a nominative slot and have a True Experiencer vs. inite verbs that cannot open a nominative slot and have an Agentive Experiencer. From the perspective of cognitive linguistics, diferent constructions are seen as diferent ways of encoding diferent sorts of relationships among elements. hus, diferences in constructional patterns can reveal diferences in semantic structure, and possibly also in conceptual structure. Whereas in Russian one expresses situations of need, requirement, frustration and the like using the impersonal v + d construction with the human experiencer in the dative case, the translation equivalents in English are personal constructions where the human experiencer is a canonical subject of the verb. It is reasonable to ask whether the grammatical diferences between the two languages correspond to a conceptual diference in terms of “thinking for speaking” (Slobin 1996). Perhaps the English preference for a nominative subject corresponds to a tendency to view the human experiencer as the instigator of a situation, whereas the Russian preference for a dative-marked (agentive) experiencer corresponds to a tendency to assign the cause to an object or circumstance? In order to probe this possibility, we designed a discourse cohesion experiment in which Russian and English speakers rated the compatibility of pairs of sentences. he irst sentence in each pair contained a v + d construction in Russian (for Russian speakers), or (for English speakers) its translation equivalent in English. he second sentence asserted that the instigator of the situation was either the experiencer, another object, or a circumstance. We did not ind any statistically signiicant diference between the Russian and the English responses. At this point, we can neither airm that there is a conceptual diference, nor exclude the possibility that such a diference might exist; it may be that a diferent kind of test or experimental design could still reveal a diference.

315

316 Laura Janda and Dagmar Divjak

Abbreviations a d g i inf l n pp v

Accusative Dative Genitive Instrumental Ininitive Locative Nominative Prepostitional phrase Verb

References Apresjan, Jurij Derenikovič (1967). Eksperimental’noe issledovanie russkogo glagola. [An experimental investigation into the semantics of the Russian verb]. Moskva: Nauka. Browne, Wayles (1987). Classiication of subordinate clauses in a grammar of Serbo-Croatian for foreign users. he Zagreb English-Serbo-Croatian Contrastive Project. Contrastive Analysis of English and Serbo-Croatian, 3, 165–191. Butler, I. G. (1967). Sočetanie ininitiva s modal’nymi i bezličnymi glagolami. Učenye zapiski MOPI, 167(3), 41–53. Crot, William (1998). he structure of events and the structure of language. In Michael Tomasello (Ed.), he new psychology of language: cognitive and functional approaches to language structure (pp. 67–92). Mahwah, NJ: Lawrence Erlbaum Associates. Crot, William (1999). Some contributions of typology to cognitive linguistics. In heo Janssen & Gisela Redeker (Eds.), Cognitive linguistics: Foundations, scope and methodology (pp.  61– 93). Berlin: Mouton de Gruyter. DOI: 10.1515/9783110803464.61 Crot, William (2001). Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780198299554.001.0001 Crot, William, & Cruse, Alan D. (2004). Cognitive linguistics. Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511803864 Divjak, Dagmar, & Janda, Laura A. (2008). Ways of attenuating agency in Russian. Transactions of the Philological Society, 106, 138–179. DOI: 10.1111/j.1467-968X.2008.00207.x Eggermont, Carmen, & Melis, Ludo (Eds.). (1992). he pronominal approach: From verb to noun phrase. Texts of the inal workshop PROTON I–II. Leuven: Linguistics Department, KUL. Eynde, Karel van den (1995). Methodological relections on descriptive linguistics. Knud Togebys principles and the pronominal approach. In Lene Schösler & Mary Talbot (Eds.), Studies in Valency I. (pp. 111–131). Odense University Press. Rask supplement vol. 1. Feldman, Jerome A. (2006). From molecule to metaphor: A neural theory of language. Cambridge, MA: MIT Press. Geeraerts, Dirk (1995). Representational formats in cognitive semantics. Folia Linguistica, 29, 21–41. DOI: 10.1515/lin.1995.29.1-2.21 Goldberg, Adele (1995). Constructions: A construction grammar approach to argument structure. Chicago: Chicago University Press.

he role of non-canonical subjects in the overall grammar of a language

Goldberg, Adele (2006). Constructions at work: he nature of generalizations in language. Oxford: Oxford University Press. Janda, Laura A. (2008). Transitivity in Russian from a cognitive perspective. In Galina Kustova (Ed.), Dinamičeskie modeli: Slovo. Predloženie. Tekst. Sbornik statej v čest’ E. V. Padučevoj (pp. 970–988). Moscow: Jazyki slavjanskoj kul’tury. Janda, Laura A., & Clancy, Steven J. (2002). he case book for Russian. Columbus, Ohio: Slavica. Lakof, George (1987). Women, ire and dangerous things. Chicago: University of Chicago Press. DOI: 10.7208/chicago/9780226471013.001.0001 Lakof, George (1993). he contemporary theory of metaphor. In Andrew Ortony (Ed.), Metaphor and throught (pp. 202–251). Cambridge: Cambridge University Press. DOI: 10.1017/CBO9781139173865.013 Lakof, George, & Turner, Mark (1989). More than cool reason. Chicago: University of Chicago Press. DOI: 10.7208/chicago/9780226470986.001.0001 Langacker, Ronald W. (1987). Foundations of cognitive grammar: heoretical prerequisites. Stanford: SUP. Langacker, Ronald W. (1991). Concept, image, and symbol: he cognitive basis of grammar. Berlin/New York: Mouton de Gruyter. DOI: 10.1515/9783110857733 Levin, Beth (1993). English verb classes and alternations: A preliminary investigation. Chicago, IL: University of Chicago Press. Lewandowska-Tomaszczyk, Barbara (2007). Polysemy, prototypes and radial categories. In Dirk Geeraerts, & Hubert Cuyckens (Eds.), Handbook of cognitive linguistics (pp. 139–169). Oxford: Oxford University Press. Schøsler, Lene, & Van Durme, Karen (1996). he Odense valency dictionary. An introduction. Odense working papers in language and communication 13. Odense: Institute of Language and Communication, OU. Slobin, Dan I. (1996). From “thought and language” to “thinking for speaking”. In J. J. Gumperz & S. C. Levinson (Eds.), Rethinking linguistic relativity (pp. 70–96). Cambridge: Cambridge University Press. Smith, Michael B. (1994). Agreement and iconicity in Russian impersonal constructions. Cognitive Linguistics, 5, 5–56. DOI: 10.1515/cogl.1994.5.1.5 Turner, Mark (1996). he literary mind. Oxford: Oxford University Press.

317

Construction index

C conditional clause 5, 48, 73, 75–77, 82, 88, 92–94 consequency construction 193–196, 199 copulative construction 26, 33 D da-ininitive construction (in Estonian) 73, 74, 79, 88, 94 Dative Adjective Nominative (D ADJ N) 259, 260, 263, 265, 272, 277, 282–286 Dative Adverb (D ADV) 260, 262, 266 (fn.), 276, 278, 280, 282–285 Dative Adverb Ininitive (D ADV INF) 267, 278, 282–285 Dative Ininitive (D INF) 258, 261, 265, 266 (fn.), 267, 268, 271, 274, 275, 277–280, 282–286 Dative Verb Ininitive (D V INF) 258, 269, 276–278, 282–285, 301 Dative Verb-SJA (D V[SJA]) 259, 260, 267–272, 275, 277, 278, 280, 284, 285 ditransitive construction 7, 258 E existential construction 4, 8, 22, 24, 31, 36–38, 103, 131 F Finnish agent participle construction 235, 236, 238, 243, 245

Finnish necessive construction 235–238, 240, 243–245 Finnish permissive construction 235, 236, 238–241, 243–245, 247 Finnish referative construction 235–238, 242, 244, 247 Finnish temporal construction 235–238, 243, 245 I if-clause 73, 78, 82, 83, 86–89 impersonal construction 3, 4, 7, 19, 22, 26, 79, 80, 104 fn., 109, 110 fn., 114, 117, 119, 131 fn., 136, 144, 253, 254, 259, 265, 266, 275, 276, 280, 296, 297, 304, 305, 311, 312, 313, 315 intransitive construction 210, 218, 257, 264, 294 M meteorological construction 8 Argument-predicate type meteorological construction 206–209, 215, 218 Argument type meteorological construction 206–209, 215, 218 Atransitive predicate type meteorological construction 207 Expletive predicate type meteorological construction 207

Intransitive predicate type meteorological construction 208 Predicate type meteorological construction 206–209, 214, 218 Transitive meteorological construction 218 N Nominative Verb Accusative (N V A) 259, 260, 264, 270, 275, 277, 278, 284–286 Nominative Verb (N V) 259–260, 264, 265, 269–288 Nominative Verb Accusative Dative (N V A D) 259, 260, 262, 263, 265, 266, 269, 270, 271, 273, 274, 276, 277, 279, 281, 284, 285, 288 Nominative Adjective Ininitive (N ADJ INF) 259, 260, 265, 271, 277, 282–286 non-prototypical trajector construction 6, 175, 176, 178 gram-driven non-prototypical trajector constructions 6, 178–185, 187, 188, 190, 194, 199 lexeme-driven non-prototypical trajector constructions 6, 175, 178, 187, 190, 191, 194 syntax-driven non-prototypical trajector constructions 6, 175, 178, 185–187, 190, 194, 198–199

320 Subjects in Constructions

P passive construction 6, 19, 22, 26, 73, 79, 81, 114, 142, 144–150, 155, 157, 160, 163–167, 260, 278, 279, 281, 283, 286, 298 participial construction 6, 146 possessive construction 8, 16, 23, 31, 32, 165, 168–169, 179 possessive perfect construction 142, 143, 146, 163, 166–168, 170

R relexive construction 7 result clause construction 34, 35 S setting subject construction 4, 8, 27–30, 38, 114–116 state clause construction 30, 32

T transitive construction 191, 195, 196, 210, 257, 258, 263, 288, 302 Z zero person construction 73–75, 79, 80, 82, 84–90, 94

Subject index

A abstract location 102, 121, 132, 136 abstract locational subject 5, 6, 101, 102, 105, 109, 110, 119, 135, 136 agent 3, 4, 6, 15, 19, 36, 38, 53, 55, 56, 58–60, 64, 67, 70, 80, 81, 86, 104 fn, 110, 111–114, 116, 141–143, 145–150, 156, 157, 162, 164–167, 169, 177, 181–184, 189, 192, 206, 209, 210, 215–217, 219, 223, 224, 226, 232, 235–238, 241–243, 245, 246, 248, 249, 293, 294, 296, 301, 302, 305, 309, 310–312 proto-agent 177–179, 181, 182, 185, 257, 262 agent orientation 4, 15, 36, 38 agentive(ity) 2, 5, 6, 15, 19, 20, 22, 23, 26, 36, 86, 88–90, 142, 146, 153, 157, 184, 189, 205, 226, 246, 305, 306, 310, 312, 313, 315 agreement 3, 13–18, 20, 28, 37, 63, 79, 103 fn, 104 fn, 105, 119 fn, 134, 143–145, 160, 176, 180 fn, 182, 194, 205, 212, 213, 220–222, 226, 232, 234, 244–249, 253, 254–256, 263, 265, 266, 279, 282, 285, 286, 295, 296, 306 alignment 16, 113, 218, 253, 258, 259, 263 animacy 2, 3, 23, 150, 157, 162, 170, 246, 259–261, 282, 283, 285–287 antecedent (sub)event 6, 175, 188–191, 193–195, 197–199 aspect(ual) 24–26, 78, 177, 234 fn, 276–278, 286

atransitive 207, 213 autonomy 2, 33, 262, 263, 282, 286 B basic reality 101, 122, 124, 126, 132, 136 behavioral properties 166, 169, 176, 206, 208, 222 bounded(ness) 24–26, 111, 234 fn C case (marking) 3, 6, 8, 13–18, 20, 22, 24–26, 28, 29, 31–35, 37, 43, 108 fn, 141–146, 150, 152–156, 160, 163–165, 168, 169, 175–181, 183–190, 192, 196, 198, 211, 212, 218–220, 231–235, 239, 240, 244–249, 254, 259–261, 265, 266, 270, 271, 279, 280, 286, 293, 295–299, 301, 302, 304, 305, 310, 312, 315 causal chain 19, 20, 247, 263 causativization 224, 225 causee 189, 191, 192, 224, 225 circumstancial subject (see setting subject) 216, 221 clause type 5, 13, 17, 27, 28, 32, 73, 94, 113, 158, 262 clustered structure 255, 256, 281, 284–288 coding properties 176, 208, 212, 256 cognitive (linguistics) 1, 2, 3, 8, 19, 20, 88, 101, 102, 113, 114, 117, 129, 131, 175, 176, 255, 277 fn, 293, 295, 315 Cognitive Grammar 1, 4–6, 13, 14, 18–20, 37, 102, 110, 263

conceptualization 1, 2, 6, 15, 19, 20, 26, 30, 34, 38, 111, 113, 114, 121, 122, 129, 175, 176, 190–194, 196, 211, 215, 217–219, 302 conceptualizer 1, 2, 5, 6, 101, 111, 113, 115, 117–119, 121, 123, 124, 127, 130, 132, 133, 135, 136, 181 conditional clause 5, 48, 73, 75–77, 82, 88, 92–94 conditional mood 4, 5, 43, 45–47, 49–53, 56–58, 67–70, 75, 79 conjunction 75, 76, 269, 274 consequent subevent 188–193 construal 1, 2, 4, 13, 14, 20, 21, 36, 38, 113, 114, 116, 176, 179, 187, 190, 196, 197, 278, 301 construction grammar 2, 3, 7, 8, 18–20, 27, 176, 193, 253, 255, 256, 294–296, 314 continuity 22, 23, 50, 56, 70 control 6, 112, 113, 116, 142, 143, 150, 157, 160–162, 165, 166, 169, 181–185, 188–191, 193, 195, 199, 208, 209, 217, 222–226, 244, 247, 255, 256, 260, 269, 275, 302, 304, 307 coordination 269, 270, 274, 282, 286 D dative 7, 165, 166, 168, 169, 181, 186, 189, 191, 236, 243–246, 248, 253–255, 258–263, 265–274, 276, 278–280, 283, 286–288, 293, 296–299, 301, 302, 304–306, 308, 310–312, 315 deity as subject 217, 218 determiner 45 directive 88, 93

322 Subjects in Constructions

discourse/discoursive 1–4, 8, 13–15, 18, 22–24, 26, 36–38, 43–45, 50, 57, 69, 78, 81, 106, 108, 121, 132–134, 160, 181, 190, 194, 205, 210, 211, 256, 262, 268, 311, 314, 315 E e-NP 4, 13–15, 18, 20–38, 231 fn elaborated reality 122 elaboration 108, 109 emergence 248, 249 entail(ment) 3, 122, 124, 175, 177, 178, 180, 182–185, 187, 189, 190, 193–195, 198, 257, 286–288 evidential(ity) 180, 181, 183, 187–190, 193, 197 exclusive 5, 43, 50, 59, 67, 70, 90, 91, 94, 96, 144, 190 expletive (subject) 105, 205, 207, 212–218, 220, 222, 224, 226 F ield 5, 6, 101, 111, 112, 116–120, 123–125, 128–130, 132 frame (semantics) 2, 5, 67, 69, 136, 180, 244, 274, 301 G generic 5, 45, 48, 49, 53, 59–61, 67, 68, 70, 73–75, 77–80, 82–94, 185, 194 genitive 7, 16, 17, 141, 145–149, 180, 182–184, 231–249, 298, 299, 301, 302 genitive-under-negation 183, 184, 198 gerund(ive) 131, 275, 276, 283, 286 giving 240, 244, 245, 301, 302 grammatical subject 1, 2, 4, 15, 19, 30, 31, 36, 38, 45, 59, 110 fn, 153, 231, 233, 254, 296–298, 304–308, 310, 311 grounding 187, 189, 191 group referent 5, 43, 44, 48–52, 55, 62–65, 67–70

I imperative 225, 226, 263, 281, 283, 286 impersonal 3–6, 8, 19, 22, 26, 43, 45, 46, 59–70, 73–75, 79–82, 84, 85, 90–95, 104–106, 109, 110, 114, 116–119, 121–125, 131 fn, 133, 136, 137, 141, 143–146, 212, 215, 254, 255, 258, 259, 262, 265, 266, 269, 272, 275, 276, 280, 296, 297, 304, 305, 310–315 impersonal pronouns 5, 104–106, 118, 121, 125, 136, 215 implied subject 5, 73, 74, 75 fn, 77–79, 85, 86, 92–94 inclusive 5, 43, 50, 58, 65, 67, 70, 94 information structure 20, 29, 158, 167, 261, 262 informational status 44 instantiation 19, 267–270, 272, 275, 276, 282, 285, 286–288 instrument 19, 146, 150, 154, 157, 267–270, 272, 275, 276, 282, 285–288, 302 instrumental (case) 22, 271, 278, 279, 298, 299, 301 insubordination 197–199 interaction(al) 1, 2, 15, 36, 62, 77, 78, 110–113, 116, 257, 288, 293 intransitive 6, 7, 14, 15, 23, 24, 27, 81, 114, 116, 143, 153, 166, 167, 181, 183, 206, 208, 210, 211, 213, 216, 218, 219, 223, 257, 258, 264, 280, 294 inversion 160, 264, 265, 282, 287, 294 L landmark 19, 21, 22, 36, 37, 113, 116 fn, 176, 191, 196 linear order 255, 261, 262, 264, 265, 275, 282, 285, 286, 299 linking 2, 178, 195, 196, 269 location 5, 15, 19, 22, 27–30, 32–34, 36, 38, 102, 109–111, 114–116, 120–124, 131, 136, 164, 183, 184, 264

locational subject 5, 6, 101, 102, 105, 109, 110, 115, 116, 119, 135, 136 long-distance relexive pronoun 272, 285 M meteorological (predicates) 6, 7, 8, 205, 206, 208, 209, 212–219, 223, 226 N narrative 5, 43, 61, 65, 67–70, 83 fn, 108, 311, 312 natural path 20, 21 necessity 16, 129, 180, 181, 187, 189, 190, 198, 218, 236, 239, 244, 245, 269, 309, 311 necessity modals 180, 181, 187, 189, 198, 309 negation / negative 14, 25, 26, 32, 34–35, 43, 45, 55, 76, 79, 86, 177, 178, 180, 183, 184, 187, 198, 241 fn network 2, 7, 8, 195, 231, 254, 278, 282, 293–296, 298, 299, 301–303, 314 nominative 14–18, 20–22, 24–26, 29, 31, 32, 35, 37, 143, 144, 152, 153, 158, 160–162, 166, 176, 180, 181, 183, 194, 196, 205, 218, 219, 226, 231, 232, 236, 237, 244–249, 253, 254, 259, 260, 261, 263, 266, 270, 271, 273, 274, 278, 280 fn, 282, 285–287, 296–299, 301, 304–313, 315 non-canonical subjects 1–4, 7, 8, 101, 141, 157, 158, 166, 169, 175, 176, 249, 293, 294, 296, 298, 305, 307, 310, 311, 315 non-prototypical trajector 6, 175, 176, 178, 179, 187, 189, 190 non-referential subject 14, 27, 32, 35, 176, 205, 210, 214, 220, 226 non-topical subject 205 number 8, 14, 16, 17, 25, 28, 59, 79, 81, 194, 205, 213, 220, 226, 235, 254, 265, 277–279, 306

Subject index

O object 15, 16, 18, 20, 23–30, 37, 81, 82, 115, 143, 144, 152–154, 158, 160, 176, 177 fn, 180, 181, 183, 185, 206, 212, 213, 218–220, 223–225, 231, 233, 234, 236, 241–244, 246–248, 253, 255–257, 260, 261, 263, 282, 286, 294, 298, 303, 311, 312, 315 oblique 3, 5, 6, 16, 28, 114, 115, 142, 143, 158, 175–177, 179, 185, 187–189, 198, 218, 223, 224, 253–256, 271, 279, 281 oblique subject 5, 16, 253, 254, 255, 256 open reference 5, 73–75, 79, 81–83, 90 P paradigmatic feature 276, 277, 286 passive 6, 15, 17, 19, 22, 26, 43, 59, 60, 67 fn, 73, 74, 75 fn, 79–86, 90–92, 94, 114, 141–146, 148, 152–167, 169, 170, 180, 181, 223, 224, 237, 238 fn, 243, 260, 265, 278, 279, 281, 283, 286, 298 passivization 116 fn, 158, 216, 223, 224 patient 6, 18, 19, 36, 38, 86, 111, 113, 114, 141–144, 177, 180, 182, 206, 209, 210, 220, 224, 244, 257, 262, 302 perfect 46, 49, 59, 141, 143, 144, 146, 163, 166, 167, 170, 180, 182, 183, 187, 189, 193 permitting 240 person(al) 3, 4, 14, 17, 23, 28, 43, 44–47, 49 fn, 50, 51, 53, 56, 57, 59, 61, 63, 65, 66, 70, 73–90, 92, 94, 103 fn, 119, 120, 136, 143, 144, 150, 151, 183, 189, 191, 194, 205, 213, 214, 220, 221, 226, 234, 243, 245 fn, 248, 254, 258, 259, 265, 277, 281, 297, 304, 306, 308, 311–315 plural 16, 17, 21, 25, 26, 28, 35, 46, 49 fn, 51, 53, 56, 59, 64, 65, 81–84, 103 fn, 144, 220–222, 266, 270, 281

possessive 7, 8, 16, 23, 31, 32, 79, 141, 143, 146, 160, 161, 163, 165–170, 179, 231, 232, 243–249, 270, 271, 273 predicate nominal 15, 23–28, 30, 32–35, 37 pro-drop 45, 79, 213, 258, 266 fn, 267, 268 process 2 fn, 17, 19, 20, 22, 82, 88, 93, 114–119, 121, 158, 192, 198, 199, 223, 249 prominence 19, 113, 176, 179, 181, 194 pronoun 5, 6, 16, 28, 43–45, 49 fn, 50–67, 69, 70, 74, 79–81, 83–85, 101 fn, 102, 103 fn, 104–110, 118–128, 131 fn, 132–136, 143, 151, 156, 160–162, 165, 169, 183, 213–215, 226, 234, 270–275, 281, 285, 306–308, 312 purpose 82, 240, 272, 274–276, 283, 286 Q quantity

24–26

R radial category 6, 8, 177, 195, 199, 293–295, 298, 301, 314 Radical Construction Grammar 3, 7, 255, 256, 295 raising 103 fn, 185, 255, 256, 279, 280, 283, 286, 287 reanalysis 107, 231, 232, 247–249 reference 5, 21, 27, 43, 45, 47, 48, 50, 52, 53, 56, 58, 59, 64, 68, 70, 73–75, 77–83, 86, 90–92, 107, 108, 111, 131, 136, 162, 177, 189, 190, 199, 220, 221, 223, 237, 259, 262 reference point 31, 279 referential chain 23, 24, 44, 46, 50, 56, 58, 60–62, 64, 65, 67, 70 referential(ity) 2–6, 14, 16, 17, 22, 23, 27, 30, 32, 33, 36, 37, 43–45, 47, 50, 55, 56, 58, 60–62, 64, 65, 67, 69, 70, 104, 160, 175, 181–183, 189–191, 193, 205, 210, 211, 213–215, 220, 222, 223, 234, 245 fn, 259, 263, 270, 272–274

role (syntactic) role 5, 7, 13, 15, 16, 18, 20, 23, 24, 37, 255, 256, 260, 294, 296, 306, 309, 310 (semantic) role 2, 4, 6, 13–15, 18, 19, 23, 28, 34, 36, 38, 64, 86, 104 fn, 111, 113, 114, 146, 152, 153, 176, 180, 206, 209, 215, 219, 232, 244–249, 254, 256, 259, 262, 285, 287, 305, 310, 311 (discursive/conversational) 44, 45, 64, 70, 81 S semi-predicate 263, 266, 280, 282, 285, 286 setting 5, 22, 27, 28, 30, 101, 108, 110–113, 115, 116, 118–122, 126, 127, 132, 133, 136 setting subject 4, 6, 8, 19, 22, 27–30, 38, 114–116, 118, 226 short-distance relexive pronoun 271–274 similarity 8, 119 fn, 129, 193, 254, 284, 287, 288 singular 17, 46, 49 fn, 65, 80–86, 89, 103 fn, 119 fn, 191, 205, 220–222, 226, 234 fn, 235, 266, 270, 281, 296, 304, 306, 308, 312 fn starting point 2, 4, 5, 13–15, 20–24, 33, 36, 37, 108, 209, 285 subjectless 15, 38, 74, 84, 143, 212, 213, 216, 225 T theme 4, 17, 18, 23, 36, 37, 38 theme orientation 4, 5, 15, 26, 27, 36, 38 topical(ity) 2–4, 6, 22, 23, 28, 142, 143, 146, 150, 157, 161, 162, 166, 167, 169, 182, 205, 226 tracking (of referents) 4, 14, 23, 24, 37, 58, 67, 211, 256, 259 trajector 1, 4, 6, 13, 14, 19–22, 26, 30, 36–38, 113–115, 116 fn, 118, 120, 127, 175, 176, 178, 179, 183, 185, 187–191, 193, 196, 198

323

324 Subjects in Constructions

transitive 6, 7, 14, 15, 20, 23, 24, 27, 29, 81, 113, 114, 143, 144, 152, 153, 160, 166, 167, 177, 183, 191, 195, 196, 206, 209, 210, 213, 216, 218, 223, 247, 257, 258, 263, 264, 288, 293, 294, 296, 298, 302, 303 typology 4, 6, 8, 15, 22, 180 fn, 198, 205, 206, 215, 218, 221, 232, 281, 295

U unbounded(ness) 24–26 undergoer 28, 34–36 usage-based (approach / linguistics) 1, 248, 261, 273 V valency 206, 223, 248 vantage point 2, 4, 190 verbs of speech 242 volitional(ity) 38, 111, 141, 142, 146, 150, 156, 157, 162, 164, 169, 179, 185, 224, 257, 301

W weather predicates (see meteorological predicates) 7, 206, 210, 212, 220, 221 word order 13, 15, 17, 18, 20, 157, 158, 162, 167, 179, 181, 182, 196, 247, 261, 294, 295 Z zero person 73–75, 79, 80, 82, 84–90, 92, 94