Unraveling the complexity of SE (Studies in Natural Language and Linguistic Theory, 99) 3030570037, 9783030570033

Citation preview

Studies in Natural Language and Linguistic Theory 99

Grant Armstrong Jonathan E. MacDonald  Editors

Unraveling the complexity of SE

Studies in Natural Language and Linguistic Theory VOLUME 99

Series Editors Marcel den Dikken, Research Institute for Linguistics and Department of English Linguistics, Eötvös Loránd University, Budapest, Hungary Liliane Haegeman, University of Gent, Belgium Maria Polinsky, University of Maryland, College Park, USA Editorial Board Member Guglielmo Cinque, University of Venice, Italy Jane Grimshaw, Rutgers University, New Brunswick, USA Michael Kenstowicz, Massachusetts Institute of Technology, Cambridge, MA, USA Hilda Koopman, University of California, Los Angeles, USA Howard Lasnik, University of Maryland, College Park, MD, USA Alec Marantz, New York University, New York, USA John J. McCarthy, University of Massachusetts, Amherst, MA, USA Ian Roberts, University of Cambridge, UK

Studies in Natural Language & Linguistic Theory provides a forum for the discussion of theoretical research that pays close attention to natural language data, offering a channel of communication between researchers of a variety of points of view. Like its associated journal Natural Language & Linguistic Theory, the series actively seeks to bridge the gap between descriptive work and work of a highly theoretical, less empirically oriented nature. The series editors invite proposals for monographs and edited volumes.

For more information on how to submit your proposal, please visit the series’ webpage http://www.springer.com/series/6559

Grant Armstrong • Jonathan E. MacDonald Editors

Unraveling the complexity of SE

Editors Grant Armstrong Department of Spanish & Portuguese University of Wisconsin–Madison Madison, WI, USA

Jonathan E. MacDonald Department of Linguistics, Department of Spanish & Portuguese University of Illinois at Urbana-Champaign Champaign, IL, USA

ISSN 0924-4670 ISSN 2215-0358 (electronic) Studies in Natural Language and Linguistic Theory ISBN 978-3-030-57003-3 ISBN 978-3-030-57004-0 (eBook) https://doi.org/10.1007/978-3-030-57004-0 © Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

We dedicate this book to our friend and colleague Paula Kempchinsky. Paula sadly passed away in March 2021 and was unable to see this book in its published form. In addition to the chapter she contributed, she had an important role in making this volume a reality through both her scholarship and mentorship. We in the Romance Linguistics community are deeply thankful for Paula’s work and hope that this volume serves as a small testament to her legacy in the field.

Preface

The SE clitic in Romance is one of the most studied topics in Linguistics. Indeed, the literature on SE is so vast and intimidating that it is sometimes difficult to think of how anything new and insightful could be said about it. A few years ago, after our separate research had led us both into the SE wormhole, we were struck by the scarcity of new work on SE aimed at investigating the big questions with respect to the different kinds of grammatical contexts where it may appear (i.e., its many “uses” or “functions”). While there is a range of stellar work on the functions of SE in individual Romance languages as well as excellent descriptive and theoretical overviews of all its different uses, an answer to the simple question Why is SE so complex? has remained elusive. As a step toward seriously engaging with this question, we organized The Workshop on Romance SE/SI in 2016 at the University of Wisconsin-Madison. The excellent papers presented at the conference provided the raw material for us to start thinking about how we might forge a path toward answering the elusive question posed above. When we asked if the conference participants would be interested in contributing to a volume about the complexity of SE with a piece based on each of their papers, we received an unanimously positive response. We were fortunate to receive very detailed and insightful analyses of a wide range of phenomena in different Romance languages from both diachronic and synchronic perspectives. This forced us to think deeply about how each contribution fit into the larger puzzle that the SE clitic represents. While it would be naïve to think we have definitively answered the question Why is SE so complex? we believe that the present volume provides a structured framework in which this question can be properly situated and through which it can be profitably researched. By choosing the title Unraveling the complexity of the SE clitic, we hope not only to show how one might approach all the different uses of SE

vii

viii

Preface

but also to inspire future work to further untangle the strands that tie these uses together. We would like to thank all the contributors to this volume for being patient and remaining with us while we crafted the appropriate narrative and structure that would make sense of the entirety of the contributions. We know it took a long time for this to come to fruition, and we greatly appreciate your willingness to remain on board during the process. We would also like to thank the anonymous reviewers of the Studies in Natural Language and Linguistic Theory, whose comments and questions helped improve the final version of the volume. Madison, WI, USA Champaign, IL, USA

Grant Armstrong Jonathan E. MacDonald

Contents

A Guide to Understanding SE Constructions: Where They Come from and How They Are Connected . . . . . . . . . . . . . . . . . . . . . . . . . . . . Grant Armstrong and Jonathan E. MacDonald Part I

1

Diachronic Perspectives

The Development of Se from Latin to Spanish and the Reflexive Object Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthew L. Maddox

33

Null-Subjects and se Revisited: What Medieval Romance Varieties Reveal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anne C. Wolfsgruber

57

Part II

Voice/Little v and Above

On (Un)Grammatical Clitic Sequences in Impersonal se Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jonathan E. MacDonald and Jeriel Melgares

87

Implicit Agents and the Person Constraint on SE-Passives . . . . . . . . . . . 111 Carmen Dobrovie-Sorin On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 Francisco Ordóñez Personal se with Unergatives in Romanian . . . . . . . . . . . . . . . . . . . . . . . 161 Monica Alexandrina Irimia and Virginia Hill

ix

x

Part III

Contents

Voice/Little v and Below

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John washes his hands’ . . . . . . . . . . . . . . . . . . . . 187 Alexandra Cornilescu and Alexandru Nicolae Causative SE: A Transitive Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Grant Armstrong and Paula Kempchinsky Light Verbs and the Syntactic Configurations of se . . . . . . . . . . . . . . . . 243 Alfredo García-Pardo The Role of SE and NE in Romance Verbs of Directed Motion: Evidence from Catalan, Italian, Aragonese and Spanish Varieties . . . . . 265 Anna Pineda Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 Margot Vivanco Part IV

A Unifying Perspective

Spanish se as a High and Low Verbalizer . . . . . . . . . . . . . . . . . . . . . . . . 321 David Basilico

About The Editors

Grant Armstrong is an Associate Professor of Spanish Linguistics at the University of Wisconsin-Madison. His research is focused on the Morphology and Syntax of Spanish and Mayan languages. He has published numerous articles on these topics in journals such as The Linguistic Review, Probus, and Studies in Hispanic and Lusophone Linguistics and as chapter contributions to edited volumes in series published by Oxford University Press, Cambridge University Press, John Benjamins, and Springer. Jonathan E. MacDonald is Associate Professor of Linguistics and of Spanish and Portuguese at the University of Illinois at Urbana-Champaign. His general research area lies in the domain of theoretical syntax, with a specific focus on Romance SE constructions, and inner aspect. His work has appeared in journals such as Linguistic Inquiry, Syntax, Journal of Linguistics, and Probus, as well as numerous contributions in volumes edited by Oxford University Press, Cambridge University Press, John Benjamins, among others.

xi

A Guide to Understanding SE Constructions: Where They Come from and How They Are Connected Grant Armstrong and Jonathan E. MacDonald

Abstract In this chapter, we outline a framework for understanding the syntactic and semantic complexity of the SE clitic. We adopt the idea that the source of the multiple uses of SE is reflexive SE itself. Moreover, assuming that reflexive SE is syntactically tied to Voice/little v, we argue that we can profitably understand SE’s diverse uses as a result of two directions of change, one where SE takes on novel syntactic and semantic functions related to Voice/little v and above; the other where SE takes on novel syntactic and semantic functions in positions below Voice/little v. We also summarize the remaining chapters in the volume and situate them within these two broad divisions. Finally, we highlight directions for further research that naturally emerge from this study. Keywords Romance SE constructions · Morphosyntactic change · Expletive SE · Pronominal SE · Paradigmatic SE · Non-paradigmatic SE

1 Introduction The Spanish sentences in (1) contain an instance of the reflexive clitic pronoun (in bold), which appears when the external argument is coreferential with an internal argument, in this case a direct object. Observe that the form of the clitic changes as a function of the person and number of the subject; it is paradigmatic.1

1

Note that with a plural antecedent, as in (1d,e,f) a reciprocal interpretation is available as well, since the reciprocal clitic pronoun is homophonous with the reflexive. Reciprocals are not discussed in this volume. G. Armstrong (*) · J. E. MacDonald University of Wisconsin-Madison, Madison, WI, USA University of Illinois Urbana-Champaign, Champaign, IL, USA e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_1

1

2

(1)

G. Armstrong and J. E. MacDonald

a. Yo me veo. I 1s see.1s ‘I see myself’ b. Tú te ves. You 2s see.2s ‘You see yourself’ c. Él/Ella se ve He/she 3s see.3s ‘He/she sees him/herself’

d. Nosotros nos vemos. We 1p see.1p ‘We see ourselves’ e. Vosotros os veis You 2p see.2p ‘You all see yourselves’ f. Ellos/Ellas se ven. They.masc/They.fem 3p see.3p ‘They see themselves’

All Romance languages have a reflexive clitic pronoun, and all Romance languages have appropriated the third person reflexive se (Reflse) for some non-reflexive uses, which we refer to as SE.2 Some of these uses are impersonal SE (Impse), passive SE (Passse), anticausative SE (AntiCse), aspectual SE (Aspse), antipassive SE (AntiPse), middle SE (Midse), inherent SE (Inherse), figure reflexive SE (FigReflse), agentive reflexive clitic SE (Arcse), and causative SE (Causse), each of which are discussed to varying degrees in this volume.3 The bulk of the data come from (varieties of) Spanish and Romanian, however, contributions also carry out comparisons with other languages, including Dutch, Russian, Chukchi, Bantu, Icelandic, French, Italian, Catalan, Aragonese, as well as Old French, Old Spanish and Latin. The main focus of this book is on the SE morpheme and its associated underlying configurations. Taken together as a whole, the contributions in this volume clearly highlight the internally complex syntax and semantics of SE. Syntactically, almost every projection along the syntactic spine is invoked: T, Voice/little v, Appl, PP, and SC. Semantically, SE is shown to be linked to discourse features, non-referentiality, human reference, agency, cause, affectedness, telicity, scalar structure, and figureground semantics. We ask two questions: Why is SE so complex? And, why has it been appropriated for so many distinct constructions? One goal of this introduction is to sketch a framework within which we can begin to provide an answer to these questions. We hope to isolate some strands that run through these SE constructions in order to connect them, and thereby unravel some of its internal complexity. Succinctly, our approach is one that combines recent work on grammaticalization from a generative perspective (van Gelderen 2011) with recent advances in the 2

The languages that have a reflexive that has been extended to non-reflexive uses go beyond Romance, and Indo-European. See Geniušiené 1987, Haspelmath 1990. Some discussion of non-Romance languages can be found in this volume. See especially Basilico (this volume). 3 Another SE in Spanish is so-called spurious SE (Perlmutter 1971; Bonet 1991; Nevins 2007) in which the first third person clitic in a sequence of two third person clitics changes to SE. Spurious SE will not feature in this volume. General overviews and analyses of SE constructions that mention a number of these uses include Burzio (1986), Wehrli (1986), Cinque (1988), Dobrovie-Sorin (1998, 2006), Mendikoetxea (1999a, b), Otero (1999), Sánchez López (2002) and Pujalte and Saab (2012). Other studies limit the discussion to one or two types. See, for example, Raposo and Uriagereka (1996), D’Alessandro (2007) and MacDonald (2017b) on Impse and Passse; Labelle (1992), Folli (2002), Schäfer (2008) and Koontz-Garboden (2009) AntiCse and Midse; Armstrong (2013) on Arcse and MacDonald (2017a, b) on Aspse, among many other studies.

A Guide to Understanding SE Constructions

3

syntax of clitics and argument/event structure. Following others, we take Reflse as the starting point for the other SE constructions and build on this by suggesting that one evolutionary strand traces SE constructions from Voice/little v and above, and another traces SE constructions from Voice/little v and below. We provide more details on these two paths in Sect. 2 of this chapter. Nevertheless, before getting there, we first begin with a description of the key contributions of each chapter in the volume, which we have organized into four groups: (i) diachronic perspectives, (ii) syntactic and semantic effects of SE above Voice/little v, (iii) syntactic and semantic effects of SE below Voice/little v and (iv) a unifying perspective. Even though the individual phenomena investigated in each chapter are quite different, both syntactically and semantically, we aim to show that this grouping allows for the extraction of interesting patterns that can be used as a way to shed light on the questions raised above. Moreover, we hope that combining insights from both diachronic and synchronic perspectives points towards a way of resolving an outstanding question in research on SE constructions regarding the nature of SE itself.

1.1

Diachronic Perspectives

Given that the complexity of SE constructions is intimately connected to diachrony, we think it is important to start with this perspective. In chapter 2 “The Development of Se from Latin to Spanish and the Reflexive Object Cycle”, Matthew Maddox tackles a key issue in the history of Romance clitics. Maddox provides evidence from Latin and different stages of Spanish in support of the idea that SE originated as a full DP, then was reanalyzed as a moved D head, and finally has been reanalyzed as an object agreement morpheme, generated in transitive Voice, following the path of the object agreement cycle proposed in van Gelderen (2011). The three stages of this change are shown in (2). (2) Stage (a) Stage (a) ! (b) Stage (b) Stage (b) ! (c) Latin Old Spanish Middle Spanish Modern Spanish sē ¼ DP se ¼ DP se ¼ DP > D se ¼ Voice He focuses primarily on reflexive uses of SE and the status of SE itself. However, he mentions that at every stage of development, SE could also be used as an anticausative, middle or passive marker. In (3), we see an anticausative use of SE in Stage (a) ! (b). (3) este algodon es atal que se non quema por fuego this cotton is such that SE.3 not burns by fire ‘This cotton is such that it does not burn from fire’

4

G. Armstrong and J. E. MacDonald

Maddox also arrives at the conclusion that structures with SE in Old Spanish appear to be transitive, based in part on auxiliary selection facts. As is well-known, the modern Romance languages French and Italian have two possible auxiliary verbs in periphrastic perfect constructions: BE and HAVE. The equivalent of BE appears with some intransitive verbs (usually unaccusatives), while HAVE appears in all other contexts (see Kayne 1993; Sorace 2000 for overviews and different analyses). Old Spanish (in contrast to Modern Spanish) showed the same auxiliary selection patterns as modern Italian and French, with one contrast. In Modern Italian and French, any verb that has a SE clitic, regardless of transitivity, surfaces with BE. In Old Spanish, however, when SE was present, HAVE surfaces, not BE (see Aránovich 2003). As illustrated in (4a), the auxiliary BE (essora) surfaces with an unaccusative verb, while in (4b), the auxiliary HAVE surfaces with reflexive SE. (4) a. Minaya Alvar Fáñez essora es llegado Minaya Alvar Fáñez then is arrived ‘Minaya Alvar Fáñez then arrived’ b. mas es necesario que quando el se ha echado en tierra but is necessary that when he SE.3s has thrown on ground ‘But it is necessary that when he has cast himself to the ground’ Based on the generalization that only clitic SE triggers the appearance of BE, while full DP SE does not, as observed for Italian and German (McFadden 2007), SE patterns like a full DP in (4b). Moreover, since nominative el is the external argument, the only other position available for the full DP SE to merge (before moving to its final position) is the internal argument position, resulting in these constructions in Old Spanish being analyzed as transitive. The topic of reanalysis is also at the forefront of understanding the evolution of impersonal SE constructions and their relation to passives. In chapter 3 “Null-Subjects and Se Revisited: What Medieval Romance Varieties Reveal”, Anne Wolfsgruber claims that the transition from passive to impersonal SE involves reanalyzing a Voice morpheme (passive SE) as a realization of T, which restricts the referential properties of the implicit argument that is licensed by T. She discusses a variety of factors in Old Spanish and Old French related to the null subject parameter, which likely led to the modern divergence with respect to the presence in Spanish of impersonal SE constructions and their absence in French. These different evolutionary stages of SE constructions correspond roughly to the variety of synchronic theoretical analyses of clitics currently on offer. Romance clitics have either been treated as D(P) heads that raise to a verbal functional projection, as part of the extended functional projection of the verb in an agreement relation with a DP or pro, or as verbal morphemes that affect the argument structure properties of the verb without establishing a relation with DP or pro (see Roberts 2010a; Ormazabal and Romero 2013 for recent overviews). These are summarized in (5) below.

A Guide to Understanding SE Constructions

5

(5) a. SE to Head (D(P) head moves to verbal projection)4 b. SE Head & pro (clitic is an agreement morpheme) c. Head only SE (clitic is a verbal head) The analytical options in (5), which are arguably a product of how clitics evolved from full DPs to heads, represent one factor in understanding the complexity of SE constructions. The differences that are observed within SE constructions can in part be attributed to this grammaticalization path. A separate factor that is independent of the nature of SE itself concerns where the primary syntactic and semantic effects of SE are localized. For instance, Maddox discusses a change in the status of SE whose effect can be observed below Voice/little v, in the direct object position, while Wolfsgruber discusses a change in which the primary effects are localized above Voice/little v, in the licensing of an implicit external argument by T. In what follows, we describe the other contributions making use, where appropriate, of these two important factors.

1.2

Voice/Little v and Above

The second set of contributions all treat varieties of SE constructions whose defining syntactic and semantic properties are localized at Voice/little v or above. In chapter 4 “On (un)grammatical clitic sequences in impersonal se constructions”, Jonathan E. MacDonald and Jeriel Melgares provide a novel account of the presence of the object clitic le(s) in Impse constructions, where lo(s) and la(s) are expected. The main empirical generalization to be explained is shown in (6).5 (6) Se *lo(s)/*la(s)/le(s) ve por aquí mucho SE.3 masc.(pl.)/fem.(pl.)/dat(pl.) sees for here much ‘You see her/him/it/them around here a lot’ Adopting the [D] feature account of the referential properties of consistent null subjects in Holmberg (2005, 2010) and Roberts (2010b), MacDonald & Melgares claim that SE is the spell out of a valued [D] feature in T. The consequence, they 4

See D’Alessandro (2007), Roberts (2010a) for recent accounts along this line. This behavior would reflect one aspect of bare phrase structure, where the difference between an XP and X is collapsed (Chomsky 1995). 5 The patterns in (6) come from Honduran Spanish, in which any accusative clitic that surfaces in Impse gives rise to ungrammaticality. Speakers resort to le(s), which is common in most varieties of Latin American and Peninsular Spanish, with some exceptions. In Rioplatense Spanish, in contrast, all accusative clitics in Impse constructions are grammatical. There are varieties that show a mixed pattern. For instance, varieties of Peninsular Spanish that Ormazabal and Romero (2013) call central Peninsular leísta dialects, pattern like Honduran Spanish with respect to masculine accusative clitics in Impse constructions, while they pattern with Rioplatense Spanish with respect to feminine accusative clitics in Impse constructions.

6

G. Armstrong and J. E. MacDonald

claim, is that only a non-referential interpretation of the implicit external argument is available in these constructions. They argue that lo(s)/la(s) is specified as [D (efinite)], while le(s) is underspecified, i.e., it lacks a [D] feature. Importantly, only when Impse and lo(s)/la(s) form a clitic cluster does the change from lo(s)/la(s) to le(s) obtain. Observe in (7a) that when the clitic is in the embedded clause lo surfaces grammatically, while le does not. Only in (7b), when the clitic climbs to form a cluster in (7b), le surfaces grammatically while lo does not. (7) a. En Navidad, se suele poder abrazar {lo/*le} him.DO/DO In Christmas, Impse tends can hug ‘In Christmas, one tends to be able to hug him’ b. En Navidad, se {*lo/le} suele poder abrazar As they illustrate, there are no syntactic or interpretive effects of the change from lo(s)/la(s) to le(s), which suggests a post-syntactic approch. They provide a Distributed Morphology account in which the clitic cluster containing two [D] features is filtered out by impoverishment, leading to deletion of [D] on lo(s)/la(s). The result is that le(s) is the only vocabulary item with a subset of features available for vocabulary insertion resulting in the sequence of Impse plus le(s). In chapter 5 “Implicit Agents and the Person Constraint on Se-Passives”, Carmen Dobrovie-Sorin offers a novel proposal of the Person Constraint (PC) on Passse. The PC states that (i) the subject of a Passse construction cannot be first/second person, nor (ii) a third person pronoun, proper name, or a possessive DP. The first part has been shown to hold in Romanian, Spanish, Italian, French and European Portuguese, and the second part has been shown to hold in Romanian and Spanish, languages that have differential object marking (DOM). The PC on Passse is illustrated below in (8) from Romanian. (8) a. M-am invitat ieri la petrecere SE.1s-have invited yesterday to party ‘I invited myself to the party yesterday’ ‘*I was invited to the party yesterday’ b. La noi întotdeauna se întâmpină {musafirii/*Ion/*el} la gară at us always SE.3 welcome.3sg guests.the/ Ion/ he at station ‘In our family/department. . ., guests/Ion/he are/is always welcomed at the station’ c. *S-a convocat profesorul tău SE.3-has summoned teacher your ‘Your teacher was summoned.’ Dobrovie-Sorin builds on Legate’s (2014) work on passives, which makes a key claim that passives may differ from one another based on how the implicit agent/ initiator variable is introduced into the syntactic derivation, modified and ultimately saturated. Perhaps the most important aspect of this work for Passse is that the implicit agent variable of a passive may be introduced in Voice and restricted by

A Guide to Understanding SE Constructions

7

voice morphology. Restriction is a type of semantic operation in which a variable is modified rather than saturated (see Chung and Ladusaw 2004), and Legate claims that Voice may be restricted by a morpheme that limits the possible referents of the implicit agent variable to third person, for example. Dobrovie-Sorin builds on this idea in order to account for the PC in SE passives. Her proposal is outlined in (9). (9) a. [TP T{uφ} [BeP Be [VoiceP Voice{φ:ARB} [VP V [DP {φ} ]]]]] (Participle Passives) Person  b. [TP T-SE{φ:ARB} [VoiceP Voice [VP V [DP {φ} ] ] ] ]

(SE Passives)

Person 

What (9a) illustrates is that participle passives have an inherently valued person feature (¼ ARB) on Voice. This enables T to enter a Person agreement relation with the internal argument DP.6 In Passse, Dobrovie-Sorin claims that SE is base generated in Tense, following Kayne (2000), and thereby supplying a valued ARB feature in T. Thus, it can only enter a number agreement relation with a DP in the internal argument position. The Voice head in this construction is an expletive, the same as in anticausatives (see Schäfer 2008). This proposal extends the ideas in Legate in that it claims that the implicit agent can be introduced (and restricted) in T and not only in Voice, leading to a plausible explanation of the PC in SE passives. In chapter 6 “On the Nature of the Impersonal Se: Why Italian Is Not Like Catalan and Spanish”, Francisco Ordóñez outlines a way to account for differences in the interpretation of impersonal SE sentences in Spanish and Catalan on the one hand, and in Italian, on the other. A basic contrast between these languages can be observed in episodic contexts like past perfective with periphrastic passive verbs (and other unaccusatives). In Catalan (and Spanish, not shown here), Impse constructions are not possible in these contexts, whereas they are in Italian. (10) a. *Ahir es va ser castigat Catalan Yesterday SE.3 was punished Intended: ‘One/someone was punished yesterday’ b. Si é stati invitati anche noi Italian SE.3 was been invited also us ‘We were also invited’ (from D’Alessandro 2007) Ordóñez builds on previous observations by Cinque (1988) and D’Alessandro (2007), showing that the distinction in (10) correlates with the inability of Catalan (and Spanish) to have first person plural interpretations in these contexts, while in Italian, the first person plural interpretation is obligatory. He claims that this contrast 6

Technically, the DP sister to V in (9a) moves to Spec,Voice where it agrees with T.

8

G. Armstrong and J. E. MacDonald

can be accounted for if we treat impersonal SE as a clitic that requires a null pronominal binder with nominative case as in (11). (11) [XP [proi] sei . . .] Variation among languages depends on the nature of the pronominal binder and the restrictions imposed on it within the clause. All Romance languages with Impse constructions have an arbitrary pronoun called ϕ (see Holmberg 2005, 2010; Mendikoetxea 2008), which must be bound by a generic (12a) or existential operator (12b). Italian differs from Catalan and Spanish in that it has an additional null pronoun, referred to as WE (12c), which gives rise to a first person plural interpretation and does not need to be bound by an operator. (12) a. GEN OPi [XP [ϕi] sei . . . ] (Available in all impersonal SE constructions) b. EX OPi [XP [ϕi] sei . . . ] (Available in all impersonal SE constructions) c. [XP [WEi] sii . . . ] (Only available in Italian impersonal SE constructions) He accounts for the contrasts in (12) above by assuming that existential operators are introduced within the domain of Voice, or little v. Passives, copulas and unaccusatives have no Voice/little v, thus SE is not licensed in this constructions in Spanish and Catalan. Italian is special, however, since WE is a pronoun behaving more like a DP that needs no special licensing. In these three chapters, we observe that the SE clitic in Impse and Passse constructions is linked to an implicit arbitrary argument that is licensed through a formal mechanism (e.g. Agree, binding) or interpretative constraint (e.g. restrict) above Voice. While the authors make use of SE Head & pro (MacDonald & Melgares; Ordóñez) and Head only SE (Dobrovie-Sorin), other analyses of the same phenomena have opted for a SE to Head analysis, assigning argumental status to SE itself, which is generated in spec Voice/little v (Raposo and Uriagereka 1996; D’Alessandro 2007). The latter analytical option is utilized by Monica Irimia and Virginia Hill in chapter 7 “Personal Se with Unergatives in Romanian”. This work adds to the empirical range of SE constructions usually grouped with Inherse, Antipse and Arcse. They observe that this is a marked construction within Romance, and possibly appears in Romanian due to contact with Slavic.7 That is, some verbs that would be unergative in Romanian displayed the -sja ending in Slavic (e.g. boretsja “to fightREFL” in Russian), and were reanalyzed as an optional SE in these cases, dubbed redundant SE in traditional texts. They discuss several of these cases of optional SE with unergatives but focus on (se) râde (laugh, 13a), concluding that the most

7 Although an anonymous reviewer observes that French also has these constructions, which thereby raises questions about its Slavic origins.

A Guide to Understanding SE Constructions

9

adequate analysis of the Romanian data is to treat SE as a part of a big DP subject of unergatives as in (13b). (13) a. Maria se râde (de mine) Maria SE.3 laughs at me ‘Maria is laughing at me’ b. [vP [Dmax [D se [DP Maria]]] [v râde [VP PP-goal [VP ]]] This innovative type of clitic doubling of an external argument is linked to topicalization of the subject, its obligatory pre-verbal position and a constraint that the argument must be human. They observe a contrast where only human external arguments can occur with optional SE, as illustrated in (14). (14) a. Mircea (s)-a greşit Mircea SE.3-has erred ‘Mircea has erred’ b. Calculatorul (*s)- a greşit computer.the SE.3-has erred ‘The computer has erred’ The ungrammatical instance of a post-verbal subject with optional SE indicates an obligatory preverbal subject position in the presence of SE, which they analyze as being in a topic or focus position. (15) (*?S)-a greşit Mircea (SE.3)-has erred Mircea ‘Mircea has erred’ Unergative Personal SE constructions like these share with Impse and Passse the restriction that the external argument must be human. However, their obvious differences are captured by the fact that SE is part of a big DP in the external argument position here, while it never is with Impse or Passse. This particular use of SE, which possibly emerged from contact with Slavic, is not easily situated within the grammaticalization path from anticausative to passive to impersonal SE, as discussed in MacDonald and Maddox (2018), Wolfsgruber (this volume), and in Sect. 2 below. As can be appreciated by the range of proposals in chapters 2–7, all of the analytical options in (5) regarding SE itself are utilized to explain the properties of the constructions under investigation. Differences between them can be understood in terms of the relation the clitic establishes with another head, operator or pronoun above Voice.

10

1.3

G. Armstrong and J. E. MacDonald

Voice/Little v and Below

The third set of contributions treat a range of SE constructions whose defining syntactic and semantic properties are localized at Voice/little v or below. In chapter 8 “On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John Washes His Hands’”, Alexandra Cornilescu and Alexandru Nicolae discuss the syntax and semantics of a class of sentences in Romanian that are marked with an accusative SE clitic and contain a prepositional body-part object as in (16). (16) S-a spălat pe spate / pe mâini/ pe obraz/ în urechi SE.3-has washed on back on hands on cheek in ears ‘He washed his back/his hands/his cheek/his ears’ They present a detailed argument in favor of treating such sentences as figure reflexives (Wood 2014, 2015). In these constructions, the SE clitic occupies the specifier position of a prepositional small clause that mediates a figure-ground relation and is bound by an agent DP that merges in Spec,Voice as shown in (17). (17) [VoiceP [DP AGENT]i Voice [vP v [pP SEi p [PP P GROUND]]]] The paper treats SE as a DP and uses a relatively new position within the extended projection of PP in order to explain its interpretation. They take the ability of the accusative reflexive SE to double, as in (18), as an argument that SE merges as an internal argument. (18) Ion nu se mai poate spăla pe mâini nici pe el însuși, John not SE.3 still can wash.inf on hands neither DOM him himself nici pe altcineva nor DOM someone.else ‘John is no longer able to wash his own hands, let alone someone else’s’ In chapter 9 “Causative SE: A Transitive Analysis”, Grant Armstrong and Paula Kempchinsky provide a novel syntactic account of causative SE constructions like (19), which exist in various Romance and other languages with SE anaphors, like Dutch (see Rooryck and Vanden Wyngaerd 2011; Zribi-Hertz 1982). (19) Juan se afeita en la barbería para parecer más elegante Juan SE.3 shaves at the barbershop for to look more elegant ‘Juan gets a shave at the barbershop to look more elegant’ The most natural interpretation of (19) is that someone other than Juan does the shaving, i.e. he has gone to a barbershop and someone else shaved him, “he got a shave”, as the translation indicates. Thus, the agentive component is implicit and attributed to the barber, which is not syntactically present. Armstrong &

A Guide to Understanding SE Constructions

11

Kempchinsky claim that causative SE constructions are generated in the following transitive configuration. (20) [vP DPi vcause [SC SEi √Root]] They propose that the peculiar interpretation of causative SE constructions arises from a combination of the external argument-introducing head vCAUSE and pragmatics. They claim that vCAUSE introduces a causer rather than an agent, and in certain contexts, defined by conventions associated with the event in question, an intermediate agent is implicit (see Rooryck & Vanden Wyngaerd (2011) for a discussion along these lines for Dutch examples). The proposal shares with Cornilescu & Nicolae the idea that SE itself is merged in an internal argument position, and thus falls within the “SE to head” type of analysis discussed above. However, causative SE constructions do not permit doubling like figure reflexives in Romanian. Armstrong & Kempchinsky claim that this is a type of intrinsic reflexivity, which is syntactically transitive, yet semantically intransitive. Essentially, this means that even though there are two argument positions (external and internal), they are obligatorily linked to the same argument DP (see Labelle 2008; Wood 2014, 2015 for discussion), and the SE clitic must fill one of those two positions. Intrinsic reflexivity plays a key part in chapter “Light Verbs and the Syntactic Configurations of SE”, by Alfredo García Pardo. He provides a proposal for the role that SE plays in Spanish light verb constructions with the verbs poner(se) and quedar(se) (both typically translated as ‘become’) and APs like those in (21), an understudied set of predicates. (21) a. Berta se puso {fuerte/nerviosa/ rígida} Berta SE.3 put strong nervous rigid ‘Berta became {strong/nervous/rigid}’ b. Juan se quedó {pálido/ tranquilo/ delgado} Juan SE.3 stayed pale calm thin ‘Juan became {pale/ calm/ thin}’ García Pardo makes innovative use of the idea that SE lexicalizes a verbal head in a decomposed VP, utilizing the richly decomposed VP proposed in Ramchand’s (2008) first phase syntax framework in which the verb consists, potentially, of the projections init (initation), proc (process) and res (result), which represent distinct types of subevents. Following previous work, he proposes that both of these light verbs lexicalize slightly different structures. Quedar-se is associated with two first phase syntactic structures, which are shown below in (22) and (23).

12

G. Armstrong and J. E. MacDonald

(22) a. Pedro se quedó despierto Pedro SE.3 stayed awake ‘Pedro stayed awake’ b. [initP Pedro init-se quedó [resP Pedro res-quedó [AP despierto]]] (23) a. La falda se quedó anticuada the skirt SE.3 got old-fashioned “The skirt became old-fashioned” b. [procP la falda proc-se quedó [resP la falda res quedó [AP anticuada]]] (22) is a stative causative relation between a causing event and a result state, in which Pedro is bringing about, or maintaining, his own state of being awake. Garcia Pardo observes, in this respect, that they pattern with other stative causative structures like object-experiencer verbs (Pesetsky 1995; Pylkkänen 2000; Rothmayr 2009), where in some cases an overt causative marker appears. The structure of stative quedarse contrasts with change of state quedarse, where instead of initP being the higher verbal projection, ProcP is, as in (23). Ponerse, on the other hand, minimally lexicalizes a structure that contains proc and res, a change of state, as shown in (26). (26) a. Pedro se puso nervioso Pedro SE.3 became nervous ‘Pedro got nervous’ b. [procP Pedro proc-se puso [resP Pedro res puso [AP nervioso]]] The first phase syntax structures proposed above are meant to explain the lexical aspectual properties of the light verbs and a host of other properties discussed in detail by García Pardo. It is important to note that SE itself is not a telic or stative morpheme, but rather appears in an event structure that is independently telic or stative. As can be seen in the configurations above, Garcia Pardo proposes that SE surfaces in these constructions as a morphological reflex of the same argument occupying two different positions in the articulated event structure (in the spirit of Cuervo 2003, 2014). In this sense, these constructions are intrinsically reflexive in the same way that the causative SE constructions analyzed by Armstrong & Kempchinsky are. What is different about this proposal is that a DP argument may move from one thematic position to another, and when this happens, SE surfaces as a verbal head, rather than as an argument. In chapter 12 “The Role of SE and NE in Romance Verbs of Directed Motion. Evidence from Catalan, Italian, Aragonese and Spanish Varieties”, Anna Pineda makes use of this same idea of SE as a verbal head. Her main claim is that directed motion verbs that take SE and the ablative clitic ne/en in varieties of Catalan (shown in 27a), Italian and Aragonese should be analyzed as in (27b).

A Guide to Understanding SE Constructions

13

(27) a. Ara se' n tornarà a casa Now SE.3 ABL go.back.FUT.3sg to home “Now (s)he will go back home” b. [VoiceP DPi Voice [Appl-LocP en [vP SE [√P √Root [SC DPi PP]]]]] Building on previous work, she argues that SE in directed motion constructions is the realization of a verbalizing cause head in a syntactic construction where the sole argument both initiates and changes location (see Cuervo 2014; Jiménez-Fernández and Tubino 2014 for similar proposals). While the framework employed is different from that of García Pardo, the core idea is the same: SE is the realization of a verbalizing head in a context where a single DP has two thematic roles. She further argues that ne/en is a locative applicative, providing comparative evidence from Bantu to support the claim. In chapters 8–11 we see two subtly different types of analyses used to capture the observation that SE marks a reflexive relation between an external and internal argument position (Cornilescu & Nicolae; Armstrong & Kempchinsky; Pineda) or between two internal argument positions in a more articulated event structure (García Pardo). In Cornilescu & Nicolae and Armstrong & Kempchinsky, the SE to head analysis is used, where SE occupies the internal argument position whereas both Pineda and García Pardo make use of the Head only SE analysis, in which an argument moves from one theta position to another and SE is the morphological reflex of this movement. The last chapter in this group investigates a class of SE constructions in which the clitic is claimed to be a verbal head, but does not necessarily function as a reflexive marker. In Chapter 13 “Scalar constraints on anticausative se. The aspectual hypothesis revisited”, Margot Vivanco discusses aspect in anticausative constructions and how the presence of SE correlates with a heretofore unnoticed aspectual distinction. The main claim centers on data like (28), where SE appears to optionally mark the unaccusative variant of the anticausative alternation. (28) a. El trueno despertó a Víctor the thunder woke.up DOM Víctor ‘The thunder woke Victor up’ b. Víctor (se) despertó Víctor SE.3 woke.up ‘Victor woke up’ When SE appears in a sentence like (28b), what information does it provide? Taking work on French (Zribi-Hertz 1982; Labelle 1992; Legendre and Smolensky 2010) and Italian (Folli 2002) as her point of departure, Vivanco claims that SE marks an aspectual distinction between achievements and accomplishments. She presents numerous tests demonstrating that unmarked unaccusatives are simple change of state achievements and the SE marked unaccustives are complex change

14

G. Armstrong and J. E. MacDonald

of state achievements. That is, SE spells out an intransitive verbalizing head in the context of a scale that involves multiple transitions from one state to another (i.e. an accomplishment) while the same intransitive verbalizing head is spelled out with a null morpheme in the context of a simple achievement, which describes the transition to a resultant state without iteration (i.e. an achievement). (29) a. Achievement ├ state ! Ø b. state ┤Achievement ├ state ! se Vivanco’s proposal is one in which SE is a verbalizing head that does not mark a reflexive relation between argument positions but is instead sensitive to the scalar structure of the predicate.

1.4

A Unifying Perspective

The final chapter in the volume, “Spanish se as a high and low verbalizer”, by David Basilico, is also based on the idea that SE is a verbalizing head that does not (necessarily) mark a reflexive relation. In this sense, it is similar to Vivanco in the way it treats SE itself. However, the objective of this paper is different from all others in the volume in an important respect. Basilico proposes to unify reflexives with anticausatives, antipassives and certain unergatives that contain SE in Spanish by claiming that SE may be a high verbalizing head that merges above Voice (his little v/Pred) or a low verbalizing head which merges below Voice. In order to understand the proposal, let us start with Basilico’s analysis of reflexives. In sentences like (30a) Basilico proposes that there is a null reflexive morpheme that takes a verb that is a predicate of events and turns it into a verb that is a relation between an event and an entity. The entity argument is a variable that must be bound; syntactically, he claims, it is PRO. Juan moves from Spec, vP-AG (where it receives an agent role) to Spec,Pred, creating a predication; a lamba operator is introduced and binds PRO, generating the reflexive interpretation. (30) a. Juan se lava Juan SE.3 washes ‘Juan washes himself’ b. [PredP Juan Pred [vP Juan v-AG [VP [V V-lava REFL]PRO]]] How is SE integrated into this structure? Basilico claims that it is a v head added above PredP as in (31).

A Guide to Understanding SE Constructions

15

(31) [vP v-se [PredP Juan Pred [vP Juan v-AG [VP [V V-lava REFL ] PRO]]]] The verbalizing head realized by SE serves to existentially disclose the event argument (Dekker 1993) and create a predicate of events that can interact with higher operators in the clause. While this might appear to be a somewhat unorthodox analysis of reflexives that dissociates SE itself from reflexive interpretations, Basilico argues that this particular analysis of reflexives allows us to unify these constructions with anticausatives and antipassives through the following structural generalization. (32) [vP SE [PredP Pred . . . ]] In each of the constructions analyzed, SE serves the same function: it creates a predicate of events through existential disclosure. Differences between SE-marked sentences are derived by appealing to different types of argument introducers below PredP. Anticaustives have an undergoer argument rather than an agent as in (33). (33) a. El jarrón se rompió The vase SE.3 broke ‘The base broke’ b. [vP v-se [PredP el jarron Pred [vP el jarron v-UND [VP V-rompió]]] Antipassives have an agent argument but lack the null REFL morpheme that combines directly with the verb. Instead, they take an internal PP argument. (34) a. Juan se confiesa de sus pecados Juan SE.3 confesses of his sins ‘Juan confessed his sins’ b. [vP v-se [PredP Juan Pred [vP Juan v-AG [VP V-confiesa [PP de sus pecados ]]]] As support for this unified approach to reflexives, anticausatives, and antipassives, Basilico draws attention to nearly identical patterns in Russian, Chukchi and Icelandic where the same morpheme appears in each of the contexts described above. While this use of SE is compatible with a variety of argument structure configurations, an additional use is more restricted. Basilico claims that certain unergative verbs like reír-se (laugh-SE) are derived by combining a verbalizing head with a nominal root. This is so-called low verbalizing SE. As is the case with the high verbalizer in (32), the primary role of SE is to make a predicate of events out of a predicate of individuals. The difference here is that SE converts a nominal root into a predicate of events by introducing a DO predicate as in (35). (35) [vP SE [√REIR]]

λxλe[DO(e, x, LAUGH)]

16

G. Armstrong and J. E. MacDonald

As with high verbalizing SE, Basilico illustrates how denominal verbs marked with a reflexive clitic from languages like Icelandic and Brazilian Portuguese can be analyzed in this fashion. Basilico’s contribution contrasts with the others in that he attempts to unify many different types of SE constructions under the same analysis, which leads necessarily to a deeper level of abstraction regarding the nature of SE itself.

2 Voice/Little v and Diverging Paths As the brief summary of the chapters indicates, SE is complex both syntactically and semantically. In this section, we sketch a framework in which to situate this complexity. As noted above, Reflse is often taken to be the source of many, if not all SE constructions, an idea that we adopt here. We suggest, nevertheless, that two different paths stemming from Reflse have led to, on the one hand, SE constructions above Voice/little v and, and on the other, SE constructions below Voice/little v. Part of this conclusion is predicated on the idea that Reflse itself is intimately linked to Voice/little v, a point we first address.

2.1

Voice/Little v and Reflse

Consider the following observation, which appears to implicate the external argument, and consequently, the external argument introducing head Voice/little v: Reflse only surfaces when an internal argument is coreferential with the external argument (Fontana and Moore 1992; Otero 1999), as illustrated in (36a, b). On the other hand, if two internal arguments are coreferential and the external argument is not one of the coreferential arguments, as in (36c), Reflse cannot surface. (36) a. Juan *(se) mandó un regalo a sí mismo Juan (SE.3) send a present DAT himself ‘Juan sent himself a present’ b. Juan *(se) vio a sí mismo en el espejo Juan (SE.3) saw DOM himself in the mirror ‘Juan saw himself in the mirror’ c. Juan (*se) enseñó a Mariai a sí mismai en el espejo Juan (SE.3) showed DOM Maria DAT herself in the mirror ‘Juan showed María to herself in the mirror’ Given that there are two coreferential arguments, SE appears to be, minimally, a marker of reflexivity (Reinhart and Reuland 1993). But where is SE in the syntax? One option, following Armstrong & Kempchinsky (this volume) is to say SE spells out an Agree relation between either a null pro or the double sí mismo in an internal

A Guide to Understanding SE Constructions

17

argument position, which itself is bound by the external argument. Since SE reflects an Agree relation with an internal argument bound by the external argument, SE will only surface when the external argument is involved. On this approach, Reflse constructions are transitive (Torrego 1995; Ormazabal and Romero 2013; Maddox this volume as well). An alternative, from Marantz (1984) and McGinnis (2004), is to place SE itself in the external argument position. This has two consequences: (i) SE will surface only when the external argument is one of the coreferential arguments; and (ii) The DP internal coreferential argument will move to Spec,T to bind SE, resulting in unaccusative syntax (Pesetsky 1995). Evidence for the unaccusativity of Reflse constructions comes from auxiliary selection in languages that show a HAVE-BE alternation, like French and Italian. With Reflse, BE surfaces, the same auxiliary that surfaces with unaccusative verbs. Recall, however, that Maddox (this volume) discusses the patterns in Old Spanish, where the reflexive was a full DP and patterned as an internal argument. As he observes, Old Spanish has auxiliary selection parallel to Italian, yet, in Reflse constructions, HAVE surfaces, not BE. Reflse constructions in Old Spanish pattern with transitives. Now consider the perspective of a native learner presented with a Reflse construction in Spanish in (37a) and the related (37b). (37) a. Juan se vio a sí mismo Juan Reflse saw DOM self same ‘Juan saw himself’ b. Se vio Reflse saw ‘Juan saw himself’ In a manner parallel to the analyses above, the learner very well could analyze (37b) in a number of ways, given the position of SE relative to the verb, and given that there is no overt direct object or subject. SE could absorb the external argument slot (i.e. be a subject), and if so, the construction could be analyzed as unaccusative, with a null subject. SE could also absorb the internal argument slot (i.e. be a direct object, which has moved), in which case the construction could be analyzed as unergative, with a null subject. In either case, SE could be reducing the argument structure of the predicate. The structure underlying Reflse is rather opaque from a learner’s perspective. This is precisely the situation that lends itself to reanalysis (Roberts and Roussou 2003 among others). It is not surprising, then, that it is often assumed that the source for the non-Reflse appropriated uses of SE is Reflse itself, often thought to mark intransitivity (Monge 1954; Kemmer 1988; Cennamo 1999, and Pountain 2000). It is in this respect, we suggest that Reflse is the locus of the two diverging paths of SE constructions, one which leads above Voice/little v, and the other, which leads below Voice/little v. In the following sections, we discuss how the

18

G. Armstrong and J. E. MacDonald

proposals in this volume relate to these diverging paths and also link these grammaticalization paths to the elusive nature of SE itself.

2.2

The Path Above Voice/Little v

A historical relation has been observed among Reflse, AntiCse, Passse and Impse constructions, where the latter developed form the former, as illustrated in the path in (38) (not limited to Romance nor to Indo-European languages; see Geniušiené 1987; Haspelmath 1990; Cennamo 1993; among others). (38) Reflse > AntiCse > Passse > Impse One motivation for the path in (38) comes from implicational relations. A language that has Impse also has Passse; a language that has Passse also has AntiCse, etc. In fact, this is the locus of one point of variation among Romance languages. While Spanish, Italian, European Portuguese, and Catalan have all of these SE constructions, Romanian and French do not. They both lack Impse. The step from Reflse to AntiCse is typically thought to reflect the reanalysis of SE as a marker of intransitivity (Monge 1954; Kemmer 1988; Cennamo 1999, and Pountain 2000). We adopt the conclusions from Maddox (this volume) in which SE was reanalyzed as the head of Voice in Reflse constructions (although we return to this point below shortly) and the transitive syntax of Reflse from Armstrong & Kempchinsky (this volume) and represent Reflse as in (39a). We also assume there is an expletive Voice head in AntiCse (Schäfer 2008) which SE heads, with an empty Spec (see also MacDonald and Maddox 2018), and no accusative case.8 (39) a. [VoiceP pro Voicese [VP DP]] b. [VoiceP Voicese [VP DP]]

[Reflse] [AntiCse]

The lack of external argument makes AntiCse intransitive (and unaccusative). The next step from AntiCse to Passse is not entirely surprising given ambiguous sentences like those in (40), illustrated in Spanish, but present in all major Romance languages.

8

In recent literature there is some discussion regarding whether AntiCse can be treated as fundamentally Reflse. See Koontz-Garboden (2009) and Schäfer and Vivanco (2016) for opposing views. This is overall consistent with our approach, yet, our aim is to go a bit beyond and include other SE constructions.

A Guide to Understanding SE Constructions

19

(40) a. Se abrió la ventana SE.3s opened the window ‘The window opened./The window was opened.’ b. Se rompió la mesa SE.3s broke the table ‘The table broke./The table was broken.’ This ambiguity is well known and can be distinguished by the use of certain modifiers (Mendikoetxea 1999a; Schäfer 2008; Koontz-Garboden 2009; Vivanco this volume among others), illustrated in (41).9 (41) a. Se abrió la ventana por sí sola. SE.3 opened the window by self alone ‘The window opened by itself.’ b. Se abrió la ventana para airear la habitación. SE.3 opened the window for air.out the room ‘The window was opened to air out the room’

[AntiCse]

[Passse]

MacDonald and Maddox (2018) claim that the development of Passse from AntiCse results in the emergence of a non-referential third person pro in Spec, Voice, giving rise to (42), but where there is still no accusative case on Voice. (42) [VoiceP pro Voicese [VP DP]] As far as we are aware, there has been no formal syntactic proposal regarding how Impse has developed from Passse. Wolfsgruber (this volume), nevertheless, offers a promising initial idea that ties the development of Impse to being a consistent null subject language as a way to explain why French has not developed an Impse construction.10 She suggests that SE was reanalyzed at some point as T as a way to neutralize the referential properties of null subject constructions in consistent null subject languages. Given recent accounts of differences between consistent null subject languages and partial null subject languages (Holmberg 2005, 2010; Roberts 2010b), we might implement Wolfsgruber’s suggestion by assuming that SE heads T and thereby “neutralizes” the [D] feature, which, as argued by Holmberg (2005, 2010) is responsible for the referential interpretation of null subjects in consistent null subject languages. Assuming that SE is in T in Impse and assuming that SE is in Voice in 9 A reviewer observes that (41b) is ungrammatical in French. For us, this raises the question of whether French has Passse or not. Consider, however, Il s’ est vendu plus de 300 livres anciens pour aider les pauvres. “There have been more than 300 old books sold to help the poor.” from Dobrovie-Sorin (this volume). Crucially, in this case, the post-verbal DP is indefinite. Minimally, Passse in French patterns differently from Passse in other Romance language in this respect. 10 Cinque (1988) also assumes that French did not develop an Impse construction because of its status as a non-null subject language.

20

G. Armstrong and J. E. MacDonald

Passse, we can then treat the development of Impse to Passse as an instance of “grammaticalization” up the tree, a common diachronic situation as often observed (Roberts and Roussou 2003; van Gelderen 2011). On this approach, since SE is within the T domain where we see morphosyntactic clitic clustering effects, such as Spurse (Perlmutter 1971; Bonet 1991; Nevins 2007) in which two third person clitics give rise to SE, it would not be surprising to find morphosyntactic clitic effects involving Impse plus other clitics as well. As discussed in MacDonald & Melgares (this volume), this is exactly what we find (see also Ordóñez and Treviño 2016).11 Moreover, as pointed out in Ordóñez (this volume) and Dobrovie-Sorin (1998, this volume), among others, Impse is not restricted by the argument structure of the predicate, while Passse is. Observe in (43), that Impse can occur with all predicate classes.12 (43) a. Siempre se nace con poco pelo Unaccusative Always SE.3 is born with little hair ‘One is always born with little hair’ (De Miguel Aparicio 1992) b. Se baila bien en este pueblo.13 Unergative SE.3 dances well in this town ‘People dance well in this town’ c. Se es feliz o no se es feliz Copular SE.3 be.3sg happy or not SE.3 be.3sg happy ‘People are happy or they are not’ d. En este país se es perseguido por la policía Passive In this country SE.3 is chased by the police ‘In this country, one is chased by the police’ ‘In this country, one is chased by the police’ (Jaeggli 1986) e. Se avisó a los bomberos Transitive SE.3 alerted DOM the firemen ‘Someone alerted the firemen’ (Mendikoetxea 1999b) Dobrovie-Sorin (1998) argues that languages like Romanian and French lack Impse, although they have Passse, based, in part, on the following data. We only present the Romanian data here, from Dobrovie-Sorin (1998). Data in (44d) is based on Dragomirescu (2013: 169).

11

Most likely it is in this domain where sequences of Impse plus Inherse are ruled out as well, since there is a sequence of identical clitics (see discussion in Martins and Nunes 2016). 12 In (44a,c,d), there is a restriction to a generic interpretation, as discussed in De Miguel Aparicio (1992) and Suñer (1990). 13 Dobrovie-Sorin (1998) argues that in Romanian the se that appears with unergatives is an instance of passive se.

A Guide to Understanding SE Constructions

21

(44) a. *In şcoala asta se pedepseşte pe elevi in school this SE.3 punish DOM students. the ‘In this school, they punish the students’ b. *Adesea se este trădat de prieteni falşi frequently, SE.3 is betrayed by friends false ‘One is frequently betrayed by false friends’ c.*Nu se este niciodataă mulțumit not SE.3 is never satisfied ‘One is never satisfied’ d.Se *construieşte/construiesc locuinţe noi SE.3 *built.3sg./built.3pl houses.nom new.nom.pl ‘New houses are built’ e.*Le se construiesc Them.acc.pl. SE.3 built.3pl Intended: ‘They are built’ f. Se cîntă/doarme/munceşte/mănîncă SE.3 sings/sleeps/works/eats ‘It is sung/slept/worked/eaten’ Note that we have been assuming a SE Head & pro analysis so far, based on the structure in (39). Yet the other two possibilities are coherent as well. Raposo and Uriagereka’s (1996) (see also D’Alessandro 2007; Roberts 2010b) approach represents a SE to Head analysis in which SE itself is the external argument. DobrovieSorin (this volume) implements a Head only SE analysis in which SE heads T with a valued Arb person feature. Whatever properties are attributed to SE, whether heading T or in external argument position, can also be attributed to pro in (38). In fact, all three approaches can account for another property of Passse in which the sole DP cannot be first or second person (Cinque 1988; D’Alessandro 2007; Mendikoetxea 2008, and MacDonald 2017b), which Dobrovie-Sorin (this volume) refers to as the person constraint.14 On the one hand, Dobrovie-Sorin’s (this volume) account assumes T is valued with an Arb person feature due to the presence of SE. Thus, only third person DPs (with no person) can Agree (in number) with T. On the other two SE accounts, one can appeal to intervention effects in the presence of

14

The person constraint on Passse constructions seems to be a later development, at least in Romanian, as it has been observed in Old Romanian, first and second person in Passse was possible (Dindelegan 2013), as the data in (i) and (ii) below illustrate, from Cornilescu & Nicolae (2015: 311–312), citing Heliade-Rădulescu (1828/1980–149) and Iordache Golescu (1840: 150), respectively for the data. i. Mă învăţ de tatăl SE.1s teach by father “I am taught by my father”

ii. mă bat eu de către tine SE.1s beat I by to/toward you “I am beaten by you”

22

G. Armstrong and J. E. MacDonald

SE/pro between the sole theme argument and T (as in D’Alessandro 2007, Mendikoetxea 2008, MacDonald 2017b). Each SE analysis can also be implemented in the grammaticalization story of Impse outlined above. On a SE to Head analysis, one might suppose that in Passse, SE is in Spec,Voice then, once reanalyzed as a subject moves to Spec,T, at a time when SE was still a DP. Then it can later be reanalyzed as Voice in Passse and as T in Impse.15 This, in fact, is what Wolfsgruber (this volume) suggests may have happened. On this latter account, the movement of SE to a head position becomes irrelevant, since SE would still be a full DP at that time (see Maddox this volume). On a Head only SE analysis, if we assume that SE is in Voice (pace DobrovieSorin this volume), the same story outlined above is feasible, since SE as a Voice head is reanalyzed as SE as a T head, as a result of grammaticalization up the tree.16 What we would like to highlight here is that the elusive nature of SE itself (pronoun, agreement morpheme, or head) makes sense within the framework of grammaticalization cycles whereby a pronoun is reanalyzed as a head (e.g. expletive Voice in AntiCse) and then a head can subsequently be associated with an arbitrary null argument (Passse), and potentially reanalyzed as a new arbitrary pronoun (some accounts of Impse). We would also like to make clear that this issue is not resolved by the papers in this volume, but one may take the different analyses as a reflection of the grammaticalization cycles alluded to above. Moving forward, it is important to look into fine-grained differences between languages and varieties in order to provide clearer answers regarding what SE is. Given recent work on non-reflexive object clitics, which has argued for subtle differences between object agreement and clitic doubling (Ormazabal and Romero 2013), it may be the case that multiple treatments of SE constructions are not only possible, but actually may be supported by the facts. We think that this is an important avenue for future research that emerges from this work.

15

On this approach, the question arises as to whether both SEs, in Spec,Voice in Passse and in Spec, T in Impse are reanalyzed as Voice and T, respectively, simultaneously or not. If they were reanalyzed simultaneously, although associated with different functional projections, this would entail that, at the appropriate level of abstraction, these SEs have a shared representation, otherwise it would not be obvious how reanalysis of one would affect the other. 16 Irimia & Hill (this volume) are somewhat difficult to fit into this picture, although their construction involves Voice/little v and above. It shows different properties from Passse and Impse in several respects, one of which is that the subject is overt, and not implicit. These constructions are seemingly the same as what Geniušiené (1987) calls subjective deleted object constructions, where -sja is used, and there is no direct object. Possibly, SE surfaces almost as an accident of contact with Slavic. If so, the situation would be similar to the inheritance of Late Latin reflexive se/sibi into Romance, as discussed in Cennamo (1999).

A Guide to Understanding SE Constructions

2.3

23

The Path Below Voice/Little v

The historical relationship between Reflse, AntiCse, Passse and Impse is widely assumed, and as we have just outlined, several properties that distinguish these SE constructions can be modelled formally as grammaticalization up the tree. In contrast, beyond the observations that many of the other SE constructions seem to appear together early in the historical documents (see Cennamo 1999, Pountain 2000, Portilla 2007, Wolfsgruber this volume, Maddox this volume, Vivanco this volume, Pineda this volume), there has been no formal historical relationship established among Reflse, AntiPse, Inherse, FigReflse, Aspse, and Causse as far as we are aware.17 It is worth noting, moreover, that their presence in a language appears not to be dependent on whether or not the language has Passse or Impse, suggesting that their development is independent of those SE constructions that develop as a result of grammaticalization up the tree. An important difference between the grammaticalization path of SE constructions above Voice/little v and those below Voice/little v is that the latter cannot be conceived of as a grammaticalization cycle. In order to situate our proposal within a theoretical framework similar to the one discussed in Sect. 2.2, we begin with a short digression into Icelandic -st constructions, as discussed in Wood (2014, 2015) as shown in (45). (45) Icelandic -st constructions (from Wood 2014: 1392–1398) a. Hann troð-st inn með hópnum Icelandic figure reflexive He squeezed-ST in with group.the ‘He squeezed in with the group’ b. Dyrnar opnu-ðu-st Icelandic anticausative Door.the open-PST-ST ‘The door opened’ The -st morpheme in Icelandic appears in a range of constructions similar, but not identical, to those of the SE clitic. Wood (2014, 2015) claims that this can be understood if -st is viewed as an argument expletive that appears in the specifier position of an argument-introducing head, but fails to saturate the theta role associated with that argument. (46) a. [VoiceP [DP Hann] VoiceAGENT [vP √troð-vcause [pP [D -st ] pFIGURE [PP inn með hópnum]]]] Hann, AGENT ¼ FIGURE b. [VoiceP [D -st ] Voice{D,Ø} [vP vcause [SC [√opnu] [DP dyrnar]]] dyrnar, THEME

17 We do not have much to say about Midse here, but see Kemmer (1988) for a discussion of its historical emergence in Romance that is not inconsistent with the ideas puts forward here.

24

G. Armstrong and J. E. MacDonald

So-called -st expletivization leads to two semantic effects. In figure reflexives like (31a), -st is merged in a low thematic position corresponding to the FIGURE argument. -st itself is unable to saturate this role, but the role may be linked to the DP that merges in the specifier of Voice. This gives rise to an intrinsic reflexive relation in which both argument variables in the structure are obligatorily linked to the same DP that merges in the highest thematic position in the structure. In anticausatives like (46b), -st is merged in a high thematic position. Since -st cannot saturate a thematic role and there is no higher DP to which the role in Voice could be linked, true expletivization is the result. This means that Voice must be a true expletive in these contexts, lacking a thematic role (see Schäfer 2008 and the discussion above).18 Using this insight as our point of departure, we suggest that the range of SE constructions below Voice has likely developed from using SE as a low and high argument expletive. In Cornilescu & Nicolae, SE is treated as low argument expletive, leading to a linking of the FIGURE role with an AGENT role, as in Wood (2014). In Armstrong & Kempchinsky, SE is a low argument expletive of a different kind of role (THEME, POSS, or BEN), leading to a linking of that low role with a CAUSER. On the other hand, both the change of state and location verbs discussed in García Pardo and Pineda could be treated as high argument expletives. In this sense, they are extensions of the marked anticausative construction to new sets of verbs. An attractive characteristic of the “SE as expletive” narrative is that it provides a clear link between being an argument at one stage of the language and being a verbal morpheme with no argumental status at another. The reasoning is as follows. If a particular syntactic object like SE appears in positions in which it cannot actually perform the function that is usually ascribed to that syntactic object (e.g. saturate a theta role), then it may be reanalyzed as having an altogether different function. The types of SE sentences discussed in Vivanco are illustrative of one type of reanalysis. Since SE is not required to mark the causative alternation with the particular set of roots she investigates, it does not make sense to say that an expletive Voice is optional. In fact, an optional expletive would likely be ignored or re-appropriated by language learners. For this reason we can think of SE as having been reanalyzed as a verbal head, say vcause, that is sensitive to aspectual distinctions. Basilico, on the other hand, discusses cases of AntiPse and Inherse that could be taken to be illustrations of a reanalysis of low expletivization. There is not much work on this particular development, but we might think of AntiPse as originally an expletive in a theme

18 This idea is compatible with the updated version of this model by Wood and Marantz (2017). In this model, all argument-introducing heads are subsumed under a single syntactic head i*. Formal and interpretative differences between argument introducing heads (Voice, Appl, p, etc.) are derived by contextual allomorphy and allosemy rules at the morphophonological and semantic interfaces, respectively. For simplicity’s sake, we don’t go into the details here, but, in a nutshell, SE could be taken to be an expletive argument of i* and all different types of SE could be derived according to syntactic context. Other alternatives to argument expletivization for explaining the behavior of SE include treating it as the spell out of reflexive relation triggered by movement of a single DP from one theta position to another (Cuervo 2003, García Pardo and Pineda this volume, following Ramchand 2008). We believe the discussion in the text is compatible with any of these.

A Guide to Understanding SE Constructions

25

argument position of a verb. However, instead of being linked to the highest argument in Voice, the thematic role is linked to an oblique argument. Eventually, SE fails to be associated with an argument position at all and is re-interpreted as a marker of intransitivity or an accusative case absorber (see Wehrli 1986; Masullo 1992; Armstrong 2016; Basilico, this volume for different ideas in this regard). Finally, Inherse is obligatorily associated with a set of verbs that may be either unaccusative or unergative (see Otero 1999; Armstrong 2016), so it is plausible to think that these have either developed from AntiCse or AntiPse, respectively, or are the result of spreading by analogy. Once SE is interpreted as a marker of intransitivity, it can spread to verbs like reír (laugh). While many of the SE constructions below Voice examined here are intransitive (Armstrong & Kempchinsky is an exception), the same story outlined above can be easily extended to other types of transitive SE constructions that are not discussed in the book, but have received detailed analyses elsewhere. For instance, Aspse is a transitive SE construction in which the presence of the SE clitic has an aspectual effect on a transitive verb (see MacDonald 2017a and references therein for an extensive discussion). An example is shown below in (47). (47) Juan se comió el helado Juan SE.3 ate the ice cream ‘Juan ate up the ice cream’ MacDonald (2017a) argues that aspectual SE is derived from a double object construction in which a bound pronoun occupies the complement position of a null preposition and receives a GROUND theta role. This role is linked to the argument in the specifier of Voice, which triggers spell out of SE in Voice, as shown in (48). (48) [VoiceP [DP Juan]i VoiceSE [VP comió [PP [DP el helado] P [pro]i]]] This proposal is similar to low expletivization in that an argument introduced by a head lower in the structure is linked to the argument in the spec of Voice. The difference here is that the clitic does not occupy the low argument, but rather spells out the fact that Juan and pro are co-referent. Since the low preposition and argument are always null in (48), it seems logical that the function of SE in this construction might undergo reanalysis. In fact Folli and Harley (2005) have claimed that aspectual SE constructions in Italian are best analyzed by treating SE as the realization of transitive vcause for a certain class of roots like comer. In essence, it marks a change in event structure from an activity to a change of state. This reflects to a large extent the kind of reanalysis discussed for the intransitive sentences above. The main point is that SE can cease to become an expletive in argument position and simply take on a role related to event structure. As can be appreciated in this brief discussion, we see the same issue regarding the nature of SE itself rear its head in our attempt to characterize the different analyses of SE below Voice. SE can be treated as a pronoun in argument position (Cornilescu & Nicolae; Armstrong & Kempchinsky), an agreement morpheme that marks reflexive relation between two argument positions, one of which must be the highest in the

26

G. Armstrong and J. E. MacDonald

structure (García Pardo; Pineda) or a verbal morpheme that is not related directly to an argument position (Vivanco; Basilico). As in the above section, we have sought to show that such a varied approach to the nature of SE may not be unwarranted when SE constructions are considered from a diachronic and cross-linguistic perspective. We have suggested that a plausible evolutionary path that accurately characterizes SE constructions below Voice/little v is one in which AntiCse ceases to act as an argument expletive and takes on a function related to verbal event structure. This change is accompanied by a shift in the status of SE from pronoun to verbal head.

2.4

Summary and Open Questions

When taken as a whole, we believe that the current volume is more than just the sum of its parts. We have argued in the preceding sections that by looking at a range of seemingly unconnected SE constructions, two distinct types of evolutionary paths can be isolated. One related to a cycle involving Reflse, AntiCse, Passse and Impse and another involving distinct types of re-analyses of Reflse and AntiCse constructions that have given rise to a more heterogeneous array of SE sentences related to reflexivity as well as argument and event structure. We think that this is an important contribution of this volume beyond the empirical and theoretical analyses of each paper and serves as point of clarity in the SE quagmire. We also believe that the papers in the volume highlight the need to focus on the nature of SE itself moving forward. There are three analytical choices when it comes to the nature of the clitic SE itself: a Head only SE analysis, a SE Head & pro analysis, and a SE to Head analysis. It is not clear that any of the accounts of the specific properties of the SE constructions depend crucially on the analysis of the SE clitic itself. Moreover, it also appears that the two divergent paths, as outlined above, are coherent with each as well. In our minds, this is a contribution of this volume: it highlights an open question, thereby, pointing us in an important direction for further research on SE. It emphasizes the need to find arguments to determine which is the most adequate analysis of SE. Ultimately, by providing an answer to this question, we could eliminate another level of complexity of SE constructions and streamline the various existing analyses, potentially uncovering unforeseen results. We conclude by asking ourselves two questions to help distinguish among these possible analyses, the first of which pits the SE to Head analysis and the SE Head & pro analysis against the Head only SE analysis simply because in the former two, there are two elements in play and in the latter only one: In SE constructions, are there “split” effects, which would depend on SE being linked to two syntactic positions, as in the SE to Head and the SE Head & pro analyses? Or are there never any “split” effects, which would be expected if there were a single element, as on the Head only SE analysis? The second question we ask helps distinguish the SE to Head analysis from the SE Head & pro analysis. On the latter, while we would expect the features between SE and pro to be the same for the most part, since they are distinct syntactic elements, the possibility arises that they could have non-overlapping features that manifest in different ways in the event that there are

A Guide to Understanding SE Constructions

27

“split” effects. In contrast, since on the SE to Head analysis SE itself occupies an argument slot then moves, we expect the features not to change (by inclusivity Chomsky 1995). Thus, even though there might be “split” effects, namely, effects associated with two distinct positions, we do not expect these effects to be different. So, do the split effects pattern the same or differently? We discuss one set of facts that initially seems to support a SE Head & pro analysis, although, of course, more work is needed. The facts come from Italian past participle agreement patterns in Impse constructions, as has been observed in Cinque (1988), D’Alessandro (2007), Ordóñez (this volume) among others. Recall the data from (10b) above, repeated below in (49) which illustrate the central pattern, where the verb is in singular, but the participle is in plural. (49) Si é stati invitati anche noi SE.3 was been invited also us ‘We were also invited’ First, we clearly see “split” agreement effects, which initially is unexpected on a SE Head only account. Second, the “split” agreement effects illustrate that invitati is plural, while è is singular, showing different patterns. These immediately suggest that the SE Head & pro analysis appears to be the most amenable. Nevertheless, alternative approaches to third singular agreement on the verb undermines this approach. Since it has been proposed that third person singular verbal agreement is default with Impse, it raises the question of whether these split effects are indicative of what we are looking for. Nevertheless, this illustrates one class of phenomena that may serve to pin down the most adequate analysis of the nature of SE, and by doing so, we hope can aid in unraveling its underlying complex syntax and semantics.

References Aránovich, Raúl. 2003. The semantics of auxiliary selection in old Spanish. Studies in Language 27: 1–37. Armstrong, Grant. 2013. Agentive reflexive clitics and transitive se constructions in Spanish. Borealis 2 (2): 81–128. ———. 2016. Pronominal verbs in Spanish. Studies in Hispanic and Lusophone Linguistics 9 (1): 29–65. Basilico, David. 2010. The Se clitic and its relationship to paths. Probus 22: 271–302. Bonet, Eulalia. 1991. Morphology after syntax: Pronominal clitics in Romance. MIT: PhD Dissertation. Burzio, Luigi. 1986. Italian syntax: A government-binding approach. Dordrecht: Reidel. Cennamo, Michela. 1993. The reanalysis of reflexives: A diachronic perspective. Napoli: Liguori Editore. ———. 1999. Late Latin pleonastic reflexives and the unaccusative hypothesis. Transactions of the Philological Society 97 (1): 103–150. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.

28

G. Armstrong and J. E. MacDonald

Chung, Sandra, and William A. Ladusaw. 2004. Restriction and saturation. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1988. On si constructions and the theory of arb. Linguistic Inquiry 19: 521–581. Cornilescu, Alexandra, and Alexandru Nicolae. 2015. The Grammaticalization of passive reflexive constructions in Romanian. In Diachronic variation in Romanian, ed. G.P. Dindelegan, R. Zafiu, A. Dragomirescu, I. Nicula, A. Nicolae, and L. Esher. Newcastle upon Tyne: Cambridge Scholars Publishing. Cuervo, María. 2003. Datives at large. PhD dissertation, MIT. ———. 2014. Alternating unaccusatives and distribution of roots. Lingua 141: 48–70. D’Alessandro, Roberta. 2007. Impersonal Si constructions. Berlin: Mouton de Gruyter. De Miguel Aparicio, Elena. 1992. El aspecto en la sintaxis del español: perfectividad e impersonalidad. Madrid: U.A.M. Dekker, P. 1993. Existential disclosure. Linguistics and Philosophy 16: 561–588. Dindelegan, Gabriela, ed. 2013. The grammar of Romanian. Cambridge: Cambridge University Press. Dobrovie-Sorin, Carmen. 1998. Impersonal se constructions in Romance and the passivization of unergatives. Linguistic Inquiry 29 (3): 399–437. ———. 2006. The se anaphor and its role in argument realization. In The Blackwell companion to syntax, ed. M. Everaert and H. van Riemsdijk, vol. 4, 118–179. Oxford: Blackwell. Dragomirescu, A. 2013. Passive and impersonal constructions. By-phrases. In The grammar of Romanian, ed. G. Pana Dindelegan, 169–174. Oxford: Oxford UP. Folli, Raffaella. 2002. Constructing telicity in English and Italian. PhD dissertation, University of Oxford. Folli, Raffaella, and Heidi Harley. 2005. Consuming results in Italian and English: Flavors of v. In Syntax, semantics, and acquisition of aspect, ed. P. Kempchinsky and R. Slabakova, 95–120. Dordrecht: Springer. Fontana, Josep, and John Moore. 1992. VP internal subjects and se reflexivization in Spanish. Linguistic Inquiry 23: 501–510. Geniušiené, Emma. 1987. The typology of reflexives. Berlin: Mouton De Gruyter. Haspelmath, Martin. 1990. The grammaticization of passive morphology. Studies in Language 14: 25–71. Holmberg, Anders. 2005. Is there a little pro? Evidence from Finnish. Linguistic Inquiry 36: 533–564. ———. 2010. Null subject parameters. In Parametric variation: Null subjects in minimalist theory, ed. Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan, 88–124. Cambridge: Cambridge University Press. Jaeggli, Osvaldo. 1986. Passive. Linguistic Inquiry 17: 587–622. Jiménez-Fernández, Ángel, and Mercedes Tubino. 2014. Variación sintáctica en la causativización. Revista española de la lingüística 41 (1): 7–38. Kayne, Richard. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47: 1–31. ———. 2000. Parameters and universals. Oxford: Oxford University Press. Kemmer, Suzanne. 1988. Middle voice systems in typological and diachronic perspective. PhD dissertation, Stanford University. Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27 (1): 77–138. Labelle, Marie. 1992. Change of state and valency. Journal of Linguistics 28: 375–414. ———. 2008. The French reflexive and reciprocal. Natural Language and Linguistic Theory 26: 833–876. Legate, Julie Anne. 2014. Voice and v: Lessons from Acehnese. Cambridge: MIT Press. Legendre, G., and P. Smolensky. 2010. French inchoatives and the unaccusativity hypothesis. In Hypothesis A/Hypothesis B: Linguistic explorations in honor of David M. Perlmutter, ed. D. Gerdts, J. Moore, and M. Polinsky, 229–246. Cambridge: MIT Press.

A Guide to Understanding SE Constructions

29

MacDonald, Jonathan E. 2017a. Spanish aspectual se as an indirect object reflexive: Evidence from (a)telicity, bare nouns and leísta PCC repairs. Probus 29 (1): 73–117. ———. 2017b. An implicit projected argument in Spanish impersonal and passive Se constructions. Syntax 40 (2): 353–383. MacDonald, Jonathan E., and Matthew Maddox. 2018. Passive Se in Romanian and Spanish: A linguistic cycle. Journal of Linguistics 54 (2): 389–427. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge: MIT Press. Martins, Ana Maria, and Jairo Nunes. 2016. Passives and Se constructions. In The handbook of Portuguese linguistics, ed. W. Leo Wetzels, João Costa, and Sergio Menuzzi. Hoboken: Wiley Blackwell. Masullo, Pascual José. 1992. Antipassive constructions in Spanish. In Romance languages and modern linguistic theory, ed. P. Hirschbühler and K. Koerner, 175–194. Amsterdam/Philadelphia: Benjamins. McFadden, Thomas. 2007. Auxiliary selection. Language and Linguistics Compass 1 (6): 674–708. McGinnis, Martha. 2004. Lethal ambiguity. Linguistic Iinquiry 35 (1): 47–95. Mendikoetxea, Amaya. 1999a. Construcciones inacusativas y pasivas. In V. Demonte and I. Bosque (dirs.) Gramática Descriptiva de la Lengua Española, 1575–1630. Madrid. EspasaCalpe. ———. 1999b. Construcciones con se: Medias, Pasivas e Impersonales. In V. Demonte and I. Bosque (dirs.) Gramática Descriptiva de la Lengua Española, 1631–1722. Madrid. EspasaCalpe. ———. 2008. “Clitic impersonal constructions in Romance: Syntactic features and semantic interpretation”. In Impersonal constructions in grammatical theory, A. Siewierska (ed.) Special issue of the transactions of the philological society, 106, 2, 290–336. Oxford: Blackwell. Monge, Félix. 1954. Las frases pronominales de sentido impersonal en español. Archivo de Filología Aragonesa VII: 7–102. Nevins, Andrew. 2007. The representation of third person and its consequences for person-case effects. Natural Language & Linguistic Theory 25 (2): 273–313. Ordóñez, Francisco, and Esthela Treviño. 2016. Agreement and DOM with impersonal Se: A comparative study of Mexican and Peninsular Spanish. In The Morphosyntax of Portuguese and Spanish in Latin America, ed. M. Kato and F. Ordóñez. Oxford: Oxford University Press. Ormazabal, Javier, and Juan Romero. 2013. Object clitics, agreement and dialectal variation. Probus 25 (2): 301–344. Otero, Carlos P. 1999. Pronombres reflexivos y recíprocos. In I. Bosque & V. Demonte (dirs.) Gramática descriptiva de la lengua española, 1427–1517. Madrid: Espasa-Calpe. Perlmutter, David. 1971. Deep and surface structure constraints in syntax. New York: Rinehart and Winston Inc. Pesetsky, David. 1995. Zero syntax. Cambridge: MIT Press. Portilla, Mario. 2007. Diacronía de las construcciones con pronombres reflexivos en español. Filología y Lingüística 33: 131–149. Pountain, Christopher J. 2000. Pragmatic factors in the evolution of the Romance reflexive (with special reference to Spanish). Hispanic Research Journal 1: 5–25. Pujalte, Mercedes, and Andrés Saab. 2012. Syncretism as PF-repair: The case of se-insertation in Spanish. In The end of argument structure? ed. M.C. Cuervo and Y. Roberge, 229–260. Bingley: Emerald Press. Pylkkänen, Liina. 2000. On stativity and causation. In Events as grammatical objects: The converging perspectives of lexical semantics and syntax, ed. C. Tenny and J. Pustejovsky, 417–445. Stanford: CSLI Publications. Ramchand, Gillian. 2008. Verb meaning and the lexicon: A first phase syntax. Cambridge: Cambridge University Press. Raposo, Eduardo, and Juan Uriagereka. 1996. Indefinite SE. Natural Language & Linguistic Theory 14 (4): 749–810. Reinhart, Tania, and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24: 657–720.

30

G. Armstrong and J. E. MacDonald

Roberts, Ian. 2010a. Agreement and head movement: Clitics, incorporation and defective goals. Cambridge: MIT Press. ———. 2010b. A deletion analysis of null subjects. In Parametric variation: Null subjects in minimalist theory, ed. Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan, 58–87. Cambridge: Cambridge University Press. Roberts, Ian, and Anna Roussou. 2003. Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rooryck, Johan and Guido Vanden Wyngaerd. 2011. Dissolving binding theory. Oxford, UK: Oxford University Press. Rothmayr, Antonia. 2009. The structure of stative verbs. Amsterdam: John Benjamins. Sánchez López, Cristina. 2002. Las construcciones con ‘se’. Madrid: Visor. Schäfer, Florian. 2008. The syntax of (anti-)causatives. Amsterdam: Benjamins. Schäfer, Florian and Margot Vivanco. 2016. Anticausatives are weak scalar expressions, not reflexive expressions. Glossa: A Journal of General Linguistics 1(1), 18: 1–36. https://doi. org/10.5334/gjgl.36 Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 79: 859–890. Suñer, M. 1990. Impersonal Se passives and the licensing of empty categories. Probus 2 (2): 209–231. Torrego, Esther. 1995. From argumental to non-argumental pronouns: Spanish doubled reflexives. Probus 7: 221–241. van Gelderen, Elly. 2011. The linguistic cycle. Oxford: Oxford University Press. Wehrli, Eric. 1986. On some properties of French clitic Se. The grammar of pronominal clitics. In Syntax and semantics, ed. H. Borer, vol. 19, 263–283. San Francisco: Academic. Wood, Jim. 2014. Reflexive -st verbs in Icelandic. Natural Language and Linguistic Theory 32: 1387–1425. ———. 2015. Icelandic morphosyntax and argument structure. Dordrecht: Springer. Wood, Jim, and Alec Marantz. 2017. The interpretation of external arguments. In The verbal domain, ed. R. D’Alessandro, I. Franco, and A. Gallego, 255–278. Oxford: Oxford University Press. Zribi-Hertz, Anne. 1982. The ‘Middle-Se’ Construction in French and its Status in the Triad Middle Voice–Passive–Reflexive. Lingvisticae Investigationes 6: 345–401.

Part I

Diachronic Perspectives

The Development of Se from Latin to Spanish and the Reflexive Object Cycle Matthew L. Maddox

Abstract The Romance clitic se/si in reflexive, passive, and related constructions has been analyzed as a functional head/verbal inflection (Cuervo MC, Datives at large. PhD dissertation, MIT, 2003; Folli R, Harley H, Linguistic Inquiry 38:197– 238, 2007) or as a pronominal argument (Raposo E, Uriagereka J, Natural Language and Linguistic Theory 14:749–810, 1996; D’Alessandro R, Impersonal Si constructions. Mouton de Gruyter, Berlin, 2007), or as both depending on type (Kempchinsky P, Teasing apart the middle. In Andolin gogoan/Homenaje a Andolin Eguzkitza, eds. Itziar Laka and Beatriz Fernández, 532–547. University of the Basque Country Press, 2006). I show that Latin and Old Spanish se was within a DP argument as indicated by diagnostics of coordination, modification, and movement/interpolation. This changes in Middle Spanish, where se merges within a DP and moves as a D head. This reanalysis is supported by loss of interpolation and the development of the doubling of the tonic reflexive pronoun by se. Modern Spanish se is an inflectional morpheme, the realization of a Voice head. Following van Gelderen E, The linguistic cycle: Language change and the language faculty. Oxford University Press, Oxford, 2011, I argue that se’s current status results from a subtype of the object agreement cycle which turns object pronouns into object agreement on the verb. The reflexive object cycle turns reflexive pronouns into valency marking inflection. Key words Passive se · Impersonal se · Grammaticalization · Reflexive pronoun · Clitic doubling

M. L. Maddox (*) Department of Modern Languages and Literatures, University of Nebraska-Lincoln, Lincoln, NE, USA e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_2

33

34

M. L. Maddox

1 Introduction The Spanish third-person reflexive pronoun se occurs with a verb in configurations resulting in different readings such as reflexive (1), passive (2), impersonal, etc. (1)

Juan se lava (2) Juan Reflse washes ‘Juan washes himself’

Se destruyeron las casas Passse destroyed the houses ‘The houses were destroyed’

This property is present, with some variation, throughout all Romance. A primary question addressed by scholars regarding se-constructions is: how should we analyze se? Some argue it is a functional head/inflection (Cuervo 2003; Folli and Harley 2007) while others analyze it as a pronominal argument (Raposo and Uriagereka 1996; D’Alessandro 2007). A mixed position has also been proposed by scholars such as Dobrovie-Sorin (1998) and Kempchinsky (2006), according to whom se may be a functional head or argument, depending on type (see Armstrong and MacDonald [this volume] for discussion). In what follows, I present data from Latin, Old Spanish (OS), and Middle Spanish (MidS) that support the analysis of Modern Spanish (MS) se as a functional head.1 Following Cuervo (2003, 2014), Folli and Harley (2007), and MacDonald (2017) I take passive and impersonal se to be the spell out of a v or Voice head. I claim that the status of se as inflection is due to grammaticalization.2 In Latin and OS, se was in a phrasal reflexive argument, the head of a DP. In MidS, se can move as a D head out of DP. In MS, it is reanalyzed as a higher functional head marking valency.3 Se has undergone a subtype of the “object agreement cycle,” as discussed in van Gelderen (2011). The cycle that I propose for se is summarized below. (3) Stage (a) Latin sē ¼ DP

Stage (a) ! (b) Stage (b) Stage (b) ! (c) Old Spanish Middle Spanish Modern Spanish se ¼ DP se: DP > D se ¼ Voice

This “Reflexive Object Cycle,” turns the reflexive pronoun into valency inflection on the verb. Grammaticalization is the process that turns lexical items into grammatical items; e.g., subject pronouns becoming subject agreement on the verb, demonstratives becoming copulas, etc. This process has been studied by authors such as C. Lehmann (1995), Heine and Kuteva (2002), Hopper and Traugott (2003), Roberts

1

Latin (200 BCE–500 CE), Old Spanish (1200–1400), Middle Spanish (1400–1600), Modern Spanish (1600–present). 2 As noted by Faltz (1977, 2008) and van Gelderen (2011), this cycle has crosslinguistic parallels. 3 The change in categorial status of clitics generally from maximal to minimal projection from OS to MS has been argued by other scholars such as Rivero (1986, 1991); Fontana (1993). I build on this by incorporating Latin into the discussion and focusing on the reflexive clitic se. See also Martins (2003).

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

35

and Roussou (1999). van Gelderen (2004, 2011) approaches grammaticalization from the perspective of the Minimalist Program (Chomsky 1995). She argues that grammaticalization is cyclical and is the result of economy principles such as the Head Preference Principle (4) and the Late Merge Principle (5). (4)

Head Preference Principle (HPP) (5) Be a head rather than a phrase

Late Merge Principle (LMP)4 Merge as late as possible

The HPP in (4) motivates the reanalysis of phrases into heads while the LMP in (5) drives grammaticalization of a lexical head to a functional head or a functional head to a higher functional head. The intuition behind these principles is related to language acquisition. Based on previous studies (Bloom 1970; Diessel 2004), van Gelderen argues that children exhibit a preference for heads over phrases during acquisition. Thus, when possible they will reanalyze phrases as heads. Additionally, grammaticalization is cyclical because after an item is reanalyzed, “renewal” may take place, subsequently restarting the cycle. For example, take the case of subject pronouns becoming subject agreement. A fully phrasal pronoun becomes phonologically weakened and is then reanalyzed as just its D head and later as a subject agreement affix. After this another overt DP or pro may renew the cycle (van Gelderen 2011: 37–85). On this analysis, clitic doubling is a type of renewal. This paper is divided into six sections. In Sect. 2, I present the object agreement cycle, focusing on direct object clitics in Spanish. I propose (building on van Gelderen 2011) that a subtype of this cycle affects reflexive pronouns; i.e., the reflexive object cycle. The focus of Sect. 3 is direct object clitic doubling. This may be used to diagnose how far an element has progressed toward becoming agreement. In Sect. 4, I move on to the reflexive clitic and show that it has changed its status from being phrasal to a head on the way from Latin to Spanish based, in part, on diagnostics of coordination, modification, and movement. I then extend clitic doubling as a diagnostic to se in Sect. 5. Section 6 is a summary of the three stages of the reflexive object cycle. In Sect. 7, I present my conclusion.

2 The Object Agreement Cycle (OAC) 2.1

Object Pronouns > Object Agreement

Object pronouns can be grammaticalized into object agreement morphology. This process has taken place to a varying extent in Spanish, Southern Slavic, Bantu, and Austronesian languages (van Gelderen 2011: 87). In this case, the HPP (see example 4 above) results in a phrasal object pronoun being reanalyzed or reduced to its D

4

This principle originated in Chomsky (1995).

36

M. L. Maddox

head. The OAC takes place in three stages. At stage (a), the object pronoun is merged within a DP that can be modified, coordinated, or undergo XP movement. The HPP takes effect at stage (b), where the DP is reanalyzed as its D head. It merges in a theta-position within a DP, but then moves as a D head following Chomsky (1995). As a head, it can be picked up by the verb on the way to v. At this stage, an additional coreferential full nominal is not allowed; i.e., no clitic doubling. At stage (c), the clitic/pronoun is reanalyzed as a higher functional head (such as v) or as features of that head via the LMP. At this point, renewal can occur with a coreferential pro or a full nominal doubling the clitic in object position. If an overt pronoun merges instead of pro, the cycle may start over again. These stages are summarized in (6) below. (6) Stages of the OAC Stage (a): Pronoun in D

Stage (b): Pronoun in DP, moves as D

vP v

vP v’

VP V

D+v

DP

VP V

DP

Stage (c): Pronoun = v vP v

VP V

pro/DP

One challenge to this framework is the issue of how to distinguish phrases from heads. Van Gelderen (2011: 38) notes that “agreement markers are always heads and nominal arguments are typically phrases...Pronouns can be either heads or phrases.” She utilizes the following diagnostics based on Zwicky and Pullum (1983), Cardinaletti and Starke (1999), and Mithun (2003): (1) phrases can be coordinated, heads cannot; (2) phrases can be modified, heads cannot; (3) phrases bear theta-roles and occur in specifier position. Additionally, agreement has a fixed position and is obligatory while DP arguments are optional and can undergo XP movement. Spanish object clitics are an example of the OAC in progress. The third-person direct object clitics are derived from Latin demonstratives. In Latin, the demonstratives are phrasal since they can be coordinated, as in (7a).5 In MS, coordination is not possible (7b).

5

All diachronic data are cited in the following format: author, title, chapter/section; date.

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

(7) a.

b.

37

et illum et me vehementer ignoras6 both him and me vehemently not-know ‘Both him and me you vehemently do not know’ *no lo y me conoces not him and me know ‘You do not know him and me’

This suggests that Spanish object clitics are heads. In fact, various authors (Fontana 1993; Franco 1993) have argued that these clitics function more like object agreement; i.e., heads.

2.2

Reflexive Object Cycle

The OAC is not restricted to object pronouns; it can also affect reflexive pronouns. van Gelderen (2011: 120), based on data in Faarland (2004) and Ottosson (2004), outlines this process in Scandinavian.7 In Old Norse, reflexive sik is phrasal since it can be modified by sjalfa, as in (8): (8) Sumir hofðu sik sjalfa deydda. some had REFL.Acc self.Acc killed ‘Some had themselves killed.’

Old Norse

As (9) shows, it can also be a suffix. (9) Kalla-sk. Old Norse calls-REFL ‘He calls himself/He is called.’ In Modern Swedish, sik has become the valency marking suffix -s, as in (10). (10) Dörren öppnades door-the opens-REFL ‘The door opens’

Modern Swedish

The phrasal pronoun sik was weakened and is now a head. Two stages of this cycle are represented below in (11), adapted from van Gelderen (2011: 121). 6

Cicero, Pro Rabirio Postumo 33.2; 54 BCE See also Cennamo (1993) for a functionalist analysis of this cycle in Romance and Russian. The grammaticalization of reflexives into valency markers is well-attested cross-linguistically; v. Geniušienė (1987), Haspelmath (1990), and references in Heine and Kuteva (2002). A reflexive cycle has also been proposed by Faltz (1977, 2008). 7

38

(11)

M. L. Maddox

Stage (a)

Stage (b)

vP

vP v'

dörren öppnade + v

dörren

öppnade + -s + v

VP

sik V

v'

V'

VP

V'

V

In stage (a) in (11), sik merges in object position (Spec,V), and may then undergo XP movement.8 At stage (b), the reflexive is still merged in object position but it moves out of the DP as a D head to v. When the verb moves to v, the D head attaches to the verb. If the reflexive suffix can be doubled, Swedish would be at stage (c) of the cycle which would be followed by renewal. My central proposal is that this is the type of cycle that took place in Spanish.

3 Clitic Doubling as a Grammaticalization Diagnostic 3.1

Background

Clitic doubling serves as a diagnostic to determine how far a language has progressed in the object agreement cycle (van Gelderen 2011: 89). At stage (b) no doubling is allowed because the clitic merges within a DP in theta-position and then moves as a head. Movement into a theta-position, or into a position that an element has already moved out of is not possible (Chomsky 1995) and thus no XP may merge into the direct object position.9 However, at stage (c), doubling is permitted because the former DP is now a v head. The object position is open and needs to have something merged in order for the theta-role to be assigned. This means that if a language exhibits unrestricted doubling, it is at stage (c) in the object agreement cycle.

8 In (6) the pronoun is complement to V. In (11), the pronoun is in Spec,V, which is how it is represented in van Gelderen (2011). 9 This assumption has been challenged; see Saito (2001).

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

3.2

39

Object Clitic Doubling

Latin object pronouns were phrasal and thus there was no doubling. As pronouns become heads, doubling should increase, since the object position can now be left open. MS shows variation with respect to doubling, suggesting that varieties of Spanish are at different stages in the object cycle.10 Clitic doubling in MS has been studied extensively (see references in Anagnostopoulou 2006). Direct object clitic doubling is conditioned by definiteness, animacy, and the presence/absence of a differential object marker. The most advanced varieties lack these constraints. Consider the following data taken from Jaeggli (1982), Suñer (1988), Mayer (2003), Hill (1987), and Franco (1993), some of which is discussed in van Gelderen (2011: 102). (12) a.

[animate, +specific, +definite] (*La) Vimos la casa de Maria it we-saw the house of Mary ‘We saw Mary’s house’ b. [+animate, +pronominal, +definite] *(Lo) vimos a él him we-saw DOM he ‘We saw him’ c. [+animate, pronominal, +specific] *Pedro lo vio a Juan Pedro him saw DOM Juan ‘Pedro saw Juan’

“Standard” Spanish

Example (12a) shows that an inanimate, specific, definite object cannot be doubled. The typical case of doubling is in (12b), where a [+animate, +pronominal] object occurs. When this is the case, doubling is obligatory. This contrasts with (12c), where a [+animate, +specific] cannot be doubled. Thus, doubling is restricted in Standard Spanish. Compare the standard doubling in (12) with Rioplatense Spanish in (13) below. (13) a.

10

[+animate, +specific] Pedro lo vio a Juan Pedro him saw DOM Juan ‘Pedro saw Juan’

Rioplatense Spanish

See Maddox (2019: 69ff) for an analysis of clitic doubling at different stages of the object agreement cycle and the proposal that patterns of object movement feed the reanalysis of object clitics.

40

M. L. Maddox

b.

c.

[+animate, +specific, +definite] Las saludé a las maestras del jardín11 them I-greeted DOM the teachers of-the kindergarten ‘I greeted the teachers from the kindergarten’ [animate, +specific, +definite] *La compramos esa novela it bought that novel Intended: ‘We bought that novel’

As (13a, b) show, in Rioplatense a [+animate, +specific, +definite] object can be doubled, a situation not acceptable in standard Spanish (12c). Thus, this variety is further along in the cycle. Nevertheless, as seen in (13c), a [animate] object cannot be doubled. In Malinche Spanish and Patagonia Spanish, however, an inanimate object can be doubled, as in (14), which is indefinite, and (15), which is definite. (14) [animate, specific, definite] Lo trae un chiquihuite it he-brings a basket ‘He brings a basket’ (15) [animate, +specific, +definite] Lo agarré (a)l mate it I-took DOM-the mate ‘I took the mate (a type of tea)’

Malinche Spanish12

Patagonia Spanish13

Thus, these varieties are further along in the cycle. Furthermore, as discussed in Zdrojewski and Sánchez (2014), some varieties even allow the doubling of quantified objects, which is the most advanced type of doubling. The varieties in which this happens are at the end of stage (c), while those with more restricted doubling are in the beginning of stage (c).

3.3

Intrasystemic Diachrony

As discussed above, direct object clitic doubling is subject to dialectal variation; some varieties of Spanish are at different stages in the cycle. Indirect object clitics, on the other hand, do not exhibit such variation and are much less constrained with respect to doubling (v. Suñer 1988). Ormazabal and Romero (2013) show that not all clitics should be treated equally. The third-person DO clitics form one group, Type

11

(13b) is from Zdrojewski and Sánchez (2014: 164). Malinche Spanish, spoken in Puebla and Tlaxcala, Mexico, may be a contact variety. If so, this could be an instance where contact has accelerated a cycle already under way. 13 Example (15) is from Vilanova et al. (2016), who also discuss non-canonical types of doubling in Lima Spanish, Andean Spanish, and Judeo-Spanish. 12

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

41

1, while all other DO clitics and all IO clitics form another group, Type 2.14 Patterns of doubling suggest that the two types ought to be kept distinct. In what follows I briefly review some of the arguments in Ormazabal and Romero (2013) and I propose that the two types are at different stages in the cycle. First, consider the doubling of quantified strong pronouns, as in the following data15: (16) a.

(*Los) he comprado todos ellos them I-have bought all them ‘I have bought all of them’ b. *(Les) he pegado pegatinas a todos ellos on-them I-have stuck stickers to all them ‘I have stuck stickers on all of them’ c. *(Nos) han visto a todos nosotros us they-have seen DOM all we ‘They have seen all of us’ d. *(Os) han visto a todos vosotros you.Pl they-have seen DOM all you.Pl ‘They have seen all of you

In (16a), the Type 1 clitic (los) cannot double the quantified pronoun, while in (16b-d), doubling with a Type 2 clitics (les, nos, os) is obligatory. Additionally, Type 1 clitics are more restricted in the types of arguments they can double. For example, third-person DO clitics (Type 1) cannot double a dislocated, negative DP, while IO clitics (Type 2) can. (17) a.

*Ningún libro lo han vendido not-any book it they-have sold Intended: ‘Not any book have they sold’ b. A ninguna estudiante le han dado el título to not-any student to-him they-have given the degree ‘To not any student have they given the degree’

Furthermore, first- and second-person DO clitics (Type 2) do not show any restrictions. They can double definite DPs (18a), quantified expressions (18b), whphrases (18c), and nonreferential phrases (18d). (18) a.

b.

14 15

Os han visto a los niños you.Pl they-have seen DOM the children ‘They have seen you children’ Os vimos a algunos / muchos niños you.Pl we-saw DOM some / many children ‘We saw some/many of you children’

The “Type 1” and “Type 2” labels are my own terminology. The data in this section are adapted from Ormazabal and Romero (2013).

42

M. L. Maddox

c.

d.

¿A quiénes / cuántos os han elegido para el puesto? DOM whom / how-many you.Pl they-have chosen for the position ‘Whom/how many of you have they chosen for the position?’ No os encontraron a nadie / ninguno not you.Pl they-found DOM no-one / not-any ‘They found none/not any of you’

Thus, Type 1 clitics are very restricted when it comes to doubling, while Type 2 clitics are not, suggesting the two types should be kept distinct. Given the assumptions discussed above with respect to doubling and the stages of the object agreement cycle, the differences displayed by the two types of clitics serve as evidence bearing on their diachronic status. The further along an element is in the cycle, the less constrained it is in what it can double. Since Type 1 clitics are highly constrained, they are early on in stage (c) while Type 2 clitics have completed this stage, given their unrestricted doubling. Hence, clitics show dialectal variation but they also display diachronic intrasystemic variation. That is, even within a single variety, so-called “standard” Spanish, Type 2 clitics are more grammaticalized than Type 1. Standard Type 2 clitics are agreement morphemes. Type 1 clitics are D heads that undergo movement in doubling constructions. In conclusion, the two classes of clitics argued for by Ormazabal and Romero synchronically are also valid diachronically and are the result of the OAC. Doubling as a diagnostic is extended to se in Sect. 5 below.

4 The Categorial Status of Se and the Stages of the Reflexive Object Cycle Se changed its status diachronically from being phrasal in Latin, to moving as a D head in Middle Spanish, and finally a Voice head in Modern Spanish. Originally a pronominal argument, se is now an inflectional element. Evidence for this comes from differences with respect to modification, coordination, movement, and patterns of auxiliary selection. The changes from DP to D head to Voice head each represent a stage of the reflexive object cycle.

4.1 4.1.1

Stage (a): Latin Background

Latin had an -r suffix parallel to MS se; i.e., it could have a reflexive (19), anticausative (20), passive, or impersonal reading.

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

43

(19) Excepit Seleucus fabulae partem et ‘ego’ inquit ‘non cotidie lavor’16 took-hold Seleucus conversation part and I said not daily wash ‘Seleucus took up part of the conversation and ‘I,’ he said, ‘do not wash myself daily’ (20) Omnis liquor vapore solvitur ac frigoribus magnis conficitur17 all liquid heat dissolves and great cold congeals ‘All the liquid is dissolved by the heat and congealed by great cold’ Latin also had a reflexive pronoun sē, the ancestor of MS se, which occurred in reflexive (21) and anticausative constructions (22).18 (21) similī tālem sē vidit in aurō19 (22) dum calor sē frangat20 likewise such Reflse sees in gold until heat AntiCse breaks ‘Likewise he sees himself in the gold” ‘. . . until the heat goes down’

4.1.2

Distribution

Latin sē has the distribution of a phrase rather than a head. For example, Latin sē can be coordinated, as in (23a), while MS se cannot (23b). (23) a.

b.

mē et sē hīsce impedīvit nuptiīs!21 Me and Reflse this shackled marriage ‘He shackled me and himself in this marriage!’ *Me y se aprisionó en este matrimonio Me and himself imprisoned in this marriage

Additionally, Latin sē can be modified, as in (24a) by an intensifier, while its MS counterpart cannot (24b). (24) a.

b.

16

sē ipse sine mūnītiōne dēfenderet.22 Reflse very.M.S. without fortification defended ‘He defended his very self without fortification.’ *se mismo defendió sin municiones. Reflse very defended without fortification

Petronius, Satyricon, 42.2.1; 66 CE Columella, De re rustica, 1.6.18.7; 70 CE 18 There is no consensus in the literature as to whether there was a passive se in Latin; v. Kärde (1943), Monge (1954), Cennamo (1999), Adams (2013). There is no evidence of an impersonal se. 19 Statius, Achilleid, 1.865; 94–96 CE 20 Cicero, De Oratore 1.265; 55 BCE 21 Terence, Phormio, 2.4; 161 BCE 22 Caesar, de Bello Gallico, 20.5; 58–49 BCE 17

44

M. L. Maddox

Finally, Latin sē can be separated from the verb, as in (25). (25) apud Platonem Socrates in caelum effert laudibus Protagoram Hippiam with Plato Socrates in heaven brings praises Protagoras Hippias Prodicum ceteros, sē autem omnium rerum inscium fingit et rudem.23 Prodicus others Reflse but all things ignorant represents and coarse ‘In the writings of Plato, Socrates praises to the heavens Protagoras, Hippias, Prodicus, and others, but himself he represents as coarse and ignorant of all things’ This is not possible in MS, as (26) below shows. (26) *Juan se no lava regularmente John Reflse not washes regularly Intended: ‘Juan does not wash himself regularly’ Latin sē could be contrastively focused. In (25) above, Cicero states that Socrates praises other philosophers exceedingly, but considers himself ignorant. MS se cannot be contrastively focused, as (27) shows. sus amigos mucho pero SEi considera ignorante (27) *Juani alaba a John praises DOM his friends a-lot but Reflse considers ignorant Intended: ‘Juan praises his friends a lot but he considers himself to be ignorant’ Thus, Latin sē differs from its MS descendant in at least four ways: it can be coordinated, modified, separated from the verb, and contrastively focused. Latin sē merged in a full DP. It was at stage (a) in the reflexive object cycle.

4.2 4.2.1

Stage (a) Moving to Stage (b): Old Spanish (1200–1400) Distribution

From the earliest documents, OS had Reflse, AntiCse, and Passse (Monge 1954; Cennamo 1993). However, OS se differs from MS se. First, it can be separated from the verb by different constituents. This “interpolation,” is illustrated in (28) through (30); see also Chenery (1905), Ramsden (1963), Poole (2013), inter alios. (28) Desí mando que se non rasiessen24 thus order that Reflse not shave ‘Thus I order that they not shave (themselves)’ 23 24

Cicero, Brutus, 292.14; 46 BCE. Alfonso X, General Estoria, Primera Parte, fol. 277V; c. 1275

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

45

(29) este algodon es atal que se non quema por fuego25 this cotton is such that AntiCse not burns by fire ‘This cotton is such that it does not burn from fire’ (30) No hay guisa por que se esto diga26 not is fashion by which Passse this say ‘There is no way by which this is said’ Latin sē displayed the same behavior, as discussed above, which is evidence supporting it as a merging in a DP. That OS clitics are maximal projections has been proposed by others (Rivero 1986, 1991; Fontana 1993). Fontana claims that OS clitics are second-position clitics and thus XPs in OS, but are reanalyzed as heads in MS. However, OS interpolation is rather constrained, occurring primarily in embedded clauses and limited to specific types of interveners. This suggests that, while se was in a DP in both Latin and OS, it was undergoing reanalysis among some speakers, hence the rarity of interpolation. Thus, OS was not firmly at stage (a) of the cycle but rather moving towards stage (b), where it can move as a D head.

4.2.2

Auxiliary Selection

More evidence that se merges in a DP argument in OS comes from auxiliary selection.27 In languages like Italian, passives and unaccusatives in the compound past select the BE auxiliary, essere, as in (31) and (32), respectively. (31)

La lettera è scritta (32) the letter is written ‘The letter is written’

Giovanni è venuto Giovanni is come ‘Giovanni came’

Thus, both have intransitive syntax and, consequently, a derived subject (McGinnis 2004). With transitive verbs, the compound past is formed with the HAVE auxiliary, avere, as in (33). (33) Giovanni ha accusato Luigi Giovanni has accused Luigi ‘Giovanni accused Luigi’

25

Anonymous, Lapidario, fol. 17R; c. 1250 Anonymous, Calila e Dimna, Chap. 3; 1251 27 See McFadden (2007) for an overview of auxiliary selection. The Italian data in this section are from Burzio (1986) and McGinnis (2004). 26

46

M. L. Maddox

The same pattern applies if a verb is used reflexively with the full DP se stesso, but if the clitic si is used, the BE auxiliary is selected. This is illustrated in (34) and (35) below. (34)

Giovanni ha accusato se stesso (35) Giovanni has accused self very ‘Giovanni accused himself’

Giovanni si è accusato Giovanni Reflse is accused ‘Giovanni accused himself’

Since other intransitives select BE, (35) is evidence that Italian si-constructions have intransitive syntax; i.e., a single argument merges in object position and moves to subject position. OS, unlike MS, also exhibited auxiliary selection but the patterns do not match Italian. Aránovich (2003) shows that most OS reflexive verbs select the HAVE auxiliary. Consider the following: (36) Minaya Alvar Fáñez essora es llegado28 Minaya Alvar Fáñez then is arrived ‘Minaya Alvar Fáñez then arrived’ (37) como ninguno de los athenienos no se a vestido de negro por mi29 because none of the Athenians not Reflse has dressed of black for me ‘. . .because none of the Athenians has dressed in black for me’ (38) mas es necesario que quando el se ha echado en tierra. . .30 but is necessary that when he Reflse has thrown on ground ‘But it is necessary that when he has cast himself to the ground. . .’ The datum in (36) shows that unaccusatives selected the BE auxiliary in OS. This property is lost in Early Modern Spanish, with the latest examples occurring in the seventeenth century. According to Benzing (1931), the verb llegar continues to select BE until the sixteenth century. Examples (37) and (38) contain compound reflexives, and both select HAVE, a pattern not expected given Italian data above. This is evidence that se-constructions in OS were transitive, with se being the internal argument. In (38), it is clear that se is the internal and not the external argument, since el is the agent DP which would merge in Spec,v to receive its thetarole.

28

Anonymous, Cid, line 2449; 1207 Juan Fernández de Heredia, Traducción de Vidas paralelas de Plutarco, III, fol. 176v; 1379–1384 30 Ferrer Sayol, Libro de Pallado BNM 10211, para. 115; 1380–1385. 29

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

47

Thus, OS se was at the end of stage (a), moving toward stage (b). Being in a DP internal argument it could undergo constrained types of phrasal movement as demonstrated by the interpolation data. However, I have found no evidence that it could be coordinated or modified as it could be in Latin, which suggests that in OS se was able to move as a head for some speakers; i.e., stage (b).

4.3 4.3.1

Stage (b): Middle Spanish (1400–1600) Loss of Interpolation

At stage (b) of the cycle, the pronoun can move as a D head, which means interpolation of se should disappear, on the analysis of interpolation as XP movement. The historical record is consistent with this and interpolation is indeed lost during this period. As (39) below shows, it does occur but it is quite rare. (39) que se non ficiese deservicio de Dios e daño de la tierra.31 that Passse not do disservice of God and damage of the earth ‘. . .that a disservice not be made to God and damage to the earth. . .’ In order to quantify this loss, I conducted a CORDE search for interpolation patterns with negation, which is the most frequent type.32 The results are given in Table 1 below. Interpolation (Pattern 1) is already a minority pattern in the OS period, declining from 25.2 to 11.8% from the thirteenth to the fourteenth century. By the MidS period, it is quite rare and in the modern period it is almost nonexistent. Table 1 Loss of interpolation by century Century Old Spanish Middle Spanish Early modern Spanish

31

1200–1299 1300–1399 1400–1499 1500–1599 1600–1699 1700–1799

Pattern 1: se + Neg + V 25.2 11.8 6.2 0.5 0.1 0

Pattern 2: Neg + se + V 74.8 81.2 93.8 99.5 99.9 0

Jerónimo Zurita, Anales de la corona de Aragón, Primera Parte, para. 232; 1562 CORDE search conducted Oct. 12, 2015 at 4:00 p.m. See also Chenery (1905) and Ramsden (1963) for more quantitative data on interpolation. See Poole (2013) for quantitative data on the incidence of interpolation from the Corpus del Español. 32

48

4.3.2

M. L. Maddox

Auxiliary Selection

In MidS, auxiliary selection is still active, with unaccusatives like llegar selecting BE, as in (40). (40) a tal puncto el miserable ombre es llegado33 to such point the miserable man is arrived ‘The miserable man has arrived to such a point’ At this stage, compound reflexives continue to pattern with transitive verbs in selecting HAVE, as in (41). (41) ca el mesmo se ha cortado la lengua34 because he himself Reflse has cut the tongue ‘Because he himself has cut out his own tongue’ These patterns, along with the loss of interpolation, suggest that in MidS, se was still merging in an internal argument DP, but it could undergo subsequent head movement.

4.4

Stage (c): Modern Spanish (1600–Present)

In MS, se has been reanalyzed from a D head to Voice via the Late Merge Principle (LMP). Rather than merging within a DP and moving as a D head to Voice, it is now the spell out of Voice. Se is now a head since interpolation is no longer possible. Auxiliary selection is lost in the seventeenth century (Aránovich 2003), so it can no longer be used as a diagnostic. This suggests that MS is in stage (c) of the cycle.

4.4.1

Se as Agreement

MS se shares properties that characterize inflection. This has been independently argued by others (Cuervo 2003, 2014; Folli and Harley 2007) so I present here only a couple points. First, se can interact morphophonologically with other inflection. Consider the following data taken from Harris and Halle (2005).

(42) a.

33 34

Normative Váyan-se go-Pronse ‘Go!/Leave!’

b.

Alternative Váyan-se-n go-Pronse-n ‘Go!/Leave!’

Enrique de Villena, Traducción y glosas de la Eneida Libros I-III, para. 24; 1427–1428 Antón de Zorita, Árbol de Batallas (de Honoré Bouvet), para. 166; c. 1440–1460

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

(43) a.

Sírvan-se b. serve-Reflse ‘Serve yourselves!’

49

Sírvan-se-n serve-Reflse-n ‘Serve yourselves!’

The normative forms in (42a) and (43a) have se coming after the plural agreement affix -n. However, in different registers, speakers place an additional affix after se, as shown in (42b) and (43b). Harris and Halle (2005: 210) note that this phenomenon occurs most frequently with se, me, and le, and less frequently with third-person direct object clitics.35 Additional evidence comes from patterns of repetition in coordinated VPs. This diagnostic has been used by others such as Kayne (1975) and Culbertson (2010) to show that clitics pattern like agreement affixes. In (44) below, for example, the subject agreement affix cannot be omitted in the second conjunct. (44) a. b.

Juan com-e y habl-a ‘John eats and talks’ *Juan com-e y habl-Ø

Now consider how se behaves in this environment: (45) a. b.

Se puede y se debe comer verduras ‘One can and one should eat vegetables’ *Se puede y debe comer verduras

Example (45b) is unacceptable; i.e., se cannot be omitted. Thus, it patterns like an agreement affix. In Sect. 3.3 above, I distinguished Type 1 and Type 2 clitics, based on Ormazabal and Romero (2013). The former are D heads that are at the early part of stage (c) of the OAC, while the latter are v heads at the end stage (c) of the cycle, and thus agreement morphemes. What type is se? Since it is an inflectional element it is an agreement morpheme, a Type 2 clitic. A final piece of evidence supporting this comes from morphological simplicity. Ormazabal and Romero note that Type 1 clitics are marked for gender while Type 2 clitics are not. MS se is not marked for gender and thus patterns with Type 2, which means that it is an agreement morpheme and is in stage (c) of the reflexive object cycle. The typology of clitics, based on Ormazabal and Romero but including se, is summarized in (46) below.

35 Roca (1996) observes, based on Camacho’s (1993) data, that in Latin American dialects the thirdperson direct objects are not acceptable at all in these configurations.

50

M. L. Maddox

(46) Spanish Clitics: Type 1 (Determiner-heads) ! lo, la, los, las Type 2 (Agreement morphemes) ! me, te, nos, os, le, les, se The data from Harris and Halle (2005) and Roca (1996) also lend support to the typology proposed in (46). From the discussion above we may conclude that MS se has a distribution and behavior similar to that of agreement, which is the expected outcome of the cycle.

5 Doubling with Se Diachronically 5.1

Middle Spanish

Doubling becomes possible when se is no longer phrasal. Thus, it is not expected at stage (a); i.e., Latin and Old Spanish. At stage (b), an additional coreferential XP is not allowed because se merges within the internal argument in complement position and then moves as a head. As se loses features and becomes reanalyzed as a Voice head doubling begins. Examples of doubling start to increase in frequency during this period, as in (47), where a si mismo is doubled by se. (47) si es necçessario que el onbre se ame a si mismo mas que if is necessary that the man Reflse loves DOM self very more than a los otros onbres.36 DOM the other men ‘. . .if it is necessary that one love himself more than others.’ However, as the first clause in (48) shows, for some speakers a sí mismo does not have to be doubled by se. (48) el que a sí mismo aborrece, él se juzga a mal37 he who DOM self very abhors he Reflse judges to evil ‘He who abhors himself, he judges himself to be evil’ To determine how frequent doubling with a sí mismo was, I conducted a search in the CORDE for reflexive clitic doubling in the OS and MidS periods, the results of which are in Table 2 below.38

36

El Tostado, Alonso Fernández de Madrigal, Libro de amor e amicicia; 1440–1455 Fray Diego de Valencia, Sobre la predestinación y sobre la Trinidad y la Encarnación, para. 22 (1486–1487) 38 The CORDE search was conducted on May 30, 2020, at 3:40 p.m. I searched for a sí mismo in simplex, finite, transitive clauses, with and without the written accent, since orthography was not completely standardized yet. I only counted a sí mismo when postverbal as clitic doubling, since preverbal could be clitic-left dislocation. In the tokens column, the number on the left is the total occurrences of se; the number on the right is the number doubled. 37

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

51

In the OS period, doubling is very rare. It starts to appear with some regularity in the sixteenth century. Table 2 Clitic doubling with se by century Period/Century Old Spanish Middle Spanish Early modern Spanish

1200–1299 1300–1399 1400–1499 1500–1599 1600–1699

Tokens 177/3 46/3 172/8 529/72 439/94

Percentage 1.7% 6.5% 4.7% 13.6% 21.4%

Interestingly, I found zero tokens of doubling co-occurring with interpolation. This is to be expected if the cycle is at stage (c). Poole (2013:83) discusses data in which interpolation and putative “clitic doubling” do co-occur, as in (49) below.39 (49) Et quando Diana andaua a caça o a correr so mont & la siesta la alla and when Diana went to hunt or to run her mountain and the nap it there alla tomaua40 she-took ‘And when Diana went hunting or to climb her mountain and took a nap there. . .’ In the second clause in (49), the direct object la siesta is doubled by the clitic la. However, this appears to be clitic-left dislocation rather than true clitic doubling since the doubled constituent is in the left periphery. Poole (2013: 84) also labels this as clitic-left dislocation. The loss of interpolation, auxiliary selection of reflexives with HAVE, and increase of doubling suggest that MidS was at stage (b) in the cycle, with incipient progression toward stage (c), where doubling should be allowed.

5.2

Modern Spanish

In MS, doubling of se is common. Since se is a Voice head and not an argument, the object position is open for DPs. The data in Table 2 above show that doubling starts to increase in the sixteenth century, the last century of what I periodize as “Middle Spanish.” However, by the seventeenth century, doubling with a sí mismo increases

39 40

I thank an anonymous reviewer who made me aware of these data in Poole (2013). Alfonso X, General Estoria II; thirteenth century

52

M. L. Maddox

to 21.4%, suggesting that for many speakers, se is now a head.41 An Early Modern Spanish example of doubling is given below in (50). (50) No tiene humildad interior el que no se aborrece a sí mismo con not he-has humility interior he who not Reflse abhors DOM self very with un mortal odio. . .42 a mortal hate ‘He does not have humility who does not hate himself with a mortal hate’ For some Early Modern speakers, however, se is not yet obligatory, as the second clause in (51) below shows. (51) ni el Eterno Padre perdonó a su Hijo, ni el Hijo neither the eternal father pardoned DOM his son no the son perdonó a sí mismo43 pardoned DOM self very ‘Neither did the Eternal Father pardon his Son, nor did the Son pardon himself. . .’ The present-day state-of-affairs is that Reflse can be optionally doubled by a sí mismo, or a prepositional phrase with a pronominal complement, in which case a bound reading is induced (Torrego 1995), as is shown in (52). lava (proi / a sí / a sí mismo) (52) Juani sei John Reflse washes him DOM self DOM self very ‘John washes himself’ Crucially, Reflse cannot be omitted in the presence of a sí mismo as a complement (53).44 (53) Juan *(se) lava a sí mismo Thus, in MS, reanalysis from a D head to Voice is complete.

6 The Reflexive Object (Se) Cycle in Three Stages The stages of the reflexive object cycle are summarized in (54) below.

41

Calculated from the same CORDE search discussed above. Miguel de Molinos, Guía espiritual; 1675–1676 43 Antonio Panes, Escala Mística y Estímulo de Amor Divino; 1675 44 Labelle (2008) makes the same observation for observation for French. 42

The Development of Se from Latin to Spanish and the Reflexive Object Cycle

53

(54) Reflexive (Se) Object Cycle Stage (a) - Latin/(Old Spanish)

Stage (b) - Middle Spanish

TP

TP vP

T Caesar

vP

T v'

vidit + v

César VP

v'

se+vede+v

sē

V' V

V DPDat > DPAbsolutive: 1st /2nd person (> notates c-command) *

6

Let me note that the main point of Rezac (2006) is not the intervention-based explanation of the PCC, which had already been assumed by many theoreticians, but rather new empirical data that confirms the intervention analysis. Rezac observes that this analysis predicts that the PCC disappears if the DPAbsolutive is raised past the intervening DPDat. And indeed, this prediction is confirmed (although only for a sub-set of Basque speakers, those speakers who allow ‘absolutive displacement’, i.e., they allow the subject of applicative unaccusatives to become ergative (crucially this happens in all and only PCC contexts).

Implicit Agents and the Person Constraint on SE-Passives

119

In order to explain why Basque applicative unaccusatives allow the DPAbsolutive to be third person (this is a crosslinguistically general property of PCC configurations), Rezac (2006) follows Anagnostopoulou (2003) in assuming that AGREE in Person should be kept distinct from AGREE in Number: DPDat would be inherently valued as +Person, thus blocking AGREE in Person between little v and DPAbsolutive, but not AGREE in Number, which would be sufficient for the legitimation/Case-checking of a DPAbsolutive marked as third person (recall that according to a rich literature, third Person is to be analyzed as absence of Person, or maybe as [-Person]). Giurgea (2015) proposes that the intervention-based explanation of the PCC can be extended to cover the PC on SE-passives. Thus, in (18) AGREE in Person between T and the subject DP (carrying the Theme role) would be blocked by the intervening Initiator, arguably marked as [+Person]: (18) [T[uPerson, uNumber] [Initiator+Person [... DPTheme+Person+Number ..]]] *

Underlying this proposal is the hypothesis that the Initiator of SE-passives is projected in a syntactic position that C-commands the DPTheme.7 In the following section I will briefly explain why I am reluctant to assume that the Initiator is syntactically projected in an A-position, which will force me to look for another explanation of the PC.

3 The Initiator in Passives 3.1

Where Is the Initiator in Passives?

Descriptively speaking, the semantic interpretation of passive configurations involves an external argument—labelled Initiator here—that does not sit in the canonical subject position (say Spec,Tense) and is not marked with Nominative Case. In Nom-Accusative languages, it is the Theme of passives that is marked with Nominative Case and (therefore) is arguably in a relation with Tense for its syntactic legitimation. Well-established tests (Agent-oriented adverbs and control into purpose clauses) indicate however that the agentivity of passives is more than just a matter of agentive interpretation. This point is particularly clear if we compare SE-middles and SE-passives: although both types of constructions involve an implicit Agent, it is

7 In order to make his analysis compatible with the hypothesis that the Initiator of passives is not syntactically projected, Giurgea suggests that the offending Person feature might be a feature of the passivizing head Pass, a head supposed to existentially bind the unsaturated external argument position (Bruening 2012): since, in the case of SE-passives, this head not only binds the external argument but also introduces an animacy restriction on the variable it binds, Giurgea suggests that this head bears an interpretable Person feature.

120

C. Dobrovie-Sorin

only the Agent of SE-passives that can act as a controller, a property that is currently referred to as syntactically active Agent: (19) a. Il s’ est vendu plus de 300 livres anciens pour aider les pauvres. it se is sold more of 300 books old for help the poors ‘There have been more than 300 old books sold to help the poor.’ b. Ces livres se vendent facilement (*pour aider les pauvres)8 these books se sell.3pl easily for help.inf the poor ‘These books sell easily.’ // *These books sell easily to help the poor.’ Because of their ‘syntactic activity’, the implicit Agents of passives have been assumed to occupy A-positions, and the proliferation of functional categories in the minimalist program has offered various possibilities, e.g., Spec,vP or more recently Spec,Voice. Some authors have however resisted this line of research and have instead pursued the more traditional view according to which the implicit Agent of passives is not a null pronoun sitting in a syntactic position, but instead is present only for the interpretation, in the form of an existentially bound variable (Keenan 1980, 1985; Schäfer 2008; Alexiadou 2012; Bruening 2012; Kiparsky 2013; Legate 2014; Wood 2015, a.o.). According to a third line of research, the Agent is realized by the passive morpheme itself, in particular by -en (see Baker et al. 1989) or by the functional head itself, see Embick’s (2004) Voice head (which is assumed to lack an external argument but to contain agentive features) or the Passive head in Alexiadou, Anagnostopoulou and Schäfer (2015, 124). In this paper I will side with those who assume that the Agent of passives is not syntactically projected9 in Spec,Voice. My proposal will build on Legate’s (2012, 2014) proposal that the Voicepassive head carries interpretable (which I take to mean ‘inherently valued’) phi-features that restrict the implicit Initiator. In other words, the implicit Initiator of passives is rendered legible at the LF interface due to already valued phi-features on Voicepassive. Important differences with Legate’s assumptions will show up as we proceed.

3.2

The Phi-Features of the Initiator in Acehnese Passives: Legate (2012, 2014)

In Acehnese, the verb agrees with the external argument in both active and passive configurations: 8 French SE-verbs are ambiguous between passive and middle readings. Adverbs such as facilement ‘easily’ are only compatible with the middle reading, which in turn is incompatible with control into a final clause. If facilement is suppressed, the example becomes grammatical, but it loses the middle reading. 9 In this paper I will not be interested in by-phrases. For concreteness I assume that they are adjuncts. I will not try to explain the generalization according to which the by-phrases related to participle passives have the same th-roles as the subjects of the corresponding active verbs.

Implicit Agents and the Person Constraint on SE-Passives

(20)

a. Uleue nyan di-kap lôn. snake DEM 3FAM-bite 1SG ‘The snake bit me’.

121

Active voice

b. Lôn di-kap lé uleue nyan. 1SG 3FAM-bite by snake DEM ‘I was bitten by the snake.’ Crucial for our present purposes is Legate’s (2014) hypothesis that in Acehnese passive configurations, the Initiator is not fed into Spec,Voice. Instead, the Voice head carries valued features that modify (in other words restrict) the ϕ-features of the Initiator: 3Fam indicates that the Initiator is third person. In case a by-DP phrase is introduced, the DP must be 3Fam; in case no by-DP is merged, the variable corresponding to the Initiator in the denotation of VoiceP is existentially closed, yielding an implicit Initiator understood as 3Fam. (21)

a.

VoiceP

le.Initiator(e, the snake) & 3fam(the snake) & Biting(e) & Theme(e, me)

DP the snake

Voice’

lx.le.Initiator(e, x) & 3Fam(x) & Biting(e) & Theme(e, me)

Voice

lx.le.Initiator(e, x) & 3Fam(x)

Voice

f 3Fam

lx.le.Initiator(e, x)

lx.3Fam(x)

vP

le.Biting(e) & Theme(e, me)

v

VP

le.Biting(e) & Theme(e, me)

V bite

lx.le.Biting(e) & Theme(e, x)

b.

DP me

VoiceP

le.l x.Initiator(e, x) & 3Fam(x) & Biting(e) & Theme(e, me)

Voice

vP

lx.le.Initiator(e, x) & 3Fam(x)

Voice lx.le.Initiator(e, x)

f3Fam

lx.3Fam(x)

le.Biting(e) & Theme(e, me)

v

VP

le.Biting(e) & Theme(e, me)

V bite

lx.le.Biting(e) & Theme(e, x)

DP me

122

C. Dobrovie-Sorin

Based on the evidence provided by Acehnese passives, Legate (2014, 85) suggests that crosslinguistically, Voicepass carries semantically interpretable phi-features that restrict the Initiator th-role. I will adopt this hypothesis, but I will revise it in order to account for the difference between SE-passives and participle passives (see Sect. 4 below).

3.3

No ϕP in SE-Passives

Legate (2014, Chap. 4) proposes that what she calls ‘grammatical object passives’ (Icelandic new passive, Ukrainian -no/-to participle construction and Irish highregister autonomous) have a ‘partially projected Agent’, meaning that Spec,Voice is occupied by the phi-features (notated ϕP) of the Initiator, in contrast to ‘canonical passives’ (past participle-passives in Germanic or Romance, but also Acehnese passives), which involve an ‘unprojected’ Agent. Legate (2017) pursues this analysis and includes SE-passives among the grammatical object passives. While I agree that we need to distinguish between two types of passives, I do not think that this can be done by assuming that ϕP is merged in the Spec,Voice of grammatical object passives (SE-passives in particular). The first problem is the nature of ϕP itself. One might think that it stands for a special kind of pronoun, but there are two worries about this guess: on the one hand, minimalist theory treats pronouns on a par with DPs (pronouns are sometimes viewed as ‘intransitive D s’) and on the other hand Legate observes that the relation between her ϕP and the ϕP proposed by Déchaine and Wiltschko (2002) is unclear. Note furthermore that Legate assumes that the ϕP she postulates for grammatical object passives does not saturate Spec,Voice, its only role being exactly the same as that attributed to the phi-feature on the Voice head of canonical passives, namely that of ‘restricting’ the Initiator th-role (i.e., constraining it to refer to the speaker, hearer or some other entity, depending on the phi-features of the verb). The facts that Legate attempts to capture are quite well-known: (1) in all grammatical object passives, the Theme is not promoted to Spec,TP (for SE-passives, see Raposo and Uriagereka 1996; Dobrovie-Sorin 1986, 1994); (2) in the Icelandic new passive, the Theme is assigned Accusative Case. By assuming that the Spec,Voice is filled with ϕP we render it inaccessible for the raising of the Theme and with further assumptions one might also capture Accusative Case assignment to the Theme. This suggestion is a reincarnation of Raposo and Uriagereka’s (1996) analysis of SE-passives: although SE is not a genuine subject DP, it is nevertheless assumed to be somehow related to Spec,Tense, rendering Spec,Tense inaccessible to the Theme. Arguments against this solution can be found in Cabredo Hofherr and Dobrovie-Sorin (2010): at least in Romanian, SE is clearly an Accusative-marked clitic (rather than a subject clitic), and as such its inherent properties cannot be held responsible for the inaccessibility of the Spec,Tense position.

Implicit Agents and the Person Constraint on SE-Passives

123

Note that in Legate’s representations in (21)a–b, no Tense is present, the highest functional category being Voice. This is because, as Legate demonstrates, in Acehnese, the verbal prefix that agrees with the Initiator attaches to little v rather than to Infl (or Tense). In Romance languages, on the other hand, SE clearly attaches to Tense rather than to little v, the only debate being whether SE (as all other pronominal clitics) is externally or internally merged on Tense (i.e., whether it is directly inserted in Tense or rather moved there from a vP-internal position). Legate’s hypothesis regarding ϕP sitting in Spec,Voice requires SE to externally merge (merge first) on little v and then internally merge on (remerge, or move to) Tense. The impossibility of the Theme to get to Spec,Tense would then be due to the fact that it cannot transit through Spec,Voice.

3.4

Summary

Following Legate (2014) I will assume that crosslinguistically, the Voice head of passives is endowed with valued/semantically interpretable phi-features that restrict the Initiator even in those languages in which such features are not overtly realized on the verbal inflection. I will however depart from Legate regarding the analysis of the grammatical object passives, and in particular of SE-passives: they do not involve ϕP-in-Spec,Voice.

4 Valued Phi-Features and Implicit Initiators Across Passive Configurations According to the minimalist program, the functional categories on the verbal spine, Voice and Tense, enter the derivation with unvalued ϕ-features, which get valued via the AGREE relation with a DP, which is thereby ‘Case-checked’. In active and passive configurations, the role of the Voice head is to introduce the Initiator, but anti-causative configurations arguably have a Voice head despite their interpretation lacking an Initiator: according to Pylkkänen (1999), Alexiadou et al. (2006), Harley (2013), Schäfer (2008), a.o., the Voice head of anti-causatives introduces causation but no argument. In line with Alexiadou et al. (2006, 2015) a.o., I will assume several types of Voice heads (respectively corresponding to active, passive and anticausative configurations), all of which occupy the same syntactic position, the lowest functional head on the clausal spine. The details of my proposal are however significantly different from previous analyses. The main innovation will be to adopt Legate’s (2014) hypothesis regarding the valued ϕ-features of Voice in passives and to extend it to Tense in order to account for the passive interpretation of SE-verbs: just like the features of Voice, the features of Tense can be inherently valued, their role being the same, that of restricting the denotation of an implicit Initiator. Under this view, then, Voice is not the only functional category responsible for introducing the Initiator: Tense can also assume that job.

124

4.1

C. Dobrovie-Sorin

Phi-Features and Case-Checking in Active Configurations

My analysis of active configurations in Romance languages will be the one currently assumed in minimalist theory enriched with Kratzer’s (1996) hypothesis that the Voice head is responsible for the introduction of the Initiator. The unvalued ϕ-features on Tense and Voice (notated ϕ:) enter AGREE relations with the DPInitiator and the DPTheme, respectively, thereby Case-checking those DPs and inheriting their phi-features. The arrows indicate feature-inheritance from the valued features of the DPs to the unvalued features of Tense and Voice. The direction of Case-checking, not represented here, goes the other way round: (22)

TenseP Tense {f:}

VoiceP DPInitiator

Voice’

Voice {f:}

VP V

4.2

DPTheme

Phi-Features and Case-Checking in Participle Passives

Building on Legate’s (2014) analysis of Acehnese passives, and following her suggestion to extend the analysis crosslinguistically (note that Legate herself does not pursue this suggestion), I will assume that crosslinguistically, passive Voice is born with valued ϕ-features. Combined with the theory of AGREE, this hypothesis derives as a consequence Burzio’s generalization (the correlation between absorption of the external role and suspension of the assignment of Accusative case): because the ϕ-features of the Initiator are already valued on Voice, they cannot enter an AGREE relation with DPTheme, which means that the Theme cannot be assigned Accusative Case. Quite clearly, the participle passives of Romance (or Germanic) languages do not carry ϕ-features relating to the Initiator. This can be captured by assuming that in certain languages, passive Voice is inherently endowed with the underspecified value ARB, which at LF is interpreted as introducing an implicit, arbitrarilyinterpreted (i.e., referentially unspecified) Initiator, hence the notation Voice {ϕ: ARB} used below:

Implicit Agents and the Person Constraint on SE-Passives

(23)

125

TenseP Tense{f:}

BeP Be

VoiceP Spec,Voice DPTheme

Voice’ Voice {f:ARB} V

VP DPTheme

Voice {ϕ:ARB} behaves on a par with the Voice head of Acehnese insofar as it supplies an implicit Initiator, making it impossible for the Initiator to Merge in Spec, Voice and at the same time blocking Accusative Case assignment: since the features of Voice are already valued, no AGREE relation can be established between Voice and DPTheme and correlatively DPTheme cannot be assigned Accusative Case. This forces DPTheme to internally Merge into Spec,Voice (in GB terms, this amounts to saying that DPTheme moves from the VP-internal position to Spec,Voice; this movement is signaled in (23) by the striking out of the lower copy of DPTheme). From this position, DPTheme enters an AGREE relation with Tense, thereby valuing the unvalued feature of Tense and getting Case-checked at the same time. In sum, the proposal made here is that active and passive Voice in Romance languages is to be analyzed as Voice {ϕ:} and Voice {ϕ:ARB}, respectively.

4.3

Phi-Features and Case-Checking in Anticausatives

In order to propose an analysis for SE-passives, I first need to sketch an analysis for the anti-causative meaning of SE-verbs. Following Schäfer (2008), Alexiadou et al. (2015) and Wood (2015), a.o., I will assume that anticausatives involve an ‘expletive’ Voice, i.e., a functional head that denotes the identity function, which means that it does not introduce an Initiator and it does not assign Accusative Case. I will implement this idea in a slightly different way, by assuming that the anticausative Voice is devoid of features, which I will notate as Voice {Ø}. Unlike the authors just cited, I will not assume that anticausative markers, e.g., sich in German or SE in Romance occur in Spec,Voice. I will instead assume sich to be an ‘expletive’ object, comparable to expletive subjects insofar as it is a mere place-holder:

126

C. Dobrovie-Sorin

(24)

TenseP Tense{f:}

VoiceP

Spec,Voice Voice’ DPTheme Voice {Ø}

VP

sich

V

Being devoid of unvalued features, Voice {Ø} is unable to enter an AGREE relation with sich. Nothing goes wrong, because sich has no ϕ-features but does have inherent Case features (as expletives in general), hence it does not need to be Casechecked via an AGREE relation with Voice. Because sich does not have phi-features, it does not saturate V, which means that VP has the same denotation as V, i.e., a function that looks for a Theme argument; and as usual, the denotation of VP is transferred unchanged up to VoiceP. Since Voice {Ø} has no unvalued features, it does not introduce an Initiator into Spec,Voice and since the internal argument (DPTheme) has not yet been fed into the derivation, it will be externally merged in Spec,Voice. This position is accessible to AGREE with Tense{ϕ:}, allowing DPTheme to be Case-checked. SE crucially differs from sich in being a clitic rather than a pronoun. In line with Kayne (2000) and subsequent literature, I assume that SE is base-generated (externally merged) in Tense,10 which allows DPTheme to merge VP-internally. (25)

TenseP SE Tense {f:} VoiceP Voice {Ø} V

VP DPTheme

Since Voice {Ø} is unable to introduce the Initiator, and since DPTheme has already been merged VP-internally, the Spec position of Voice is not projected in those configurations in which DPTheme remains in its external Merge position. Note however

10

Kayne (2000) assumes this hypothesis for all object clitics in Romance. This view is most plausible for SE, the pronominal analysis of which is debatable (see Burzio 1986, who treats it as an ‘affix’): see in particular the fact (established by clear tests) that the postverbal subjects of SE-verbs sit in the direct object position. Compare sich-verbs in German, which do not allow their postverbal subjects to sit in the direct object position.

Implicit Agents and the Person Constraint on SE-Passives

127

that SE-anticausatives do allow preverbal subjects, in which case, because of locality constraints, DPTheme will transit through Spec, Voice on its way to Spec,Tense: (26)

TenseP Spec,Tense Tense’ DPTheme Tense{f:} VoiceP Spec,Voice Voice’ DPTheme Voice {Ø}

VP

V

4.4

DPTheme

Phi-Features and Case-Checking in SE-Passives

Let us now go back to SE-passives, which by definition differ from anticausatives by their agentive interpretation. The question is what is the piece of a SE-configuration that is responsible for agentivity. The first option, which is most frequently— probably unanimously—assumed is that it is the Voice head that introduces the Initiator. Under this view, two distinct Voice heads are assumed for SE-verbs, an anticausative and a passive one. Modulo different labels, such a homonymical analysis can be found in Schäfer (2008), Alexiadou et al. (2015) or Wood (2015). Given this type of analysis, SE-passives and participle passives are syntactically indistinguishable, which bars the way towards explaining the differences between these two types of passives. My proposal is represented in the tree below. The innovation is to assume that implicit Initiators can be introduced not only by Voice {ϕ:ARB}, as in participle passives, but also by Tense {ϕ:ARB}, the Voice head itself being the same as that of anticausatives, i.e., devoid of any features: (27)

TenseP SE-Tense VoiceP {f:ARB} Voice{Ø} V

VP DPTheme

128

C. Dobrovie-Sorin

In line with the rich literature on the PCC, I will assume that the PC on SE-passives signals a limitation on the AGREE relation between the legitimating head (in our case Tense) and the DPTheme that needs to be Case-checked. The difference between my proposal and most accounts of the PCC11 is the reason that blocks AGREE in Person. Since I assume that the Initiator is not projected in Spec, Voice, the blocking of AGREE in Person cannot be due to the Initiator intervening (sitting on the path) between Tense and DPTheme. I will instead propose that the ϕ-valuation of Tense by DPTheme is limited to Number because the Person feature of Tense is already valued as ARB. The derivation crashes whenever DPTheme is marked with Person. Note that according to this analysis the success of a derivation does not exclusively depend on the Probe’s features being valued, but also on the Goal’s features being Case-checked via an AGREE relation with the Probe’s unvalued features. We can now understand why participle passives are not subject to the PC. In participle passives (see (23)), the implicit Initiator is introduced by the Voice head (due to Voice’s features being valued as ARB); because Voice is already partially valued, it need not participate in any AGREE relation. Because Tense has unvalued features for both Person and Number, it can enter an AGREE relation with a DPTheme that carries not only Number but also Person features.

4.5

Summary

In sum, I have proposed an account of the PC on Romance SE-passives without assuming that the Initiator is syntactically projected in passive configurations (be they participial passives or SE-passives). I have also explained why participial passives (in Romance as well as in Germanic languages) are not subject to the PC. The core assumption was an extended and refined version of Legate’s hypothesis that the implicit Initiators of passives are introduced due to Voice being inherently valued for features. In order to be able to extend this assumption crosslinguistically, I had to weaken it in two ways: (1) the Voice head of passives need not be fully valued, it can be only partially valued as ARB; (2) the implicit Initiator can be supplied not only by a valued Voice, but also by a valued Tense. According to my proposal, those Voice morphemes that uniquely encode passive meanings are to be analyzed as functional heads that are inherently valued as ARB, whereas syncretic Voice morphemes such as SE in Romance or sich in German are to be analyzed as ‘expletive’ Voice (Voice{Ø} in my implementation), their passive meaning being due to an implicit Initiator being contributed by an inherently valued Tense, Tense {ϕ:ΑRΒ}.

11

The only exception I know of is Adger and Harbour (2007).

Implicit Agents and the Person Constraint on SE-Passives

129

5 Agentivity Tests in Passives12 In this section I will briefly review some empirical observations regarding the interpretation of the Initiator in SE-passives and participle passives. The intention is to suggest ways of accounting for the data within the hypothesis adopted here, according to which the Initiator is syntactically projected neither in participle passives nor in SE-passives, but is instead introduced by Voice and Tense whenever these functional heads are born with phi-features that are valued as ARB. Some of the tests usually employed for the syntactic ‘activity’ of the Initiator give similar results for both types of passives: (28) a. Aceste haine se vând / sunt vândute pentru a ajuta săracii (Ro.) these clothes SE sell.3PL / are sold for to help poor-the b. Acest lucru s-a făcut / a fost făcut cu bună ştiinţă (Ro.) this thing SE-has done has been done deliberately The example in (a) shows that the Initiator of passives (both SE- and participle passives) is able to control into a purpose clause. This fact does not, however, constitute evidence in favor of the hypothesis that the Initiator is syntactically projected. Indeed, according to Williams (1985), purpose clauses may express some teleological relation between events, e.g., Grass is green to promote photosynthesis or The boat sank in order to impress the queen and move her to murder her husband by the end of Act III (for further discussion see Bhatt & Pancheva 2017). It is unlikely that in these examples, built with anticausatives, there is any external argument projected in the syntax. The example in (b) shows that agentive adverbs such as cu bună ştiinţă ‘deliberately’ are also allowed in both types of passives. For this test it seems more difficult to assume that no Initiator is needed for the semantic interpretation. But arguably, at least in certain cases, the controller can be a valued phi-feature, not necessarily a DP in a syntactic position. The parallelism between the two passives is preserved in examples built with reciprocals, but importantly, in this case both types of passives show degraded acceptability judgments: (29) a. % Aici nu se vorbeşte unul cu altul (Ro.) (66%)13 here not SE speaks one with another ‘Here people don’t speak with each other’ b. % Lucrările au fost copiate unii de la alţii (66%) papers-the.FPL have been copied ones.MPL from others.MPL ‘The papers have been copied (? (by the students) from each other)’ 12

Most of the generalizations and examples in this section are borrowed from Giurgea (2015). The suggested explanations are mine. 13 The indicated scores reflect judgments of ten informants, on a scale from 0 to 3 (? ¼ 2,?? ¼1; for those speakers who did not use??, the judgment? was counted as 1.5).

130

C. Dobrovie-Sorin

In order to explain the low acceptability of these examples we need to assume that (1) reciprocals require control by a syntactically projected DPInitiator and (2) the DPInitiator is projected neither in participle passives nor in SE-passives. By this analysis, the examples are ‘ungrammatical’ (i.e., not generated by the computational system) but nevertheless acceptable due to their being interpretable. Experimental work would be needed in order to propose more than what I suggested above. There is, however, an important empirical generalization that is independent of more fine-grained experimental results: the contrast between the full acceptability of (28) and the degraded acceptability of (29) strongly suggests that only in the latter cases is the syntactic projection of the Initiator required. And again, both passives behave alike, which suggests that the Initiator is projected in neither of them. Let us now turn to a number of differences between the two passives. Giurgea (2015) observes that the subject-oriented anaphor sine ‘self’ is more easily controlled by the Initiator of SE-passives: (30) a. Nu aşa se vorbeşte/scrie despre sine (Ro.) (100%) not so SE speaks/writes about self ‘One does (should) not speak/write like this about oneself’ b. %?? Cartea a fost de fapt scrisă despre sine (60%) book-the has been actually written about self ‘The book has actually been written about oneself’ c. %? Nu se acordă premii sie însuşi / sieşi (80%) not SE awards prizes self.DAT EMPH 3REFL.DAT ‘One does not award prizes to oneself’ d. ?? /* Nu sunt acordate premii sie însuşi/ sieşi (25%) not are awarded prizes self.DAT EMPH/3REFL.DAT ‘Prizes are not awarded to oneself’ Similarly for (adjectival) secondary predicates: (31) a. ?? Scrisoarea pare a fi fost scrisă beat (Ro.) letter-the seems to have been written drunk.MSG ‘The letter seems to have been written drunk’ b. Scrisoarea asta a fost scrisă beat letter-the this has been written drunk c. %? Nu se conduce beat not SE drives drunk ‘One does not drive drunk’ d. %?? Aşa ceva nu se scrie beat such something not SE writes drunk.MSG ‘Something like that should not be written drunk’ e. %? Până la plajă se poate merge gol to beach SE can walk naked

(18%)

(18%) (83%)

(56%)

(80%)

Implicit Agents and the Person Constraint on SE-Passives

131

‘One can walk to the beach naked’ Note that the examples with participial passives are eventive whereas those with SE have a generic interpretation. The eventive counterparts of the examples with SE are unacceptable, which indicates that the contrasts observed above are not between SE-passives and participial passives, but rather between generic and eventive passives. A well-known peculiarity of the interpretation of SE-passives is that the Initiator is restricted to animates (Burzio 1994; Cornilescu 1998)14: (32) Oraşul {a fost distrus/* s-a distrus} de (către) cutremur (Ro.) city-the has been destroyed SE-has destroyed by earthquake ‘The city was destroyed by the earthquake’ My proposal sheds some light on this generalization: in SE-passives, the only feature that restricts the Initiator is Person:ARB, the Number features being involved in the AGREE relation with DPTheme. A tentative suggestion would be that an Initiator that is exclusively identified by Person:ARB necessarily refers to humans; an Initiator that is identified by Number (in addition to Person:ARB, or maybe without that feature) would be able to refer to non-humans, hence to Cause Initiators. Since in participle passives, Voice{ϕ:ΑRΒ} does not enter an AGREE relation with DP, the whole set of features (i.e., Number, in addition to Person) are available for the identification of the Initiator, which therefore is not necessarily interpreted as human. MacDonald (2017) observed that only SE-passives allow a definite DPTheme to be interpreted as a body-part of the DPInitiator15:

14

This generalization, which also holds in Italian and Spanish, is not invalidated by examples of the type in (1), uttered in a context in which an owner talks to a pet dog (example due to Kańsky 1992). Crucially, the dog is a discourse participant (the addressee) and as such it is assimilated to a human: (1)

No se ladra a los invitados. No se barks to the guests. “One does not bark at guests.”

MacDonald’s data are confronted with two problems. The first one is maybe marginal, being raised by an ill-chosen example, the version of (33)a that contains an indefinite DP, unas manos. This example is problematic, since unas manos cannot occur with a body-part meaning in active configurations: 15

(1)

Los profesores levantaron unas manos. The professors raised some hands.

The second problem is that the German counterpart of (33)a is acceptable in German (Florian Schäfer, p.c. January 2016), but German does not have SE-passives.

132

C. Dobrovie-Sorin

(33) a.

El profesor hizo una pregunta. Se levantaron unas/las manos. the professor made a question. SE raised.3PL some/the ‘The professor asked a question. Some of their/their hands raised.’ b. La cabeza fue levantada (por Juan) 6¼ Juan raised his head the head was raised by Juan c. Profesorul a pus o întrebare. professor-the has put a question S-au ridicat/#Au fost ridicate nişte mâini SE-have raised/have been raised some hands (The continuation with a part. passive does not have the sensible interpretation “some people raised their hands”)

Assuming that the body-part DP must be locally c-commanded by the Initiator, MacDonald (2017) concludes that the Initiator of SE-passives is projected in the syntax, as opposed to the Initiator of copular passives. This type of data will be left open for investigation. It seems to me that the problem is not so much control, but rather the fact that in order for the body-part interpretation to be possible, the DP must stay in the VP-internal position, a constraint that can be satisfied in SE-passives. Recall that according to Raposo and Uriagereka 1996, the DP of SE-passives sits either in the object position or in a leftperipheral position, but never in Spec,Tense. In order to explain the unacceptability of the examples with participle passives, we may argue that this type of passive forces DPTheme to be internally merged in Spec,Voice, which violates the constraint on body-parts.

6 Conclusions and Consequences The account of the PC on SE-passives proposed here changes the perspective on the analysis of this phenomenon. Whereas previous work assumed the separation of Number and Person (Anagnostopoulou 2003), I have proposed that these are not independent functional heads but rather features that occur on Tense and Voice. Crucial for my account is not the separation of Number and Person, but rather the opposite: the PC arises when a single head, e.g., Tense, has valued Person features, so that it is only Number that can enter AGREE with a Goal. According to my proposal, intervention is irrelevant for the PC on SE-passives. What matters is the (non-)availability of Person-features for AGREE. It would be interesting to see whether the type of proposal made here for the PC on SE-passives could be of some relevance for the analysis of the PCC phenomena (clitic clusters, clusters of AGR morphemes, Basque applicative unaccusatives, etc.). Acknowledgments This article is a revised version of a presentation at Workshop on Romance SE/SI, University of Wisconsin-Madison, April 21–22 2016. I would like to thank Artemis Alexiadou, Ion Giurgea and Omer Preminger for comments on previous versions of this paper.

Implicit Agents and the Person Constraint on SE-Passives

133

References Adger, David, and Daniel Harbour. 2007. Syntax and syncretisms of the person case constraint. Syntax 10: 2–37. Albizu, Pablo. 1997a. Generalized person-case constraint: A case for a syntax-driven inflectional morphology. In Theoretical issues on the morphology-syntax interface, ed. Myriam UribeEtxebarria and Amaya Mendikoetxea, 1–33. Donostia: UPV/EHU. ———. 1997b. The syntax of person agreement. Los Angeles, CA: Ms., University of Southern California. Alexiadou, Artemis. 2012. Non-canonical passives revisited: Parameters of non-active voice. Linguistics 50: 1079–1110. Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2006. The properties of anticausatives crosslinguistically. In Phases of interpretation, ed. Mara Frascarelli, 187–211. Berlin: Mouton de Gruyter. ———. 2015. External arguments in transitive alternations. Oxford: Oxford University Press. Anagnostopoulou, Elena. 2003. The syntax of ditransitives: Evidence from clitics. The Hague: Mouton de Gruyter. ———. 2005. Strong and weak person restrictions. In Clitic and affix combinations, ed. Lorie Heggie and Francisco Ordóñez, 199–235. Amsterdam: John Benjamins. Baker, Mark. 2008. The syntax of agreement and concord. Cambridge: Cambridge University Press. Baker, Mark, Kyle Johnson, and Ian Roberts. 1989. Passive arguments raised. Linguistic Inquiry 20: 219–251. Béjar, Susana, and Milan Rezac. 2003. Person licensing and the derivation of PCC effects. In Romance linguistics: Theory and acquisition, ed. Anna-Teresa Pérez-Leroux and Yves Roberge, 49–62. Amsterdam: John Benjamins. ———. 2009. Cyclic agree. Linguistic Inquiry 40: 35–73. Bhatt, Rajesh, and Roumyana Pancheva. 2017. Implicit arguments. The Wiley Blackwell Companion to Syntax, Second Edition 1–35 Blanche-Benveniste, Claire. 1978. A propos des traits sémantiques utilisés en syntaxe: Critique du trait ‘+/humain’. Cahier de linguistique 8: 1–15. Bonet, Eulàlia. 1991. Morphology after syntax: Pronominal clitics in romance. PhD diss., MIT. ———. 1994. The person-case constraint: A morphological approach. In MIT workinng papers in linguistics 22: The morphology-syntax connection, ed. Heidi Harley and Collin Phillips, 33–52. Cambridge, MA: MITWPL. ———. 1995. Feature structure of romance clitics. Natural Language and Linguistic Theory 13: 607–647. Bruening, Benjamin. 2012. By phrases in passives and nominals. Syntax 16: 1–41. Burston, Jack L. 1983. Clitic object sequences and cooccurrence restrictions in French. Linguistic Analysis 11: 247–275. Burzio, L. 1986. Italian syntax. A government-binding approach. Dordrecht: Reidel. Burzio, Luigi. 1994. Weak anaphora. In Paths toward universal grammar: Studies in honor of Richard Kayne, ed. Guglielmo Cinque, Ian Koster, Jean-Yves Pollock, et al., 59–84. Washington, D.C.: Georgetown University Press. Cabredo Hofherr, Patricia, and Carmen Dobrovie-Sorin. 2010. Implicit agentivity without agents in the syntax. A crosslinguistic analysis of se-verbs. In Edges, heads and projections: Interface properties, ed. Anna Maria Di Sciullo and Virginia Hill, 47–66. Amsterdam: Benjamins. Cinque, G. 1977. The movement nature of left dislocation. Linguistic Inquiry 8: 397–412. Cinque, Guglielmo. 1988. On Si constructions and the theory of ARB. Linguistic Inquiry 19: 521–581. Collins, Christopher, and Paul Postal. 2008. Imposters. New York: Ms., New York University. Cornilescu, Alexandra. 1998. Remarks on the syntax and the interpretation of Romanian middle passive se sentences. Revue roumaine de linguistique 43: 317–342.

134

C. Dobrovie-Sorin

D’Alessandro, Roberta. 2007. Impersonal “si” constructions: Agreement and interpretation. Berlin: Mouton de Gruyter. Déchaine, Rose-Marie, and Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33: 409–442. Dobrovie-Sorin, Carmen. 1985. Redoublement clitique, relatives et interrogatives en roumain et espagnol. Linguisticae Investigationes 9: 269–307. ———. 1986. A propos du contraste entre le passif morphologique et se moyen dans les tours impersonnels. Linguisticae Investigationes X: 289–330. ———. 1990. Clitic doubling, Wh-movement and quantification in Romanian. Linguistic Inquiry 21: 351–397. ———. 1994. The syntax of Romanian. In Comparative studies in romance. Berlin: Foris/Moutonde Gruyter. ———. 1998. Impersonal SE in romance and the passivization of unergatives. Linguistic Inquiry 29: 399–439. Embick, David. 2004. Unaccusative syntax and verbal alternations. In The unaccusativity puzzle: Explorations of the syntax–lexicon interface, ed. Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, 137–158. Oxford: Oxford University Press. Fábregas, Antonio. 2013. Differential object marking in Spanish: State of the art. Borealis–An International Journal of Hispanic Linguistics 2 (2): 1–80. Giurgea, Ion. 2015. On the person constraint on romance se-passives. Talk given at the Workshop on Non-local dependencies in the nominal and verbal domain. Lisbon, CLUNL, November 2015. Harley, Heidi. 2013. External arguments and the mirror principle: On the distinctness of voice and v. Lingua 125: 34–57. ———. 2000. Parameters and universals. Oxford: Oxford University Press. Kański, Z. 1992. Impersonal constructions as a strategy for second-order predication. In Meaning and grammar: Cross-linguistic perspectives, eds. M. Kefer & J. van der Auwera, 95–122. Berlin: Mouton de Gruyter Keenan, Edward L. 1980. Passive is phrasal (not sentential or lexical). In Lexical grammar, ed. Teun Hoekstra, Harry van der Hulst, and Michael Moortgat, 181–214. Dordrecht: Foris. ———. 1985. Passive in the world’s languages. In Language typology and syntactic description, volume 1: Clause structure, ed. Timothy Shopen, 243–281. Cambridge: Cambridge University Press. Kiparsky, Paul. 2013. Towards a null theory of the passive. Lingua 125: 7–33. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase structure and the lexicon, ed. Johan Rooryck and Laurie Zaring, 109–137. Dordrecht: Kluwer. Laenzlinger, Christopher. 1993. A syntactic view of romance pronominal sequences. Probus 5: 241–270. Legate, Julie Anne. 2012. Subjects in Acehnese and the nature of the passive. Language 88: 495–525. ———. 2014. Voice and v. lessons from acehnese. Cambridge, MA: MIT Press. ———. 2017. Personal passives, grammatical object passives, and impersonals. Talk given at the syntax & morphology workshop. University of Chicago. MacDonald, Jonathan E. 2017. An implicit projected argument in Spanish impersonal and passive se constructions. Syntax 40: 353–383. Mendikoetxea, Amaya. 2008. Clitic impersonal constructions in romance. Transactions of the Philological Society 106: 290–336. Ormazabal, Javier, and Juan Romero. 2007. Object agreement restrictions. Natural Language and Linguistic Theory 25: 315–347. Perlmutter, David. 1971. Deep and surface constraints in syntax. New York, NY: Rinehart & Winston Inc. Postal, Paul M. 1989. Masked inversion in French. Chicago, IL: University of Chicago Press. Preminger, Omer. 2014. Agreement and its failures. Cambridge, MA: MIT Press.

Implicit Agents and the Person Constraint on SE-Passives

135

Pylkkänen, Liina. 1999. The syntax of internal and external causation. In Proceedings of the Texas linguistics society 1999: Conference on perspectives on argument structure. Austin: University of Texas Department of Linguistics. Raposo, Eduardo, and Juan Uriagereka. 1996. Indefinite SE. Natural Language and Linguistic Theory 14: 749–810. Rezac, Milan. 2006. Agreement displacement in Basque. Ms., University of the Basque Country. www.umr7023.cnrs.fr/spip.php?article675 (accessed on September 9, 2010). ———. 2008. The syntax of eccentric agreement: The person case constraint and absolutive displacement in Basque. Natural Language and Linguistic Theory 26: 61–106. ———. 2011. Phi-features and the modular architecture of language. Dordrecht: Springer. Sandfeld, Kristian. 1928. Syntaxe du français contemporain I: Les pronoms. Paris: Librairie Honoré Champion. Schäfer, Florian. 2008. The syntax of (anti-)causatives. External arguments in change-of state contexts. Amsterdam: John Benjamins. ———. 2015. Medio-passives within a formal typology of voice. Paris: Ms., ENS. Williams, Edwin. 1985. PRO and subject of NP. Natural Language and Linguistic Theory 3: 297–315. Wood, Jim. 2015. Icelandic morphosyntax and argument structure. Dordrecht: Springer. Zribi-Hertz, Anne. 1982. The ‘middle-Se’ construction in French and its status in the triad middle voice–passive–reflexive. Lingvisticae Investigationes 6: 345–401. ———. 2008. Le médiopassif à accord riche en français: pour une approche multifactorielle. Actes (en ligne) du 1er CMLF. http://www.linguistiquefrancaise.org/articles/cmlf/pdf/2008/01/. cmlf08083.pdf.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish Francisco Ordóñez

Abstract I argue that impersonal se is bound by empty pronouns in Catalan, Spanish and Italian. The impersonal pronouns always receive nominative case (contra Dobrovie-Sorin C. Linguistic Inquiry 29: 399–347, 1998) since they trigger Differential Object Marking (DOM), allow bare DP objects, and are incompatible with dative experiencer psychological verbs. In the second part of the paper I examine the interpretative properties of the empty pronominals. On the one hand, these constructions allow generic and existential interpretations, and I attribute them to an empty pronoun ϕ as in Holmberg (Holmberg A. The null generic subject pronoun in Finnish: A case of incorporation in T. Parametric variation: Null subjects in minimalist theory. Biberauer T, Holmberg A, Roberts I, Sheehan M. Cambridge: Cambridge University Press. 200–230, 2010) and Mendikoetxea (Mendikoetxea A. Transactions of the Philological Society 106: 290–336, 2008). However, I argue that Italian has an additional interpretation of impersonal se as first person plural WE (D’Alessandro R. Impersonal SI constructions. Berlin: Mouton de Gruyter, 2007). Thus, I assume that Italian has a WE/ϕ pronoun, but Catalan and Spanish only have a ϕ pronoun. This double nature of impersonal se in Italian is parallel to what is found in other languages such as French on. Finally, I show that certain unexamined differences between se in Catalan and Spanish on the one hand, versus Italian, on the other, can be explained under this new approach. Keywords Impersonal se · ϕ pronoun · Nominative Case · Binding · DOM · First person plural WE

F. Ordóñez (*) Stony Brook University, Stony Brook, NY, USA e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_6

137

138

F. Ordóñez

1 Introduction Impersonal se constructions in Catalan, Italian and Spanish must contain an unstressed clitic es, se, and si respectively (referred to hereafter as se) and no overt full DP counterpart can have such an interpretation (2): (1) Se habla de Juan.1 se speak-3sg about Juan ‘one-someone speaks about Juan’ (2) *Sí mismo habla de Juan se self speaks about Juan’ ‘Oneself speaks about Juan’ Also, no overt quantified impersonal subject is possible: (3) *Uno se habla de Juan One se speaks about Juan ‘One speaks about Juan’ The empty subject is always interpreted as bearing a theta role with human interpretation as in (4).2 Se is impossible in contexts in which no theta role is available as for instance in the syntactic structures that involve expletive empty subjects as in (5) and (6)3: (4) Cuando se es importante When se is important ‘When oneself is important’. (5) *Cuando se es importante estudiar este tema When se is important to study this topic ‘When it is important to study a topic’

1

In this paper Spanish examples such as this one are not marked specifically for languages, whereas Italian ones will be indicated with (It), and Catalan ones with (Cat). 2 As pointed out by Sánchez López (2002: 57), it is possible that the arbitrary interpretation could refer to an animal as in (i). This seems to indicate that the relevant feature is animacy. However, these cases have a very restricted distribution, more like an arbitrary second person directed to an animal. In other contexts, the interpretation has to be human. As pointed out by a reviewer, this interpretation only occurs with higher animals. It is impossible to get it with an insect: (i) No se ladra a los invitados. (From Sánchez López 2002) Not se bark to the guests ‘One does not bark to the guests.’ (addressing a dog) 3 Pujalte and Saab (2012) and Saab (2014) propose a view in which se saves the derivation of an unsaturated external argument position. Their view of se is more of a PF repair strategy.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

139

(6) *Se hay mucho dinero en este pueblo se have-3ps a lot of money in this village ‘One there is a lot of money in this village’. Impersonal se can be merged with transitive, unaccusative, unergative, and passive predicates as shown in the following examples: (7) Se ha estudiado mucho a Cervantes en esta universidad. (Transitive) se has studied a lot Cervantes in this university ‘One has studied too much Cervantes in this university’ (8) Se trabaja en las vacaciones. (Unergative) se works during vacation ‘One works during the vacation’ (9) Aquí se llega tarde al concierto. (unaccusative) here se arrives late to the concert ‘Here one arrives late to the concert’ (10) Se es castigado por los jueces. (passive) se is punished by the judges ‘One is punished by the judges’ (11) Cuando se es rico. (copula) when se is rich ‘When one is rich’

2 On the Single Nominative Impersonal Se I will maintain with Kayne (2000) and Sportiche (2014) that there is a unique monomorphemic se clitic in Romance. There are no differentiated reflexive or reciprocal ses. The different interpretations of se are due to the different syntactic configurations in which the single clitic se appears. In the case at hand, se appears with an empty impersonal pronoun that yields the impersonal interpretation. In this respect the term nominative se or accusative se is not adequate, since nominative could be referring to reflexive se, middle se etc. Case is not the unique defining property of this construction (contra Dobrovie-Sorin 1998). We assume that se is just a unique morpheme that needs to be valued. In terms of the recent theory of feature valuation in the minimalist program, se has an unvalued person feature. The impersonal empty pronoun is responsible for binding the clitic se4:

4

MacDonald and Maddox (2018) propose that this is a pro with a D feature in Spanish contrary to Romanian. Their proposal can be adopted here as well. For Italian as we will see below it can also have a 1person plural feature Sect. 4.2.

140

F. Ordóñez

(12) ϕ se [i person] Se is not an l- clitic but belongs to the same class of clitics as 1p, 2p me, te, which are characterized by not having plural or gender specifications. It is impossible to have plural se-s as it is impossible to have plural me-s or te-s. Another important feature of impersonal se is that it does not have any gender marker as the l clitics might have. In previous seminal articles, Dobrovie-Sorin (1998) and Cinque (1988) propose that se is ambiguous in languages like Catalan, Italian and Spanish. For Cinque se is ambiguous between argumental and non-argumental se (Cinque 1988). DobrovieSorin updates this distinction and divides se into nominative versus accusative se (Dobrovie-Sorin 1998). Part of the reason for assuming this dichotomy is the restriction on the distribution of se in Romanian, which does not allow se in copula and passive constructions (Dobrovie-Sorin 1998). According to Dobrovie-Sorin, Romanian does not have a nominative se.5 By contrast, I will assume a single non-ambiguous nominative impersonal se for Spanish. This se is bound by an empty pronoun with an impersonal interpretation. The most important reason for such a claim is the distribution of impersonal se and DOM in Spanish. As is well known, Spanish is a DOM language because certain specific, animate objects must be preceded by the preposition a. The appearance of a has been linked to the assignment of accusative case to the object. Under an ambiguous view of se one might expect that DOM should be optional according to whether accusative case is or not assigned to the clitic se. Nominative se would be compatible with accusative DOM, whereas accusative se would ban DOM. However, as pointed out by Ordóñez and Treviño (2016) and Sánchez López (2002), there is no such optionality: DOM must be deployed for impersonal se as shown in the following contrasts (13a,b) below6: (13) a. Se castigó a los niños Se punish-3psg a the boys b. *Se castigó los niños Se punish-3psg the boys ‘The boys were punished’ Another interesting variable to examine is number agreement with animate objects. Number agreement with objects has been linked to nominative case. If the ambiguity theory is correct, one might expect that verbal agreement with animate objects without DOM could yield a grammatical derivation. However, example (14) shows that agreement with animate objects is not grammatical. In conclusion, 5

We think that case is not the aspect that explains the defining property of se in Romanian. What is important in Romanian is argument structure. Romanian only allows se in construction with small v. That includes verbs which do not assign accusative case, but do allow se in Romanian. 6 As pointed out by Sánchez López (2002), this requirement is relaxed for animate plurals with kind denotation (Sánchez López, page 57). I will leave this important point aside.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

141

the fact that DOM must be deployed and that agreement with the verb does not rescue the derivation with animate objects as in (14), indicates that ambiguity of case (Dobrovie-Sorin 1998) over-generates in Spanish.7 The only possible answer is that accusative case is always deployed because se is associated with an empty category with nominative case: (14) *Se castigaron los niños8 Se punish-3ppl the boys ‘The boys were punished’. This brings the point of discussion to non-DOM objects. We observe that in general the agreement and non-agreement possibilities are still available. (15) Se vieron los libros se saw-3ppl the books (16) Se vio los libros se saw-3ps the books ‘The books were seen’ The fact that non-DOM objects allow agreement could be problematic for the idea that se is associated with an empty category with nominative case. This reasoning is correct if number agreement with the object is treated as nominative. However, in a recent paper, Romero and Ormazabal (2019) have argued convincingly that this agreement in number with the object is not nominative. First, it immediately explains why the object cannot be a nominative pronoun. This is independent of whether the verb agrees or not with the object: (17) * se vimos nosotros se saw-1ppl we ‘One saw us’ (18) * se vio nosotros se saw-3ps we ‘One saw us’ 7

As pointed out by Ordóñez and Treviño (2016, a.o.) some dialects do allow such agreement. An important point of these dialects is that this agreement with DOM can only take place postverbally, which indicates that it is not nominative case in the normal sense: (i) a. se castigaron a los niños se punish-3ppl a the boys b. *a los niños se castigaron a the boys se punish-3ppl ‘They boys were punished’ 8 This sentence is grammatical with a reflexive interpretation, which is not the focus in here. (i) Se vieron los niños en el espejo se saw-3pl the boys in the mirror ‘the boys saw themselves in the mirror’

142

F. Ordóñez

All the examples are perfectly grammatical with the corresponding accusative clitic9: (19) Se nos ve Se 1ppl-acc see-3ps ‘We were seen’ Therefore, there is no distinction between agreeing and not agreeing se in terms of case. Nominative case is assigned to the empty category associated with it, independently of whether the object agrees or not with the verb. I follow Ordóñez and Treviño (2016), Rigau (1997) and Romero and Ormazabal (2019) in claiming that this is another example of number agreement with objects. I assume that in example (15) there is a split between person and number agreement for third person. The agreement on –n for verbs in Spanish indicates agreement in number with the object.10 Concretely, I assume that tense is responsible for number agreement with the object. Thus, I adopt the following analysis in which se is valued by the ϕ pronoun in person, the object can receive number agreement from tense and nominative case is assigned to ϕ. In this construction there is a division of labor between person agreement with se and number agreement with the object11: (20) ϕNOM se[i person] T [u-number,u-person, u-nom] [verb object [i-number]]

3 Evidence for an Empty Pronoun in Impersonal se: Case, Control and Adverbs In terms of case, the fact that DOM is required indicates that the ϕ pronoun with impersonal se must be nominative. This is clear when we contrast this impersonal se construction with periphrastic passives. As we see in the following examples, DOM

9 As pointed out by various linguists (e.g., Fernández Ordóñez 1999; Ordóñez and Treviño 2016), the use of clitic third person can change to le, the dative in the masculine in many dialects. I will not deal here with this interesting phenomenon, but see MacDonald & Melgares (this volume) for an account. 10 If objects are licensed by number in T, it would not be surprising that in some dialects DOM objects might permit agreement, which is exactly what is seen in Mexican Spanish (Ordóñez and Treviño 2016). 11 Romero and Ormazabal (2019) indicate that Φ is quirky case. The crucial point for us is that they show that there is no ambiguity of case for se (contra Dobrovie-Sorin 1998). From the nominative perspective advocated here, it remains to be understood why agreement with the object in number can occur. This might be linked to the fact that the category associated with se must be empty and it is unable to check number.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

143

is not permitted with periphrastic passives, but it is obligatory with impersonal se as we saw in the previous section. (21) Se golpeó *(a) los niños (impersonal se) Se hit a the boys ‘the boys were hit’. (22) Fueron golpeados (*a) los niños (periphrastic passive) were-3pp hit-3pp a the boys ‘The boys were hit’ This ϕ with nominative case explains why no other overt DP can appear in these impersonal se constructions in the subject position. Particularly no other DP with arbitrary meaning such as uno ‘one’ can appear in such position: (23) * Uno se llega One se arrive-3psg ‘One arrives’. (24) *Uno se es golpeado One se is hit-3psg ‘One is hit’ (25) *Uno se fue golpeado One se was hit-3p.sg ‘One is hit This observation is extended to transitive verbs (26) *Uno se vio el libro One se saw-3ps the book ‘One saw the book’ Another argument in favor of ϕ being associated with nominative case comes from the incompatibility of se constructions with dative experiencer psych verbs. Belletti and Rizzi (1988) propose that there are various types of psych verbs that are classified by the case assigned to the experiencer theta role. With some psych verbs the experiencer receives nominative case like amar, ‘love’, aborrecer ‘abhor’, odiar ‘hate’. For those verbs impersonal se is permitted, and the ϕ pronoun corresponds to the nominative experiencer. (27) Aqui ϕ(Nom,experiencer) se ama a Juan Here se loves-3ps P Juan ‘Here Juan is loved’.

144

F. Ordóñez

(28) Aqui ϕ(Nom,experiencer) se odia a Juan Here se hates-3ps P Juan ‘Here Juan is hated’ However, dative-experiencer psych verbs like gustar ‘like’, fascinar ‘fascinate’, aburrir ‘bore’ contrast with the nominative-experiencer ones above. The dativeexperiencer cannot correspond to the ϕ pronoun in (29) and (30). Therefore, we conclude ϕ must carry nominative case12: (29) *ϕ(Dat,experiencer) se aburren las novelas de ciencia ficción ϕ se bore-3ppl science fiction novels ‘Science fiction novels bore one’ (30) *ϕ(Dat,experiencer) se gustan los aventureros ϕ se like-3ppl the adventurer.pl-Nom ‘One like the adventurers’ In this dative-experiencer psych verbs, the nominative case corresponds to the theme as shown below: (31) a.

b.

Le gusto yo 3p-dat like-1psg I-Nom ‘I fascinate him/her’ Le fascino yo 3p-dat fascinate-1psg I-Nom ‘I fascinate her/him’

Since the theme is nominative, the expectation is that impersonal se might be associated with it. Nevertheless, there is a very clear contrast between nominative experiencers versus nominative themes with psych verbs. Nominative themes with psych verbs are ungrammatical with impersonal se: (32) a.*? ϕ(Nom,theme) se te gusta a ti se 2p please-3ps dat-2p ‘One pleases you’ b.*? ϕ(Nom,theme) se te fascina a ti se 2p fascinate-3ps dat-2p ‘One fascinates you’

12

Observe that an arbitrary interpretation is available when the dative pronoun is not pronounced. We assume there is an empty null pro. (i) Los aventureros gustan prodat The adventurers please-3ppl pro ‘adventurers please everybody’

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

145

The examples clearly contrast with those in which se is a nominative experiencer: adora a ti (33) ϕ(Nom,experiencer) se te se dat-2p adores-3ps dat-2p ‘One adores you’ This ungrammaticality of se with nominative themes extends to other constructions that take dative experiencers as their highest argument. For instance, in (34) a lover is the nominative theme and the dative experiencer is to her. Se can be neither a dative experiencer as in (34b) nor the nominative theme (34c): (34) a. Aquí le falta un amante a ella Here 3p lacks a lover dat-3ps ‘She lacks a lover here’ b. *Aquí ϕ(Dative, experiencer) se falta un amante Here se lacks a lover ‘One lacks a lover’ c. *?Aquí ϕ(Nom, theme) se le falta a ella Here se 3p lacks 3ps ‘One is lacked by her’ Importantly, these facts show that nominative case is a necessary but not sufficient condition to license impersonal se. In order to explain the contrast above we will appeal to the idea of Belletti and Rizzi (1988), according to whom in all psych structures the experiencer is the highest projected theta role independently of its case. Thus, while nominative case is assigned to the theme in dative psych verbs, it is nevertheless not projected as the highest theta role. The thematic and case structures of nominative experiencer and dative experiencer psych verbs are presented (35) and (36). Both structures involve little v, and the thematic hierarchy is the same. The difference is that nominative is assigned to the experiencer in (35) and to the lower theme in (36).13 (35) [vP[ACC] θ role (experiencer) v [VP θ role (theme)]] (nom. exp psych verbs) (36) [vP[DAT] θ role (experiencer) v [VP θ role (theme)]] (dative exp psych verbs)

13

As pointed out by a reviewer, there is question as to how nominative case is assigned to the theme in dative experiencer constructions. This is more general question about how case is assigned in this construction. If we assume the same thematic hierarchy above this could probably be thought of in terms of long-distance nominative case assignment, but such an analysis would require a long tangent into the syntax of these psych verbs.

146

F. Ordóñez

The licensing conditions for impersonal se are the following14: (37) Licensing conditions for impersonal se: a. Se must be associated with a ϕ pronoun which receives nominative case. b. Se must be associated with the highest theta role projected in the thematic structure. From this perspective, examples (29, 30) violate condition (37a), and examples (32, 34b, 34c) violate condition b). However, an experiencer in a nominative experiencer construction, as in (27) and (28), complies with both conditions in (37). To summarize, in impersonal se constructions, ϕ must receive nominative case and nominative case must correspond to the highest theta role.15 In all constructions in which nominative case does not coincide with the highest theta role projected, impersonal se is barred. Since the ϕ pronoun acts as a nominative subject in impersonal se constructions, we expect the object to behave like a canonical object. One important test is the distribution of bare plural nouns. Bare plural nouns are permitted with objects but are mostly ungrammatical with nominative subjects of transitive verbs. Nominative case assignment is not compatible with bare plural nouns in preverbal or postverbal position with transitive verbs as in (38) and (39).16 (38) *(Los) detectives encontraron huellas del asesinato the detectives found traces of the murder ‘Detectives found traces of the murder’ (39) Encontraron huellas del asesinato *(los) detectives found traces of the murder the detectives ‘Detectives found traces of the murder’ As shown in the example (40), bare plural nouns are grammatical with impersonal se constructions, contrary to bare plural nouns in periphrastic passive in (41) (also Sánchez López 2002). Therefore, the difference is due to the fact that the object in periphrastic passive receives nominative case, but not in impersonal se constructions. Thus, the initial object of periphrastic passives is behaving like a nominative subject like in transitive verbs above in (38) and (39). However, the object in impersonal se does not receive nominative case (see discussion in 20). In conclusion, only

14

I leave for further research the explanation for why a nominative with a lower theta role is not sufficient to license se in the context of psych verbs. If this turns to be a general property of languages it would be interesting to derive from deeper properties of the language faculty. 15 This generalization should also be derived. At this point I can just stipulate as a hierarchical condition linked to reflexive pronouns which receive the impersonal interpretation. It has to be related to the null subject parameter and reflexives. (See Wolfsgruber this volume and MacDonald and Melgares this volume for a discussion of the relation between impersonal se constructions and the null subject parameter.) 16 I am leaving aside the context of newspaper headlines in many Latin American countries.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

147

impersonal se licenses bare plurals, because from our perspective they receive accusative case17: (40) Se encontró huellas del asesinato se found-3ps traces of the murder ‘It was found fingerprints of the murder’ (41) ?*Fue encontrado huellas del asesinato was found fingerprints of the murderer ‘Fingerprints of the murder were found’ Apart from its properties with respect to case, the ϕ pronoun can be shown to be in the syntax because it acts as a binder. As we mentioned in the introduction, introducing this ϕ in impersonal se constructions allows us to distinguish reflexive and middle se from impersonal se; what defines impersonal se is the fact that it is bound by ϕ. MacDonald (2017) and Mendikoetxea (2008) show that the empty pronoun is responsible for control of the infinitive in rationale clauses as below: (42) No ϕi se escriben novelas para PROi contar la vida No se write novels to explain life ‘Novels aren’t written to explain one’s life’(from MacDonald 2017) Another binding test that shows the existence of an empty pronominal is given by MacDonald (2017) for inalienable possessive constructions. As it has been observed, an inalienable possessive POSS must have an antecedent DP: (43) El estudiantei levantó la POSSi mano. (from MacDonald 2017) The student raised the hand ‘The student raised his/her hand’ (44) Sofíai tiene las POSSi piernas largas. (from MacDonald 2017) Sofia has the legs long ‘Sofia has long legs’ The possessor POSS must be c-commanded by the DP antecedent as shown in (45). Juan in (45) does not c-command the empty POSS and therefore cannot be the antecedent. There is also a locality requirement on the binder as shown in (46). The c-commanding subject Juan cannot be the antecedent because the closer antecedent María intervenes:

17

As we already mentioned, number agreement is available with these special objects. We already argued that this should not be taken as a sign of nominative case: (i) Se encontraron huellas del asesinato se found-3ppl traces of the murder ‘It was found fingerprints of the murder’

148

F. Ordóñez

(45) El hermanoi de Juanj cerró los POSSi/*j ojos The brotheri of Johnj closed the POSS i/*j eyes ‘Juanj’s brotheri closed his i/*j eyes’(MacDonald 2017) (46) Juani dijo que Maríaj había cerrado los POSS *i/j ojos Juani said that Mariaj had closed the POSS *i/j eyes ‘Juan said that Maria had closed *his/her eyes’ (MacDonald 2017) Inalienable possession constructions are a good test to find binding properties of ϕ in impersonal se constructions (MacDonald 2017). As shown below the binder in all cases is the impersonal pronoun ϕ: (47) Aquí ϕi se levanta la POSSi mano para hacer una pregunta en clase here ϕi se raises the hand to ask a question in class ‘Here you-arb raises your hand to ask a question’. (48) Cuando ϕi se pierde la POSSi vida de ese modo, es muy triste when se loses the life of that way, is very sad ‘When you lose your life in that way it’s very sad’ All the binding facts and the distributions confirm that the empty pronoun ϕ is active in the syntax.18 In the next section we turn to what the interpretation that the ϕ pronoun might be and the different distributions of se in the various Romance languages.

4 The Interpretation of Impersonal se in Catalan, Spanish and Italian Now that the basic syntactic premises for the empty pronoun binding se are established, we will proceed to examine the possible interpretations that this pronoun has. The puzzle is that Catalan and Spanish, on the one hand, and Italian, on the other, differ in their distributional possibilities. Se allows both generic and existential interpretations in all three languages, but Italian allows a first person plural interpretation, which I will call WE. Catalan and Spanish lack this interpretation. This lack of WE interpretation in Catalan and Spanish explains why impersonal se with copula, passive and unaccusatives verbs is more restricted. Furthermore, it also explains why impersonal se has become a substitute for the first person pronoun ‘noi’ we in Italian—unlike Catalan and Spanish. Below are the possible interpretations available for se in the different Romance languages.

18

One reviewer suggests that the fact could equally obtained if we assume that se is nominative. Thus, the reviewer is suggesting more of a featural alternative in impersonal se than the one assumed here. The main reason not to pursue such an approach is to keep the clitic se simple. We attribute the different uses of se to other syntactic elements active in the syntactic configurations and derivations. Moreover, nominative case does not distinguish between reflexive, middle se and impersonal se, which are all presumably nominative. See Armstrong and MacDonald (this volume) for a discussion of the debate between se as pronoun vs. se being linked to an empty pronoun.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

4.1

149

Generic Interpretation

According to the literature se can have a generic interpretation in all three languages (Cinque 1988; Chierchia 1995; Mendikoetxea 2008; Sánchez López 2002; Suñer 1990). Generic interpretations arise in present tense, imperfective contexts, ifclauses, and temporal clauses. This interpretation is compatible with transitives, unergatives, passives, and unaccusatives. Therefore, the argument structure of the verb does not condition the distribution of generic interpretations: (49) Generalmente, no se enseñan bien las matemáticas (Sp) transitive Generalment, no s’ensenyen bé les matemàtiques (Cat) ‘Generally, math isn’t taught well’ (50) Generalmente no se trabaja bien aquí. (Sp) unergative Generalmenent no es treballa bé aquí. (Cat) Generally, no se work well here ‘Generally, they don’t work well here’ (51) Cuando se es catigado sin motivos. (Sp) passive Quan s’és castigat sense motius. (Cat) When se is punished without reasons ‘When someone is punished without any reason. . .’ (52) Cuando se desaparece sin avisar. (Sp). . . .Unaccusative Quan es desapareix sense avisar. (Cat) When se disappears without warning ‘When someone disappears without warning. . .’ (53) Cuando se está contento, el mundo parece mejor (Sp). . .copula Quan s’ és content, el món sembla millor. (Cat) When se is happy, the world looks better ‘When someone is happy, the world looks better’ In all these examples a generic operator must be responsible for licensing the ϕ pronoun.

4.2

Existential Interpretation or Episodic Interpretation

Another possible interpretation is the existential one as pointed out by Cinque (1988), Chierchia (1995), D’Alessandro (2007) and Sánchez López (2002). This interpretation is similar to an existential quantifier reading, and it does not require a generic context. Episodic interpretation is more evident with punctual aspect. This interpretation arises with transitive, unergative verbs very naturally:

150

F. Ordóñez

(54) Ahir, es va estudiar el tema de la Guerra Civil (Cat) Ayer, se estudió el tema de la Guerra Civil.(Sp) yesterday, se studied the topic of the Civil War ‘Yesterday, we/they studied the Civil War’ (55) Ahir, es va ballar molt. (Cat) Ayer, se bailó mucho. (Sp) yesterday, se danced a lot ‘Yesterday, we/they danced a lot’ (56) Ahir, es va parlar de matematiques. (Cat) Ayer, se habló de matemáticas. (Sp) ‘yesterday, se spoke about mathematics ‘Yesterday, we/they spoke about math’ However, punctual episodic aspect is ungrammatical with copula, passive and unaccusative verbs (Suñer 1990) in Spanish and Catalan: (57) *Aquí, ahir es va ser castigat (Cat) *Aquí, ayer se fue castigado (Sp) Here, yesterday se was punished ‘One was punished here yesterday’ (58) ? *Ahir es va estar a la reunió a les set (Cat) ? *Ayer se fue a la reunión a las siete. (Sp) Yesterday es aux-pas be in the meeting at 7 am ‘Yesterday one was in the meeting’ (59) ?? *Ayer se estuvo contento (Sp) ?? * Ahí es va estar content. (Cat) Yesterday se was happy ‘Yesterday, one was happy’ (60) ?*Ayer se desapareció a las 7 am ? * Ahir es va desapareixer a les 7 am ‘Yesterday, one disappeared at 7 am The important question is why this quasi-existential interpretation should be limited by argument structure in these two languages or why it is impossible with copulas and passives. Interestingly, the clear contrasts between these two classes of verbs do not hold in Italian. D’Alessandro (2007) and Cinque (1988) show that se is perfectly available in the contexts of passive with punctual past tenses. The examples are extracted from their studies: (61) Si é stati invitati anche noi. (It) (D’Alessandro 2007) se be-past been invited also us ‘We were invited’

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

151

(62) Come guardiani della legge, si è stati obbligati a controlare As guardians of the law, se have been-past obliged to follow l’osservanza.’ (It) the order ‘As guardians of the law we were obliged to follow the orders’ (63) Quando si e tornati alla pensione, si serviva gia la zuppa (It) When se be-past returned to the hotel se served already the soup ‘When we returned , we were served soup.’ (D’Alessandro 2007) According to Cinque (1988) and D’Alessandro (2007) the examples above are only possible under a reading of impersonal se similar to first person plural WE; the speaker is necessarily included in the reference of the pronoun. This reading is made clear when se co-appears with a 1ppl noi as in (64a) and (65a). Again, these same examples are ungrammatical in (Cat) and (Sp): (64) a. Si mangia la pasta anche noi (It) b. *Se come pasta incluso nosotros (Sp) c. *Es menja la pasta nosaltres també (Cat) Se eats pasta also us ‘We also eat pasta’ (65) a. Noi si riusciva a pagare la meità. (from La vita Agra p.15) b. *Nosotros se conseguia pagar la mitad c. *Nosotros s’aconseguia pagar la meitat we se managed to pay half ‘We managed to pay half’ Also, Spanish and Catalan impersonal se cannot be connected to a first person plural dislocated pronoun: (66) a. Noi, ha detto che non si è stati invitati b.*Nosotros, ha dicho que no se fue invitado. (Sp) c.* Nosaltres, ha dit que no es va ser invitat. (Cat) ‘Us, has said that not se has been invited’ ‘He/she said we’re not invited’ Another context in which the first person plural reading is manifested in Italian is with exhortative and imperatives with impersonal se. In those contexts, speaker and addressee must be involved. The equivalents are ungrammatical in Catalan and Spanish: (67) Si va! (Italian, from D’Alessandro 2007) se go-3 ps ‘Let’s go!’

152

F. Ordóñez

Si mangia la pizza! (it) se eat-3 ps the pizza ‘Let’s eat pizza’ (68) a. *Se va! (Sp) se go-3psg ‘Let’s go’ b. *Es va! (Catalan) se go-3 ps ‘Let’s go’ (69) a. *?Se come pizza!19 (Sp) b. *?Es menja pizza! (Cat) se eat-psg pizza ‘Let’s eat pizza!’ Unlike Italian this first person plural interpretation is not available in Catalan and Spanish. The important question to answer is what the source of this variation is. The reading in question is one in which the speaker plus another individual form a plurality as WE. This reading should be kept apart from the reading in which the impersonal pronoun ϕ is speaker oriented or inclusive. Thus, impersonal pronouns like ‘one’ uno in Spanish or hom in Catalan can be speaker oriented or inclusive. This is just a pragmatic fact that probably is not encoded in the grammar: (70) Cuando a uno+speaker no le hacen caso (Sp) When to one not pay attention ‘When no one pays attention to you.’ (could imply speaker)’ (71) Si uno+speaker va al médico cada día. (Sp) Si hom va al metge cada dia. (Cat) ‘If you go to the doctor every day (could imply speaker). . .’ Thus, we crucially distinguish between inclusivity of the speaker, on the one hand, and the grammatical encoding of first person plural in impersonal se, on the other. Inclusivity can be given by the pragmatics. For instance, in certain generic contexts the most prominent reading is speaker oriented, but it is not necessary:

19

The example is intended to involve the speaker and addressee together. In a generic context, not exhortative, however the sentence is grammatical as in (i). Observe that the sentence could have an inclusive reading. This reading is not obligatory since it can refer to a second person, not including the speaker. Thus, this should be distinguished from imperative contexts, which must include the speaker. The WE interpretation of Italian includes the speaker in the imperative necessarily. (i) Hoy se come pizza. (Sp) Today we/you eat pizza

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

153

(72) Cuando se es catigado sin motivos When one+speaker is punished without reasons ‘When you’re (could imply me) is punished for no reason’ (73) Aquí se puede defender este proyecto Here one+speaker can defend this project ‘Here’s where you (could imply me) can defend the project’ (74) Sí se puede si one+speaker can ‘If you (could imply me) can’ To summarize thus far, the question of generic versus existential interpretations and first person interpretations must be kept apart from the question of speaker inclusivity, which is a purely pragmatic fact. The WE interpretation discussed for Italian entails speaker inclusivity and is not pragmatic, but grammatical. Cinque (1988) and D’Alessandro (2007) point out that in Italian this inclusive reading has developed into a specified first person plural. As reported in D’Alessandro in some dialects that reading has supplanted we ‘noi.’ There is similarity between the use of this impersonal WE-reading se in Italian and the use of a gente in Brazilian Portuguese (BP) also (D’Alessandro 2007). A gente typically has a first person plural reading in which the speaker is obligatorily included with someone else. Thus, it involves a reading in which a plurality includes the speaker. That reading is not permitted in Catalan, Italian and Spanish. (75) A gente fala muito (Brazilian Portuguese) The people speak a lot ‘We/one speak a lot’ As in Italian, the plural denotation of a gente is clearly shown by the fact that it allows plural agreement in predicative contexts (dos Santos Lopes and Brocardo 2016): (76) A gente esta cansados People is tired-M-Pl ‘We’re tired’ (77) A gente andamos de bicicleta People go-1ppl by bike ‘We go by bike’

154

F. Ordóñez

The facts recall the cases of disagreement found in Spanish with person as follows20: (78) Los estudiantes tenemos razón the students be-1pl right ‘We students are right’ We will assume that the construction in (78) involves the doubling of unpronounced WE in a more complex DP. We can assume that unpronounced WE is in an outer layer above the DP. The empty unpronounced pronoun in this outer layer is responsible for the agreement with the verb in (79). A similar analysis can be adopted for the cases of a gente in BP. (79) [DPWE [DPLos estudiantes]] tenenos razón WE the students be-1pl right (80) [DPWE/ϕ [DPA gente]] andamos de bicicleta WE People go-1ppl by bike We assume that the Italian empty pronoun could be represented as WE/ϕ in a similar example. It also yields the double possibility found in Italian: (81) WE/ϕ si va in bicicleta We/One si go-3ps by bicycle ‘We/One goes by bicycle’ Similarly, this is also found with the clitic on in French: (82) WE/ϕ On va à vélo. (French) We/One on go-3ps by bicycle ‘We/One goes by bicycle’ Both constructions in French and Italian involve a clitic, contrary to the BP construction. Probably, for this reason, plural agreement with the verb is not permitted as in (83). (83) * WE/ϕ si andiamo in bicicleta. (It) WE/ϕ se go-1ppl by bike To summarize, I postulate that there is NO unique way in which impersonal se is licensed in all these constructions in the three languages. There are two different types of pronouns that bind se. There is a null impersonal pronoun corresponding to

20

I would like to thank an anonymous reviewer for pointing this out.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

155

ϕ, which approximates uno in a generic and existential (Cinque 1988; Chierchia 1995). This is possible in Catalan, and Spanish. This pronoun will always need a generic operator or an existential operator to be licensed. On the other hand, there is a null pronoun WE/ϕ which allows first person plural. I will assume that it behaves like a DP pronoun and does not require the licensing of an operator; therefore, it is less restricted in its distribution. This is the pronoun WE/ϕ proposed above. In the next section, I review the different licensing conditions for ϕ and WE/ϕ and how the differences in distribution in Catalan and Spanish versus Italian might be explained.

4.3

Distribution of Different Readings: Generic, Existential and WE

The distribution of generic readings can be explained if ϕ can be licensed by a generic operator. This is permitted in the context of copula, passive and unaccusative verbs in all three languages. Only generic contexts with a generic operator will be able to license the ϕ pronoun as follows Suñer (1990) (84) GEN OPi [ ϕi Se es inteligente] OP se is intelligent ‘One is intelligent’ Existential, episodic readings are always permitted in transitives, unergatives, and other bi-argumental verbs in Catalan and Spanish. In all these cases se is licensed with an existential reading (Cinque 1988; Chierchia 1995). However, the existential reading is not licensed in passives in Catalan and Spanish. We repeat the example with the passive: (85) *Aquí, ahir es va ser castigat (Cat) *Aquí, ayer se fue castigado (Sp) here, yesterday se was punished ‘Here, they were/one was punished yesterday’ I argue that existential readings are licensed by an existential operator and that the existential operator is sensitive to the argument structure of the verb. Only existential operators can be introduced by little v. This was pointed out by Borer (2005), Cinque (1988) and Jaeggli (1986); arbitrary interpretations require the verb to be transitive or unergative. We are claiming that that operator must be licensed in the context of little v under the existential interpretation. Since passives, copulas and unaccusatives have no little v, then no existential operator is available, and the sentence is rendered ungrammatical. No existential operator is licensed in (87) since there is no little v. This example contrasts with the transitive verb with little v in (86):

156

F. Ordóñez

(86) Ayer [vp Ex Op i ϕi se compraron varios libros] transitive verbs (Sp) yesterday se bought various books ‘Yesterday, some books were bought’ (87) * Aquí, ahir [VP Ex Op i ϕ es va ser castigat] Passive Catalan here, yesterday ϕ se was punished ‘One was punished yesterday’ On the other hand, there is a third reading available in Italian, the plural WE reading. This reading is encoded grammatically in Italian, but not in Catalan or Spanish. As discussed earlier, this reading is available with passive verbs as in (88). Since this WE/ϕ does not require the licensing by any operator, the argument structure of the verb is not so crucial. Accordingly, Italian WE readings arise in copula, unaccusative and unergative verbs in non-generic contexts: (88) WE/ϕ i Sii è stati invitati. (passive) WE si was invited ‘We were invited’ (89) WE/ϕ i sii è spariti come persone. (unaccusative) WE si has disappeared like people ‘We disappeared like people’ (90) WE/ϕi sii va! (first person plural imperative) WE se go ‘Let’s go!’ In some examples the pronoun can be overt: (91) Noi si e stati contenti dopo quella festa (copula) We, se has been happy after that party ‘We were happy after that party’ (92) Noii sii è stati invitati (passive) WE si was invited ‘We were invited’ In conclusion, I have postulated two different kinds of pronouns that bind impersonal se in Catalan, Italian and Spanish. The pronoun ϕ must be bound by a generic or existential operator in all three languages, and existential operators are only permitted when there is little v. WE/ϕ,21 however, is possible in Italian and does not require the licensing by any operator. Therefore, it is not sensitive to argument

21

Following MacDonald and Maddox (2018) we have to propose that this empty pronoun will have D feature and also a 1person feature in Italian and therefore we predict that this WE interpretation is incompatible with by phrases in Italian, which seems to be the correct prediction.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

157

structure. WE/ϕ behaves like a DP and is free to appear when there is no operator as long as it receives the first person interpretation. In contrast, ϕ is a defective pronoun which is only licensed by a generic or existential operator. The WE/ϕ pronoun in Italian triggers plural agreement in number and agreement in gender in adjectives and past participles. Thus, there is a division of labor in Italian between the clitic se, which triggers third person singular agreement as we saw in (83), and WE/ϕ which triggers plural agreement with adjectives and past participle agreement: (93) WE/ϕ plural si3rdperson è3person statiplural invitatiplural WE/ϕ plural si3rdperson was3person beplural invitedplural ‘We were invited’ Because there is no WE/ϕ in Catalan and Spanish, plural agreement is ungrammatical with past participle or adjectives. Only singular is permitted22: (94) a. Si è contenti. (It) b. * s’és contents (Cat) c. *se está contentos (Sp) ‘se is happy-plural’ ‘One is happy’ Similarly, impersonal se in Italian licenses floating quantifiers. This is expected because of the plurality of WE/ϕ; however, it is impossible with ϕ in Catalan and Spanish. As in (95) the pronoun is licensing the plural floating quantifier in Italian. However, this is not possible in Catalan and Spanish because ϕ is singular: (95) WE/ϕ plural Si è arrivatiplural tuttiplural se arrived-pl all-pl (96) *ϕi SEi llegó todosi (Sp) se arrived all-pl (97) *ϕi ES va arribar totsi (Cat) se arrived all-pl Finally, impersonal se can bind a reciprocal pronoun in Italian but not Catalan or Spanish. This is again the prediction made by our approach since reciprocals require plurality. Thus we find the following contrasts represented below23:

22

This raises the interesting issue of the distribution of this plural agreement in Italian. Does it have to be interpreted as WE? The answer seems to be no. This is why we label it WE/ϕ. 23 A simple google search of both sentences in Italian and Spanish shows the difference. While Spanish yielded 9 results, Italian yielded more than 9000 for sentences with impersonal se and reciprocals. (found June 20, 2020)

158

F. Ordóñez

(98) WE/ϕ i si era parlato [l’uno con l’ altro]i (It) (99) a.*ϕi es parlava [l’un amb l’ altre]i (Cat) b. *ϕi se hablaba [el uno con el otro]i (Sp) ‘se talked one with the other’ The reason why (99) is ungrammatical is the same reason why generic singular pronouns cannot be antecedents for reciprocals (100). (100) * Unoi hablaba [el uno con el otro]i (Sp) ‘One spoke the one with the other’ Impersonal se in Catalan and Spanish have no instances with reciprocals, contrary to Italian.24 Another consequence of this WE/ϕ interpretation is the fact that it is compatible with predicates which require plurality interpretation in the subject and are coindexed with a reciprocal-reflexive first person pronoun (Cinque 1995). Under the WE/ϕ possibility this is perfectly available as shown below. Thus, the Italian sentence is ambiguous between a reciprocal and a non-reciprocal interpretation of the first person plural clitic: (101) WE/ϕi Cii si trova in Piazza WE 1ppl si find in the square ‘We meet each other in the square’ (102) WE/ϕi Cij si trova in Piazza WE 1ppl si find in the square ‘One meets us in the square’ However, Catalan and Spanish lack WE/ϕ and therefore cannot have impersonal se with a first reciprocal interpretation for the first person clitic. The sentences are OK in Catalan and Spanish with a non-reciprocal interpretation for the clitic in first person plural, parallel to what we see in Italian in (102). (103) *ϕi se nosi encuentra en la plaza ϕ si 1ppl find in the square ‘We meet each other in the square’ (104) ϕi se nosj encuentra en la plaza ϕ si 1ppl find in the square ‘One meets us in the square’

24

I leave this interesting semantic difference here. A much deeper semantic analysis is needed, which I leave for further research. What is crucial is the fact that such reciprocals are permitted in Italian but not in Catalan or Spanish.

On the Nature of the Impersonal SE: Why Italian is not like Catalan and Spanish

159

5 Conclusion This paper has explored the syntax and interpretation of impersonal se constructions in Spanish, Catalan, and Italian. We have argued for a syntactic approach in which se is bound by an empty pronoun. We have argued for two different kinds of empty pronouns. One is a ϕ pronoun, which is licensed either by a generic operator or by existential closure. The other interpretation, limited to Italian, consists of a silent WE/ϕ pronoun. There is a clear parallel between Italian se and the clitic on in French. On might be interpreted as WE/ϕ. That WE/ϕ exists in Italian explains why it has a wider distribution. Thus, it is permitted in copula and passive constructions in non-generic contexts, in imperatives with first person interpretation, can be bound by reciprocal pronoun, can be bound by reciprocal clitics, allows floating quantifiers and can be doubled and be coindexed by a topicalized first person pronoun. None of these possibilities are available in Catalan or Spanish. Finally, this paper shows how a detailed comparative analysis of these languages is able to uncover new differences not well understood until now. Acknowledgments I would like to thank Grant W Armstrong and Jonathan E MacDonald, two anonymous reviews and the ARC fellowship at the Graduate Center CUNY for making this research possible.

References Belletti, Adriana, and Luigi Rizzi. 1988. Psych verbs and theta theory. Natural Language and Linguistic Theory 6: 291–352. Borer, Hagit. 2005. Structuring sense, The normal course of events. Vol. II. Oxford: Oxford University Press. Chierchia, Gennaro. 1995. The variability of impersonal subjects. In Quantification in natural languages, ed. E. Bach, E. Jelinek, A. Kratzer, and B. Partee, 107–143. Dordrecht: Kluwer. Cinque, Guglielmo. 1988. On Si construction and the theory of Arb. Linguistic Inquiry 19: 521–581. ———. 1995. Chapter 5. Ci si. In Italian syntax and universal grammar. Cambridge: Cambridge University Press. D’Alessandro, Roberta. 2007. Impersonal SI constructions. Berlin: Mouton de Gruyter. Dobrovie-Sorin, Carmen. 1998. Impersonal Si constructions in romance and the passivization of unergatives. Linguistic Inquiry 29: 399–347. dos Santos Lopes, Celia Regina, and Maria Teresa Brocardo. 2016. Main morphosyntactic changes and grammaticalization process. In The handbook of Portuguese linguistics, ed. W. Leo Wetzels, Joao Costa, and Sergio Menuzzi, 471–487. Hoboken: Wiley-Blackwell. Fernández Ordóñez, Inés. 1999. Leísmo, laísmo y loísmo. In Gramática descriptiva de la lengua española, ed. Ignacio Bosque and Violeta Demonte, 1317–1398. Madrid: Espasa. Holmberg, Anders. 2010. The null generic subject pronoun in Finnish: A case of incorporation in T. In Parametric variation: Null subjects in minimalist theory, ed. Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan, 200–230. Cambridge: Cambridge University Press. Jaeggli, Osvaldo. 1986. Passive. Linguistic Inquiry 17: 587–622.

160

F. Ordóñez

Kayne, Richard. 2000. Chapter 8. Person morphemes and reflexives in Italian, French and related languages. In Parameters and universals. Oxford: Oxford University Press. MacDonald, Jonathan E. 2017. An implicit projected argument in Spanish impersonal and passive se constructions. Syntax 20: 353–383. MacDonald, Jonathan E., and Matthew Maddox. 2018. Passive se in Romanian and Spanish: A linguistic cycle. Journal of Linguistics 54: 389–427. Mendikoetxea, Amaya. 2008. Clitic impersonal constructions in romance: Syntactic features and semantic interpretation. Transactions of the Philological Society 106: 290–336. Ordóñez, Francisco, and Esthela Treviño. 2016. Agreement and DOM with impersonal se: A comparative study of Mexican and Peninsular Spanish. In The morphosyntax of Portuguese and Spanish in Latin America, ed. Mary Kato and Francisco Ordóñez, 236–258. Oxford: Oxford University Press. Pujalte, Mercedes, and Andrés Saab. 2012. On se insertion in Spanish. In The end of argument structure? ed. M.C. Cuervo and Y. Roberge, 229–260. Bingley: Emerald Press. Rigau, Gemma. 1997. Locative sentences and related constructions in Catalan: Ésser/Haver alternation. In Theoretical issues at the morphology-syntax interface, ed. Amaya Mendikoetxea and Miriam Uribe-Etxebarria, 395–421. Bilbao/San Sebastián: Universidad del País Vasco/ Diputación Foral de Gipuzkoa. Romero, Juan and Javier Ormazabal. 2019. The formal properties of non paradigmatic SE. Borealis, An International Journal of Hispanic Linguistics 8(1): 55–84. Romero, Juan, and Javier Ormazabal. 2019. The formal properties of non paradigmatic ‘se’. Borealis—An International Journal of Hispanic Linguistics 8: 55–84. https://doi.org/10.7557/ 1.8.1.4704. Saab, Andrés. 2014. Syntax or nothing: Some theoretical and empirical remarks on implicit arguments. Borealis: An International Journal of Hispanic Linguistics 3: 125–183. Sánchez López, Cristina. 2002. Las construcciones con se. Madrid: Visor Libros. Sportiche, Dominque. 2014. French reflexive se: Binding and merge locality. In Locality, ed. E. Aboh, M. Guasti, and I. Roberts, 104–137. Oxford: Oxford University Press. Suñer, Margarita. 1990. Impersonal Se passives and the licensing of empty categories. Probus 2: 209–231.

Personal se with Unergatives in Romanian Monica Alexandrina Irimia and Virginia Hill

Abstract Romance se has long been taken to be an indicator of unaccusativity (Rooryck and Vanden Wyngaerd, Dissolving binding theory. Oxford: Oxford University Press, 2011, a.o.). This generalization has already been questioned for certain types of reflexives, and it was found to be too strong (Reinhart and Siloni, Linguistic Inquiry 36:389–436, 2005; Sportiche D, Linguistic Inquiry 45:305–321, a.o.). An outstanding question in current studies concerns the status of se that does not obviously mark reflexivity or anticausativity (Armstrong G, Borealis 2:81–128, 2013, a.o.). In this paper, we signal the presence of such constructions in Romanian, and propose that lack of reflexivity in the presence of se may follow from the direct merge of this element inside a DP instead of a vP. The hypothesis is that, if se merges on the nominal (versus clausal) spine, it is orthogonal to the mapping of internal arguments or to Accusative Case, and may occur with unergatives. Keywords Personal se unergatives · Animacy · Differential marking · Redundant se

1 Introduction Romance se has long been taken to be an indicator of unaccusativity (Rooryck and Vanden Wyngaerd 2011, a.o.). This generalization has already been questioned for certain types of reflexives, and it was found to be too strong (Reinhart and Siloni 2005; Sportiche 2014, a.o.). An outstanding question in current studies concerns the status of se that does not obviously mark reflexivity or anticausativity (Armstrong 2013, a.o.). In this paper, we signal the presence of such constructions in Romanian,

M. A. Irimia (*) University of Modena and Reggio Emilia, Reggio Emilia, Italy e-mail: [email protected] V. Hill University of New Brunswick, Saint John, Canada © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_7

161

162

M. A. Irimia and V. Hill

and propose that lack of reflexivity in the presence of se may follow from the direct merge of this element inside a DP instead of a vP. The hypothesis is that, if se merges on the nominal (versus clausal) spine, it is orthogonal to the mapping of internal arguments or to Accusative Case, and may occur with unergatives. The data under discussion is best illustrated with the unergative verb râde ‘laugh’, which may display Accusative se, apparently on an optional basis and without any change in the interpretation. More precisely, this verb may be used by itself, as in (1a); or with an Accusative form of se, as in (1b); or with a Dative reflexive form of se, as in (1c). In Modern Romanian, (1a) and (1c) are used in all language registers, whereas (1b) is limited to sub-standard varieties. All three variants are productive, as can be easily seen from internet searches. (1) a.

b.

c.

Maria râde (de mine) Maria laughs at me ‘Maria is laughing at me’ Maria se râde (de mine) Maria SE.3ACC laughs at me ‘Maria is laughing at me’ Maria îşi râde de mine Maria SE.3ACC laughs at me ‘Maria is laughing at me/is mocking me’

Traditional grammars notice that the use of se in (1b) is redundant, since it does not bring any change in the argument structure or in the interpretation compared to (1a), except for signaling a different language register. From this point on, we shall refer to this se as redundant se, in order to differentiate it from the garden varieties of reflexive se. We also must clarify that this is personal se (i.e., it agrees with the lexical subject), not the impersonal se discussed in Dobrovie-Sorin (1998) (see Armstrong and MacDonald [this volume], MacDonald and Melgares [this volume], and Ordóñez [this volume] for more on impersonal se constructions). The latter type applies to unergatives (or other verb classes), has a generic/arbitrary interpretation and excludes co-occurrence with lexical subjects. We pick up on the intuition of traditional grammarians that this personal se is redundant, and question its function in (1b), within a feature based derivational framework, where redundant elements are unlikely. The main point of our proposal is that se has been reanalyzed in a language contact situation, with some unergative verbs, from being both reflexivizing and mediator for Clitic Left Dislocation (CLLD) (which is the case for se in regular reflexive structures in Romanian), to being used only for CLLD. This specialization entails a change in the merge position, from the vP domain to the DP domain. The paper is organized as follows: Section 2 introduces the data. Section 3 presents the readings arising from the various structures with râde. Sections 4 and 5 explore the current analyses of Romance se and conclude that none can capture the properties of constructions such as (1b). Section 6 presents tests on Romanian data that indicate a tight relation between redundant se and the human nature of the

Personal se with Unergatives in Romanian

163

relevant subject. Section 7 puts forth a proposal where redundant se functions as a clitic personal pronoun that mediates CLLD. Section 8 summarizes the findings.

2 Data The verbs that qualify for redundant se display this element on an “optional” basis, in the sense that they are not inherently reflexive. Historical linguists working on Romanian (e.g., Pană Dindelegan 2014) consider that redundant se entered the language through borrowings from Old Slavonic, where some verbs that would count as unergative in Romanian displayed the reflexive ending -sja in Slavonic (e.g., boretsja ‘to fight-REFL’ in Russian; see also Basilico 2017 for a discussion of Slavic sja). The Romanian verbs that fall in this category may not necessarily have a Slavonic etymology (e.g., greşi ‘err’ comes from Slavic, but jura ‘swear’ comes from Latin) but were used in doublets with a Slavic -sja counterpart which was calqued. Although -sja may have had an impact on the argument structure of such verbs in Slavic languages (e.g., see Medová 2009 for an inherently reflexive analysis of Czech se smál ‘he laughs’), that effect did not transfer to Romanian, where -sja was reanalyzed as an optional redundant se in these contexts.1 The list of verbs that display redundant se depends on borrowings/calques, and it is much longer in Old Romanian than in Modern Romanian, since the Slavic influence on the lexicon decreased after the eighteenth century. The verbs that survived with redundant se in Modern Romanian are unergative, a couple of which may also double as transitives: • Unergatives: greşi ‘err’; jura ‘swear’; gândi ‘think’; lupta ‘fight’; plânge ‘complain’; tăcea ‘keep quiet’ • Transitives: jura ‘swear’; gândi ‘think’ In addition, Modern Romanian shows an increased use of redundant se with râde ‘laugh’ (it is not clear to what extent this was present in Old Romanian) and extended it to the French borrowing risca ‘risk’. A puzzling observation is that there are no unaccusative verbs with redundant se. Constructions with se râde, se greşi, se risca occur in sub-standard Romanian and are condemned by prescriptivists, although (se) jura, (se) gândi and (se) lupta are accepted. The use of se with risca, which is a new borrowing, indicates that this particular analysis of se is still active in the language.

1

As correctly pointed out by a reviewer, constructions similar to (1b) are also seen in French (e.g., Marie se rit de moi), as well as in other Romance languages. However, it is not clear whether the construction has the same structural source as in Romanian. What we know for sure is that most verbs with redundant SE are borrowed or calqued after Slavonic reflexive verbs, and they are more productive in Old Romanian, where Slavonic vocabulary had more impact, than in Modern Romanian. In the latter period many Slavonic borrowings have been lost, the language becoming more inclined to lexical borrowing from French.

164

M. A. Irimia and V. Hill

For the purpose of this paper, which is to sort out the syntactic status of redundant se, we focus on one relevant verb, namely (se) râde ‘laugh’, illustrated in (1), so we can maintain consistency in tests. Other verbs are tested as needed.

3 Râde: (In)voluntary Reading Since the most tested verb in this paper is (se) râde ‘laugh’, we begin with a description of its use and interpretation. Basically, there are two readings for râde ‘laugh’, as shown in (1): (1) an involuntary one, which excludes the PP goal de mine ‘at me’, and denotes a physical state – optionality of PP-de in (1a) is marked through round brackets; (2) a voluntary one, which obligatorily involves the presence of the PP-de goal indicating the disadvantaged party, and yields a cognitive reading of intent. The contrast concerns the willful intent, in the sense of Al Zahre and Boneh (2010), as further shown in (2), where the constituent ‘without intending to’ yields anomalous readings in the presence of a PP goal. (2) a.

a'. b. c.

Maria (se) râde fără să vrea Maria SE.3ACC laughs without SUBJ wants ‘Maria is laughing without intending to’ Maria râde de mine # fără să vrea Maria laughs at me without subj wants Maria se râde de mine # fără să vrea Maria SE.3ACC laughs at me without SUBJ wants Maria îşi râde de mine # fără să vrea Maria SE.3DAT laughs at me without SUBJ wants Intended:‘Maria laughs at me without intending to’

Only (2a) is felicitous in the presence of ‘without intending to’, irrespective of the presence of se, as long as the PP goal is absent. This is the involuntary reading. On the other hand, the presence of the PP goal entails a reading close to (but not as strong as) ‘to mock’, which is semantically incompatible with lack of intent. Accordingly, the question is whether râde ‘laugh’ has the same argument structure irrespective of its forms or interpretive outcomes, or whether the difference in interpretation reflects changes in the (in)transitivity frames for this verb. This is important in light of the reflexivization with either Accusative or Dative se in Romanian, a distinction that is not visible in other Romance languages.

Personal se with Unergatives in Romanian

165

4 Antipassives The clarification of the argument structure of râde with or without se will take into account previous analyses of Romance se, especially the antipassive structure in Masullo (1992). Thus, this section introduces the antipassive analysis of reflexives proposed for Spanish, and then considers its application to Romanian. The main point is that significant differences arise between the two languages in the type of reflexivization of similar classes of verbs.

4.1

Masullo (1992)

For verbs of the same class as ‘mock’, Masullo (1992) proposes an antipassive analysis, which is based on the contrast in (3). (3) a. b.

Juan confiesa sus pecados John confesses his sins Juan se confiesa (de sus pecados) John se.ACC confesses about his sins ‘John confesses his sins.’ (Masullo 1992: 175)

In (3a), the verb has a transitive frame that seems to turn intransitive in (3b), through reflexivization. This mechanism arises for several Spanish verbs that denote a mental attitude (e.g., ‘mock’; ‘complain’; ‘take pride’; ‘refuse’; ‘realize’ etc.). Masullo’s hypothesis is that reflexivization in these cases involves the incorporation of se into the verb, which mimics the effects of an unergative frame. That is, se absorbs the verb’s Case, so the object needs to merge as a PP and passivization is disallowed. The formalization of this proposal replicates the mechanism for passivization in Baker (1988): the passive marker –en incorporated in the verb absorbs a theta-role so the agent is mapped through adjunction, as a PP-by.2 This analysis makes several predictions, among which, the argument sharing between se and PP-de in (3b): both items share in the checking of the theme/patient theta-role. Presumably, the PP is selected here by se-V, instead of being adjoined to vP. The point is that the theta-role can be checked only by se, so the PP is optional. Accordingly, no DP or CP are expected to occur in these structures as direct objects, which mimics the argument structure of unergatives (Hale and Keyser 2002).

2

Note that Masullo’s antipassives are labelled in other studies as reflexive unergatives (e.g., Alsina 1996; Reinhart & Siloni 2005; Alboiu et al. 2004). These are different from the verbs with redundant se discussed in this paper, which are bona fide unergative in their active form.

166

M. A. Irimia and V. Hill

Most tests for antipassivization are not available in Romanian.3 Hence, in order to check how Masullo’s proposal fares for Romanian, we can look only at the PP that presumably shares the theta-role with se, and at its extraction across wh-islands. For the latter, the idea is that the wh-island constraint can be violated when the PP-de is extracted because extraction from a selected position is always better than from an adjunct position (relativized minimality—Rizzi 1990). This is shown in (4). (4) De quéj of what

quieres want.2

saber know

quién sej confesó who refl. 3.acc confessed

tj?

We shall supplement these two criteria for antipassivization with other tests, as needed.

4.2

Antipassives in Romanian

The first point of variation is that, unlike Spanish, Romanian has distinct forms for Dative and Accusative reflexive pronouns. Thus, the list of antipassive verbs provided in Masullo (1992) decreases for Romanian, since some of them come as Dative reflexives, as in Table 1. Table 1 Reflexives of mental attitude

‘refuse’ ‘remember’ ‘become aware’

Spanish negarse acordarse percartarse

Romanian a-şi/(*se) refuza a-şi/(*se) aminti a-şi/(*se) da seama

Importantly, the Dative reflexive pronoun is a clitic that spells out the indirect object and does not interfere with the projection of a direct object position. Thus, these verbs occur with direct objects, or with PP-de, as in (5). The alternation between DP and PP-de in the complement position confirms that PP-de is selected in these constructions. (5)

Elena şi-a amintit poezia/de mine Helen se.DAT-has remembered poem.the/of me ‘Helen remembered the poem/me’

Romanian verbs that correspond to Masullo’s antipassive list show the following peculiarities: 3 The tests proposed in Masullo (1992) that are not valid for Romanian data: I-subjects & locative subjects, because Romanian is VSO by default; causatives cannot block the clitics on the embedded T/V because they are bi-clausal control structures with different CPs; infinitives are able to license subjects in various contexts.

Personal se with Unergatives in Romanian

167

• Some are inherently reflexive and transitive (vs. antipassive); e.g., se spovedi ‘confess’, so a PP-de is excluded • Some do not have an active counterpart; e.g., *(se) mândri ‘take pride’ is inherently reflexive • Most antipassives differ in meaning from their active counterparts, insofar as they denote mental processing and/or intent, which shows that se incorporation has an impact on the root; e.g., se ţine de ‘persist’ vs ţine ‘hold’; se prinde de ‘realize’ vs prinde ‘catch’. The two criteria for assessing the antipassive status are applied in (6) to se mândri ‘take pride’. (6) a.

b.

c.

Maria se mândrea (cu/de rezultatele Maria SE.3ACC took.pride with/of results.the fetei.) daughter.the.GEN ‘Maria was taking pride in her daughter’s results’ ?Cine cu ce vrei să ştii dacă sej who with what want.2SG SUBJ know whether refl. 3.acc mândrea tk tj took pride ‘Who would you like to know took pride in what?’ ?Cinek pe cinej vrei să ştii dacă who DOM whom want.2SG SUBJ know whether a invitat tk tj? has invited ‘Who do you want to know whether they invited whom?’

In (6a), the PP is optional since it shares the theta-role with se. In (6b), extraction of the selected PP is acceptable across ‘whether’, as predicted from (5). We slightly modified the extraction pattern in (6b) to cater to the multiple wh-movement that applies in Romanian, but the criterion is the same (i.e., a CP with a wh-phrase in its Spec). Thus, the extraction of the PP from an antipassive construction has the same degree of grammaticality as the extraction from argument positions in active constructions, as seen in (6c). Masullo’s class of antipassives includes ‘mock’, which comes close to the voluntary interpretation of râde in Romanian. Thus, the question arises whether se râde is antipassive.

168

M. A. Irimia and V. Hill

5 (In)transitivity Frames This section proposes tests that clarify the argument structure of râde; namely, past participle forms, anaphoric DP and telicity. The result is that the verb remains unergative, even in the presence of se; that is, there is no complement position, although the number of theta-roles mapped to other positions may vary. The analysis refers to the paradigm in (2), so we repeat it here for convenience, as (7). (7) a.

Maria (se) râde fără să vrea Maria SE.3ACC laughs without SUBJ wants ‘Maria is laughing without intending to’ a’. Maria râde de mine # fără să vrea Maria laughs at me without SUBJ wants b. Maria se râde de mine # fără să vrea Maria SE.3ACC laughs at me without SUBJ wants c. Maria îşi râde de mine # fără să vrea Maria SE.3DAT laughs at me without SUBJ wants Intended:‘Maria laughs at me without intending to’

5.1

Involuntary Râde

Involuntary active râde ‘laugh’, illustrated in (7a), is predictably unergative, on par with other verbs of the same semantic group, such as ‘cough’ or ‘sneeze’, mapping the only theta-role in the subject position. The standard tests for deciding whether a verb is unergative or unaccusative are: the option for ‘have’ or ‘be’ auxiliaries; extraction of partitive pronouns such as Italian ne; object-verb agreement (see Sportiche 2014 for an overview). From these, only the alternation ‘have’/‘be’ is available in Romanian, and only in passivized structures, as in (8a). (8) a.

b.

*Maria Maria

era was

râsă. VS laughed

ok

Maria era plecată/căzută Maria was gone/fallen ‘Maria was gone/fallen’ *O consider râsă VS ok O consider plecată/căzută her consider.1SG laughed her consider.1SG gone/fallen ‘I consider her gone/fallen’

In (8a), râde rules out the structure, whereas an unaccusative like ‘go’ or ‘fall’ allows for it. This contrast is further confirmed in (8b), where a selected past participle predicate is possible with unaccusatives but not with the unergative râde ‘laugh’.

Personal se with Unergatives in Romanian

169

In (7a’), we also see the possibility of having the involuntary se râde. Can se change the argument structure, so that ‘laugh’ becomes transitive? For that, we compare (9a) and (9b). Transitive reflexive verbs license anaphoric DPs in object position, as in (9a). That is ruled out in (9b), with se râde, so we have to exclude a transitive frame for this verb. (9) a.

Dan sej priveşte pe sine însuşij în Dan SE.3ACC watches DOM him.self in ‘Dan looks at himself in the mirror’ b. *Dan sej râde pe sine însuşij la Dan SE.3ACC laughs DOM him.self at Intended: ‘Dan amuses himself at the cinema’

oglindă mirror cinema cinema

Accordingly, one may consider se râde as antipassive, with se incorporated in the verb, which would block the direct object position for the anaphoric DP. This is, however, problematic, because, as mentioned in the previous section, antipassive se may optionally co-occur with the related PP it shares the theta-role with, with no change in the meaning of the verb. This is not the case with se râde, as mentioned for (7): the presence or the absence of PP-de causes a change in the interpretation (from involuntary to voluntary) which indicates that PP-de spells out a different theta-role than se. Furthermore, se râde is used in free alternation with the active form, and se is rather agentive versus theme/patient. In this respect, there are other verbs showing a similar alternation in form for a similar reading, namely, (se) greşi ‘err’ and (se) risca ‘risk’. Consider (10). (10) a.

b.

El (s)-a greşit când a spus asta4 he SE.3ACC-has erred when has said this ‘He erred/made a mistake when he said this’ http://www.timpul.md/articol/nu-te-rade-cand-te-gresesti-34992.html Ce zici, (mă) risc să copiez what SAY.2 SE.3ACC risk.1 SG subj copy.1SG la examen? at exam ‘What do you say, should I risk cheating on the exam?’ http://mariana-badea.ro/?p¼113

In (10), the presence of se makes no difference in the reading. Let us look closer at greşi, which can have either an intransitive frame, as ‘err’ in (10a), or a transitive frame elsewhere, such as ‘make a mistake’/‘mess up’ in (11).

Se greşi is not the equivalent of se enganar ‘to be wrong’, for which Romanian has the verb se înşela.

4

170

M. A. Irimia and V. Hill

(11) Maria a greşit formula //*l-a greşit pe el Maria has mistaken formula him-has mistaken DOM him ‘Maria messed up the formula’ vs ‘Maria messed him up’ Crucially, the transitive frame allows only for inanimate objects (e.g., ‘formula’), so the personal pronoun with human reference does not qualify for its direct object position in (11). Thus, there is no passive version of transitive ‘err’ with a human subject. Hence, there are two different lexical entries for greşi, one with a transitive grid, and one with an intransitive/unergative grid. The latter has only one theta-role, and the presence of se does not modify the theta-grid. This is similar to se râde, the two verbs being used as if they were active unergatives, as in (12). (12) Nu te râde când te greşeşti! not SE.3ACC laugh when SE.3ACC err.2SG ‘Don’t laugh when you err!’ http://www.timpul.md/articol/nu-te-rade-cand-te-gresesti-34992.html Furthermore, the type of reflexivization we see with ‘laugh’ and ‘err’ also applies to a transitive verb like ‘risk’ in (10b). Since there is a direct object for ‘risk’ in (10b), and this is not a double object verb, se cannot be considered incorporated in the verb. Again, se cannot be in the direct object position and the construction cannot be antipassive.

5.2

Voluntary Râde

The voluntary reading of râde arises when a PP is present in the derivation to indicate the party affected by the derision, as in (7a’). Redundant se may also occur in this context, as in (7b). Furthermore, the same context allows for the Dative reflexive se, as in (7c). The Dative reflexive pronoun indicates the beneficiary, which is also the agent of the derision. A tertiary theta-role with îşi râde is not surprising, considering cross-linguistic evidence that willful intent is a general characteristic of non-argumental reflexive Datives. For example, Al Zahre & Boneh (2010: 3–4) identify the implicature obtained with the Hebrew Reflexive Datives, which have an “isolating effect for the referent of the subject DP. . . .the subject can be seen as engaging in the activity for her/his own pleasure or sake”. For Romanian, the consulted speakers agreed that the reflexive variant with îşi has this implicature at a high degree: in (1c), Maria is doing the laughing activity for her own sake and enjoyment, and I am negatively affected as a result. The mocking dimension and willful intent of râde are stronger in the presence of îşi than in the other versions. The question is: How are these theta-roles mapped to syntax? In particular, is there a difference in the argument structure of the active unergative râde and of se/îşi râde?

Personal se with Unergatives in Romanian

171

The first point we make is that voluntary râde, in all its variants, is incompatible with a regular transitive structure. This follows not only from the lack of a DP in direct object position but also from telicity effects: as shown in (13), none of its variants pass telicity tests when submitted to the Vendlerian diagnoses (Vendler 1957) (13a). However, this test is successful with unaccusative verbs of the type ‘fall’, as in (13b). (13) a.

b.

(S-/ Şi)au râs de noi timp SE.3ACC/ SE.3DAT¼ have laughed at us time de o oră// *într-o oră of an hour in an hour ‘They laughed at us for an hour’//*’They laughed at us in an hour’ Maria căzu într-o oră// *timp de o oră Maria fell in an hour time of an hour ‘Maria fell in an hour’//*’Maria fell for an hour’

The point of this test is that the ‘in XP’ reading fails when there is no direct object. Accordingly, the verb is either unergative (as argued in the previous section for the involuntary (se) râde) or antipassive, in case se is incorporated in the verb and blocks the object position. The next test concerns the status of the PP: Is it selected or adjoined? If the PP is selected, it is obligatory and the verb has a different argument structure than the unergative with no internal argument. If the PP is adjoined, then it maps an optional theta-role to the unergative. For the constructions with (se) râde, PP-de qualifies as an adjunct: its presence forces a voluntary reading, and this is orthogonal to the presence of redundant se. Therefore, se does not share the theta-role with PP-de in this construction; se is not incorporated into V, nor is PP-de a sister of V. Evidence in this respect comes from ungrammaticality upon extraction of PP-de across a wh-island, as shown in (14b). This test was shown to work for the Romanian antipassive in (6). (14) a.

b.

Elena se râde *(de mine) Elena REFL. 3.ACC laughs at me ‘Elena laughs at me’ *Cine de cine vrei să ştii dacă who at whom want.2SG SUBJ know whether se râde? REFL.3.ACC laughs

Therefore, we cannot adopt an antipassive analysis for se râde, with neither voluntary nor with involuntary readings, and have to maintain an unergative analysis for this verb. The next task is to verify if other explanations, namely, complex predicate structures, may grasp the use of se with râde.

172

5.3

M. A. Irimia and V. Hill

Problems with Other Complex Predicate Analyses

In this section, we discuss the proposals by Armstrong (2013), and MacDonald (2004, 2008), and show that their incorporation analyses for se-verbs are not adequate for our data.

5.3.1

Armstrong (2013)

The novel observation made by Armstrong (2013) is that, despite surface structure identity, Spanish non-selected se spells out two distinct structures, both non-doubling: the Transitive SE Clitic (TSC) and the Agentive Reflexive Clitic (ARC).5 TSCs are more relevant for our discussion, given that, similarly to Romanian, no aspectual construal related to telicity is obtained. The properties of TSC are given in (15). (15) a. b.

TSCs do not systematically assign one particular role to their subject TSCs have no set aspectual value associated with them

An indication that the configuration obtained with the TSCs is not an accomplishment comes from their co-occurrence with bare nouns. (16) El niño se comió veneno. The child SE.3 ate poison ‘The child ate poison’

(Armstrong 2013: ex. 25)

The hypothesis is that TSCs are structures in which the main predicate is formed from a verb and se, as in (17). The internal argument is merged to this complex, and v is further added. (17)

vP v´

EA/DP v

VP IA/

´

SE

5 ARCs are constructions which introduce obligatory accomplishments, thus not further discussed here.

Personal se with Unergatives in Romanian

173

In the case of Romanian se râde ‘laugh’ (as well as ‘err’ or ‘risk’), we cannot generate the counterpart of (17) (note that ‘risk’ takes a CP not a DP as complement). We can, on the other hand, add a PP-de goal, which is not possible in (17). Furthermore, there are certain predictions arising from (17) that are not borne out in the Romanian constructions: • V + SE does not always preserve the meaning of V, since it has access to the verb root. Contrary to this prediction, in our data, redundant se obligatorily preserves the meaning of the verb, which indicates that it does not access the verb root. • V +SE might impose different selectional restrictions on its internal and external arguments than V alone (without drastically changing the meaning of the verb). This is not the case either: the thematic grid remains the same, irrespective of whether V does or does not display redundant se (e.g., răde ‘laugh’ is constantly unergative). The mismatch between our data and the analysis in (17), together with its predictions, is unsurprising considering that the tests we performed indicate that the unergative frame for râde ‘laugh’ and greşi ‘err’ (as well as the transitive frame for risca ‘risk’) remains unchanged in the presence of se. So the configuration in (17) is unsuitable to our data.

5.3.2

MacDonald’s Studies

MacDonald (2004, 2008, 2017) and MacDonald and Huidobro (2010) also employ a complex predicate analysis, focusing on the on/with denotation introduced by non-argumental se constructions (labeled Aspse). If the predicate is eventive, there is obligatory telicity in these constructions, as shown in (18a). However, as MacDonald (2017) also shows, stative Aspse does not give rise to a telic VP with the same Aspse. (18)

a.

b.

Juan se lavó el coche en/# durante una hora an hour Juan ASPse.3 SG washed the car in/ for ‘Juan washed the car in/# for an hour’ Juan se sabe la lección desde el martes Juan ASPse.3 SG knows the lesson since the Tuesday ‘Juan has known the lesson since Tuesday.’ (MacDonald 2017, ex. 9a)

In MacDonald (2017), Aspse instantiates a type of a double-object construction, where se functions as an indirect object reflexive. The relevant structure is illustrated

174

M. A. Irimia and V. Hill

in (19), where pro is base generated as a complement of a prepositional head merged with V.6 (19) [VoiceP DPi Voicese [ApplP proi App [VP DP V [PP PØ proi]]]] (MacDonald 2017, ex.5)

In (19), V and the null prepositional phrase form a complex predicate, to which the object is introduced compositionally. Another important component of the analysis is that pro further moves to a Spec, Appl through a process similar to indirect object raising. Se is located in Voice, and is coreferential with pro in Spec, Appl. One important result is that pro in Spec, Appl is predicted to behave like run-of-the-mill indirect objects under a variety of diagnostics. MacDonald (2017) demonstrates that this is indeed the case by analyzing a leista variety where PCC effects are observed in this construction PCC effects have been connected to animacy, in the sense that two arguments with a [+Person] feature compete for valuation in a configuration that contains only one licensor (see, for example, the detailed discussion in Ormazabal and Romero 2007). Given the humanness restriction we see with redundant se in Romanian, which is always personal, we can proceed to an approach of the Romanian data in light of the Aspse configuration in (19). The first problem is that, as explicitly mentioned in MacDonald (2017), Aspse itself is unspecified for animate features. Thus, the PCC effect arises only when an animate object is added into the derivation (as shown by Ormazabal and Romero 2007, in the relevant leista dialect the morphological shape of animate direct object clitic is homophonous with that of the dative). The examples we are discussing, on the other hand, point to the conclusion that personal se is restricted to an animacy specification. This is clearly demonstrated in the examples (25)–(26) discussed in the next section. This type of sensitivity strengthens the unergative status of the predicates we are examining. Furthermore, although MacDonald’s (2017) analysis can avoid the problems with antipassivization, it is still not adequate for Romanian from other points of view: (1) Romanian does not exhibit the ‘on/with’ entailment, hence a P with this type of semantics is not warranted; (2) there is no DP object; (3) telicity is systematically absent in Romanian, although the predicate râde cannot be analyzed as stative; (4) as we noticed, the clitic varies between an Accusative and a Dative form, and both preserve the humanness restriction. Thus, animacy cannot be equated with the indirect object status of se. All these observations indicate that the two constructions have radically distinct natures, and Aspse cannot be equated with the non-selected se under scrutiny in this paper.

6

The presence of the PP can capture the similarities between the aspectual properties of Aspse and telicizing particles.

Personal se with Unergatives in Romanian

175

A few remarks are also in order with respect to another possible analysis. The observation that redundant se is obligatorily linked to humanness also demonstrates that an account connecting its structure to that of unergative impersonal se, as discussed by Dobrovie-Sorin (1998) cannot be on the right track either—the constructions are clearly different. It is however, not surprising that redundant se exhibits these special properties which are very hard to accommodate under most accounts available on the market. The general pattern of the construction is an old Slavic one, as mentioned in the first section. The analyses described above, however, are mostly restricted to Romance data. Given that Romanian displays a typological mix, it is not surprising that the data cannot be accommodated under a unitary pattern.

6 Derivations For ease of presentation, in this paper we consider that a verb is transitive if it has an internal (direct object) and an external (subject) argument; unergative if it has only an external argument; and unaccusative if it has only an internal argument. Other positions are possible with all these verbs for mapping theta-roles, but they are orthogonal to the verb classification. We also assume that external arguments are DPs/CPs merged in the Spec, vP, whereas internal arguments are DPs/CPs merged as sisters to V. Along these lines, we find that (se) râde ‘laugh’ is consistently unergative, although it allows for variation with respect to an adjunct PP-de.

6.1

Râde without se

For this active verb, the tests of the previous sections indicated an unergative structure: it fails passivization and telicity and there is no se to justify alternative analyses. In its involuntary reading, this verb is monadic and maps its theta-role (the experiencer) to the subject position, as in (20b), which underlies the clause in (20a). (20) a.

b.

Elena râde Helen laughs ‘Helen is laughing’ [vP DP-subject [v râde [VP < râde>]]]

The voluntary reading of râde involves the semantic addition of willful intent, amounting to a diadic structure, with an agent and a goal. The agent is mapped to the subject position and is checked by a directly merged DP. The goal is necessarily a PP, since the verb remains unergative. We consider the PP-de an adjunct to VP, since the subject in situ can c-command it and bind its anaphoric items, as in (21c).

176

M. A. Irimia and V. Hill

(21) a.

Elena râde de mine Helen laughs at me ‘Helen is laughing at me’

b. [vP DP-subject [v râde [VP PP-goal [VP < râde>]]] c. Râdea Elena de sine însăşi laughed Helen at her self ‘Helen was laughing at herself’ These configurations are cross-linguistically unsurprising, at least for Romance languages.

6.2

Dative Reflexive râde

In this version, râde is triadic: the beneficiary of the willful intent is not just inferred as in (21), through the principle of compositionality, but it is marked through a dedicated theta-role. The beneficiary theta-role is mapped as an indirect object, within a Larson (1988) shell, as in (22b), which underlies (22a). (22) a.

b.

Mirceaj îşij râde de elk Mircea se.DAT laughs at him ‘Mircea laughs at him’ [vP Mircea [v´ îşi [v râde [VP [PP de el] [VP râde]]]]]

Alternatively, the goal theta-role can be mapped through an indirect object instead of a PP-de. This is shown in (23a), where Mariei is the Dative marked indirect object that undergoes clitic doubling. In this case, the Dative reflexive and the Dative DP/indirect object are in complementary distribution: although they map different theta-roles, they compete for the same syntactic position. (23) a.

b.

Mircea îik râde Marieik în faţă Mircea her.DAT laughs Maria.DAT in face ‘Mircea laughs at Maria to her face’ *Mirceaj îşij îik râde Marieik în faţă Mircea se.DAT her.DAT laughs Maria.DAT in face

Therefore, there is no indication that the unergative status of the verb has changed from (21) to (22). Variations can occur with respect to whether the beneficiary/ maleficiary is mapped, but they do not involve the direct object position, which remains absent.

Personal se with Unergatives in Romanian

6.3

177

Râde with Redundant Se: Human Subjects and CLLD

Section 5.2 showed that the tests for an antipassive analysis fail for se râde with either voluntary or involuntary readings. Hence, the question is where does se merge and why? In this section, we show that redundant se is not free; instead, it is conditioned by the [human] feature of the subject. Furthermore, the human subjects have a topic/focus reading, hence they tend to be preverbal. A clarification is in order with respect to the position of subjects in Romanian (for an ample discussion, see Alboiu 2002). This is a VSO language that spells out the subject in situ, in Spec, vP, irrespectively of the noun class of the subject. SVO is derived from subject movement to discourse positions in the CP (i.e., to TopP or FocP). Other word orders are possible, according to constituent or verb movement to the discourse field, which is prolific in this language and may involve CLLD among other operations. Against this background, the first observation is that the presence of redundant se coincides with an obligatorily human subject for the unergative verb. For example, râde ‘laugh’ may have either human or non-human subjects in its active form, as shown in (24). However, when redundant se is present, only human subjects are allowed, as in (24a), whereas non-human subjects reject se, as in (24b, c). In (24d), we see that a non-human subject is compatible with reflexive structures in general (here, an antipassive), so the problem in (24b, c) does not arise from reflexivization but from the special status of se, which is not reflexive. (24) a. Studenţii se râd de profesori students.the SE.3ACC laugh.PL at professors ‘The students are mocking the professors’ b. (*Se) râde ciob de oală spartă SE.3ACC laughs crock at vase broken Intended: ‘The cracked crock laughs at the broken vase.’ (proverb) c. Primăvara (*se) râde din nou spring.the SE.3ACC laughs of new ‘The spring is laughing again’ d. Primăvara se mândreşte cu lalele spring SE.3ACC prides with tulips ‘The spring boasts its tulips’ The same results are replicated with greşi ‘err’ in (25) and risca ‘risk’ in (26), so the constraint is systematic. (25) a.

b.

Mircea (s)a greşit Mircea SE.3ACC has erred ‘Mircea has erred’ Calculatorul (*s)a greşit computer.the SE.3ACC has erred

178

M. A. Irimia and V. Hill

c.

‘The computer has erred’ Calculatorul sa referit la datele de ieri computer.the SE.3ACC has referred at data.the of yesterday ‘The computer referred us to the yesterday data’

(26) a.

Alpiniştii (se) riscă să cadă de pe stâncă climbers SE.3ACC risk SBJV fall from on mountain ‘The climbers risk to fall from the mountain’ b. Piatra (*se) riscă să cadă de pe stâncă rock.the SE. 3ACC risk SBJV fall from on mountain ‘The rock risks to fall from the mountain’ c. Piatra se eliberează de hingile care o ţineau rock.the SE.3ACC free from straps.the that her held ‘The rock freed itself from the straps that held it’

These data indicate that [human] is a formal feature that constrains the merging of redundant se in narrow syntax. The second observation is that the human subject is overwhelmingly preverbal and has a topic or focus reading. For example, out of the first 100 different occurrences of se râde ‘laugh’ that appeared on the screen in an internet search, we found only one with VS, whereas the same amount of occurrences with râde ‘laugh’ without se had around 50% VS. In fact, VS may give ungrammatical results, as in (27a), or, when accepted, it entails a topic/focus reading on the low subject, as in (27b, c). We graphically signal the emphatic reading through commas around the low subject, although breaks in intonation do not apply. (27) a. b.

c.

(*?S)-a greşit Mircea SE.3ACC -has erred Mircea Se râd, studenţii, de profesori SE.3ACC laugh students.the at professors ‘The students laugh at the professors’ Se riscă, alpiniştii, să cadă de pe stâncă SE.3ACC risk climbers.the SBJV fall from on mountain ‘The climbers risk falling from the mountain’

In (27), the low linearization of the subject does not entail its location in Spec, vP, since the interpretation indicates a discourse position. It is rather the case that the verb or constituents move above TopP/FocP, whereas the subject is constantly in Spec, Top/FocP. The intonation on the preceding material is not neutral either. The non-argumental, discourse status of subjects in the presence of redundant se can also be checked by a test with non-D-linked indefinites. As pointed out in Alboiu (2002), in Romanian such subjects remain in situ since they are incompatible with a

Personal se with Unergatives in Romanian

179

topic reading. This scenario yields good results with unergative verbs only if redundant se is not involved, as shown in (28).7 (28) a.

b.

(*Se) râd studenţi prin holurile facultăţii SE.3ACC laugh students through halls.the faculty.the.GEN ‘There are students laughing through the hall ways in the university’ (*Se) greşesc contabili şi directori SE.3ACC err accountants and directors ‘Accountants and directors may err’

The restrictions on se in (28) confirm that, even when a VS linearization occurs with redundant se, as it was seen in (27), the subject is not in situ and the reading is not neutral.

7 Proposal The data and tests discussed in this paper indicate that redundant se is tightly related to the licensing of subjects/external arguments: only human subjects are licensed in Spec, vP in the presence of redundant se (and then move). This finding is in line with the observation made in Sect. 2, namely that redundant se can be seen with unergatives, but not with unaccusatives, as further indicated in (29). (29) a.

b.

*Elena s-a căzut pe scări Elena SE.3ACC-has fallen on stairs Intended: ‘Elena fell on the stairs’ *Elena s-a venit la Paris Elena SE.3ACC-has come to Paris Intended: ‘Elena came to Paris’

Unaccusative verbs do not license DPs as external arguments in Spec, vP, so the absence of redundant se with this class of verbs is expected, if its trigger depends on DPs merged as external arguments (in Spec, vP). The problem is understanding the connection between [human] and external arguments in the presence of redundant se. There are precedents in the literature for identifying a formal feature [animate/human] associated with the vP/VP (see feature lists in Wiltschko 2014). Thus, the straightforward approach is to consider that redundant se spells out [human] on v, irrespective of what may have triggered the mapping of this feature. This would be in line with several studies on inherent reflexives that argue for the merge of se in v as a verbal operator/morpheme

7

This configuration is hard to obtain with the active form of ‘risk’, so we did not include it here.

180

M. A. Irimia and V. Hill

(Zubizarreta 1987), or as a Voice marker (Folli and Harley 2005; Labelle 2008; Basilico 2010; Armstrong 2013; MacDonald 2017). Following this path, we can argue that redundant se merges in Voice/v head to check [human] and agrees with the DP in the local Spec, vP, which checks the [theta-role] feature. This se has nothing to do with the aspectual feature of Voice/v, whose values are determined by the properties of the verb (e.g., no telicity with unergatives, and se is ineffectual for changing this situation). This yields the configuration in (30). (30) [vP DPj [v sej [VP. . .]]] However, (30) fails to predict two important properties of redundant se: 1. The exclusion of unaccusatives remains unjustified: a human DP should be able to move from object to subject position and further to Spec,TopP, while licensed by se. 2. The obligatory topic/focus reading on the subject would remain unjustified, and this is a very strong requirement. In light of these problems, we explore an alternative analysis, which takes into consideration the topic/focus reading of the human subject. In Romanian, like in other Romance languages, DP arguments (i.e., direct and indirect objects) that receive such readings undergo Clitic Left Dislocation (CLLD) (see Delfitto 2002 for theoretical justification). It is, then, reasonable to suspect that redundant se also mediates CLLD for the DP argument, which would account for the reading peculiarities as well as for the fact that this se does not trigger reflexivization. In this respect, it is important to notice that reflexive pronouns can always (given appropriate contexts) fulfil double tasks, for reflexivization and for CLLD or clitic doubling (CD), the latter being a sub-case of CLLD (Delfitto 2002).8 This transpired earlier in our examples and is further illustrated in (31). (31) a.

b.

Gena se acuză pe sine însăşi Gena SE.3ACC accuses DOM her self ‘Gena accuses herself’ Pe sine însăşi se acuză Gena (nu pe altcineva) DOM her self SE.3ACC accuses Gena (not DOM someone.else) ‘It is herself that Gena accuses, not someone else’

8 The [agr] feature on T involved in CLLD reflects discourse agreement, and is thus different from phi-agreement. For Delfitto (2002), Top (in the CP field) has a predication [Pred] feature reflected through [agr] on T, and the clitics spell out this [agr] feature. For Miyagawa (2010), the discourse features are a type of agreement at C, so they are subject to C-to-T feature transfer, which may or may not happen cross-linguistically. This analysis has the advantage of not discriminating between Top and Foc when it comes to CLLD, which suits the Romanian data.

Personal se with Unergatives in Romanian

181

In Modern Romanian, differential object marking (DOM), as in (31a), obligatorily triggers CD. Thus, in (31a), se is responsible not only for reflexivization but also for ensuring the needed CD. Furthermore, (31b) shows CLLD of the DOM-ed object to FocP, for which no supplementary clitic pronoun is merged, which means that se also mediates CLLD. Along these lines, redundant se performs only one of the two tasks reflexive pronouns are used for in Romanian grammar: that is, it specializes for CLLD/CD while being inactive for reflexivization. This specialization arises upon reanalysis of Slavic reflexive unergatives in a Romance pattern, where unergatives do not reflexivize: in such contexts, se is reanalyzed as a spellout of discourse [agr] that mediates CLLD, with no further reflexivizing function. Since the subject DP is the only argument with unergative verbs, se and CLLD are forced on this DP. The unavoidable consequence of this analysis is that redundant se must conform to the syntactic pattern of clitic personal pronouns that regularly trigger CLLD. According to Delfitto (2002), clitic personal pronouns select the DP argument, so together they form a nominal constituent, e.g., Dmax. If redundant se follows this pattern, then it merges in the DP domain, not in the vP domain, as in (32). (32) a. b.

[Dmax [D se [DP . . .]]] [vP Dmax [VP. . .]]

In (32), Dmax checks [human] on v upon merge, and undergoes CLLD mediated by se. With objects, Dmax receives Case upon merge with V, and this is reflected in the choice of the clitic form, that can be either Accusative or Dative. With the subject in (32), Dmax in Spec, vP receives Nominative Case through long distance Agreement with T (Alboiu 2002), and that automatically extends to se. Hence, the Accusative form of se is irrelevant to the actual structural Case assigned to it in syntax. Romanian has no Nominative Case forms for clitic pronouns in the morphological paradigm, so there is no other choice for the spell out of se at PF than the Accusative form.9 From this perspective, (32) allows for a configuration where the CLLD of subject DPs has a spell-out for the mediating [Agr] on T, which is unusual in the language. The default situation in Romanian is that subject DPs are not doubled by clitics when moving to Top/FocP, because there are no Nominative clitic personal pronouns in the morphological paradigm. Crucially, this default configuration does not constrain the type of subjects or the type of verbs involved in the CLLD. On the other hand, CLLD through redundant se tightly restricts the type of subjects to those with a human feature and the type of verb to unergatives (the few transitive verbs are

9 Some Italian dialects display clitic doubling with subjects. However, those subject clitics spell out phi-agr for lexical subjects that occupy an argumental Spec, TP position, whereas redundant se spells out discourse-agr for lexical subjects that occupy a non-argumental Spec, TopP position. Therefore, comparison is not warranted.

182

M. A. Irimia and V. Hill

derived from unergative frames). So constructions with redundant se display a type of marking on human subjects. Different justifications are proposed in the literature for the speaker’s tendency to syntactically mark human arguments in Romanian (see Mardale 2015 for an overview); e.g., to differentiate subjects from objects in a VSO language; to confer semantic prominence to the respective argument in relation to other arguments; to promote the argument on the agentivity scale; to ensure referential stability. It is not clear that any of these trigger redundant se, since there is no generalized subject marking in Romanian, and the phenomenon is limited to a sub-class of unergative verbs. One may say that this is a dying construction, insofar as many of the Slavic borrowings were lost towards Modern Romanian, and each of these verbs is also productive without se. However, as we have already mentioned, we do see the extension of redundant se to the new French borrowing ‘risk’, so it may be the case that redundant se has sufficient occurrences in the primary linguistic data, in the colloquial register, to justify its low but steady spreading.

8 Conclusions This paper aimed to identify the underlying structures of Romanian constructions with unergative verbs in the presence of the reflexive pronoun se. The tests on se râde ‘laugh’ showed that this se does not affect the argument structure or the aspectual properties of v (telicity results are unchanged), nor does it have any impact on the interpretation of the verb. Further tests showed that constructions with se râde do not qualify as antipassive or as part of complex predicates proposed for non-argumental se. In fact, the data indicated that this is a redundant se, not a reflexivizing se, which was also noticed in traditional grammar. The tests also established that redundant se correlates with human subjects that undergo CLLD. Hence, [human] was identified as a formal feature on v/Voice and two approaches to redundant se were considered: one has se merged directly on v to check [human], the other has se merged in Dmax and spells an [Agr] feature on T. The latter analysis was adopted, since it explains the obligatory CLLD with the relevant subject DPs. Theoretically, this analysis points out that se is not necessarily involved in reflexivization, nor in the spell out of features on little v. Se, in fact, may also occur as a property of nominal domains, where it functions on a par with clitic personal pronouns, for CLLD, without any involvement with theta-grids, Case assignment or aspectual modification. This type of nominal merge can explain a gap in the syntactic distribution of se: Most accounts merge this element in Voice/v or other verbal projections (see, for instance, García Pardo [this volume], Pineda [this volume], Vivanco and Basilico [all this volume]) but it is not clear what theory internal requirement would block its merge DP-internally. The examples we have analyzed could be taken to fill in this distributional gap.

Personal se with Unergatives in Romanian

183

References Al Zahre, Nisrine and Nora Boneh. 2010. Coreferential dative constructions in Syrian Arabic and Modern Hebew. Brill’s Annual of Afroasiatic Languages and Linguistics 2: 248–282. http:// pluto.huji.ac.il/~bonehn/CD_AlZahre_Boneh.pdf/. Accessed 18 Jan 2015. Alboiu, Gabriela. 2002. The features of movement in Romanian. Bucharest: University of Bucharest Press. Alboiu, Gabriela, Michael Barrie, and Chiara Frigeni. 2004. SE and the unaccusative-unergative paradox. In Antwerp papers in linguistics 107, ed. Martine Coene, Gretel de Cuyper, and Yves D’Hulst, 109–139. Antwerp: Universiteit Antwerp. Alsina, Alex. 1996. The role of argument structure in Grammar. Evidence from Romance. CSLI lecture notes 62. Stanford: CSLI Publications. Armstrong, Grant. 2013. Agentive reflexive clitics and transitive SE constructions in Spanish. Borealis: An International Journal of Hispanic Linguistics 2 (2): 81–128. Baker, Mark. 1988. Incorporation. A theory of grammatical function changing. Chicago: University of Chicago Press. Basilico, David. 2010. The se clitic and its relationship to paths. Probus 22: 271–302. ———. 2017. Russian–sja as a verbalizer. In Proceedings of the 34th west coast conference on formal linguistics, ed. Aaron Kaplan et al., 62–69. Somerville: Cascadilla Proceedings Project. Delfitto, Dennis. 2002. On the semantics of pronominal clitics and some of its consequences. Catalan Journal of Linguistics 1: 29–57. Dobrovie-Sorin, Carmen. 1998. Impersonal se constructions in romance and the passivization of unergatives. Linguistic Inquiry 29: 399–437. Folli, Raffaella, and Heidi Harley. 2005. Consuming results in Italian and English: Flavors of v. In Aspectual inquiries, ed. Paula Kempchinsky and Roumanya Slabakova, 95–120. Dordrecht: Springer. Hale, Kenneth, and Samuel J. Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge: MIT Press. Labelle, Marie. 2008. The French reflexive and reciprocal se. Natural Languages and Linguistic Theory 26: 833–876. Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19: 335–391. MacDonald, Jonathan E. 2004. Spanish reflexive pronouns: A null preposition hypothesis. In Proceedings of WCCFL 23, ed. Vineeta Chand, Ann Kelleher, Angelo J. Rodríguez, and Benjamin Schmeiser, 528–540. Somerville: Cascadilla Press. ———. 2008. The syntactic nature of inner aspect: A minimalist perspective. Amsterdam/Philadelphia: John Benjamins. ———. 2017. Spanish aspectual se as an indirect object reflexive: The import of atelicity, bare nouns and the PCC. Probus 29: 73–118. MacDonald, Jonathan E. and Susana Huidobro. 2010. The lack of Spanish nonargumental clitic doubling, ed.Claudia Borgonovo, Manuel Español-Echevarría, and Philippe Prévost. Selected Proceedings of the 12th Hispanic Linguistics Symposium. 50–62. Somerville, MA: Cascadilla Proceedings Project. Mardale, Alexandru. 2015. Differential object marking in the first original Romanian texts. In Formal approaches to old Romanian DP, ed. Virginia Hill, 200–245. Leiden: Brill. Masullo, Pascual José. 1992. Antipassive constructions in Spanish. In Romance languages and modern linguistic theory. Selected papers from the XXth linguistic symposium on romance languages, ed. Paul Hirschbuhler and Konrad Koerner, 75–195. Amsterdam: John Benjamins. Medová, Lucie. 2009. Reflexive clitics in the Slavic and Romance languages. A comparative view from an antipassive perspective. PhD dissertation, Princeton University. Miyagawa, Shigeru. 2010. Why agree? Why move? Unifying agreement-based and discourse configurational languages. Cambridge: MIT Press. Ormazabal, Javier, and Juan Romero. 2007. The object agreement constraint. Natural Language and Linguistic Theory. 25: 315–347.

184

M. A. Irimia and V. Hill

Pană Dindelegan, Gabriela. 2014. Variaţie în construcţia verbului în româna veche. In Diacronie si sincronie în studiul limbii române, ed. Rodica Zafiu, Adina Dragomirescu, and Alexandru Nicolae, vol. 1, 155–175. Bucharest: EUB. Reinhart, Tanya, and Tal Siloni. 2005. The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36: 389–436. Rizzi, Luigi. 1990. Relativized minimality. Cambridge: MIT Press. Rooryck, Johan, and Guido Vanden Wyngaerd. 2011. Dissolving binding theory. Oxford: Oxford University Press. Sportiche, Dominique. 2014. Assessing unaccusativity and reflexivity: Using focus alternatives to decide what gets which theta-role. Linguistic Inquiry 45 (2): 305–321. Vendler, Zeno. 1957. Verbs and times. The Philosophical Review. 66 (2): 143–160. Wiltschko, Martina. 2014. The universal structure of categories. Cambridge: Cambridge University Press. Zubizarreta, Maria Luisa. 1987. Levels of representation in the lexicon and in syntax. Dordrecht: Foris.

Part III

Voice/Little v and Below

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John washes his hands’ Alexandra Cornilescu and Alexandru Nicolae

Abstract The aim of the paper is to discuss a class of transitive-prepositional reflexive constructions in Romanian, tied to the sphere of inalienable possession. It is argued that their properties are best explained by the presence of a prepositional phrase ( pP), generated as the complement of V, which encodes a figure-ground relation between a possessor and an inalienably possessed object. Transitive variants of these constructions have a thematic Voice head and the reflexive clitic is merged as the figure argument and bound by the DP in spec, Voice. Unaccusative variants, on the other hand, have an expletive Voice head. In these cases, the DP subject merges in the position of the figure argument and raises to Tense while the reflexive clitic is a purely case-absorbing element that does not saturate a theta role. Keywords Figure-ground · Expletive voice · Derived Unaccusatives · Reflexive constructions · Inalienable possession · Romanian

1 Introduction 1.1

Figure Reflexive Constructions

The aim of the paper is to discuss a class of transitive-prepositional reflexive constructions, tied to the sphere of inalienable possession. The characteristic property of this pattern is the sometimes implicit, but often overt and obligatory presence of a locative PP which embeds a body-part nominal in a VP which is otherwise monotransitive. Typical verbs occurring in this pattern are those in (1); the class can be extended by taking into consideration ‘extended inalienable possession’ nouns—

A. Cornilescu · A. Nicolae (*) University of Bucharest, Bucharest, Romania “Iorgu Iordan—Al. Rosetti” Institute of Linguistics, Bucharest, Romania e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_8

187

188

A. Cornilescu and A. Nicolae

i.e., nouns which represent metonymic extensions of body parts (see (2)) and which may also appear inside the PP. Metaphorical extensions in idiomatic constructions are also available. (1)

a spăla ‘wash’, a tăia ‘cut’, a lovi ‘hit’, a scărpina ‘scratch’, a scobi ‘scoop’,a freca ‘rub’, a împușca ‘shoot’, a curăţa ‘clean’, a defecta ‘spoil, go bad’, a răni ‘to injure’, a bate ‘beat’, a strica ‘spoil’, sterge ‘dry, wipe’, a arde ‘burn’, a frige ‘burn, fry’, a înţepa ‘pierce’, a uda ‘get wet, water’, etc.

(2)

cămașă ‘shirt’, mânecă ‘sleeve’, rochie ‘dress’, palton ‘overcoat’, bluză ‘blouse’, etc.

Here are some relevant examples. In (3) the PP is optional,1 in (4) it is obligatory, in (5) the PP is obligatory and the interpretation is idiomatic. (3) a.

S-a SE.ACC¼AUX.PERF.3SG

b.

c.

(4) a.

b.

c.

(5) a.

1

spălat (pe spate/pe mâini/ pe obraz/ washed on back on hands on cheek

în urechi) in ears ‘He washed his back/his hands/his cheek/his ears’ S-a tăiat (la deget/ pe obraz) SE.ACC¼AUX.PERF.3SG cut at finger on cheek ‘He has cut his finger/his cheek’ Ion se vopsește (la păr) Joh SE.ACC dyes at hair ‘John dyes his hair’ Ion s-a stricat la stomac John SE.ACC¼AUX.PERF.3SG spoiled at stomach ‘John (has) upset his stomach’ *Ion s-a stricat John SE.ACC¼AUX.PERF.3SG upset Ion se freacă la ochi/*Ion se freacă John SE.ACC rubs at eyes John SE.ACC rubs ‘John is rubbing his eyes’ Ion se curăţă în urechi /*Ion se curăţă John SE.ACC cleans in ears John SE.ACC cleans ‘John is cleaning his ears’ Ion s-a stricat la cap John SE.ACC¼AUX.PERF.3SG gone.bad at head ‘John has gone mad’

Some verbs are specified for particular body parts and therefore the overt expression of the body part is not required and may be understood in context: a se rade (pe obraz) (‘shave (one’s cheek)’), a se vopsi (la păr) (‘dye (one’s hair)’).

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

189

*Ion s-a stricat John SE.ACC¼AUX.PERF.3SG gone.bad b. Bate-te peste gură! beat.IMP.2SG¼CL.REFL.2SG.ACC on mouth (approx.) ‘Touch wood!’ c. Ion se bate (cu pumnii) în piept John SE.ACC beats (with fists) in breast ‘John is showing off’ The set in (6) features examples of extended inalienable possession; the PP is obligatory in this configuration. (6) a. Ion s-a încheiat la cămașă John SE.ACC¼AUX.PERF.3SG buttoned at shirt ‘John buttoned up his shirt’ b. Maria se desface la bluză Mary SE.ACC undoes at shirt ‘Mary is undoing the buttons on her blouse’ Many of the properties of this construction have been identified by Manoliu-Manea (1993, 1996). In agreement with Talmy (1978), she establishes the important result that at a more abstract conceptual level, the reflexive SE and the PP stand in a FigureGround configuration, hence the term ‘figure-reflexive’ used in this paper and elsewhere in the literature (e.g., Wood’s 2012 discussion on Icelandic). In this interpretation, the reflexive possessor is conceptualized as a Figure, i.e., as a more dynamic entity which undergoes a change along some dimension (e.g., in an example like (1a), it gets to be cleaner by washing). On the other hand, the locative PP is a Ground, i.e., a reference entity, specifying which part or aspect of the Possessor is operated on.

1.2

Goals of the Paper

The paper further examines the properties of the construction and proposes a syntactic interpretation of the data. From a syntactic perspective, two properties stand in need of an explanation: (1) the first is the status of the PP (an argument or an adjunct), given that the verb is monotransitive; (2) the second is to explain how the possessive interpretation arises in a structure where a possessive constituent is not present, and, as will be seen, is not welcome either. The second aim of the paper is to extend the category of figure reflexives from transitive to unaccusative verbs. We will mostly discuss change of state constructions (Ion s-a albit la față ‘John turned white in the face’), but the analysis may be naturally extended along these lines to change of location constructions (Ion s-a urcat în copac să culeagă cireșe ‘John climbed the tree to pick cherries’).

190

1.3

A. Cornilescu and A. Nicolae

Proposal

Figure reflexive constructions are unified by the fact that their internal argument is conceptualized as a prepositional small clause, which introduces the Figure-Ground relation. In the analysis we propose that all of these constructions are also Voiced, in the sense that an external argument position is projected even if it may be non-thematic. In other words, the proposal is that transitives and SE unaccusatives differ only in the kind of Voice head (thematic or expletive) that they project (Schäfer 2008). What looks like the internal argument of the transitive verb is actually the subject of the Figure-Ground small clause. Transitive verbs project a thematic Voice head which introduces an Agent in the structure. Unaccusatives introduce an expletive, i.e., non-thematic, Voice projection, whose role is simply to signal that the event structure of transitives and derived unaccusatives is the same, consisting of a causing subevent (an activity) and a result state. The discussion of unaccusatives is theoretically relevant in that it strengthens the suggestion that there is a connection between the expletivization of Voice and derived reflexive unaccusativity. Change of state unaccusatives share the syntactic configuration of transitives, differing through the presence of expletive Voice. This assumption is sufficient to account for the difference between basic and derived reflexive unaccusatives. We thus agree with Wood and Marantz (2017) who suggest that all argument-introducing heads have thematic and expletive variants. The paper is structured as follows: in Sects. 2 and 3 we discuss the transitive reflexive construction (its empirical properties and its syntactic analysis), followed by change of state unaccusatives in Sect. 4; Sect. 5 is devoted to the anticausative alternation.

2 Transitive Figure Reflexives: Empirical Properties 2.1

The Dative Paraphrase

Most of the inalienable possession accusative-reflexive structures above (e.g., (7a)) may be paraphrased by a transitive possessive dative-(reflexive) construction (7a’), generally available in Romance, but extensively used in Romanian to cover inalienable, as well as alienable possession. Romanian is also different in that it possesses specific dative reflexive morphology, while elsewhere in Romance the same SE/SI clitic is used (see MacDonald 2015 for a recent discussion). Accusative figurereflexive idioms usually resist this dative alternation, as shown by the unacceptability of (7b’) below. However, when the dative paraphrase is possible, the two alternatives are truth-conditionally equivalent, but they are used in different contexts, and evince semantic and pragmatic properties which follow from their very different syntax. Some of these differences will be discussed below.

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

Ion se spală pe mâini John SE.ACC washes on hands a’. Ion își spală mâinile John SE.DAT washes hands.DEF ‘John is washing his hands’ b. Ion s-a stricat John SE.ACC¼AUX.PERF.3SG gone.bad ‘John has gone mad’ b’. *Ion și-a stricat John SE.DAT¼AUX.PERF.3SG gone.bad

191

(7) a.

la cap at head capul head.DEF

The dative reflexive is, thematically, a Possessor and, in alienable possession contexts may be replaced by a DP-internal genitive Possessor without destroying grammaticality, as in (8b). The Possessor dative is not part of the argument structure of the verb, rather, it is launched by the direct object Possessee (Landau 1999). (8) a. b.

2.2

Ion îşi rezolvă problemele cu John SE.DAT solves problems.DEF with Ion rezolvă problemele lui cu John solves problems.DEF his.GEN with ‘John solves his problems successfully’

succes success succes success

Transitivity

The dative structure in (7a’) or (8a) is clearly transitive and its SVO structure shows the verb to be monotransitive (as explained, the dative is not an argument of the verb). While in French and Italian SE/SI is the hallmark of an unaccusative construction, as shown by auxiliary selection (Fr. être/It. essere in reflexive constructions), in Romanian, some of the uses of SE are transitive, i.e., SE is a marker of reflexivization, showing that the two arguments of the transitive verb are coreferential, without, however, reducing one of the two arguments. As suggested above, we claim that the accusative figure-reflexive construction is equally monotransitive, except that its internal argument is the small clause which encodes the Figure-Ground configuration. The reflexive pronoun merges as the subject of the small clause and raises into the main clause, becoming a derived direct object. Since Romanian is a clitic-doubling, DOM language, the derived direct object clitic may be doubled by a strong emphatic (9b) or reflexive pronoun (9c). The possibility of DOM and clitic doubling proves that the main verb is transitive beyond reasonable doubt.

192

A. Cornilescu and A. Nicolae

(9) a.

Ion se poate încă spăla el însuși pe picioare John SE.ACC can still wash.INF he himself on feet ‘John himself is still able to wash his feet’ Ion nu se mai poate spăla pe mâini John not SE.ACC still can wash.INF on hands nici pe el însuși, nici pe altcineva neither DOM him himself nor DOM someone.else ‘John is no longer able to wash his own hands, let alone someone else’ Ion nu se mai poate spăla pe sine / singur John not SE.ACC still can wash.INF DOM himself by.himself nici pe mâini, darămite pe spate not.even on hands let.alone on back ‘John his no longer able to wash his hands on his own, let alone wash his back’

b.

c.

2.3

Agentivity

The subject is agentive, in as much as agent-oriented adverbs (10a), instrumental adjuncts (10b) and control into infinitive purpose clauses (11) are all possible. However, intentionality is not required (12a), as has also been noticed for other transitive structures. Lack of intentionality is more obvious when one compares the SE construction with its dative paraphrase: intended action is expressed by the dative structure (12b), not the accusative one (12a). (10)

a.

Ion nu s-a spălat de bunăvoie John not SE.ACC¼AUX.PERF.3SG washed willingly pe spate, pentru că îl durea rana on back because CL.ACC.3SG.M hurt.IMPERF.3SG wound.DEF ‘John didn’t willingly wash his back, because his wound still hurt him’ b. Ion s-a spălat pe cap cu șampon John SE.ACC¼AUX.PERF.3SG washed on head with shampoo ‘John washed his hair with shampoo’ (11) Ion s-a spălat pe cap John SE.ACC¼AUX.PERF.3SG washed on head pentru a face o impresie bună Mariei for A.INF make.INF an impression good Mary.DAT ‘John washed his hair to impress Mary’ (12) a. Ion s-a rănit John SE.ACC¼AUX.PERF.3SG wounded la picior fără să vrea at leg without SĂ.SUBJ want.SUBJ.3SG ‘John unwillingly hurt his leg’

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

??Ion și-a rănit John SE.DAT¼AUX.PERF.3SG wounded piciorul fără să vrea leg.DEF without SĂ.SUBJ want.SUBJ.3SG

b.

2.4

193

Restriction to Inalienable Possession

The accusative Figure-Ground construction is restricted to inalienable and extended inalienable possession (13), sharply contrasting with the dative construction which accommodates any possessive relation (14). With the accusative Figure-Ground construction, non-alienable possession nouns are out. (13)

a.

b.

(14)

a.

b.

*Ion se spală pe mașină / la maiou John SE.ACC washes on car at T-shirt intended: ‘Johnwashes his car /T-shirt’ *Ion se defectează la mașină John SE.ACC spoils at car intended: ‘John spoils his car’ Ion își spală mașina/maioul John SE.DAT washes car.DEF shirt.DEF ‘John washes his car / T-shirt’ Ion își defectează mașina în fiecare iarnă John SE.DAT spoils car.DEF in every winter ‘John spoils his car every winter’

The more restricted domain of the accusative figure-reflexive construction will be shown to follow from its constrained syntax.

2.5

Possession Expressed Metonymically

As already mentioned, the accusative figure-reflexive construction shows no overt possessive or genitive constituent. In particular, the body part in the PP cannot be modified by a possessive or genitive constituent (15a). Inherent possession is however configurationally expressed in the small clause, through the metonymic relation (inclusion) that holds between the accusative clitic, expressing the Possessor/Figure, and the locative PP body-part expressing the Possessee/Ground (see also Niculescu 2013, 190). Yet, the essential property of the construction is the integration of the small clause in the thematic and conceptual structure of the main verb, through raising of the small clause subject into the derived object position of the main verb (see below). This allows the small clause subject (i.e., the Possessor/

194

A. Cornilescu and A. Nicolae

Figure clitic) to establish a more local head-complement connection with the main verb (cf. also the following English contrast: ?? He cut on the chin/He cut himself on the chin). The Possessor is foregrounded through raising and must always be overtly expressed. The body-part Possessee is backgrounded, appearing as no more than an aspect of the Possessor, more specifically, a location with respect to which the Possessor is being modified through the Agent’s activity. The Possessee body part is also syntactically backgrounded remaining in the lower small clause. Hence the PP may remain implicit if it is stereotypically associated with the verb (see footnote 1). Expectedly, the PP is obligatory for body parts which are non-stereotypical for a particular verb (e.g., compare a se spăla pe burtă (‘wash one’s belly’)/a se spăla (pe mâini) (‘wash one’s hands’)), as well as in more idiomatic readings (e.g., a se zgâria pe ochi (‘eat one’s heart out’, lit. ‘to scratch one’s eyes’)). The PP is conceptualized as a Ground, expressing the aspect with respect to which the Figure is being changed. The body-part is not conceptualized as an independent referential phrase. Hence, possessive modifiers (15a), demonstrative determiners (15b), qualifying adjectives (15c) and modifying relative clauses are all ruled out. The only admissible determiners are the articles (15d). Typically, the PPbody-part semantically incorporates forming a complex predicate with the verb. ‘Washing one’s hands’, ‘washing one’s feet’ etc. represent stereotypical activities, and simply denote particular manners of washing. Referential interpretation of the body part are also contextually possible (15d). (15)

a.

b.

c.

d.

*Ion (se) spală (pe) mâinile sale John SE.ACC washes on hands.DEF his.POSS.ADJ intended: ‘John washes his hands’ *Ion se spală pe mâinile astea John SE.ACC washes on hands.DEF these intended: ‘John washes these hands’ *Maria se spală pe mâinile frumoase Mary SE.ACC washes on hands.DEF beautiful Intended: ‘Mary is washing her beautiful hands’ A reușit să se spele AUX.PERF.3SG managed SĂ.SUBJ SE.ACC wash.SUBJ.3SG pe un picior / piciorul stâng, dar nu și pe celălalt, on a leg leg.DEF left but not also on the.other care era rănit which be.IMPERF.3SG wounded ‘He managed to wash one leg / his left leg, but not the other one, which was hurt’

As insightfully noticed by one of the reviewers, the restrictions true of the bodypart PP in the Romanian accusative figure-reflexive construction are also true of the body-part DO in the possessive dative construction of other Romance languages, as extensively documented in Guéron (2006). The resemblance is not accidental since both the Romance possessive dative construction and the Romanian construction

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

195

under analysis are restricted to expressing inalienable possession and, as discussed above, this imposes certain restrictions on the semantics of the body-part constituent (see Guéron 2006 for details). These restrictions do not hold, however, for the Romanian possessive dative construction, which can express both alienable and inalienable possession. The internal argument accepts any determiner, and may be qualified by possessive and qualifying adjectives, relatives, etc., i.e., the direct object is treated as a fully independent referential phrase. (16)

2.6

a.

Spală-ţi superbele mâini // wash.IMP.2SG¼SE.DAT.2SG superb.DEF hands mâinile astea/ tale murdare și hands.DEF these/ your.POSS.ADJ dirty and vino la masă come.IMP.2SG to dinner ‘Wash your beautiful hands // these dirty hands of yours and come to dinner’ b. Spală-ţi cămașa albă pentru petrecere wash.IMP.2SG¼SE.DAT.2SG shirt.DEF white for party ‘Wash your white shirt for the party’ c. Ion își spală mâinile rănite John SE.DAT washes hands.DEF injured ‘John is washing his injured hands’ c’. ??Ion se spală pe mâinile rănite John SE.ACC washes on hands.DEF injured ‘John is washing his injured hands’

Selection of the Body-Part Preposition

As mentioned, the PP is a place-designating phrase. This also accounts for the selection of the preposition, which is not determined by the verb, but by the bodypart. The preposition is chosen so as to denote a particular type of location, a point (la ‘at’), a surface (pe ‘on’), a volume (în ‘in’), etc. A degree of idiomaticity is also apparent, as well as possible idiosyncratic differences of conceptualization. (17)

a.

b.

S-a tăiat SE.ACC¼AUX.PERF.3SG cut ‘He cut his finger’ S-a tăiat SE.ACC¼AUX.PERF.3SG cut

la deget/ at finger

*?pe deget on finger

pe / *în obraz on in cheek

când when

196

A. Cornilescu and A. Nicolae

c.

2.7

se bărbierea SE.ACC shave.IMPERF.3SG ‘He cut himself on the cheek/?? in the cheek while shaving’ S-a scărpinat în/ ?? la ureche SE.ACC¼AUX.PERF.3SG scratched in at ear ‘He scratched himself in the ear’

Affectedness

The main semantic difference between the two patterns under analysis (accusative figure-reflexives and their dative paraphrases) regards affectedness. Affectedness has been loosely defined as a persistent change in or impingement on an event participant. A more precise definition considers it to be “a relationship between a theme participant that undergoes a change and a scale participant that measures the change” (Beavers 2011, 335, highlight ours). This definition allows one to successfully speak of degrees of affectedness, as a result of monotonically weakening the truth conditions about the result state of Theme on the scale. That affectedness is a matter of degree is apparent in examples like: (18)

a.

b.

c.

d.

Băieţelul a mâncat mărul până la capăt. ‘The little boy ate the apple up’ [apple consumed, no apple left] Băieţelul a tăiat mărul ‘The little boy cut the apple’ [apple cut, not necessarily to a particular degree] Băieţelul a lovit mărul cu piciorul. ‘The little boy kicked the apple’ [apple impinged, not necessarily affected] Băieţelul a atins mărul ‘The little boy touched the apple’ [apple manipulated, not necessarily affected]

Verbs usually contain, in their lexical conceptual structure, some scalar dimension along which the Theme may or must undergo a change, as in the following types of affectedness proposed by Beavers (2011, 339): (19)

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

x changes in some observable property (clean/paint/fix/break x) x transforms into something else (turn/carve/change/transform x) x moves to and stays at some location (move/push/roll x) x is physically impinged (hit/kick/punch/rub/slap/wipe/scrub/sweep x) x goes out of existence (delete/eat/consume/reduce x) x comes into existence (build/design/construct/create/fashion x)

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

197

The properties in (19) are not unrelated; they are all prototypical properties of direct objects. Almost all are also conditioning factors on middle and passive formation. As far as degrees of affectedness are concerned, an important difference holds between verbs which express a potential for change (touch, sweep, rub), and verbs which entail a change in the Theme (Undergoer) (build, eat, destroy). Verbs whose internal argument is conceptualized as Path (e.g., cross the desert, read a book) fall into the first category. Thus, Path objects do not pass the entailment test, i.e., in their case, change is potential, not entailed: The settlers just crossed the desert, but nothing is different about it (Beavers 2011: 346). A consideration of the accusative vs. the dative possessive constructions from the point of view of their event structure and of how affectedness is expressed reveals significant differences between them. We propose that in the accusative figure-reflexive construction the internal argument is ultimately conceptualized as a Path. This is suggested by the fact that the body part Possessee, included in the (raised) Possessor, is syntactically realized as a locative or goal PP, partly or fully travelled along by the Agent (see examples (20a), (20b)). The verbs in (1) are conceptualized as verbs that express surface contact with an object x, which is physically impinged on (hit, kick, punch, rub, slap, wipe, scrub, seep x). The accusative object is analysed as a Path, which may but need not be affected by the event; compare example (20c) with (20a), (20d), (20e). Potentially affected objects do not, or rather, do not always, measure out the event, hence the VP containing them may be interpreted as an activity (20a) or as an accomplishment (20b), (20e). Notice example (20b) in particular, where a stereotypical body part (i.e., the body) may be left out, while the (second) PP explicitly says that the whole of the Possessor’s body has been cleaned. The (second) PP is a delimiter which measures out the event securing an accomplishment reading. (20)

a.

b.

c.

d.

Ion s-a încheiat (la camașă) pe jumătate John SE.ACC¼AUX.PERF.3SG buttoned (at shirt) on half ‘John half buttoned (his shirt)’ Ion s-a spălat (pe corp) din cap John SE.ACC¼AUX.PERF.3SG washed on body from head până în picioare to feet ‘John washed (his body) from top to bottom’ Ion se freacă la ochi de cinci minute John SE.ACC rubs at eyes for five minutes ‘John has been rubbing his eyes for five minutes’ Ion s-a tăiat la picior John SE.ACC¼AUX.PERF.3SG cut at leg și îl doare rana and CL.ACC.3SG.M hurts wound.DEF ‘John cut his leg and his wound hurts’

198

A. Cornilescu and A. Nicolae

e.

S-a descheiat la palton în cateva secunde SE.ACC¼AUX.PERF.3SG unbuttoned at overcoat in a-few seconds ‘(S)he unbuttoned her/his overcoat in a few seconds’

Through its very different syntax, the possessive dative construction presents a different event structure of the same event. In this case it is the body-part, a direct object DP, which is the internal argument interpreted as a Path, potentially being changed by the Agent’s activity. So, while in the accusative figure reflexive construction, it is the reflexive clitic (Possessor) that gets the Path interpretation, in the dative construction, it is the Possessee direct object which is understood as a Path, and which is potentially affected in the event. The Possessor dative is only indirectly affected, in as much as, if one part of a body is affected, the whole body will also be (Lambert 2010). The Path interpretation is confirmed by examples like (21). (21) a.

b.

Ion și-a încheiat cămașa pe jumătate John SE.DAT¼AUX.PERF.3SG buttoned shirt.DEF on half ‘John half buttoned his shirt’ Maria și-a încheiat bluza până la gât John SE.DAT¼AUX.PERF.3SG buttoned blouse.DEF up to neck ‘John buttoned her blouse to the neck’

Given its syntax, the possessive dative construction of these and similar verbs also allows for a re-conceptualization of the event. Namely, the body-part direct object may be understood as a Theme or Undergoer, which is expected to be affected as a result of the Agent’s activity. The main effect of affectedness for inalienable possession constructions is that the Agent’s operation on the body-part Possessee may end up destroying the inclusion relation Possessor/Possessee typical of inalienable possession, in which case the body part is separated from the body (ManoliuManea 1993, 1996), as in the sentences below. (22) a.

b.

Și-a tăiat singur degetul SE.DAT¼AUX.PERF.3SG cut by.himself finger.DEF care se infectase which SE had.infected ‘He cut off his finger which had got infected’ Și-a scos singur un dinte SE.DAT¼AUX.PERF.3SG removed by.himself a tooth care îl durea that CL.ACC.3SG.M hurt ‘He pulled out a tooth of his that hurt’

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

199

Interestingly, the accusative figure-reflexive construction simply cannot express an event of separating the body part from the body. In the accusative construction the body (i.e., reflexive constituent) is an obligatory constituent, and the body-part PP is either overt or implicit (3c); the fact is that the accusative construction, which does not have an alternative Agent/Undergoer interpretation, cannot express the separation of the part from the whole. The contrast between the two patterns is illustrated in (23). In such examples, the two sentences are clearly not mutually entailing any more, having different truth conditions. (23) a.

b.

Ion s-a tăiat la deget [acc-reflexive] John SE.ACC¼AUX.PERF.3SG cut at finger ‘John cut his finger’ Ion și-a tăiat degetul [dat-reflexive] John SE.DAT¼AUX.PERF.3SG cut finger.DEF ‘John cut his finger off’

Verbs which conceptualize their object only as an Undergoer and where change is entailed, rather than potential appear only in the dative construction and have no accusative figure-reflexive paraphrase. (24) a.

a-și rupe piciorul A.INF¼SE.DAT break.INF leg.DEF a’. *a se rupe la picior A.INF SE.ACC break.INF at leg ‘to break one’s leg’ b. a-și frânge gâtul A.INF¼SE.DAT fracture.INF neck.DEF b’. *a se frânge la gât A.INF SE.ACC fracture.INF at neck ‘to fracture one’s neck’

[dat-reflexive, DO] [acc-reflexive, PP]

[dat-reflexive, DO] [acc-reflexive, PP]

With such verbs, the use of the SE pattern may be a mark of an idiomatic figurative reading, as shown by the following contrast: (25) a.

b.

a-și scrânti piciorul A.INF¼SE.DAT dislocate.INF leg.DEF ‘to sprain one’s leg’ a se scrânti la cap A.INF SE.ACC dislocate.INF at head ‘to go crazy’

[dat-reflexive, DO]

[acc-reflexive, PP]

Having discussed the empirical properties of the accusative figure reflexive construction, we may turn to its syntactic analysis.

200

A. Cornilescu and A. Nicolae

3 Transitive Figure-Ground Reflexive: Syntactic Analysis The derivation of accusative figure reflexive constructions has the features depicted below, stemming from their established empirical properties. 1. The accusative reflexive construction is a means of expressing inalienable possession. The inalienable possession component is syntactically analyzable as a prepositional small clause, which is the internal argument of the verb and which introduces the Figure-Ground relation. The idea that a prepositional small clause is a natural configuration for the Figure-Ground relation goes back to Talmy (1978). Svenonius (2007) proposes that Figures merge in the specifier of the functional prepositional head p: (26)

[pP

FIGURE

[p’p [PP P

GROUND]]]

Arguments merged in Spec, pP will generally A-move to a higher position such as Spec,vP, Spec, TOP, Spec, VoiceP, etc. 2. As seen above, the accusative figure-reflexive construction is transitive and SE exhibits argumental properties (an independent θ-role, accusative case, doubling). Taking into account the argumental nature of SE in this construction, it should merge in a θ-position and be θ-interpreted. We propose that SE merges as the specifier of the prepositional small clause, and it is assigned the Figure role. 3. Romanian SE always has an unvalued accusative case feature (Dobrovie-Sorin 1994). As the configuration is transitive, a TOP (¼ Tense-Object phrase, as defined in Pesetsky and Torrego 2004) is licensed, projected below the VoiceP which introduces the external argument, but above the lexical VP. SE will raise from its merge position to Spec, TOP where it values its Case feature. Finally, SE goes up to the Person-Tense domain, and agrees with the (coreferential) external argument, which completes its ϕ-features (i.e., Number, assuming that SE is inherently provided only with a third person and accusative case feature). 4. Since the construction is Agentive, the Voice head is θ-marking and assigns the Agent role. The configuration observing the properties established above has the derivation in (27). (27)

Ion se spală John SE.ACC washes ‘John washes his hands’

pe mâini on hands

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

201

(a)

VoiceP DPAGENT

Voice’

Voice Ion [AGENT]

TOP TO

VP pP [FIGURE]

V spală

p’

DP SEFIGURE

p Æ

PP[GROUND] pe mâini

(b)

VoiceP DPAGENT/FIGURE Voice’ Ion

Voice [AGENT]

ToP DP

To’ To SE

VP pP [FIGURE]

V spală

p’

DP SE

p Æ

PP[GROUND] pe mâini

The derivation is straightforward and accounts for the properties identified above.

202

A. Cornilescu and A. Nicolae

4 Extending the Analysis: Unaccusative Figure Reflexive Constructions In what follows, we extend the analysis to unaccusatives, claiming that the same structure underlies these configurations, except that the Voice projection is expletive.

4.1

The Data

The same inalienable possession Figure-Ground component is also found in constructions with a large number of change of state unaccusative verbs. Even if these are otherwise one-argument verbs, a locative PP is possible and often required to express a particular meaning. Thus, as with the transitive construction, an extra argument is apparently added to the one argument of change of state predicates. As with reflexive transitive verbs, the additional PP contains a Possessee DP, while the one argument of the unaccusative predicate expresses the Possessor. The unaccusatives found in this pattern (28) are mostly adjective-based, with a few noun-based exceptions (a se îngloda ‘sink into mud’, a se afunda ‘sink to the bottom’). (28) a se albi ‘whiten’, a se îngălbeni ‘(turn) yellow’, a se înroși ‘redden’, a se lumina ‘brighten, light up’, a se întuneca ‘darken’, a se închide ‘close’, a se deschide ‘open’, a se rări ‘thin out’, a se împuţina ‘lessen’, a se împlini ‘gain/put on weight’, a se îngroșa ‘thicken’, a se îngloda ‘sink into mud’, a se afunda ‘sink’, a se însănătoși ‘become healthy’, a se înzdrăveni ‘become stronger’, a se betegi ‘grow weaker’, a se ofili ‘wither’, a se ologi ‘became lame/crippled’, a se ascuți ‘become sharper’, a se subţia ‘grow thinner’, etc. All of these verbs have transitive counterparts (a închide ceva/a se închide ‘close (something)’), and some also have bare unaccusative counterparts (a albi/a se albi ‘whiten’; a îngălbeni/a se îngălbeni ‘yellow’), so that there are also triplets, composed of a transitive verb, a bare unaccusative verb and a complex unaccusative verb (see Dragomirescu 2009, 2010; Dragomirescu and Nicolae 2017 for the full inventory of verbs occurring in the transitive/unaccusative alternation): (29) a.

b.

c.

Maria albește rufele Mary whitens laundry.DEF ‘Mary is bleaching the laundry’ Părul albește cu vârsta hair.DEF whitens with age ‘One’s hair grows grey with age’ Părul se albește cu vârsta hair.DEF SE.ACC whitens with age ‘One’s hair grows grey with age’

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

(30) a.

b.

c.

203

Maria înroșește ouăle Mary.NOM reddens eggs.DEF ‘Mary is reddening the eggs’ Ouăle se înroșesc eggs.DEF SE.ACC redden ‘The eggs are reddening’ *Ouăle înroșesc eggs.DEF redden

From the point of view of the current discussion it is of interest that change of state verbs require a supplementary PP to express the inalienable possession meaning component of the sentence. Note that the PP is obligatory unless the idiomatic interpretation is otherwise signalled. The examples have been grouped into cases where the PP is optional (31), cases where the PP is obligatory and the sentence may get a different reading (or the variant without the PP is ungrammatical) (32), and finally examples where the interpretation is idiomatic and the PP is an obligatory part of the idiom (33).2 (31) a.

Ion s-a înroșit (la faţă) John SE.ACC¼AUX.PERF.3SG reddened (in face) ‘John turned red in the face’ b. Maria s-a ofilit (la chip) Mary SE.ACCRAUX.PERF.3SG withered (in face) ‘Mary’s face has withered/Mary has withered in the face’ c. De o vreme Ion s-a împuţinat/ for a while John SE.ACC¼AUX.PERF.3SG lessened s-a împlinit (la trup) SE.ACC¼AUX.PERF.3SG become.fuller (in body) ‘For a while John has grown thinner/fuller (in the body)’ (32) a. Ion se înrosește în gât când are febră John SE.ACC reddens in throat when has fever ‘John turns red (¼sore) in the throat when he is feverish’ a’. Ion se înrosește John SE.ACC reddens ‘John turns red’ (a 6¼a’)

2

Other extensions of the locative pattern, which do not include possession, also exist:

(1)

a. Ion se

înglodează în datorii / *Ion se

John SE.ACC sinks

in debts

înglodează

John SE.ACC sinks

‘Ion is sinking into debt’ b. Ion se afundă în lectură / *Ion se afundă John SE.ACC plunges into reading / John SE.ACC plunges ‘John is plunging into reading’

204

A. Cornilescu and A. Nicolae

b.

b’. c.

c’. (33) a.

a’. b.

b’.

4.2

Ion s-a luminat la faţă John SE.ACC¼AUX.PERF.3SG brightened at face când a intrat Maria when AUX.PERF.3SG entered Mary ‘John’s face brightened up when Mary entered’ *Ion s-a luminat John SE.ACC¼AUX.PERF.3SG brightened Ion s-a cam îngroșat la obraz John SE.ACC¼AUX.PERF.3SG rather thickened at cheek ‘John’s cheek has thickened (i.e., He has become rather rude)’ *Ion s-a cam îngroșat John SE.ACC¼AUX.PERF.3SG rather thickened Ion s-a închis în sine John SE.ACC¼AUX.PERF.3SG closed in himself ‘John has shrunk into himself’ *Ion s-a închis [starred on the intended reading] John SE.ACC¼AUX.PERF.3SG closed Ion s-a cufundat în sine John SE.ACC¼AUX.PERF.3SG plunged in himself ‘John has plunged into himself’ *Ion s-a cufundat [starred on the intended reading] John SE.ACC¼AUX.PERF.3SG plunged

On the Role of the PP in the Anticausative Construction

Even if the verb is basically a change of state one, the locative PP is obligatory. This raises the problem of the integration of the PP into the argument structure of the anticausative verb. To understand the role of the Figure-Ground component in the anticausative configuration, consider the following systematic paraphrase of the extended Figure-Ground pattern: (34) a.

Ion s-a luminat la faţă John SE.ACC¼AUX.PERF.3SG brightened at face ‘John’s face brightened up’ a’. Faţa lui Ion s-a luminat face.DEF.NOM GEN John SE.ACC¼AUX.PERF.3SG brightened ‘John’s face brightened up’ b. Ion s-a înroșit în gât John SE.ACC¼AUX.PERF.3SG reddened in throat ‘John’s throat has turned red (¼sore)’ b’. Gâtul lui Ion s-a înroșit throat.DEF GEN John SE.ACC¼AUX.PERF.3SG reddened ‘John’s throat has turned red (¼sore)’

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

205

The paraphrases clearly show that the PP actually specifies which subpart of the subject entity is actually affected by the change, so that the locative PP is in fact (part of) a possessive/partitive construction. Thus the Figure-Ground structure indicates which aspect of the Figure (Possessor) actually undergoes the change. This interpretation is confirmed by the existence of one argument sentences with a subject body part and a DP-internal Possessor (i.e., the existence of pairs like (34a–a’), (34b–b’)). Since what actually changes is the subpart, the body part in the examples above, it is hardly surprising that the body part must be overt rather than inferred. From a semantic perspective, all the verbs in (28) are degree achievements and describe changes in the degree/extent to which an individual possesses some property specified by the adjective-based verb. In the simple cases, illustrated so far, degree achievements express change of an individual over time. The locative PP small clause metonymically introduces the object that undergoes the change (the Possessee) promoting the Possessor to the role of subject of the main verb. Events of change in an individual over time (along some dimension specified by the verb) form the basis for the semantic analysis of this degree achievement verb class. It may be noticed, however, that the use of the metonymic (Possessor-Possessee) configuration for specifying the entity which is affected/described in a degree achievement sentence goes far beyond the body-part structures already illustrated, which only constitutes the most common type of degree achievement: changes of an individual over time. In addition, there are also so-called spatial extent readings (35), with similar possessive paraphrases (35a–a’). (35) a.

Cărarea se îngustează spre vârf trail.DEF.NOM SE.ACC narrows towards summit ‘The trail narrows towards the summit’ a’. Partea dinspre vârf a cărării se îngustează part.DEF.NOM towards summit GEN trail SE.ACC narrows ‘The part of the trail towards the summit narrows’ b. Râul se lărgește mult la capăt river.DEF.NOM SE.ACC widens a.lot at end ‘The river widens a lot at the end’ c. Drumul se îmbunătățește între Iași și Vaslui road.DEF.NOM SE.ACC gets.better between Iași and Vaslui ‘The road gets better between Iași and Vaslui’

Such readings do not involve measuring change over the temporal extent of an event, but rather over the spatial extent of an individual—in the cases cited, changes affect a particular subpart of the trail (35a), the river (35b) or the road (35c). Finally, Deo, Francez and Koontz-Garboden (2013) also comment on a class of ‘functional readings’ of degree achievements, where no change in some individual is involved, whether over time or space. Here are two examples of this type, where the entity which is described, rather than affected, is again specified through the metonymic relation between the subject and the locative PP, as shown by the same possessive paraphrase.

206

A. Cornilescu and A. Nicolae

(36) a.

Intriga se dezintegrează în partea a doua a romanului plot.DEF SE.ACC disintegrates in part second GEN novel.DEF.GEN ‘The plot disintegrates in the second part of the novel’ a’. Intriga părții a doua a romanului se plot.DEF part.GEN second GEN novel.DEF.GEN SE.ACC dezintegrează disintegrates ‘The plot of the second part of the novel disintegrates’ b. Acțiunea se complică în actul al II-lea al plot.DEF SE.ACC complicates in act.DEF.ACC second GEN piesei play.DEF.GEN ‘The plot becomes more complex in the second act of the play’ b’. Acțiunea actului al II-lea se complică plot.DEF act.DEF.GEN second SE.ACC complicates ‘The plot of the second act becomes more complex’

Such examples can be construed as expressing differences with respect to some attribute (coherence, complexity) between subparts of an extended individual (the plot, the script) along some abstract structuring dimension. The full range of uses of degree achievements thus requires moving from a semantics based on change in an individual over the course of an event to a more abstract semantics, based on comparing values of functions at different points along an ordered domain (for a detailed semantic analysis of degree achievements, see Deo et al. 2013). From the present perspective, what counts is the pervasiveness of metonymically defining the entity which is affected, always viewed as a subpart of some more salient whole (appearing in subject position).

4.3

Non-agentivity

The verbs discussed in this section are unaccusative. Unlike the transitive figurereflexive structures discussed in the previous sections, the examples in (31)– (33) above are non-agentive. This can be seen in the absence of the by-phrase (37a), the impossibility of agent-oriented adverbs (37b), and of control into purpose clauses (37c). (37) a.

b.

c.

*Ion s-a luminat la faţă de către Petru John SE.ACC¼AUX.PERF.3SG brightened at face by Peter intended: ‘John’s face brightened up by Peter’ *Ion se înrosește în gât cu bună știință John SE.ACC reddens in throat with good awareness. intended: ‘John’s throat willingly turned red (¼sore)’ *Ion se înrosește în gât pentru a-i John SE.ACC reddens in throat for A.INF¼CL.DAT.3SG

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

207

face plăcere doctorului do.INF pleasure doctor.DEF.DAT intended: ‘John turns red (¼sore) in the throat to please his doctor’

4.4

No Dative Paraphrase

Lack of agentivity is also suggested by the absence of an adequate dative paraphrase. Recall that the dative paraphrase clearly shows the monotransitive argumental structure of the verb. Thus, with unaccusatives, the dative construction is clearly ill-formed in the intended reading. (38) a.

b.

(39) a.

b.

Maria s-a albit la faţă Mary SE.ACC¼AUX.PERF.3SG whitened at face când l-a văzut when CL.ACC.3SG.M¼AUX.PERF.3SG seen ‘Mary turned white in the face when she saw him’ *Maria și-a albit faţa Mary SE.DAT¼AUX.PERF.3SG whitened face.DEF când l-a văzut when CL.ACC.3SG.M¼AUX.PERF.3SG seen actual reading: ‘Mary whitened her face when she saw him’ ≠ (38a) Maria se înroșește în gât când are febră Mary SE.ACC reddens in throat when has fever ‘Mary turns red (¼sore) in her throat when she is running a fever’ *Maria își înroșește gâtul când are febră Mary SE.DAT reddens throat.DEF.ACC when has fever actual reading: ‘Mary reddens her throat when she is running a fever’≠(39a)

It may safely be concluded that the verbs under analysis fall into the class of complex reflexive unaccusatives (as defined, for instance, in Schäfer 2009).

4.5

Syntactic Derivation

Taking stock of the properties reviewed above, we now turn to the syntactic analysis of unaccusative figure reflexives.

208

A. Cornilescu and A. Nicolae

1. The unaccusative configuration shares the Figure-Ground component with the transitive one, and this common element will have the same prepositional small clause representation. The prepositional small clause is the internal argument of the anticausative verb, just as it was the internal argument of the transitive verb. 2. More recent research on the causative alternation (Harley 1995; Alexiadou, Anagnostopoulou and Schäfer 2006; Schäfer 2008, 2009; Martin and Schäfer 2014 i.a.) has led to the view that both causative and anticausative predicates have the same bi-eventive structure (as different from older studies, e.g., Dowty 1979). This point of view has been voiced time and again at least since Harley (1995). Both causative and anticausative verbs subcategorize the same internal argument and differ as to whether or not they also overtly express the second argument. Lexical causatives and anticausatives involve the same nominal or adjectival root (expressing a resultant state predicated of the Theme), which is incorporated into the V, but anticausatives involve a become projection (leaving the causer out) (40), while causatives involve a Cause projection, which introduces the Causer (41). (40)

vP vbecome

VP V

(41)

DPTheme

VoiceP DPCauser

Voice’

Voice

VP V

DPTheme

More needs to be said about languages that have not only bare anticausatives, but also about the reflexive component of other types of morphologically marked anticausatives. Bare and complex unaccusatives represent different syntactic configurations in spite of the fact that they largely encode the same meaning (see below). Alexiadou et al. (2006) view the alternation as a Voice alternation. Bare unaccusatives have no Voice component. Reflexive anticausatives have one more layer of structure, an expletive VoiceP. The reflexive is assumed to merge in Spec,VoiceP, a proposal which cannot be adopted for Romanian, where SE is a ‘low clitic’ (Cornilescu and Nicolae 2015). We follow the literature (Schäfer 2008 for German; Wood 2012 for Icelandic; Labelle and Doron 2010 for French i.a.) in assuming that derived SE unaccusatives have an expletive Voice projection, so that, syntactically, they look like transitive, rather than unaccusative configurations, but we propose a slightly different implementation of this view, in agreement with the properties of Romanian SE. We continue to treat expletive Voice as a head that may accommodate a DP, but may not assign it a θ-role. Rather, in the course of the derivation, this expletive phrase will be filled by movement.

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

209

Before deciding on the proper representation, let us notice that the anticausative has very different properties from transitive SE, as discussed above. In the unaccusative configuration the reflexive clitic cannot be doubled by a strong reflexive and is not assigned a theta role, i.e., it is not an argument. However, in Romanian at least, it still contributes an unvalued accusative feature. We have shown elsewhere (Cornilescu and Nicolae 2015) that in a model like D’Alessandro’s (2007), where SE/SI is an argument which merges or passes through different functional positions before reaching the Tense/Person field, Romanian SE never moves higher than the inner Aspect/accusative-assigning functional projection (TOP in the notation proposed by Pesetsky and Torrego 2004). In other words it never occupies the Spec, ApplP or Spec, VoiceP positions. Voice selects TOP as in the transitive configuration. Anticausative SE merges in Spec, TOP to value its accusative feature. The configuration proposed is given in (42) below. It differs from the transitive configuration in that SE is not an argument and VoiceP is expletive. SE is a syntactic word which blocks the assignment of the Acc case to another DP. On the other hand, a bare unaccusative configuration contains neither a Voice phrase nor an accusative licensing projection. The specifier of the vP will accommodate the Specifier of the prepositional small clause ( pP). SE

(42) Ion se luminează John SE.ACC brightens ‘John’s face is brightening’

la față at face

VoiceP DP Ion

Voice’ Voice [-θ]

TOP TO’

DP se

TO

VP DP

V’ V luminează

pP DP

p’ p Æ

PP la faţă

The lexical argument merges in the Spec, pP where it is assigned the Figure role in the Figure-Ground configuration. Next, it raises to Spec,vP, probably being assigned a second Theme role since it is the specifier of a verb which lexically

210

A. Cornilescu and A. Nicolae

encodes a Path. From this position, the lexical argument can easily target Spec, TP. When both the lexical argument and the reflexive pronoun reach the T-domain, there is agreement in ϕ-features between the derived subject and the reflexive clitic.

5 More on the Anticausative Alternation 5.1

Older Confirmed Results

Some of the verbs listed in (28) have bare unaccusative counterparts. This raises the problem of accounting for the possible semantic differences between them. The difference between bare and reflexive change of state unaccusative verbs (as in (43a)/(43b)) has generally been described as having to do with internal vs. external causation (see Labelle 1992; Dobrovie-Sorin 2015 a.o.; see Vivanco [this volume] for an alternative based on scale structure) and with differences in the level of spontaneity, assuming that externally caused events and less spontaneous events prefer SE-marked unaccusatives (Haspelmath 1993). Haspelmath (1993) actually claims that there is a universal scale of spontaneity. (43) a.

Ion a albit la păr cu vârsta John AUX.PERF.3SG whitened at hair with age a’. ?Ion s-a albit la păr cu vârsta John SE.ACC¼AUX.PERF.3SG whitened at hair with age ‘John’s hair has whitened with age’ b. Ion s-a albit la faţă când John SE.ACC¼AUX.PERF.3SG whitened at face when a văzut-o has seen¼CL.ACC.3SG.F ‘John’s face has whitened when he saw her’

The examples in (43) show that the natural process of one’s hair whitening with age is suitably described by the bare unaccusative, while the unexpected, externally caused whitening of somebody’s face prefers the derived unaccusative. Opinions often diverge on the interpretation of this difference in other languages, as well as in Romanian (see Dragomirescu 2009, 2010; Dragomirescu and Nicolae 2017 for a review of the literature on Romanian and an analysis of all verb classes that show the reflexive alternation). At one end of the spectrum, Labelle and Doron (2010) find systematic contrasts between the two structures and attribute different event structures to simple vs. complex reflexive anticausative; in their view, the reflexive anticausatives only have the result sub-event, while the bare unaccusatives exhibit the process + result small clause configuration and focus on the process. At the other end of the spectrum, Martin and Schäfer (2014) claim that marked and unmarked anticausatives compete, but do not differ in meaning (i.e., in their truth conditional semantics). Possible differences can be derived from the lexical meaning of the verb (the nature of the process described), as well as from more general

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

211

semantic-pragmatic principles (see Vivanco [this volume] for a discussion of these issues in Spanish, and a novel take).

5.2

More Recent Results

Given what has been said so far, we will consider significant only differences in the meaning of the same verb when used in the alternating anticausative structures. Verbs which have only one form have a wider, unmarked interpretation, as can be seen in the examples below: (44) a.

b.

Universul se mărește în întindere. universe.DEF.NOM SE.ACC extends in surface ‘The surface of the Universe is extending’ (reflexive anticausative, only internal cause) Casa a explodat și este în ruină house.DEF.NOM AUX.PERF.3SG exploded and is in ruin ‘The house has exploded and it is in ruin’ (bare anticausative, only external cause)

We assume that, for alternating verbs, semantic differences should follow from the absence/presence of the (expletive) VoiceP, and have the status of suspendable implicatures. Thus, reinterpreting Haspelmath’s (1993) proposal from a syntactic perspective, his claim that reflexive anticausatives describe events of lower spontaneity which require a (sometimes additional) external causer is plausibly related to a configuration which requires an external argument—or at least a placeholder of it. Furthermore, as noticed by Beavers and Koontz-Garboden (2013), since reflexive anticausatives syntactically suggest the existence of a causer, it is possible to negate the causer, without also negating the change of state, i.e., the caused event: (45) A:

B:

A:

B:

Ce s-a întâmplat aici? what SE.ACC¼AUX.PERF.3SG happened here ‘What happened here?’ S-au spart niște geamuri. SE.ACC¼AUX.PERF.3PL broken some windows ‘Some windows broke’ Cum adică S-AU spart? how exactly SE.ACC¼AUX.PERF.3PL broken ‘What do you mean by they broke?’ Le-a spart cineva și am CL.ACC.3PL.F¼AUX.PERF.3SG broken someone and have impresia că tu. impression.DEF that you ‘Someone has broken them and I think it was you’

212

A. Cornilescu and A. Nicolae

The same authors remark that simple unaccusatives do not allow the negation of the causer, separately from the caused event (unless some modifier is used): (46) Apa a fiert. #NU A fiert. water.DEF AUX.PERF.3SG boiled not AUX.PERF.3SG boiled Vrei să spui că tu ai fiert-o. want SĂ.SUBJ say.SUBJ.2SG that you AUX.PERF.2SG boiled¼CL.ACC.3SG.F ‘The water boiled. No, it DID NOT boil. What you mean to say is that you have boiled it’ Such data naturally follow from the assumption that complex anticausatives project an expletive VoiceP.3

3

Still further, if reflexive anticausatives have the same syntactic configuration as causatives, in particular as true reflexives, then in context their derived subject should be interpreted as an AgentTheme, deriving a true reflexive reading:

(i)

a.

Din păcate, Ion

s-a

înecat

pentru că

unfortunately John

SE.ACC¼AUX.PERF.3SG

drowned

because

nu

știa

să

înoate [anticausative]

SĂ.SUBJ swim.SUBJ.3SG not know.IMPERF.3SG ‘Unfortunately, John got drowned because he did not know how to swim’

b.

S-a

înecat

ca

SE.ACC¼AUX.PERF.3SG

drowned

in.order

salveze

copilul.

save.SUBJ.3SG

child.DEF

să-și SĂ.SUBJ¼SE.DAT

[reflexive causative]

‘He drowned in order to save his child’

Conversely, potential Agents may be reinterpretable as Themes, if the subject position is focused and alternatives need to be supplied: (ii)

a.

Astăzi copilul

a

today child.DEF.NOM AUX.PERF.3SG spele

reușit

să

se

managed

SĂ.SUBJ

SE.ACC

singur pe cap!

wash.SUBJ.3SG alone on head ‘Today, the child managed to wash his hair by himself!’ b.

Nici vorbă! Nu S-A no word not SE.ACC¼AUX.PERF.3SG

spălat, ci washed but

L-AM

spălat

eu!

CL.ACC.3SG.M¼AUX.PERF.1SG

washed

I

‘No way! He did not wash, I washed him’

Thus, the presence of the VoiceP facilitates the rebracketing of the construction as a causative reflexive, as well as the negation of the causer, distinct from the negation of the caused event.

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

5.3

213

No External Argument

At the same time, SE-unaccusatives also clearly show that their Voice position is expletive, i.e., no DP merges in this position. As known, Wood and Marantz (2017), following Bruening (2013), bring solid evidence that the presence of a by-phrase is not the only indicator of a thematic external argument position. Rather there is a family of adjuncts which correlate with the presence of an external argument: in addition to the by-phrase, this family of adjuncts includes instrumental agents and external argument comitatives. These adjuncts are licensed by SE-verbs, but are not licensed by bare unaccusatives, as expected. Crucially, when a SE-verb occurs with instrumental agents (47) or external argument comitatives (48), the interpretation cannot be anticausative, but it shifts to a true reflexive reading or a passive reflexive (cf. (49)). (47) a.

Ion s-a albit la faţă cu pudră John SE.ACC¼AUX.PERF.3SG whitened at face with powder ‘John has whitened his face with powder’ b. *Ion a albit la faţă cu pudră John AUX.PERF.3SG whitened at face with powder (48) a. Ion s-ar putea albi la păr John SE.ACC¼AUX.COND.3SG can whitened at hair cu un machior bun. with a make-up.artist good ‘John may whiten his hair with (the help of) a good make-up artist’ b. *Ion ar albi la păr cu un machior John AUX.COND.3SG whiten at hair with a make-up.artist bun good (49) Ouăle s-au înroșit cu vopsea [SE-passive] eggs.DEF SE.ACC¼AUX.PERF.3PL redden with dye ‘The eggs have been reddened with dye’

5.4

More on the Anticausative Alternation with Degree Achievements

Restricting ourselves to the verbs under analysis, and more generally to the class of reflexive degree achievements, some possible implicatures of the reflexive vs. the bare variant of the same verb may follow from the fact that the IA moves to Spec, VoiceP, occupying the position of an external argument. Hence, as suggested in traditional grammars, the IA is more “like a subject” (a similar effect is found in

214

A. Cornilescu and A. Nicolae

middles, for the same reason). The semantic effect is that of an intense participation of this derived subject in the event. Higher involvement may lead to higher responsibility for the change resulting from the event. Differences of involvement and responsibility are clearly visible in the examples below. (50) a.

b.

(51) a.

b.

(52) a.

b.

(53) a.

b.

Ion obosește când deschide mereu ușa la musafiri John gets.tired when opens always door.DEF at guests ‘John gets tired when he keeps opening the door to the guests’ În ultima vreme s-a obosit prea mult lately SE.ACC¼AUX.PERF.3SG tired too much ‘Lately he has tired himself out too much’ Ion ostenește când udă florile John gets.tired when waters flowers.DEF ‘John gets tired when he waters the flowers’ Ion nici nu s-a ostenit John even not SE.ACC¼AUX.PERF.3SG tired să ude florile SĂ.SUBJ water.SUBJ.3SG flowers.DEF ‘John didn’t even bother to water the flowers’ ??Ion înverzește la față când aude ce John turns.green at face when hears what și-a mai cumpărat SE.DAT¼AUX.PERF.3SG more bought fratele lui brother.DEF his.GEN Ion se înverzește la față când aude ce John SE.ACC turns.green at face when hears what și-a mai cumpărat fratele SE.DAT¼AUX.PERF.3SG more bought brother.DEF his.GEN lui ‘John turns green in the face when he hears what his brother bought’ ??Ion a albit la față când John AUX.PERF.3SG whiten at face when a aflat vestea AUX.PERF.3SG found news.DEF Ion s-a albit la față când John SE.ACC¼AUX.PERF.3SG whiten at face when a aflat vestea AUX.PERF.3SG found news.DEF ‘John turned white in the face when he found out the news’

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

215

Generally, a review of the available examples shows that changes in the selectional restrictions, often resulting in idiomatic readings, are associated with the complex reflexive structure as also noticed by Labelle and Doron (2010), presumably because the subject is felt to be responsible for the ensuing results. (54) a.

Mașina grăbea spre casă [non-reflexive] car.DEF hurry.IMPERF.3SG towards house a’. *Ion grăbea spre casă John hurry.IMPERF.3SG towards house b. Mașina se grăbea spre casă [reflexive] car.DEF.NOM SE.ACC hurry.IMPERF.3SG towards house b’. Ion se grăbea spre casă John SE.ACC hurry.IMPERF.3SG towards house ‘The car hurried towards the house/John hurried towards the house’ (55) a. În ultima vreme, Ion a știrbit lately John AUX.PERF.3SG jagged de tot [non-reflexive] completely ‘Lately, John has completely lost his teeth’ b. Autoritatea lui s-a știrbit/ authority.DEF his.GEN SE.ACC¼AUX.PERF.3SG diminished *a știrbit de tot [reflexive: OK /non-reflexive: *] AUX.PERF.3SG diminished completely ‘His authority completely diminished’ (56) a. Trenul a pornit devreme / de la sine train.DEF AUX.PERF.3SG set.out early by itself ‘The train set out early / by itself’ [non-reflexive] b. Ion s-a pornit cu noaptea în cap / John SE.ACC¼AUX.PERF.3SG set.out with night in head *de la sine [reflexive] by himself ‘John set out very early / *by himself’

6 Conclusions 1. Figure reflexive constructions are unified by the fact that their internal argument is a prepositional small clause, which introduces the Figure-Ground relation. The analysis is economical, being based on one syntactic structure, the difference lying in the presence/absence of a θ-feature on the Voice functional head (cf. The Chomsky-Borer Conjecture, as defined in Baker 2008).

216

A. Cornilescu and A. Nicolae

2. All of the figure reflexives are Voiced constructions (i.e., they contain an external argument position). The Voice head is thematic in transitive constructions, but non-thematic in unaccusative configurations. The thematic VoiceP is filled by External Merge, the non-thematic VoiceP is filled by Internal Merge. The VoiceP always selects an accusative case projection (TOP), as proposed in Pesetsky and Torrego (2004). There is evidence that Romanian SE is best viewed as a DP-phrase, which merges in a specifier position and must value its accusative feature before reaching the Person/Tense domain (see Dobrovie-Sorin 2015). 3. The discussion of unaccusatives strengthens the suggestion that there is a connection between the expletivization of Voice and derived reflexive unaccusatives. Change of state unaccusatives share the syntactic configuration of transitives, differing through the presence of expletive Voice. This assumption is sufficient to account for the difference between basic and derived reflexive unaccusatives. Thus, we may conclude with Wood and Marantz (2017) that all argumentintroducing heads have thematic and expletive variants. 4. The analysis of change of state unaccusatives may be naturally extended to change of place unaccusatives (see Pineda [this volume]). Change of place unaccusative verbs expectedly develop reflexive doublets using the same mechanism, with semantic effects similar to those produced in the case of change of state unaccusatives. Here are examples of such doublets: a urca/a se urca (‘to climb (up)’), a coborî/a se coborî (‘climb down’), etc. Semantic effects are less apparent with change of location verbs since the PP is an obligatory constituent in this case. Acknowledgements We would like to express our gratitude to both the editors of the volume and the reviewers of this paper. We were particularly sensitive to the sharp questions and substantive issues raised by the reviewers, which have led to a careful rewriting of the paper in the hope of increasing clarity and relevance. We are also indebted to the participants in the Workshop on Romance SE/SI, held on April 21st and 22nd, 2016 at the University of Wisconsin-Madison. Thanks are also due to Adina Dragomirescu for the comments on an earlier version of the paper. The editors of the volume deserve our warmest gratitude for their kindness, patience and wonderful help.

References Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2006. The properties of anticausatives crosslinguistically. In Phases of interpretation, ed. Mara Frascarelli, 175–199. Berlin: Mouton. Baker, Mark C. 2008. The macroparameter in a microparametric world. In The limits of syntactic variation, ed. Theresa Biberauer, 351–373. Amsterdam/Philadelphia: John Benjamins. Beavers, John. 2011. On affectedness. Natural Language and Linguistic Theory 29 (2): 335–370. Beavers, John, and Andrew Koontz-Garboden. 2013. In defense of the reflexivization analysis of anticausativization. Lingua 131: 199–216. Bruening, Benjamin. 2013. By-phrases in passives and nominals. Syntax 16 (1): 1–41.

On a Class of Figure Reflexives in Romanian: Ion se spală pe mâini ‘John. . .

217

Cornilescu, Alexandra, and Alexandru Nicolae. 2015. The grammaticalization of a constraint on passive reflexive constructions in Romanian. In Diachronic variation in Romanian, ed. G. Pană Dindelegan, R. Zafiu, A. Dragomirescu, I. Nicula, and A. Nicolae, 309–361. Newcastle upon Tyne: Cambridge Scholars Publishing. D’Alessandro, Roberta. 2007. Impersonal si constructions. Agreement and interpretation. Berlin: Mouton de Gruyter. Deo, Ashwini, Itamar Francez, and Andrew Koontz-Garboden. 2013. From change to value difference in degree achievements. Proceedings of SALT 23: 97–115. Dobrovie-Sorin, Carmen. 1994. The syntax of Romanian. Berlin: Mouton de Gruyter. ———. 2015. Reflexive-marking in romance: Voice and feature deficiency. Ms., CNRS-LLF, to appear in Syn-Com. Dowty, David. 1979. Word meaning and montague grammar. Dordrecht: Reidel. Dragomirescu, Adina. 2009. Consecinţele sintactice ale variaţiei reflexiv/nonreflexiv în cazul verbelor inacuzative. In Limba română: teme actuale, ed. R. Zafiu, G. Stoica, and M.N. Constantinescu, 85–95. Bucharest: Editura Universităţii din Bucureşti. ———. 2010. Ergativitatea: Tipologie, sintaxă, semantică. Bucharest: Editura Universității din București. Dragomirescu, Adina, and Alexandru Nicolae. 2017. Semantic constraints on the reflexive/nonreflexive alternation of Romanian unaccusatives. In Contrastive studies in verbal valency, ed. L. Hellan, A. Malchukov, and M. Cennamo, 407–430. Amsterdam/Philadelphia: John Benjamins. Guéron, Jaqueline. 2006. Inalienable possession. In The Blackwell Companion to syntax, ed. Martin Everaert and Henk van Riemsdijk, 589–638. Oxford: Wiley-Blackwell. Harley, Heidi. 1995. Subjects, events and licensing. PhD diss., MIT. Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Causatives and transitivity, ed. Bernard Comrie and Maria Polinsky, 87–120. Amsterdam/ Philadelphia: John Benjamins. Labelle, Marie. 1992. Change of state and valency. Journal of Linguistics 28 (2): 375–414. Labelle, Marie, and Edit Doron. 2010. Anticausative derivations (and other valency alternations) in French. Probus 22 (2): 303–316. Lambert, Silke. 2010. Beyond recipients: towards a typology of dative uses. PhD diss., University at Buffalo, State University of New York. Landau, Idan. 1999. Possessor raising and the structure of VP. Lingua 137: 1–37. MacDonald, Jonathan E. 2015. A case of multiple agree: Accusative, not dative se. In Romance linguistics 2012. Selected papers from the 42nd linguistic symposium on romance languages (LSRL), Cedar City, Utah, 20–22 April, ed. Jason Smith and Tabea Ihsane, vol. 2012, 275–288. Amsterdam/Philadelphia: John Benjamins. Manoliu-Manea, Maria. 1993. Gramatică, pragmasemantică și discurs. Bucharest: Litera. ———. 1996. Inalienability and topicality in Romanian: Pragma-semantics of syntax. In The grammar of inalienability: A typological perspective on body part terms and the part–whole relation, ed. Hilary Chappell and William McGregor, 711–743. Berlin: Mouton de Gruyter. Martin, Fabienne and Florian Schäfer. 2014. Anticausatives compete but do not differ in meaning: A French case study. In Proceedings of Congrès Mondial de Linguistique Française 2014, FU Berlin, 2485–2500. Niculescu, Dana. 2013. The possessive dative structure. The possessive object. In The grammar of Romanian, ed. G. Pană Dindelegan, 183–190. Oxford: Oxford University Press. Pesetsky, David, and Esther Torrego. 2004. Tense, case, and the nature of syntactic categories. In The syntax of time, ed. J. Guéron and J. Lecarme, 495–538. Cambridge, MA: MIT Press. Schäfer, Florian. 2008. The syntax of (anti-)causatives. External arguments in change-of-state contexts. Amsterdam/Philadelphia: John Benjamins. ———. 2009. The causative alternation. Language and Linguistics Compass 3 (2): 641–681.

218

A. Cornilescu and A. Nicolae

Svenonius, Peter. 2007. Adpositions, particles and the arguments they introduce. In Argument structure, ed. E.J. Reuland, T. Bhattacharya, and Giorgos Spathas, 63–103. Amsterdam/Philadelphia: John Benjamins. Talmy, Leonard. 1978. Figure and ground in complex sentences. In Universals of human language, ed. J. Greenberg, C. Ferguson, and H. Moravcsik, 627–649. Stanford, CA: Stanford University Press. Wood, Jim. 2012. Icelandic morphosyntax and argument structure. PhD diss., New York University. Wood, Jim, and Alec Marantz. 2017. The interpretation of external arguments. In The verbal domain, ed. Roberta D’Alessandro, Irene Franco, and Ángel Gallego, 255–278. Oxford: Oxford University Press.

Causative SE: A Transitive Analysis Grant Armstrong and Paula Kempchinsky

Abstract In this paper we offer an analysis of the so-called causative SE construction exemplified by sentences like Juan se afeita en la barbería (‘Juan gets a shave at the barbershop’). Even though these sentences look like ordinary reflexive constructions, the subject is not interpreted as an agent, but as a causer who initiates an event whose agent is implicit. For instance, in the sentence above, the barber is the agent of the shaving event, not Juan. We propose that this unique interpretation arises only in certain transitive configurations that involve change of state semantics and intrinsic reflexivity. This syntactic configuration, coupled with extra-linguistic knowledge about causal chains, accounts for the syntactic and semantic properties of causative SE constructions. Keywords Causative SE · Change of state · Implicit agent

1 Introduction In this paper we offer an analysis of one particular construction in Spanish with the clitic SE, the so-called causative SE construction. These SE constructions are transitive verbs whose subject has a causer reading. Though these look identical to reflexive sentences, there is a crucial difference in interpretation that is illustrated below in (1), and captured somewhat in the English translations. (1) a. Juan se afeita cuidadosamente todos los días Juan SE.3 shave.PRES.3SG carefully all the days ‘Juan carefully shaves every day’ G. Armstrong (*) University of Wisconsin-Madison, Madison, WI, USA e-mail: [email protected] P. Kempchinsky University of Iowa, Iowa City, IA, USA © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_9

219

220

G. Armstrong and P. Kempchinsky

b. Juan se afeita en la barbería para parecer Juan SE.3 shave.PRES.3SG at the barbershop for look.INF más elegante more elegant ‘Juan gets a shave at the barbershop in order to look more elegant’ In (1a), a reflexive construction, Juan is the agent who does the shaving and is also the patient. Cuidadosamente ‘carefully’ modifies the agentive component of this reflexive sentence. (1b) illustrates the causative construction; the subject DP Juan is a causer rather than an agent. The natural interpretation of this sentence is that Juan has gone to the barbershop in order to get a shave from the barber. Thus, the agentive component in (1b) is implicitly attributed to the barber, which is not syntactically expressed. Causative SE is one of the least studied types of constructions in which the SE clitic appears. It is briefly mentioned in Masullo (1992: 236–240) and in Sánchez López (2002: 79) and has been the subject of debates on normative use in the past (Alcaraz 1976; Sabatini 1977) and in various descriptive grammars cited in these works (e.g. Alonso 1968; Gili Gaya 1964). Nonetheless, the Nueva gramática de la lengua española (RAE 2009) does set apart this construction from other sentences with SE, using as a diagnostic the possibility of a paraphrase with the periphrastic hacer causative: Una variante de la interpretación causativa es la llamada factitiva, en la que el referente del sujeto no ejecuta de modo efectivo la acción expresada por el verbo, sino que hace que otros la realicen en su lugar. El que dice Me hice un traje no está manifestando necesariamente que lo confeccionó él mismo, lo que da lugar a la paráfrasis ‘Me lo hice hacer’ (op.cit. Vol. II. p. 2624). [A variant of the causative interpretation is the so-called factitive, in which the referent of the subject does not perform the action expressed by the verb, but rather causes that others carry it out instead. He who says Me hice un traje lit. ‘I made myself a suit’ is not saying necessarily that he himself sewed the suit, which gives rise to the paraphrase ‘Me lo hice hacer’ ‘I had it made for me’].1

In Sect. 2 below, we will return to this diagnostic. Examples of this construction in Spanish are not hard to find; below are instances from a rather cursory search on the internet:

Note that this differs from the use of the term “factitive” in English, where the standard definition of a factitive verb is one with a small clause complement which represents the result of the verbal action, as in The idiots elected him president.

1

Causative SE: A Transitive Analysis

221

(2) a. El 56% de la población mayor de 65 años se vacunó contra la gripe este año ‘56% of the population older than 65 got vaccinated against the flu this year’ http://www.abc.es/fotonoticias/fotos-sociedad/20151002/poblacion-mayoranos-vacuno-1622111674113.html b. El Chapo se operó para mejorar su desempeño sexual ‘El Chapo had an operation to improve his sexual prowess’ http://www.elmundo.es/internacional/2016/01/15/ 56991a6b268e3eb17b8b460f.html c. Yo me examiné para ser músico callejero ‘I took an exam to be a street musician’ http://www.lavanguardia.com/cultura/20131204/54395813748/yo-meexamine-para-ser-musico-callejero.html d. Y yo me peiné en la peluquería Parra y me maquillé en casa de mi esteticién ‘And I had my hair combed in the Parra hair salon and had my make-up done at my esthetician’s house’ http://www.eldiariodelanovia.com/nuestro-gran-dia-la-boda-de-n/ e. Se bautizó para demostrar que había decidido hacer la voluntad de Dios ‘He got baptized in order to show that he had decided to follow God’s will’ http://wol.jw.org/es/wol/d/r4/lp-s/2012247 The examples above all show the verb in a conjugated form (in fact, in all cases, the preterite). Nevertheless, infinitival examples can also be found: (3) a. Dado el gran índice de peticiones para examinarse del permiso de conducir tipo B. . . ‘Given the high number of requests to take the exam for type B of the driving license. . . http://www.dgttest.org/ b. ¿Cuánto cuesta teñirse en la peluquería? ‘How much does it cost to get (your hair) dyed at a hair salon?’ https://es.answers.yahoo.com/question/index/ c. Afeitarse es un placer en la barbería ‘Getting a shave is a pleasure at the barbershop’ http://www.labarberiadesde1966.com/ Finally, all of the examples presented thus far contain verbs in which the causer is also the patient, but the causer may also be interpreted as the possessor of the direct object of the verb. The most common examples of this type are ones in which the direct object is a body part and the causer is the inalienable possessor (4a) but there are also examples where a benefactive possessor is possible (4b).

222

G. Armstrong and P. Kempchinsky

(4) a. Creen que Lady Gaga se operó la nariz . . . Es una de las cirugías estéticas más conocidas ‘They think that Lady Gaga got a nose job . . . It’s one of the most widely known esthetic surgeries’ http://www.chismevlog.com/creen-que-lady-gaga-se-opero-la-nariz/ b. Una de mis experiencias en cada uno de mis viajes a Asia ha sido el hacerme un traje a medida por un sastre local ‘One of my experiences in each of my trips to Asia has been to Have a suit custom-made by a local tailor’ http://www.3viajes.com/hacerse-un-traje-a-medida-en-asia/ The main objective of this paper is to bring causative SE constructions into the theoretical spotlight by proposing a novel syntactic analysis of them. In Sect. 2 we present the descriptive properties of causative SE constructions in Spanish. Section 3 discusses the difference between complex and simplex reflexive constructions and presents evidence against extending the unaccusative analysis of causative reflexive sentences proposed for Dutch by Rooryck and Vanden Wyngaerd (2011) to Spanish. In Sect. 4 we present the details of our analysis. We propose that causative SE sentences involve special interpretations of a transitive change of state construction in which the grammatical subject is an external causer of a change of state that is represented as a small clause. The predicate of the small clause is the verbal root while the subject can either be SE itself or a DP to which SE is linked through a possessive relation. Finally, in Sect. 5 we offer some concluding remarks with a brief cross-linguistic and cross-dialectal comparison of causative SE constructions.

2 Properties of the Causative se Construction The causative SE construction overlaps in a number of ways with other SE constructions, but the particular constellation of properties that it displays are unique. First, like all constructions with SE except true (‘extrinsic’) reflexives, it disallows doubling of the clitic: (5) a. Juan se afeita cuidadosamente a sí mismo Juan SE.3 shave.PRES.3SG carefully DOM himself b. Juan se afeita (*a sí mismo) en la barbería Juan SE.3 shave.PRES.3SG (*DOM himself) at the barbershop (6) a. Me teñí el pelo a mí misma por primera vez SE.1SG dye.PRET.1SG the hair DOM myself for first time la semana pasada con ese tinte the week last with that dye ‘I dyed my own hair for the first time last week with that hair dye’

Causative SE: A Transitive Analysis

223

b. Me teñí el pelo (*a mí misma) en la peluquería SE.1SG dye.PRET.1SG the hair (DOM myself) at the salon en preparación para la ceremonia in preparation for the ceremony’ ‘I had my hair dyed at the hair salon in preparation for the ceremony’ In the prohibition against doubling, the causative construction overlaps, for example, with anticausative SE and passive SE, among others: (7) a. Se secó la ropa (*a sí misma) a pesar de las SE.3 dry.PRET.3SG the clothing (DOM itself) in spite of the nubes clouds ‘The clothing dried in spite of the clouds’ ANTICAUSATIVE b. Se construyen muchas casas (*a sí mismas) por la zona SE.3 build.PRES.3P many houses (*DOM themselves) in the area ‘Many houses are being built in the area’ PASSIVE Secondly, as noted above, the DP subject of the causative SE construction, like that of a reflexive SE construction, is interpreted in relation to two thematic roles: causer and theme in the case of the former, and agent and theme in the case of the latter. In the association with the role of theme, there is overlap with inchoative SE, middle SE and passive SE, but in the case of these constructions, the grammatical subject is not associated to any other thematic role. Thirdly, the causer subject can clearly control the implicit subject of a purpose adjunct, as in (1b) and (2b,c) above. Since in the presence of the causer subject the actual agent of the action becomes implicit, there does appear to be some overlap with passive SE constructions. Nevertheless, there are two important differences between the causative construction and the passive construction. One is the fact that the causative SE construction is completely paradigmatic, while the passive SE construction is limited to third person arguments. The other is that the implicit agent in the passive SE construction is syntactically available to control a purpose adjunct, as in (8a) (Mendikoetxea 1999),2 which is completely impossible with the causative SE construction: (8) a. Se construyeron esas casas con materias baratas para SE.3 build.PRET.3P those houses with materials cheap for después vender.las a unos precios módicos after sell.INF.them at some prices reasonable ‘Those houses were constructed with cheap materials in order to be able to sell them afterwards at reasonable prices’

2

There has been some debate over the years as to whether the implicit agentive argument in passive SE constructions is syntactically projected; for an analysis which proposes that this is indeed the case, see MacDonald (2017).

224

G. Armstrong and P. Kempchinsky

b. Isabel se tiñó el pelo en la peluquería para Isabel SE.3 dye.PRET.3SG the hair in the salon for experimentar con ese nuevo tinte experiment.INF with that new dye ‘Isabel got her hair dyed at the hair salon to try out that new dye’ In (8b), the controller of the infinitival purpose adjunct can only be Isabel, not the semantically implicit agent of the hair-dyeing. In the introduction we noted the diagnostic employed in the RAE (2009) grammar of paraphrase with the periphrastic causative construction. Interestingly enough, however, there is an important difference between the causative SE construction and the corresponding periphrastic hacer causative: the latter allows the presence of a syntactically active implicit agent, while the causative SE construction does not: (9) a. #Juan se afeita cuidadosamente en la barbería Juan SE.3 shave.PRES.3SG carefully at the barbershop para parecer más elegante for look.INF more elegant Intended: Juan gets a shave carefully at the barbershop in order to look more elegant’ b. Juan se hace afeitar cuidadosamente en la Juan SE.3 make.PRES.3SG shave.INF carefully at the barbería para parecer más elegante barbershop for look.INF more elegant ‘Juan has himself carefully shaved at the barbershop in order to look more elegant’ (9a) at best is semantically anomalous; the presence of the adverb cuidadosamente ‘carefully’ forces the agentive reading of the subject Juan, which clashes with the locative en la barbería. In (9b), in contrast, cuidadosamente modifies the action of the implicit agent, while the causer subject Juan controls the purposive infinitival adjunct para parecer más elegante ‘to look more elegant’. The causative paraphrase diagnostic brings us to another issue: the question of whether these interpretations are necessarily linked to the presence of SE or not. Some researchers claim that there is not necessarily a link between a causer interpretation like the one described above and SE. This claim is based on the observation that both agent and causer are also available for verbs that lack SE as in (10). (10) Carlos III construyó la Puerta de Alcalá ‘Carlos III built the Puerta de Alcalá’ (Sánchez López 2002: 79) In principle, so the argument goes, (10) may have an agentive interpretation or a causer interpretation. Pragmatics dictates which one is more natural, thus explaining why we get the causer interpretation in (10). This line of argumentation claims, then,

Causative SE: A Transitive Analysis

225

that the fact that verbs like afeitarse ‘to shave’ are reflexive makes it appear that SE is responsible for a causative reading when in reality it is not. The clitic SE is there to mark reflexivity rather than a causative reading. The causative reading is something that arises independently of SE. The Nueva gramática de la lengua española (RAE 2009) explicitly addresses this issue, noting that it is, at best, controversial (“discutible”, op.cit. Vol. II. p. 2624) whether verbs such as construir ‘to build’ should be analyzed in a parallel fashion to examples such as operarse ‘to have an operation’ and cortarse el pelo ‘to get one’s hair cut’. Two reasons are presented for not grouping verbs like construir with the causative SE construction. First, the RAE claims that for these verbs the salient interpretation is that the subject is a causer, not an agent, which is not the case for the verbs with SE, noting that there are many actions “which can be attributed to the person who is in charge or responsible for them, without implying that this person realizes the action with his own hands” (pueden atribuirse a quien se las encarga o es responsable de ellas, sin implicar que las lleve a cabo con sus propias manos). Secondly, they fail the translation test: they are directly translatable to other Romance languages (and, we would add, English) without requiring the periphrastic hacer causative paraphrase. In terms of the division of transitive verbs proposed by Armstrong (2011), verbs such as construir fall into his Class 1, for which v has the value or ‘flavor’ (Folli and Harley 2005) of CAUSE, while verbs such as operar fall into Class 2, for which v has the value of DO. The question of the value of v with the verbs in the causative se construction will be crucial in our analysis. For the moment, the conclusion which we want to emphasize here is that the presence of SE is indeed crucial for the causative interpretation to arise, overriding, as it were, the agentive interpretation. That being so, it is not the case that any verb with an agentive subject and which can appear in a reflexive SE construction allows the causative interpretation, and even those that do may not always freely yield that interpretation: (11) a. Yo me examiné (a mí mismo) cuidadosamente I SE.1SG examine.PRET.1SG (DOM myself) carefully para asegurarme de que el gato no me había for assure.INF.me of that the cat NEG me have.IMP.3SG mordido bitten ‘I examined myself carefully to make sure that the cat had not bitten me’ b. #Yo me examiné (*a mí mismo) en la clínica para I SE.1SG examine.PRET.1SG (DOM myself) at the clinic for asegurarme de que la herida no se había infectado assure.INFme of that the wound NEG SE.3 have.IMP.3SG infected ‘I had myself examined at the clinic in order to make sure the wound hadn’t become infected’

226

G. Armstrong and P. Kempchinsky

c. Yo me examiné (*a mí mismo) al final del I SE.1SG examine.PRET.1SG (DOM myself) at.the end of.the curso en matemáticas y en inglés course in math and in English ‘I took exams at the end of the academic year in math and in English’ (12) a. El paciente se observó (a sí mismo) en el espejo The patient SE.3 observe.PRET.3SG (DOM himself) in the mirror mientras hacía los ejercicios de rehabilitación while do.IMP.3S the exercises of rehabilitation ‘The patient observed himself in the mirror as he was doing the rehabilitation exercises’ b. El paciente se observó en la clínica para The patient SE.3 observe.PRET.3SG at the clinic for asegurarse de que la herida no se había infectado assure.INF.SE of that the wound NEG SE.3 have.IMP.3SG infected ‘The patient observed himself at the clinic to be sure that the Wound had not become infected’ # ‘The patient got observed at the clinic to be sure that the wound had not become infected’ (11c) represents a typical example of the verb examinar in the causative SE construction (see also (3a)). However, in (11b) with the same verb this interpretation is judged to be infelicitous. In (12b) with the verb observar ‘to observe’, the causative reading is unavailable, yielding a pragmatically odd interpretation.3 Rooryck and Vanden Wyngaerd (2011) (henceforth R and VW) describe a similar phenomenon in Dutch simplex reflexive verbs, whereby for a subset of these verbs a causative interpretation is also available, as in (13): (13) a. Frans heeft zich gevaccineerd tegen de griep Frans has REFL vaccinated against the flu ‘Frans had himself vaccinated (by a doctor) against the flu’ b. Frans heeft zichzelf gevaccineerd tegen de griep Frans has REFL.self vaccinated against the flu ‘Frans vaccinated himself against the flu’ (op. cit. 104–105, ex. 140) In (13a), a simplex reflexive marked by zich, a causer interpretation is the most natural one whereby the vaccination itself is carried out by a specialist such as a doctor, although the agentive interpretation is also available. In (13b), a complex 3

Note that (3a) is from peninsular Spanish (the official website of the Dirección General del Tráfico), and the speakers who were consulted on the felicity of (11b) vs. (11c) are all peninsular speakers. In contrast, at least some of the Latin American speakers we consulted also rejected (11c). See Sect. 5.

Causative SE: A Transitive Analysis

227

reflexive marked by zichzelf, only the agentive interpretation is available, in which Frans is the one who performs the vaccination. R and VW list a number of verbs for which this alternation holds: (14) aanmelden ‘register’, bekeren ‘convert’, bevrutchten ‘impregnate’, bijspijkeren ‘brush up’, castreren ‘castrate’, dopen ‘baptize’, inloten ‘draw a place by lot’, inschrijven ‘register’, intekenen ‘subscribe’, inwerken ‘break in’, inwijden ‘initiate’, klaarstomen ‘cram’, kleden ‘dress’, klonen ‘clone’, kwalificeren ‘qualify’, omscholen ‘retrain’, ontmaagden ‘deflower’, oplappen ‘doctor up’, overtuigen ‘convince’, schminken ‘apply make-up’, wassen ‘wash’, scheren ‘shave’, steriliseren ‘sterlize’ (op cit: 106, ex. 143) They claim that the class can be subdivided semantically into those verbs that describe events of ‘ritual activities’ (initiations, etc.), ‘medical interventions’ (vaccinations, etc.), ‘(re)training’ (brushing up on something, etc.) and ‘grooming’ (dressing, washing, etc.). What the entire class of verbs shares is that they are compatible, modulo certain cultural differences, with a situation in which certain agentive properties of the event can be distributed between the syntactically expressed subject and an implicit “specialist”, who actually carries out the action. We return to this observation in Sect. 5. Provisionally, a sample list of verbs in Spanish for which we have found relevant data (primarily, the Corpus del español (Davies 2016) and internet searches) is given in (15): (15) afeitar ‘to shave’, bautizar ‘to baptize’, cortar (el pelo) ‘to cut (one’s hair)’, dar de alta/baja ‘to register/unregister’, examinar ‘to examine’, hacer ‘to make’, maquillar ‘to put on make-up’, operar ‘to operate’, peinar ‘to comb’, teñir (el pelo) ‘to dye (one’s hair)’, vacunar ‘to vaccinate’ These verbs clearly fall into R and VW’s classes: ritual activities (bautizar), medical interventions (vacunar), training (examinar) and personal grooming (afeitar). The fact that the Dutch example of (13a) has two possible interpretations—the causative and the agentive—while (13b) does not, raises the question of whether the causative interpretation is dependent in some way on the nature of the reflexivity: simplex vs. complex. That is the focus of the next section.

3 Complex/Simplex Reflexive Constructions and Causative se Above we observed that the causative SE construction, unlike reflexive SE, disallows doubling of SE. However, this is a simplification of the situation. Consider the data below:

228

G. Armstrong and P. Kempchinsky

(16) a. El médico se vacunó a sí mismo contra la gripe ‘The doctor vaccinated himself against the flu’ b. El médico se vacunó contra la gripe. REFLEXIVE AGENTIVE: ‘The doctor vaccinated himself against the flu’ CAUSATIVE: ‘The doctor got vaccinated against the flu’ (16b), without doubling of SE, has two interpretations, as is true for (13a)—and now, given the referent of the subject, the causative interpretation and agentive interpretations are equally available, in pragmatic terms. Since Reinhart and Reuland (1993) it has been generally assumed that there is a structural difference between simplex reflexives, such as zich in Dutch or non-doubled SE in Spanish, and complex reflexives, represented by zichzelf and doubled SE. In R and VW’s account of Dutch examples such as (13a), they explicitly derive the causative interpretation from the structure which they propose for simplex reflexives in general, represented in (17b), in which an Agree relation established between zich in the specifier of the RP and the DP Milo in the complement position serves to value the [uϕ] on zich: (17) a.

Milo heeft zich bezeerd Milo has self hurt

b.

VP V bezeer

RP R¢

DP1 zich[uf] R

DP1 = possessum DP2 Milo[if]

DP2 = possessor

(17b) shows an unaccusative syntax for sentences with the simplex reflexive zich, with a R(elator)P (den Dikken 2006) structurally identical to a RP in inalienable possession structures; the DP Milo eventually merges with T to satisfy the EPP. R and VW argue that the DP is an Experiencer rather than an Agent (on the basis of, among other pieces of evidence, the distribution of certain kinds of adjunct phrases). In contrast, with the full reflexive zichzelf, there is a transitive construction with an agentive subject. More specifically, they argue that the DP argument in a structure such as (17b), like the subject of any unaccusative verb, is not necessarily not an agent, given that if the verb above were wassen ‘to wash’, the most pragmatically felicitous interpretation would be that the subject voluntarily carried out the act of washing. They observe that even with core cases of unaccusative verbs such as ‘to arrive’, an agentive reading for the subject is possible if such a reading is pragmatically

Causative SE: A Transitive Analysis

229

coherent, based on the semantic properties of the DP argument (cf. The students/ #The train arrived late on purpose). However, given that the structure does not force an agentive reading for the subject (in R and VW’s terms, the subject has reduced agentivity), the possibility then arises, with verbs which express specialized activities, that the DP argument can be understood as the initiator of the action, while the non-syntactically expressed specialist is in fact the agent or ‘doer’. This possibility does not arise with zichzelf in (13a) because in this case there is a transitive structure whose external argument is a full-fledged agent. In a nutshell, the essence of the R and VW analysis is that the ‘specialist’ reading is a pragmatic consequence of the interaction between the unaccusative syntax of sentences with zich, entailing the possibility of “reduced agentivity”, and speaker knowledge of which activities can possibly be realized by a specialist. The presence of the simplex reflexive form is necessary to give rise to the syntactic structure which then lends itself to the pragmatic interpretation. Taking this analysis as a point of departure, we need to examine two questions: (1) what is the structure of complex and simplex reflexives in Spanish, and (2) is the possibility of a causative interpretation directly linked to the possibility of a given verb appearing in an independently motivated simplex reflexive structure. We assume, following Torrego (1995) and subsequent developments outlined in Ormazabal and Romero (2013), that complex reflexives are full transitive structures in which the direct object position is occupied by a sí mismo and the clitic SE is the reflex of an agreement relation between transitive v and the direct object. In (18b) below the agreement relation is represented with a dotted line and the arrow “⇨” is meant to represent the spell out of this agreement relation. (18) a.

El niño se vistió a sí mismo. the child SE.3 dress.PRET.3SG DOM himself vP

b.

v'

DP vDO  SE

V

VP DP a sí mismo

We also assume that this particular agreement relation is limited to agentive v, or vDO, which accounts for why transitive sentences with a sí mismo have agentive readings and are usually limited to contrastive focus contexts in which a sí mismo is the focus, representing one alternative to other non-reflexive direct objects that can also appear in the same syntactic position. Turning to Spanish simplex reflexives, we could simply adopt R and VW’s proposed structure in (17b); they in fact specifically extend their unaccusative

230

G. Armstrong and P. Kempchinsky

analysis of Dutch simplex zich to French reflexive se (see Sportiche 2014 for an alternative view). However, a few problems with extending this approach to Spanish and other Romance languages have been observed in previous research. The main issue raised by the unaccusative analysis of simplex reflexives is that many simplex reflexive constructions either do not pass unaccusativity diagnostics or do not behave like other unaccusative predicates in certain grammatical environments. Mendikoetxea (2012) explicitly rejects an unaccusative analysis of all simplex reflexive SE sentences based on the failure of some of them to comply with a number of unaccusative diagnostics, such as the possibility of bare postverbal subjects in Spanish or of partitive ne-cliticization in Italian (Burzio 1986). She notes further that simplex reflexives like (19) resist an unaccusative analysis because of the possibility of reflexive datives which co-occur with accusative-marked direct objects, as in (19): (19) a. Juan se ha comprado un coche Juan SE.3 have.PRES.3SG bought a car ‘Juan has bought himself a car’ (Mendikoetxea 2012: 489 ex. 23) b. El pelo me lo corto cada dos meses the hair SE.1SG it.ACC cut.PRES.1SG each 2 months ‘My hair, I get it cut every 2 months’ A second reason not to adopt the unaccusative analysis of certain simplex reflexives is due to observations regarding verbs of personal grooming by Otero (1999) (and earlier by Saltarelli 1994). Such verbs in effect have the possibility of appearing in either the complex or the simplex reflexive structure, and in many cases the simplex structure is ‘preferred’, and can be forced, for example, by the presence of manner or instrumental adjuncts in the sentence: (20) a. Juan se lava con jabón La Toja (??a sí mismo) Juan SE.3 wash.PRES.3SG with soap La Toja (??DOM himself) ‘Juan washes (??himself) with La Toja soap’ b. Juan se viste siempre modestamente (??a sí mismo) Juan SE dress.PRES.3SG always modestly (?? DOM himself) ?? ‘Juan always dresses ( himself) modestly’ (Otero 1999:1467 ex. 111a,d) He suggests that the existence of such examples raises some doubt about analyzing all simplex reflexives as unaccusative structures since these examples do not exhibit any characteristics of reduced agentivity and contain instrumental and manner modifiers that require the presence of an agent. Following these observations we propose that at least some simplex reflexives like vestir-se ‘to dress’, afeitar-se ‘to shave’ and bañar-se ‘to bathe’ are essentially transitive structures, but instead of acting as the reflex of an agreement relation between an internal argument position and v, the clitic itself is generated in the internal argument position and is licensed by an agreement relation with v. Once this agreement relation is established, the clitic moves to clitic position and adjoins to the

Causative SE: A Transitive Analysis

231

verb (see Uriagereka 1995 for a classic reference and Kramer 2014 for a more recent proposal regarding details about agreement and cliticization). This is shown in (23). (23)

vP DP

v' vDO V

VP D SE

Our proposal is that the clitic SE can either be treated as a reflex of agreement with an internal argument position, in which case doubling results, or it may be treated as the argument itself, in which case there is no doubling. In the former scenario, the internal argument position can in principle be saturated by any DP. The reflexive interpretation arises through binding and se appears as the spell out of agreement with a bound DP. Such a relation has been labeled extrinsic reflexivity in the literature (Reinhart and Reuland 1993; Otero 1999). The latter scenario is representative of a different type of reflexive relation, known as intrinsic reflexivity. In the current system, intrinsic reflexivity is represented as a type of transitive construction in which one of the argument positions of a syntactically transitive construction can only be filled with a reflexive clitic and only a reflexive interpretation is possible. Following ideas outlined in Wood (2014) (and references therein), we might informally characterize the difference between the extrinsic reflexive construction in (18) and the intrinsic one in (23), both built from the same verb root vestir ‘to dress’, as in (24). (24) a. e is an event of x dressing y & x ¼ y (18) b. e is an event of x dressing x (23) The intuition here is that intrinsic reflexivity is semantically intransitive in that it involves a single argument variable, but it is syntactically transitive as there are two argument positions. Reflexive clitics are expletive elements that fill one of the argument positions that are obligatorily linked to the same variable (see Wood 2014 for a detailed discussion). Note now that lavar(se) ‘to wash’ and vestir(se) ‘to dress’, like other prototypical examples of simplex reflexive constructions, fall into the class of personal grooming verbs, one of the verb classes identified by R and VW (2011) as potential participants in the causative interpretation. Further, the simplex reflexive structure disallows doubling, by definition, and this is also true of the causative SE construction, as we saw above. This would appear to indicate that the causative SE construction derives

232

G. Armstrong and P. Kempchinsky

in some way from the agentive simplex reflexive structure of (23). In fact, the most straightforward analysis, assuming v to be a functor head with a limited set of values such as DO, CAUSE and BECOME (Folli and Harley 2005), would be to propose that the causative SE construction derives directly from the simplex agentive reflexive construction, but with a change in the value of v from DO to CAUSE—a change mediated in some way by the presence of SE. Part of our analysis does indeed involve a change in the value of v, but this change is not as straightforward as it might appear. Consider again example (16), repeated below as (25), with the verb vacunar ‘to vaccinate’—not a verb of personal grooming: (25) a. El médico se vacunó a sí mismo contra la gripe ‘The doctor vaccinated himself against the flu’ b. El médico se vacunó contra la gripe ‘The doctor vaccinated himself/got vaccinated against the flu’ Clearly (25a) must have the structure of (18b). Leaving aside the causative SE reading of (25b), it would seem, at first glance, that the agentive reading of (25b) should have the structure of (23): there is no doubling of SE with a sí mismo. Nevertheless, we assume that the agentive reading of (25b) in fact corresponds to the structure in (18b) without an overt a sí mismo double. That is, an extrinsically reflexive construction always has an internal argument position filled by a sí mismo or a (null) pronoun (see Torrego 1995 for details). Note that with vacunar, the presence of instrumental or manner adjuncts does not affect the acceptability of a sí mismo: (26) a. El médico se vacunó torpemente a sí mismo The doctor SE.3 vaccinate.PRET.3SG clumsily DOM himself ‘The doctor clumsily vaccinated himself’ b. El médico se vacunó con una jeringa The doctor SE.3 vaccinate.PRET.3SG with a syringe contaminada a sí mismo contaminated DOM himself ‘The doctor vaccinated himself with a contaminated syringe’ To take an example from another class of verbs that enter into the causative se construction, the same is true of the verb bautizar ‘to baptize’: (27) El ermitaño se ha bautizado con el agua de la The hermit SE.3 have.PRES.3SG baptized with the water of the fuente a sí mismo fountain DOM himself ‘The hermit has baptized himself with the water from the fountain’ A second piece of evidence against a simplex reflexive structure for the agentive reading of (25b) is related to the behavior of the clitic SE under causative hacer.

Causative SE: A Transitive Analysis

233

Kempchinsky (2004) notes that the acceptability of doubling of SE correlates negatively with the acceptability of omission of SE in Spanish under causative hacer. Specifically, omission of SE is optional in cases that resist doubling (with some variability across speakers, indicated with ‘%’ in 28), but impossible in cases that fully allow doubling,4 as in (29). In such cases, the omission of SE necessarily gives rise to a non-reflexive interpretation in which there is an implicit agent. (28) a. La madre hizo bañar (%se) al niño The mother make.PRET.3SG bathe.INF.(SE.3) DOM.the child ‘The mother made the child bathe (himself)’ b. Su novia hizo afeitar (%se) a Juan His girlfriend make.PRET.3SG shave.INF.(SE.3) DOM Juan ‘His girlfriend made Juan shave (himself)’ (29) a.

La profesora hizo criticarse a los The professor make.PRET.3SG criticize.INF.SE.3 DOM the estudiantes students ‘The professor made the students criticize themselves/each other’ b. La profesora hizo criticar a los estudiantes The professor make.PRET.3SG criticize DOM the students ‘The professor had someone criticize the students’ * ‘The professor made the students criticize themselves/each other’

In (30), we see that vacunar ‘to vaccinate’ behaves not like the simplex agentive reflexives in (28) but rather like the complex ones, like (29). Reflexive and causative SE interpretations are only possible when SE is present. When it is absent, only the implicit agent interpretation like of that of (29b) is possible. (30) a. El médico hizo vacunarse a los viajeros The doctor make.PRET.3SG vaccinate.INF.SE.3 DOM the travelers ‘The doctor made the travelers vaccinate themselves/get vaccinated’ b. El médico hizo vacunar a los viajeros The doctor make.PRET.3SG vaccinate.INF DOM the travelers ‘The doctor had someone vaccinate the travelers’ Thus, although with the class of verbs in the personal grooming category there is overlap between the possibility of appearing with the simplex reflexive SE in the agentive reflexive and the possibility of appearing in the causative SE construction, and although the causative SE construction, like the simplex agentive reflexive, disallows doubling, we conclude that the causative SE construction does not derive in some way from the simplex agentive reflexive, with some reinterpretation of the 4 This contrasts, for example, with Italian, which systematically disallows the clitic si under the causative verb fare (Burzio 1986).

234

G. Armstrong and P. Kempchinsky

subject theta-role from agent to causer or initiator.5 Further, the fact that various verbs which enter into the construction can only otherwise appear in the transitive extrinsic reflexive structure, as we have argued (e.g. vacunar ‘to vaccinate’, bautizar ‘to baptize’), is evidence against attributing the possibility of the causative interpretation to some kind of “reduced agentivity” on the part of the subject, as in R and VW’s account.

4 Analysis In order to remedy the problems associated with deriving the interpretations associated with causative SE from other simplex reflexive constructions (unaccusative or agentive), we propose that causative SE constructions are built from a causative transitive syntactic structure that involves a transitive vcause and a small clause complement (Folli and Harley 2005; Schäfer 2008; Marantz 2013; among many others). Before getting into the details of how the analysis works, let us briefly give a background of the transitive causative construction. As discussed at length in previous work, the main components of this construction consist of a causer in the specifier of the vP, a theme that is the subject of the small clause complement and the verbal root that acts as the predicate of the small clause and is interpreted as a result state. A common class of verbs that have been proposed to instantiate this construction are change of state verbs like romper (break) and abrir (open). The representation of (31a) is shown in (31b). The verbal root moves to vcause and then to the inflectional layer of the clause before being spelled out as abrió. (31) a.

Ella abrió la puerta ‘She opened the door’ vP

b.

v´

DP ella

vcause

SC DP

√abr(ir)

la puerta

5 In general terms this puts us in agreement with Otero (1999), who briefly considers the causative vs. agentive interpretation of a sentence such as Juan se afeita (en la peluquería/en casa) ‘John shaves (at the salon/at home), stating that the causative originates in a transitive structure, while the other is an intrinsic reflexive.

Causative SE: A Transitive Analysis

235

Cases like (31) typically involve direct causation whereby the causer argument is conceived as the direct cause of the result state described by the small clause. Barring extra information that specifies otherwise, the pronoun ella in (31a) is interpreted as the person who opens the door rather than someone who initiates a chain of events that causes someone or something else to open the door (some roots like construir lend themselves more naturally to such interpretations as discussed in Sect. 2 but this is an issue related to encyclopedic knowledge rather than grammar). Our proposal is that causative SE constructions should be represented using the same syntactic construction as (31b). The main difference is that there is an intrinsic reflexive relation within this construction that is realized by the se clitic. Based on the verbs we have found in our searches, we claim that there are three variants of the causative SE construction that differ with respect to where se is generated. These are shown in (32). (32) Variants of the causative se construction a.

SE = Theme bautizar-se (get baptized)

b. SE = Possessor cortar-se el pelo (get a haircut)

vP

vP v´

DP

v´

DP

vcause

vcause

SC √Root

D SE

SC

D SE

Poss´ Poss

c.

SE = Benefactor hacer-se un traje (have a suit made) vP v´

DP vcause

ApplP D SE

Appl´ Appl

SC DP

√Root

√Root

PossP

DP

236

G. Armstrong and P. Kempchinsky

In (32a), a causer initiates a change of state that is intrinsically reflexive so the theme position is realized by SE. The licensing of SE works in the manner discussed in Sect. 3 for simplex reflexives generally. It enters an agree relation with v and then moves to a clitic position higher in the clause. In (32b), a causer initiates a change of state whose theme is an inalienably possessed body part. The role of SE here is to syntactically realize the possessor role. We claim that licensing of SE follows the same mechanism as in (30a) and remain agnostic as to whether the possessed DP body part is also licensed by v through multiple agree (MacDonald 2015) or through some other mechanism (Ormazabal and Romero 2013). Finally, in (32c), the causer initiates a chain of events that leads to him/her benefiting from the resultant state such as ‘getting a tailored suit.’ Following Cuervo (2003), we claim that se in these cases is an affected applicative, which is merged between the vP and the small clause complement. We argue that the interaction of the constructional meaning of the transitive causative construction with the more specific meaning components added by SE and the VP is what gives rise to the unique causative interpretation of these constructions. There are three major factors at play. First, there is a causing event that leads to a result state. Second, there is an intrinsically reflexive relation of causing and being affected by some action that is linked to the presence of SE. Third, there is the encyclopedic knowledge associated with VP-internal components that are responsible for inferences regarding how a process typically unfolds in a causal chain of events. The first two factors are the minimal grammatical requirements for a causative interpretation of the type described here to arise while the third factor is external to the grammar but crucial to determining the possibility of a causative interpretation. With verbs like vacunar-se ‘get vaccinated’ and cortar-se el pelo ‘get one’s hair cut’, these three factors conspire to give rise to an inference that an implicit agent performed the action that is initiated by and ultimately affects the causer. In this sense, we are in agreement with R and VW (2011) that certain types of actions that are instantiated by simplex reflexives are not generally performed by whatever DP acts as the grammatical subject. Where we differ from these authors is in treating such interpretations as arising from syntactically transitive rather than unaccusative constructions. By making this move, we avoid the problems with deriving the interpretation of causative SE constructions from unaccusative or agentive simplex reflexives already mentioned in Sect. 3. Recall that not all verbs that appear in the causative SE construction can appear in other simplex reflexives. In our analysis, causative SE is not derived from another simplex reflexive but rather constitutes its own type of simplex reflexive that depends crucially on the constructional meaning of a causative change of state syntactic structure. Another important aspect of our analysis, which is also shared with that of R and VW (2011), is that the implicit agent is not projected syntactically. The implicit agent in causative SE constructions can be specified with a PP headed by con ‘with’ or it can be implicated by adding a locative PP headed by en ‘in, at’ that describes the place in which the agent typically performs the relevant procedure.

Causative SE: A Transitive Analysis

237

(33) a. Juan se vacunó con el médico Juan SE.3 vaccinate.PRET.3SG with the doctor ‘Juan got vaccinated at the doctor’s/by the doctor’ b. Juan se afeita en la barbería de la esquina Juan SE.3 shave.PRES.3SG in the barbershop of the corner ‘Juan gets shaved at the barbershop on the corner’ However, as we noted in Sect. 2, a major difference between the implicit agent in causative se constructions and passives/periphrastic causatives is that agent-oriented adverbs are not licensed by the implicit agent. In (34a), if Juan is interpreted as the causer—an interpretation pragmatically induced by the locative en la barbería— then cuidadosamente ‘carefully’ has no agent to modify. In (34b), repeated from (8b) above, only the causer, not the implicit agent, can control into a purpose clause. (34) a. #Juan se afeita cuidadosamente en la barbería para Juan SE.3 shave.PRES.3SG carefully at the barbershop for parecer más elegante look.INF more elegant b. Isabel se tiñó el pelo en la peluquería para Isabel SE.3 dye.PRET.3SG the hair at the hair salon for experimentar con ese nuevo tinte experiment.INF with that new dye ✓ Isabel wanted to experiment with the new hair dye ✗ The hairdresser (implicit agent) wanted to experiment with the New hair dye Our claim is that the data in (34) fall out from treating causative SE constructions as transitive constructions like those in (30). Note that there is no agent projected in the syntax, which is what arguably distinguishes causative SE constructions from passives. We claim further that the data in (33a) can be accounted for by treating the PP as a modifier of the vP. This is similar to how Alexiadou et al. (2006) have analyzed causer PPs that appear with anticausatives like romper-se ‘to break’ as in (35) (see Mendikoetxea 1999 for a discussion of causer PPs with anticausatives in Spanish). (35) La ventana se rompió con el viento The window SE.3 break.PRET.3SG with the window ‘The window broke because of the wind’ The idea is that in anticausatives like romper-se, there is a causative vP projected in the syntax that introduces an unspecified cause. Certain classes of PPs can modify this unspecified cause, specifying some part of the causal chain that leads to the result state. The PP headed by con in (31a) can be given a similar treatment in that it modifies part of the causal chain initiated by the causer. In this case, it specifies who the implicit agent is. In this sense, our analysis captures a shared characteristic between anticausatives and causative SE constructions: both have vcause.

238

G. Armstrong and P. Kempchinsky

5 Concluding Remarks Since causative SE constructions have hardly been explored in the literature on Romance languages, we believe that the present study opens a number of avenues of future research. The first concerns the availability of causative SE across Romance. It is not limited to Spanish; in (36) and (37) we provide some examples from Italian (36) and Portuguese (37): (36) a. Mi sono vaccinato per prevenire l’influenza ‘I got vaccinated in order to prevent the flu’ b. Maria si pettina e si trucca in un salone di bellezza il giorno delle nozze ‘Maria has her hair combed and her make-up done in a beauty Salon on her wedding day’ c. Gianni si rade dal barbiere per sembrare raffinato ‘Gianni gets shaved at the barbershop to appear refined’ (37) a. Eu vacinei-me contra a gripe ‘I got vaccinated against the flu’ b. Vai se maquiar no salão? ‘Are you going to get your make-up done at the (beauty) salon? http://www.tudosobremake.com.br/noticia/vai-se-maquiar-no-salao In contrast, French speakers generally reject the construction, almost consistently rephrasing it as the periphrastic causative faire construction:6 (38) a. *Je me suis vacciné pour prévenir la grippe ‘I got vaccinated to prevent the flu’ b. % Je me suis examiné pour obtenir un permis de conduire ‘I got tested to get a driving license’ c. *Marie se peigne et se maquille dans un salon de beauté ‘Marie has her hair combed and her make-up done at a beauty salon’ d. *Jean se rase chez le barbier pour avoir un air d’élégance ‘Jean gets shaved at the barbershop to look elegant’ It is tempting to look for some parametric difference between French and the other languages. As is well-known, constructions with SE in French have a number of differences vis-à-vis Spanish and Italian: for example, the lack of impersonal SE, and properties of the French se-moyen which do not pattern clearly with passive SE or middle SE (Kempchinsky 2006, see Wolfsgruber [this volume] for a diachronic

6 French is perhaps what the authors of the relevant section of the RAE (2009) grammar had in mind, in referring to the necessity of paraphrase “en las otras lenguas romances” for causative se.

Causative SE: A Transitive Analysis

239

treatment of this difference). We leave the question of the status of causative SE in French for future research.7 A second area of future research concerns dialectal variation within Spanish. It is intriguing to note that within Spanish there are also some cross-dialectal differences on exactly which verbs are acceptable in this construction, as we already alluded to above. The prime example is the verb examinar(se), fully acceptable (and widely used, to judge by a superficial internet search) to peninsular speakers, but rejected by Latin American speakers that we consulted (Mexican and Colombian). But, as we noted in example (11), even in peninsular Spanish it is acceptable in the causative SE construction only as a member of the “training” set of verbs, and not in the set of “medical intervention” verbs. Although in generative syntactic analysis we do not typically take into account usage factors, including sociocultural effects, we do want to comment on such effects here, in the light of the differences we have observed. The motivation for this comes from an entirely different study, Cole’s (2016) investigation of serial verbs in Lao, as in (39):8 (39) Nòòj4 piing2 paa5 kin3 Noy grill fish eat ‘Noy grilled the fish and (then) ate it’ (op.cit.:5, ex. 1) (39) is a consecutive serial verb construction in Lao, as captured by the English translation. In his research, Cole found that whether or not two verbs were acceptable to Lao speakers in this construction seemed to hinge on the extent to which the two actions in question were perceived to generally be part of the same macroevent or not (essentially confirming for Lao what had already been found by researchers in other serial verb languages)—i.e., whether the two subevents occurred together with enough regularity for speakers to make that generalization. Thus (40a) below was judged by his informants to be acceptable with the serial verb construction, while (40b) was not: (40) a. Nòòj4 sak1 khùang1 taak5 Noy wash thing hang.up ‘Noy washed the clothes and hung them up’ (op.cit.:127, ex. 6) b. *laaw2 um4 dêk2-nòòj4 suaj1 3SG carry child help Intended: ‘She carried the child and saved him’ (op.cit.:127, ex. 7)

7

The judgments in (38) are based on two informants, both from France and both researchers in linguistics, although not in theoretical syntax. One of them did accept the sentence in (38b)—hence the ‘%’—although she indicated that it could also be paraphrased with causative faire. 8 The numbers after each word represent lexical tones.

240

G. Armstrong and P. Kempchinsky

The claim is not that the existence of the construction in the language can be attributed to the cultural expectations; for a language to allow a serial verb construction a number of (morpho)syntactic properties must be present. The claim, rather, is that the extent to which the syntactic construction is used or the particular lexical items which are acceptable in the construction will depend on these expectations. Returning to the causative SE construction, consider the verb examinar(se), acceptable for peninsular speakers in the ‘training’ class. Examinations are a prevalent feature of many aspects of Spanish civil and educational life, from annual yearend exams in high school and the selectional exam for university study to the threestep testing process for obtaining a driver’s license to the tests administered for civil service positions. This simply leads to frequent use of this verb in a non-agentive reflexive sense, such that it becomes a widely accepted lexical item in the causative SE interpretation. This follows the spirit of R and VW’s (2011) characterization of the causative reading as a pragmatic consequence of speaker knowledge. In a different cultural context or in a different language, other otherwise transitive reflexive verbs might lend themselves to the causative interpretation—insofar as the necessary syntactic structure is available in the language. This second area of investigation is thus related to how extra-grammatical or encyclopedic knowledge of lexical items interacts with grammar in order to determine if particular interpretations are acceptable and how these may vary across speakers.

References Alcaraz, Javier. 1976. Juan se afeita en la barbería: “Sub-standard Spanish?”. Hispania 59: 491–492. Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2006. The properties of anticausatives cross-linguistically. In Phases of interpretation, ed. M. Frascarelli, 187–211. Berlin: Mouton de Gruyter. Alonso, Martín. 1968. Gramática del español contemporáneo. Madrid: Guadarrama. Armstrong, Grant. 2011. Two classes of transitive verbs: Evidence from Spanish. PhD dissertation, Georgetown University. Burzio, Luigi. 1986. Italian syntax: A government and binding approach. Dordrecht: Kluwer. Cole, Douglas. 2016. Lao serial verbs and their event representations. PhD dissertation, University of Iowa. Cuervo, María Cristina. 2003. Datives at large. PhD dissertation, MIT. Davies, Mark. 2016. Corpus del español: Web/Dialects. http://www.corpusdelespanol.org.webdial/. den Dikken, Marcel. 2006. Relators and linkers. Cambridge: MIT Press. Folli, Raffaella, and Heidi Harley. 2005. Flavours of v: Consuming results in Italian and English. In Aspectual inquiries, ed. Paula Kempchinsky and Roumanya Slabakova, 95–120. Dordrecht: Springer. Gili Gaya, Samuel. 1964. Curso superior de sintaxis española. Barcelona: Spes-Vox. Kempchinsky, Paula. 2004. Romance SE as an aspectual element. In Contemporary approaches to romance linguistics: Selected papers from the 33rd linguistic symposium on romance languages, ed. J. Auger et al., 239–256. Amsterdam/Philadephia: John Benjamins.

Causative SE: A Transitive Analysis

241

———. 2006. Teasing apart the middle. In Andolin gogoan/Homenaje a Andolin Eguzkitza, ed. Itziar Laka and Beatriz Fernández, 532–547. Bilbao: University of the Basque Country Press. Kramer, Ruth. 2014. Clitic doubling or object agreement: An Amharic investigation. Natural Language & Linguistic Theory 32: 593–634. MacDonald, Jonathan E. 2015. A case of multiple agree: Accusative, not dative, indirect object se. In Romance linguistics 2012, ed. Jason Smith and Tabea Ihsane, 275–288. Amsterdam/Philadelphia: John Benjamins. ———. 2017. An implicit projected argument in Spanish impersonal and passive se constructions. Syntax 20: 353–383. Marantz, Alec. 2013. Verbal argument structure: Events and participants. Lingua 130: 152–168. Masullo, José Pascual. 1992. Incorporation and case theory in Spanish. A crosslinguistic perspective. PhD dissertation, University of Washington. Mendikoetxea, Amaya. 1999. Construcciones con se: Medias, pasivas e impersonales. In Gramática descriptiva de la lengua española, ed. Ignacio Bosque and Violeta Demonte, 1631–1722. Madrid: Espasa-Calpe. ———. 2012. Passives and se constructions. In The handbook of Hispanic linguistics, ed. José Ignacio Hualde, Antxon Olarrea, and Erin O’Rourke, 477–502. Oxford: Blackwell Publishing. Ormazabal, Javier, and Juan Romero. 2013. Object clitics, agreement and dialectal variation. Probus 25: 301–344. Otero, Carlos. 1999. Pronombres reflexivos y recíprocos. In Gramática descriptiva de la lengua española, ed. Ignacio Bosque and Violeta Demonte, 1427–1517. Madrid: Espasa-Calpe. Real Academia Española (RAE). 2009. Nueva gramática de la lengua española. Madrid: Espasa. Reinhart, Tania, and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24: 657–720. Rooryck, Johan, and Guido Vanden Wyngaerd. 2011. Dissolving binding theory. Oxford: Oxford University Press. Sabatini, R. Nicholas. 1977. Sabatini’s observations on Alcaraz’s article. Hispania 60: 77–79. Saltarelli, Mario. 1994. Voice in Latin and romance: On the representation of lexical subjects. In Issues and theory in romance linguistics, ed. Michael Mazzola, 445–478. Washington: Georgetown University Press. Sánchez López, Cristina. 2002. Las construcciones con ‘se’. Madrid: Visor. Schäfer, Florian. 2008. The syntax of (anti-)causatives. Amsterdam/Philadelphia: John Benjamins. Sportiche, Dominique. 2014. Assessing unaccusativity and reflexivity: Using focus alternatives to what gets which θ-role. Linguistic Inquiry 45: 305–321. Torrego, Esther. 1995. From argumental to non-argumental pronouns: Spanish doubled reflexives. Probus 7: 221–241. Uriagereka, Juan. 1995. Aspects of the syntax of clitic placement in Western romance. Linguistic Inquiry 26: 79–123. Wood, Jim. 2014. Reflexive –st verbs in Icelandic. Natural Language and Linguistic Theory 32: 1387–1425.

Light Verbs and the Syntactic Configurations of se Alfredo García-Pardo

Abstract This paper offers a novel analysis of the Spanish light verbs poner(se) ‘to become’ and quedar(se) ‘to become/stay’ in combination with adjectives. These verbs have the peculiarity that they may appear with or without se in these constructions, with clear effects on argument structure and thematic and aspectual interpretation. I propose a unified formal analysis that takes event complexity and argument coindexing as the common denominator of se in these complex predicates. I further observe that these effects with se are not exclusive to these light verbs, but can also be observed in certain types of lexical verbs, strongly suggesting that both lexical and light verbs lexicalize the same eventive spine.

1 Introduction Spanish has a rich set of aspectual light verbs, also known as ‘pseudocopulative’ in traditional grammar, that are translatable as ‘to become’ in English and which appear with se, sometimes optionally and sometimes obligatorily. However, despite their surface similarity, these verbs have distinct syntactic and semantic properties that become apparent upon closer inspection, especially if we take into account the role of se in the structures. This paper will focus on two of these verbs: poner(se) (e.g. (1a)) and quedar(se) (e.g. (1b)).1 I gloss se as REFL, given that it is morphologically reflexive.

1 There are two other light verbs denoting change in Spanish, volver(se) and hacer(se) ‘to become’. The received wisdom is that these two verbs take Individual-level adjectives, whereas poner(se) and quedar(se) take Stage-level adjectives. See Demonte and Masullo (1999) and Morimoto and PavónLucero (2007) for further discussion of these four verbs.

A. García-Pardo (*) State University of New York – Purchase College, Purchase, NY, USA e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_10

243

244

(1)

A. García-Pardo

a. Berta se puso {fuerte/ nerviosa/ rígida}. Berta put strong nervous rigid ‘Berta became {strong/ nervous/ rigid}.’ b. Juan se quedó {pálido/ tranquilo/ delgado}. stayed pale calm thin Juan ‘Juan became {pale/ calm/ thin}.’

The study of these verbs is important for two reasons. First, although the status of se in the structure of the VP has been extensively discussed in the literature on Romance languages, much less attention has been paid to instances where the main verb is not lexical, but light.2 Also, the theoretical research on non-verbal predication in Hispanic linguistics has focused mostly on the copular split (i.e. ser vs. estar ‘to be’), whereas the split on light verbs has been largely neglected beyond the classic descriptive studies. From this perspective, the goal of this paper is to argue that the grammatical behavior of these light verbs, and the different effects observed with se, can be derived from the event structures they lexicalize. I propose that se surfaces as a morphological reflex when the same argument occupies two different positions in the syntactically articulated event structure (in the spirit of Cuervo 2003, 2014).3 These structures, as well as the different behavior of se, are essentially not different from that of other lexical verbs that may appear with se (eg. romper(se) ‘to break’ or morir(se) ‘to die’). I will further show how the formal aspectual interpretation of these structures straightforwardly explains the descriptive observations regarding poner(se) and quedar(se) in the traditional literature. This paper is structured as follows. Section 2 presents the data on poner(se) and quedar(se), with particular attention to the role of se. Section 3 introduces the basics of the theoretical framework I have chosen for my study as well as my proposal regarding the structure of these light verbs. Section 4 concludes the paper.

2 The term light verb was originally coined by Jespersen (1965). In Jespersen’s view, which I adopt here, a lexical verb is a conceptually rich verb that has meaning in isolation and belongs to the open class type of lexical items. Light verbs, on the other hand, belong to the closed (i.e. grammatical) class of lexical items and they are parasitic on a lexical predicate to form a syntactically and semantically coherent (complex) predicate. 3 In this respect, my work takes the stance that the so-called unaccusative se signals both reflexivity and aspectual complexity. As far as the reflexivity analysis is concerned—technical differences aside—I side with Chierchia (2004) and Koontz-Garboden (2009). In terms of the aspectual complexity of se, the reader is referred to Vivanco (this volume) and the overview therein.

Light Verbs and the Syntactic Configurations of se

245

2 Poner(se) and quedar(se): The Data As I mentioned in the introduction, the structures of the type are change of state constructions.4 They accept in x time PPs and adverbial modifiers such as little by little or progressively, as we can see in (2). (2)

a. La pantalla se puso negra {en un minuto/ poco a poco}. the screen put black {in a minute/ little by little ‘The screen went black {in a minute/ little by little}.’ b. El bar se quedó vacío {en una hora/ progresivamente}. stayed empty {in an hour/ progressively the bar ‘The bar became empty {in an hour/ progressively}.’

However, structures can also be stative. The sentences in (3) do not denote a change of state, but rather a permanence in a state. That is, (3a) means that the chair Catalina was supposed to sit on remained empty during a contextually relevant period of time. Similarly, (3b) does not mean that Pedro awakened, but rather, that he remained awake for the whole night.5 As expected, stative quedar(se) does not pass telicity tests (e.g. (3c)).6 (3)

a. Catalina no vino y su silla se quedó vacía. Catalina not came and her chair stayed empty ‘Catalina didn’t show up and her chair remained empty.’ b. Pedro se quedó despierto toda la noche. stayed awake all the night Pedro ‘Pedro stayed awake all night.’ c. Roberto se quedó despierto {*en diez minutos/ *poco a poco}. stayed awake {*in ten minutes/ *little by little Roberto ‘Roberto stayed awake {*in ten minutes/ *little by little}.’

4 For reasons of space, I will restrict my study of these light verbs to adjectival predicates. However, poner(se) and quedar(se) can also appear with other non-verbal predicates, such as PPs. Moreover, while many adjectives can take poner(se) and quedar(se) with equal ease, other adjectives can only combine with either one or the other (e.g. gordo ‘fat’ can only appear with poner(se) and débil ‘weak’ can only take quedar(se)). Focusing on these restrictions in depth goes beyond the scope of this work, but see Bybee and Eddington (2006) and Morimoto and Pavón-Lucero (2007) for discussion. 5 As Morimoto and Pavón-Lucero (2007) observe, there is also a counter-expectational presupposition involved with quedar(se). As such, it is odd to say, without further context, Se quedó despierto toda la mañana ‘He stayed awake all morning’, since the expectation is that one is typically awake in the morning. 6 The reader may wonder why a change-of-state reading is precluded in (3c). It is due to the selectional restrictions of change-of-state quedar(se) (see also ft. 4), which cannot take the adjective despierto.

246

A. García-Pardo

Se has different roles with poner(se) and quedar(se), so I will discuss them in turn, beginning with the latter verb. In general, se is optional with quedar (Demonte and Masullo 1999; Morimoto and Pavón-Lucero 2007), as (4) shows. (4)

a. Julio (se) quedó contento con mi regalo. Julio ( stayed happy with my gift ‘Julio was happy with my present.’ b. Juan (se) quedó ciego después del accidente. Juan ( stayed blind after of-the accident ‘Juan went blind after the accident.’ (From Demonte and Masullo 1999:2513)

In the descriptive literature, the difference between quedar and quedarse is taken to boil down to two factors: thematic and aspectual. With respect to the thematic factor, it is observed that quedarse allows for an agentive interpretation of the subject, whereas quedar does not (Demonte and Masullo 1999). Notice the oddity of quedar in the agentive context of (5), as well as the impossibility of having imperatives with quedarse (6a), but not with quedar (6b). (5) El hombre *(se) quedó quieto sin moverse mientras la enfermera the man *( stayed still without moving while the nurse le extraía sangre. drew blood ‘The man remained still while the nurse drew blood from him.’ (6)

a. ¡Quédate {quieto/ tranquilo}! ¡stay{still/ calm ‘Stay {still/ calm}!’ b. *¡Queda {quieto/ tranquilo}! *¡stay {still/ calm ‘Stay {still/ calm}!’ (From Demonte and Masullo 1999:2513)

Regarding the aspectual factor, Demonte and Masullo (1999) observe that quedarse emphasizes the achievement of a state, and quedar focuses on the resulting state. As evidence for this aspectual difference, Morimoto and Pavón-Lucero (2007) observe that quedarse accepts the progressive and the adverbial modifier poco a poco ‘little by little’, whereas quedar does not (e.g. (7) and (8), from Morimoto and Pavón-Lucero 2007:46).

Light Verbs and the Syntactic Configurations of se

247

(7)

a. Está quedándose sordo. is stayingdeaf ‘He is going deaf.’ b. *Está quedando sordo. *is staying deaf ‘He is going deaf.’ (From Morimoto and Pavón-Lucero 2007:46)

(8)

a. Se

quedó sordo poco a poco. stayed deaf little by little ‘He went deaf little by little.’

b. ??Quedó sordo poco a poco. ??stayed deaf little by little ‘He went deaf little by little.’ 2007:46)

(From Morimoto and Pavón-Lucero

It has also been observed that se cannot appear with quedar when its complement is a passive participle (Demonte and Masullo 1999; Morimoto and Pavón-Lucero 2007). I provide examples in (9), adapted from Demonte and Masullo (1999:2512). (9)

a. La explicación (*se) quedó cuidadosamente aclarada (por las the explanation (* stayed carefully clarified (by the autoridades). authorities ‘The explanation became carefully clarified by the authorities.’ b. Los alimentos (*se) quedaron cuidadosamente limpiados. the food (* stayed carefully cleaned ‘The food got carefully washed.’ (Adapted from Demonte and Masullo 1999:2512)

The se in poner(se), on the other hand, marks the transitive-unaccusative alternation. The transitive counterpart does not appear with se but the unaccusative version does, as in (10). This is unlike quedar(se), which does not participate in this alternation: for a transitive paraphrasing we need a different verb, dejar ‘to leave’, as (11) shows (Porroche 1990; Marín 2000; Morimoto and Pavón-Lucero 2007).

248

A. García-Pardo

(10)

a. La tormenta puso el cielo gris. the storm put the sky gray ‘The storm made the sky gray.’ b. El cielo se puso gris. the sky put gray ‘The sky turned gray.’

(11)

a. El lago se quedó seco. the lake stayed dry ‘The lake got/ stayed dry.’ b. La sequía {*quedó/ dejó} seco el lago. the drought {*stayed/ left dry the lake ‘The drought left the lake dry.’

To summarize, poner(se) and quedar(se) have differences in terms of thematic and aspectual interpretation, as well as argument structure, in which the absence or presence of se plays a key role. I summarize the empirical observations made in this section on Table 1. Table 1 Main grammatical properties of poner(se) and quedar(se) Properties Aspectual reading Transitive counterpart Optional se Agentive subject possible Aspectual focus Appears with passive participles

ponerse Telic Yes No Yes Change No

quedarse Telic/Stative No Yes Yes Change Yes

quedar Stative No Yes No Result state Yes

These properties, interestingly, are not essentially different from those found in many lexical verbs. Poner(se) patterns with change-of-state verbs like romper(se) ‘to break’, which also undergo the transitive-unaccusative alternation (e.g. (12)). Quedar(se), on the other hand, patterns with unaccusative such as caer(se) ‘to fall’. Syntactically, they do not have transitive variants (e.g. (13a)) and se is optional (e.g. (13b)). Semantically, we find similar patterns between those lexical verbs and quedar(se). First, the se variant allows for agentive readings, whereas the se-less variant does not (e.g. (14)). Second, se variants seem to focus on the change, whereas se-less variants focus on the result. Teomiro (2013:146) notes that modifiers such as de costado ‘sideways’ have different interpretations with caer(se) verbs: with se-less variants, de costado modifies the result state, whereas with se variants it modifies the process event (i.e. the manner of falling), as we can see in (15). Se, then, has aspectual effects even in intransitive lexical verbs where its presence or absence is optional.

Light Verbs and the Syntactic Configurations of se

(12)

a. Pedro rompió el vaso. Pedro broke the glass ‘Pedro broke the glass.’ b. El vaso se rompió. the glass broke ‘The glass broke.’

(13)

a. *Pedro cayó a Juan. *Pedro fell Juan (‘Pedro fell Juan.’) b. Pedro (se) cayó. fell Pedro ( ‘Pedro fell.’

(14)

Pedro *(se) cayó a propósito. Pedro *( fell on purpose ‘Pedro fell on purpose.’

(15)

249

a. Juan cayó de costado. Juan fell of side ‘Juan fell on his side.’ [Juan fell down and ended up on his side] b. Juan se cayó de costado. Juan fell of side ‘Juan fell sideways.’ [Juan’s position while falling was sideways]

3 The Proposal Having presented the main empirical data under study in Sect. 2, I will now introduce the theoretical framework and my own analysis for the syntax and semantics of these light verbs in relation to se.

3.1

Overview of the Theoretical Framework

I assume that the traditional VP is composed of three verbal projections as in (16), each of which denotes a separate subevent (Ramchand 2008). The subevent introduced by procP is dynamic whereas the ones introduced by initP and resP are stative. The aspectual meaning of the VP (i.e. the different event types) is derived by

250

A. García-Pardo

causational entailments between events in the decomposed VP configuration, read off from a head-complement relation configuration as in (17), adapted from Hale and Keyser (1993) (i.e. the higher subevent e1 causes or brings about the lower subevent e2). (16)

a. init(iation)P: initiational sub-event, introduces an argument b. proc(ess)P: dynamic sub-event, introduces an argument. c. res(ult)P: result sub-event, introduces a argument.

(17)

a.

VP

V VP b. e1 → e2

Thematic roles, in this system, are not grammatical primitives, but entailments from the aspectual structure. For instance, the subject of initP is not an Initiator because it is a thematic position per se, nor is Initiator a theta role in the classic sense: it is an Initiator, descriptively speaking, because it appears in a complex aspectual configuration in which the subevent introduced by init forms a causal chain with the events below (introduced by either procP or resP) and as such its subject, by entailment, is interpreted as the entity that causes or brings about the eventuality. The subject of resP, similarly, is a Resultee insofar as the subevent it is a subject of is caused by a causing subevent (introduced by either initP or procP). If resP where the only verbal projection in the clause, its subject would no longer be a Resultee, but a stative Holder. (I will actually argue in Sect. 3.2 that this is the case with se-less quedar.) With respect to the syntax of se, I assume that it is the morpho-phonological reflex of a co-referential relation between arguments within the VP.7,8 This relation comes about when an argument is copied onto the higher adjacent specifier within the syntactic structure of the articulated VP. I represent it schematically in (18). Note

7

For similar proposals, see Mendikoetxea (2000) and Cuervo (2003, 2014); for semantic accounts along the same lines, the reader is referred to Chierchia (2004) and Koontz-Garboden (2009). 8 An anonymous reviewer asks how the spell-out of se is restricted, given that it is not the case that all coindexed arguments trigger se. The reviewer points out the case of coindexed direct and indirect arguments (see (i), from Otero 1999:1462). (i)

Las transnacionales (*se ) vendieron la madre patria a sí misma. the multinationals *( sold the mother land to self same ‘The multinational corporations sold the mother land to itself.’

A plausible solution would be to stipulate that ApplP cannot host reflexive morphology, under the assumption that the two internal arguments of ditransitive verbs are introduced by a low ApplP complement of V (Pylkkänen 2002). I leave aside this issue for future work.

Light Verbs and the Syntactic Configurations of se

251

that the syntactic movement of arguments in Ramchand’s system is semantically meaningful, which means that an argument may have more than one aspectual role in the event decomposition. (18)

VP1 DP

V’ V se

VP2 DP

...

I will now present the structures I propose for these light verbs, as well as the evidence for each structure and the pertinent discussion.

3.2

Quedar and Change-of-State quedarse

I propose that the light verb quedar lexicalizes a single res head, which denotes a stative subevent. Since there are no higher verbal projections, resP is simply interpreted as a state, without implications of there being a previous event of change. The single argument of the verb is a subject of state (a ‘holder’, to use common terminology) and the predicative adjective is a complement of res, which provides the state denoted by res with lexical content. (19)

a. La silla quedó vacía. The chair remained empty ‘The chair remained empty.’ b.

resP res’

DP La silla

res quedó

AP vacía

Quedarse, in its change-of-state use, is built from stative quedar with an additional procP, the dynamic projection, on top of resP. The argument in (Spec, resP) moves up to (Spec, procP). Now resP, selected for by proc, denotes a result state brought about by a process, and the argument is interpreted as both the Undergoer and Resultee of the eventuality.

252

(20)

A. García-Pardo

a. La falda se quedó anticuada. The skirt got old-fashioned ‘The skirt became old-fashioned.’ b. procP proc’

DP La falda

proc se quedó

resP res’

DP La falda

res quedó

AP anticuada

These structures shed light on the descriptive literature as follows. Quedar ‘focuses’ on the result state because it is strictly stative: there might have been a previous change-of-state event, but it is not asserted (remember examples (7) and (8)). However, I side with Camacho (2012) in that quedar has inchoative semantics encoded in its lexical entry, i.e. it grammatically specifies a beginning point for the state, hence its apparently contradictory characterization in the literature as a result state, rather than a simple state. On the other hand, quedarse ‘focuses’ on the event of change simply because it is an event of change, built with procP and resP. Aside from the differences in aspectual interpretation, there is further syntacticosemantic evidence that my proposal is on the right track. As it is known, adverbs like almost and again only allow one reading with states and activities, but two readings with change-of-state verbs: with the latter, these adverbs can scope over the result state or the process sub-event, depending on where in the syntax they attach (see von Stechow 1995 for German and Cuervo (2014) for optional-se unaccusatives). As we can see in (21a) and (22a), quedar only has one reading for these adverbs, whereas with quedarse there is a scope ambiguity (e.g. (21b) and (22b)).9 The adverb casi ‘almost’ does not always deliver a clear ambiguity with these light verbs, as an anonymous reviewer rightly notes. This happens mostly when they select adjectives denoting psychological states. What is more, this adverb sounds weird with psychological predicates to begin with (e.g. (2)). A likely explanation is that it is hard to see how a psychological (change-of-) state would be compatible with any of the possible readings delivered by casi, i.e. how can one assess the degree to which a person can be said to be close to be nervous (i.e. with casi scoping below the process event and above the result state)? Or how can one assess whether the whole event of a person being or becoming nervous could have almost begun (i.e. with casi scoping over the whole event)?

9

(ii)

?Pedro casi se puso nervioso. ?Pedro almost put nervous (‘Pedro almost got nervous.’)

Light Verbs and the Syntactic Configurations of se

253

(21)

a. El restaurante casi quedó vacío. the restaurant almost remained empty ‘The restaurant almost remained empty.’ 1. The state of the restaurant being empty failed to hold (regardless of how many people ended up being in the restaurant). almost < resP b. El restaurante casi se quedó vacío. became empty the restaurant almost ‘The restaurant almost became empty.’ 1. The restaurant almost began to lose customers, but didn’t at the end. almost < procP < resP 2. The restaurant began losing customers and it was close to becoming fully empty. procP < almost < resP

(22)

a. El restaurante quedó vacío otra vez. the restaurant remained empty other time ‘The restaurant remained empty again.’ 1. The state of the restaurant being empty came to hold again. again < resP b. El restaurante se quedó vacío otra vez. the restaurant became empty other time ‘The restaurant remained empty again.’ 1. The restaurant became empty again (i.e. the change of state happened twice). again < procP < resP 2. The state of the restaurant being empty came to hold again (no implications of there being a change of state the first time). procP < again < resP

3.3

Stative quedarse

I argue that stative quedarse, just like I proposed for change-of-state quedarse, is syntactically and semantically more complex than stative quedar. Stative quedarse is built from quedar, also like its change-of-state version. In this case, the higher verbal projection is initP. The theme argument in (Spec,resP) moves up to (Spec, initP). I provide an example in (23).

254

(23)

A. García-Pardo

a. Pedro se quedó despierto. Pedro stayed awake ‘Pedro stayed awake.’ b. initP init’

DP Pedro

init se quedó

resP res’

DP Pedro

res quedó

AP despierto

The semantic consequences of this syntactic structure are the following. With respect to the aspectual interpretation of the structure, it is clear that we will not have a dynamic event, since procP does not project: the resulting predicate then cannot be a change of state, but a state. Note that, despite its stativity, the structure is nonetheless causative and resultative: there is a stative causative relation between a causing state, denoted by initP, and a result state, denoted by resP. The single argument is, in turn, both the Initiator and the Resultee of the eventuality: Pedro is bringing about (or maintaining, if you wish) his own state of being-awake. Both the stativity and the agentivity of stative quedarse are then derived by the syntactic structure and, indeed, the thematic factor alluded to in the descriptive literature, i.e. the link between se and agentivity.10 Stative quedarse is, then, a stative causative predication, rather than a simple state. The event type of stative causatives has been studied mostly in the realm of object-experiencer psychological verbs (OEPVs), i.e. verbs in which the object is the experiencer of the psychological state denoted by the verb (Pesetsky 1995; Arad 1998; Pylkkänen 2000; Kratzer 2000; Rothmayr 10

It seems that the link between agentivity and se is indeed restricted to the stative version of quedarse. While stative quedarse is clearly agentive (e.g. (iiia), and see also (5) and (6)), change-ofstate quedarse does not seem to be (e.g. (iiib)). This is not surprising, given that change-of-state quedarse does not include initP in its decomposition.

(iii)

a. Pedro se quedó despierto con la intención de estudiar. Pedro stayed awake with the intention of study ‘Pedro stayed awake with the intention of studying.’ b. ??Pedro se quedó delgado con la intención de tener mejor salud. Pedro got thin with the intention of have better health (‘Pedro became thin with the intention of having better health.’)

Light Verbs and the Syntactic Configurations of se

255

2009, a.m.o.). The main empirical evidence in favor of analyzing OEPVs as stative causatives is that many languages show overt causative markers with these verbs, as we can see in (24) (emphasis mine). (24)

a. Hyttyset inho-tta-vat Mikko-a. mosquitos. findDisgustingMikko‘Mosquitos disgust Mikko.’ (Finnish, from Pylkkänen 2000:418) b. Sono sirase-ga Tanaka-o yorokob-ase-ta. Tanakabe-pleasedthat news‘That news pleased Tanaka.’ (Japanese, from Pesetsky 1995:67) et rina. c. ha-seret hifxid rina the-film frightened. ‘The film frightened Rina.’ (Hebrew, from Sichel 2010:170)

As it turns out, Spanish stative OEPVs have causative morphology as well, in the form of affixes such as a-, en- and -izar (roughly, ‘-ize’, ‘-en’) (Marín and SanchezMarco 2012). This is illustrated in example (25). Crucially, Spanish stative OEPVs have intransitive versions that appear with se as well (Marín and McNally 2011; Marín and Sanchez-Marco 2012), as shown in (26). (25)

a. La crisis aterroriza a Vanessa. the crisis terror-izes Vanessa ‘The crisis terrorizes Vanessa.’ b. Las peleas me entristecen mucho. the fights me sadden much ‘Fights make me very sad.’

(26) asustar(se) ’to scare’, impresionar(se) ’to impress’, aburrir(se) ’to bore’, preocupar(se) ’to worry’...

In Fábregas and Marín (2015), OEPVs are analyzed as stative causative structures with a causative projection on top of a stative one that is interpreted as a result, i.e. essentially my structure for stative quedarse in (23b). I provide an example of their proposal in (27). Although none of these authors are explicit about the syntax of the se-version of these verbs, I assume that both variants are derivationally related: the intransitive version of OEPVs involves copies of the single argument in two subject positions, (Spec,initP) and (Spec,resP), thus providing a context for the appearance of se. I provide an example in (28). (27)

a. La conferencia aburrió a María. the conference bored María ‘The conference bored María.’ b. [ init P La conferencia [ init⬘ aburrió [ res P a María [ res⬘ aburrió ]]]]

256

(28)

A. García-Pardo

a. María se aburrió. María got.bored ‘María got bored.’ b. [ init P María [ init⬘ se aburrió [ res P María [ res⬘ aburrió ]]]]

There is a difference between stative quedarse and OEPVs, however: OEPVs with se also have a transitive variant (e.g. (27a) and (28a)), but quedarse does not. In this respect, quedarse behaves like so-called inherently reflexive verbs like arrepentirse ‘to regret’, desinteresarse ‘to have lack of interest’, ensimismarse ‘to be lost in thought’ and so on. I assume that stative quedarse, just like those verbs, is lexically specified for a reflexive structure. The lexical entry of quedarse should then be as in (29), which captures both the se-less and change-of-state versions of quedar: minimally, quedar lexicalizes a res head. It may also lexicalize either init (stative quedarse) or proc (change-of-state quedarse), but if it does, the subject position of initP/procP and resP must be occupied by the same argument, which is represented with a subscript i in the eventive labels in (29). (29) quedar(se): [(init /proc ),res

Providing support for the complex structure of stative quedarse, beyond the logical event structure interpretation derived from the framework, is somewhat trickier than with change-of-state quedarse. For instance, the scope ambiguity test with almost is not valid for stative quedarse, given that almost is also sensitive to the temporal sequencing of the subevents: whereas in change-of-state quedarse there is a sequencing between the process event and the result state that favors the ambiguity, in the case of stative quedarse both eventualities are temporally coextensive (i.e. the caused state holds for as long as the causer maintains it), and hence the ambiguity is vacuous. A similar problem holds for the test with the adverb again, since it is also sensitive to temporal sequencing. However, there are different scope effects with dative applicative arguments in anticausative contexts with se that support our structure for stative quedarse. I follow authors like Cuervo (2003), Schäfer (2008) and Fernández-Soriano and Mendikoetxea (2011) in that there are two kinds of applicative arguments with distinct interpretations depending on the attachment height of the Applicative Phrase (ApplP) that introduces them in the articulated VP structure. The two types are presented below: 1. High ApplP: it takes a vP complement (procP, in our system) and its subject can be interpreted as an accidental or unintentional causer. An example is given in (30). (30)

a. A

María se María

b. ApplP > procP

le her.

rompió un plato. broke a glass

Light Verbs and the Syntactic Configurations of se

257

2. Affected ApplP: its complement is a phrase that denotes a result state (a stative root in some accounts, resP in our account). Its subject can be interpreted as a beneficiary or maleficiary of the eventuality described. I provide an example in (31). (31)

a. A

Pedro se le nubló el día. Pedro him. got.cloudy the day Reading: The day got cloudy, and Pedro was affected by it. b. ApplP > resP

With respect to our proposal for quedarse, this typology predicts that quedarse can only have Affected Applicatives, which take resP as a complement, but not High Applicatives, which take procP as a complement, since procP is absent in stative quedarse. The prediction is borne out, as we can see in (32). Change-ofstate quedarse, on the other hand, is predicted to have both readings available, given that it is composed of both procP and resP. Again, the prediction is correct, as shown in (33). (32)

María se le quedó el hijo despierto toda la noche. María her. stayed the son awake all the night 1. Affected reading: María’s son stayed awake all night, and María was affected by it. 2. Accidental causer reading: Unavailable (María accidentally brought about that her son stayed awake all night) b. Affected reading: initP > ApplP > resP

(33)

a. A

3.4

a. A

María se le quedó mudo el paciente. María her. got mute the patient 1. Affected reading: María’s patient became mute, and María was affected by it. 2. Accidental causer reading: María accidentally brought about that her patient became mute (e.g. by a botched operation). b. Affected reading: procP > ApplP > resP c. Accidental causer reading: ApplP > procP > resP

Poner(se)

Poner(se), as we saw in Sect. 2, does not appear in as many syntactico-aspectual configurations as quedar(se). This verb denotes aspectually a change-of-state verb and, like many other change-of-state verbs, it participates in the transitive-

258

A. García-Pardo

unaccusative alternation. Thus, poner(se) must lexicalize minimally proc and res, and it may optionally appear with init, in which case we have a transitive structure. The proposed lexical entry is given in (34), and examples of the unaccusative and transitive version are given in (35) and (36), respectively. (34) poner(se): [(init),proc ,res ] (35)

a. Pedro se puso nervioso. Pedro put nervous ‘Pedro got nervous.’ b. procP DP Pedro

proc’ proc se puso

resP res’

DP Pedro

(36)

res puso

AP nervioso

a. María puso a Pedro nervioso. María put Pedro nervous ‘María made Pedro nervous.’ b. initP DP María

init’ init puso

procP proc’

DP a Pedro

resP

proc puso

res’

DP a Pedro

res puso

AP nervioso

Light Verbs and the Syntactic Configurations of se

259

The syntactic structures of change-of-state ponerse and quedarse are then identical, built with procP and resP. Like change-of-state quedarse, ponerse shows scope ambiguity with adverbs like almost and again, as we can see in (37) and (38), respectively. Ponerse, as expected, also shows ambiguity in terms of the interpretation of applicative arguments (e.g. (39)). (37) La comida casi se puso mala. the food almost bad ‘The food almost went bad.’ almost > procP: The food almost began to go bad, but it didn’t at the end (e.g. because I remembered to put it in the fridge). 2. procP > almost > resP: The food began going bad (e.g. it started to show signs of decay) and it got close to becoming completely bad.

1.

(38) La pantalla se puso negra otra vez. the screen put black other time ‘The screen went black again.’ 1. 2.

again > procP: The screen underwent the event of turning black twice. procP > again > resP: The state of the screen being black came to hold twice (no implications of there being a change of state the first time).

(39)

a. A

María se le puso negra la salsa. María her. put black the sauce 1. Affected reading: María’s sauce turned black (e.g. from overexposure to air), and María was affected by it. 2. Accidental causer reading: María accidentally caused the sauce to turn black (e.g. by overcooking it). b. Affected reading: procP > ApplP > resP Accidental causer reading: ApplP > procP > resP

3.5

Quedar + Participles

As we mentioned in Sect. 2, quedar may appear with passive participles, in which case we cannot have se (see (9), repeated below). Poner(se) is also impossible with passive participles (e.g. (40)).

260

A. García-Pardo

(9)

a. La explicación (*se) quedó cuidadosamente aclarada (por las the explanation (* stayed carefully clarified (by the autoridades). authorities ‘The explanation became carefully clarified by the authorities.’ b. Los alimentos (*se) quedaron cuidadosamente limpiados. stayed carefully cleaned the food (* ‘The food got carefully washed.’ (Adapted from Demonte and Masullo 1999:2512)

(40)

*El museo (se) puso inaugurado por el alcalde. *the museum ( put inaugurated by the mayor (‘The museum became inaugurated by the mayor.’)

Why are passive participles only possible with se-less quedar? I suggest that the restriction is due to the incompatibility of the argument structure of the passive participle and that of the se versions of quedarse and ponerse. Passive participles, as is known, have two crucial properties (Baker et al. 1989): (i) they have an implicit external argument (which may be overtly expressed by means of a by-phrase); (ii) the external argument is referentially disjoint from the internal argument, i.e. passive participles cannot be reflexive structures. Evidence for the existence of an implicit external argument is the possibility of having purpose clauses and agentoriented adverbs (e.g. (41a)) and evidence for the impossibility of a reflexive configuration for passive participles is shown in (41b), where an explicitly reflexive by-phrase is out. (41)

a. Las reuniones quedaron prohibidas {para fortalecer al dictador/ the meetings got prohibited to strengthen -the dictator deliberadamente}. deliberately ‘Meetings got prohibited {to strengthen the dictator/ deliberately}.’ b. Los niños quedaron vestidos {*por sí solos/ por sus padres}. the kids stayed dressed {*by themselves alone by their parents (‘The kids got dressed by themselves/ by their parents.’)

Now, if the external argument of the participle is syntactically projected (as proposed by Baker et al. 1989), the explanation reduces to a minimality effect: the internal argument of the participle cannot move to the specifier of initP or procP

Light Verbs and the Syntactic Configurations of se

261

bypassing the implicit external argument of the participle. I illustrate the deviant structure in (42b), from (42a).11 (42)

a. Pedro (*se) quedó arrestado (por la policía). Pedro (* stayed arrested by the police ‘Pedro got arrested by the police.’ b. *[ init P / proc P Pedro [ init⬘ / proc⬘ se quedó [ res P Pedro [ res⬘ quedó [ Part P (External Argument ≠ Pedro) [ Part⬘ arrestado Pedro ]]]]]]

An alternative proposal, suggested to me by an anonymous reviewer, would be that the semantics of change of ponerse and quedarse clash with the semantics of change of the participle: this is not unlike data like (43), the reviewer notes, where the verb built can only select an adjective, but not a participle. (43)

The door was built open/ *opened.

(From Embick 2004:357)

While I do not disagree with the reviewer that a semantic clash between two change-of-state configurations could also be at play, it is worth noting that stative quedarse also rejects passive participles. The participle encerrado ‘locked up’ in (44a) is ambiguous between an adjectival and a passive participle, and se is optional. However, when it is disambiguated by the addition of a by-phrase (e.g. (44b)), the se version is not possible. (44)

a. (Se) quedó encerrado en su celda toda la noche. ( stayed locked-up in his cell all the night ‘He stayed locked up in his cell all night.’ b. (*Se) quedó encerrado en su celda por la policía. stayed locked-up in his cell by the police (* ‘He stayed locked up in his cell by the police.’

I will leave the issue of light verbs and participles here, noting that more work is needed to understand their combinatorial restrictions, which will require a more

11

A reviewer points out that it is possible to have a reflexive stative verb (i.e. with se) and a participial complement, as in (iv) (example from the reviewer).

(iv)

Los padres de unos niños se sienten engañados por los grupos antivacunas. the parents of some kids feel cheated by the groups anti-vaccines ‘Some kids’ parents feel cheated by the anti-vaccination groups.’

While I agree this would be a potential counterexample for my theory, I note that the structure in (41b) is rather exceptional: there are very few inherently reflexive verbs that can take a passive participal complement whose internal argument is coindexed with the subject of such verb. Only sentirse ‘feel’ and considerarse ‘consider oneself’ come to mind. I leave aside a deeper study of the role of se in these specific constructions for future work.

262

A. García-Pardo

thorough characterization of the internal structure of the participles and their compatibility with the event structures projected by these light verbs than I can do justice here.

4 Conclusions This paper has studied the grammar of the Spanish light verbs poner(se) and quedar (se) with adjectives, focusing on the effects that se has in their argument structure properties and their aspectual and thematic interpretation. I have proposed that the common denominator of all these effects with se is a syntactic configuration in which a single argument occupies two different syntactic positions within the VP. The findings of this research are of relevance both for Hispanic linguistics and for general linguistic theory. For Hispanic linguistics, my proposal derives straightforwardly the different behavior of se with these light verbs in intransitive configurations, which had only been dealt with in descriptive terms. More particularly, I have explained why the presence of se allows both for an agentive reading and a change-of-state interpretation with quedar. I have also shown the different effects of se in the aspect and argument structure of these light verbs is not essentially different to other lexical verbs in Spanish: poner(se) behaves like verbs that alternate in transitivity (e.g. romper(se) ‘to break’), whereas quedar(se) behaves like optional-se unaccusative verbs (e.g. morir(se) ‘to die’) as well as inherently reflexive verbs (e.g. arrepentirse ‘to regret’). For the general theory, this work represents further evidence for the view that light verbs and lexical verbs are not grammatically distinct, in that they instantiate the same structures in the syntax. As such, they show identical possibilities and restrictions regarding argument structure and aspectual meaning (Hale and Keyser 2002). The difference would be that the roots that we call ‘light’ are lexico-conceptually impoverished with respect to ‘lexical’ roots, which carry rich encyclopedic content. Acknowledgements I am grateful to the audiences of the Workshop on Romance se-si, where a previous version of this work was presented, for their constructive feedback. I also thank two anonymous reviewers and María Luisa Zubizarreta for the extensive discussions I had with her on this topic. Last but not least, thank you to Grant Armstrong and Jon MacDonald for organizing the workshop and for the relentless effort they have put in making this volume a reality. Errors are my own.

References Alexiadou, Artemis, Berit Gehrke, and Florian Schäfer. 2014. The argument structure of adjectival participles revisited. Lingua 149: 118–138. Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2015. External Arguments in Transitivity Alternations: A Layering Approach. Oxford: Oxford University Press.

Light Verbs and the Syntactic Configurations of se

263

Arad, Maya. 1998. VP Structure and the Syntax–Lexicon Interface. PhD dissertation, University College London. Arad, Maya. 2002. Universal features and language-particular morphemes. In Theoretical Approaches to Universals, ed. Artemis Alexiadou, 15–39. Amsterdam: John Benjamins. Arche, Maria Jesus. 2006. Individuals in Time. Amsterdam: John Benjamins. Baker, Mark C. 2003. Lexical Categories: Verbs, Nouns and Adjectives. Cambridge: Cambridge University Press. Baker, Mark, Kyle Johnson, and Ian Roberts. 1989. Passive arguments raised. Linguistic Inquiry 20 (2): 219–251. Bybee, Joan, and David Eddington. 2006. A usage-based approach to Spanish verbs of ‘becoming’. Language 82(2): 323–355. Camacho, Jose. 2012. Ser and estar: The individual/stage level distinction and aspectual predication. In The Handbook of Hispanic Linguistics, eds. Jose Ignacio Hualde, Antxon Olarrea, and Erin O’Rourke, 453–476. Oxford: Blackwell. Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences. In The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface, eds. Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, 22–59. Oxford: Oxford University Press. Cuervo, Maria Cristina. 2003. Datives at Large. PhD dissertation, Massachussets Institute of Technology. Cuervo, Maria Cristina. 2014. Alternating unaccusatives and the distribution of roots. Lingua 141: 48–70. Demonte, Violeta, and Pascual Masullo. 1999. La predicacion: Los complementos predicativos. In Gramatica descriptiva de la lengua española, eds. Ignacio Bosque, and Violeta Demonte, 2461–2523. Madrid: Espasa Calpe. Embick, David. 2004. On the structure of resultative participles in English. Linguistic Inquiry 35(3): 355–392. Fábregas, Antonio, and Rafael Marín. 2015. Deriving individual-level and stage-level psych verbs in Spanish. The Linguistic Review 32(2): 227–275. Fernández-Soriano, Olga, and Amaya Mendikoetxea. 2011. Non-selected dative subjects in anticausative constructions. Archivio Glottologico Italiano XCVI: 87–128. Folli, Raffaella, and Heidi Harley. 2005. Flavors of v: Consuming results in Italian and English. In Aspectual Inquiries, eds. Roumyana Slabakova, and Paula Kempchinsky, 95–120. Dordrecht: Kluwer. Hale, Kenneth L., and Samuel J. Keyser. 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press. Jespersen, Otto. 1965. A Modern English Grammar on Historical Principles, Part VI, Morphology. London: George Allen and Unwin Ltd. Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27(1): 77–138. Kratzer, Angelika. 2000. Building statives. In Proceedings of the 26th Meeting of the Berkeley Linguistics Society, eds. Lisa J. Conathan, Jeff Good, Darya Kavitskaya, Alyssa B. Wulf, and Alan C. L. Yu, 385–399. Berkeley: Berkeley Linguistics Society. Luján, Marta. 1981. The Spanish copulas as aspectual indicators. Lingua 54: 165–210. Marín, Rafael. 2000. El componente aspectual de la predicacion. PhD dissertation, Universidad Autonoma de Barcelona. Marín, Rafael, and Louise McNally. 2011. Inchoativity, change of state and telicity: Evidence from Spanish reflexive psychological verbs. Natural Language and Linguistic Theory 29(2): 467–502. Marín, Rafael, and Cristina Sanchez-Marco. 2012. Verbos y nombres psicologicos: juntos y revueltos. Borealis: An International Journal of Hispanic Linguistics 2: 91–108. Mendikoetxea, Amaya. 2000. Relaciones de interficie: Los verbos de cambio de estado. Cuadernos de Linguistica del IUOG 7: 125–144.

264

A. García-Pardo

Morimoto, Yuko, and María Victoria Pavón-Lucero. 2007. Los verbos pseudo-copulativos del español. Madrid: Arco Libros. Otero, Carlos P. 1999. Pronombres reflexivos y recíprocos. In Gramática descriptiva de la lengua española, eds. Ignacio Bosque, and Violeta Demonte, 1429–1517. Madrid: Espasa-Calpe. Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press. Porroche, Margarita. 1990. Aspectos de la atribucion en español. Zaragoza: Portico. Pylkkänen, Liina. 2000. On Stativity and Causation. In Events as Grammatical Objects: The Converging Perspectives of Lexical Semantics and Syntax, eds. Carol Tenny, and James Pustejovsky. Standford: CSLI Publications. Pylkkänen, Liina. 2002. Introducing Arguments. PhD dissertation, Massachussets Institute of Technology. Ramchand, Gillian C. 2008. Verb Meaning and the Lexicon: A First Phase Syntax. Cambridge: Cambridge University Press. Rothmayr, Antonia. 2009. The Structure of Stative Verbs. Amsterdam: John Benjamins. Schäfer, Florian. 2008. The Syntax of (Anti-)causatives. Amsterdam: John Benjamins. Sichel, Ivy. 2010. Event structure constraints in nominalization. In The Syntax of Nominalizations Across Languages and Frameworks, eds. Artemis Alexiadou, and Monika Rathert, 159–197. Berlin: Mouton de Gruyter. Vivanco, Margot. 2020. Spanish anticausative se and its relation to scales. In Unraveling the Complexity of SE. Cham: Springer International Publishing.

The Role of SE and NE in Romance Verbs of Directed Motion: Evidence from Catalan, Italian, Aragonese and Spanish Varieties Anna Pineda

Abstract The purpose of this paper is to analyze the use of se with directed motion verbs in several Romance languages and varieties. Building on some observations that have been made for Spanish, we adopt a broader cross-linguistic perspective, bringing into discussion an element that has, until now, gone generally unnoticed (aside from descriptive works): the ablative locative clitic that appears, together with se, in Catalan, Italian and Aragonese varieties, as in tornar-se’n ‘go back SE NE’ from Catalan. Our data from different Romance languages and dialects allow us to refine the settings of the connection between pronominal directed motion verbs and the existence of a source component. In particular, we posit the existence of a locative head (here tentatively analysed as an applicative), which can be spelled out by an ablative locative clitic. We also argue that directed motion verbs can be conceived of by Romance speakers as simple, punctual events denoting the achievement of a particular position, but also as denoting a complex event that consists of a causing subevent and a resultant state (which is connected to achieving a new position and remaining there for some time, after having left behind the original location). In the latter case, these verbs can surface in their pronominal form, even if it does not happen always, since there is cross-linguistic and cross-dialectal variation regarding the availability of pronominal forms for these verbs, due to different lexicalization patterns. Keywords Directed motion verbs · Aspectual se · Ablative locative clitics

A. Pineda (*) Universitat Pompeu Fabra, Barcelona, Spain © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_11

265

266

A. Pineda

1 Introduction So-called aspectual se has been a broadly discussed topic in the literature on Romance languages. When examples of aspectual se are given, different types of verbs are usually mixed: from the classical examples with transitive verbs such as comerse una manzana (‘eat-SE an apple’) to unaccusative examples such as morirse (‘die-SE’) or irse (‘go-SE’). Focusing on the particular case of directed motion (from now on, DM) verbs (such as irse), we offer a contribution that is new for two reasons: first, we take a pan-Romance perspective including less studied languages and non-standard varieties; second, we focus on the occurrence of se with a particular class of verbs that has not received much attention until now from the theoretical point of view and, at best, has been included in passing when dealing with other instances of the so-called aspectual se. The outline of this paper is as follows: Sect. 2 illustrates the pronominal use of DM verbs in Spanish, Catalan, Italian and Aragonese varieties; Sect. 3 deals with the role of the clitics se and ne that appear with these verbs; Sect. 4 argues that ne may be analysed as a locative applicative head; Sect. 5 accounts for the differences across and within languages regarding the realization of se and ne; and Sect. 6 summarizes the implications of the analysis put forth in this study.

2 Pronominal Verbs of Directed Motion in Romance Languages 2.1

The Clitic se

The use of se with several types of verbs has been a topic broadly discussed in Romance linguistics, and in the case of Spanish, there is a considerable amount of diverse studies on it (for an overview, see Sánchez López 2002). Many authors have claimed that se, with a variety of transitive and intransitive verbs, contributes an aspectual value, claiming that the clitic only occurs with bounded events, e.g. comerse una manzana ‘eat-SE an apple’, but not comerse manzanas ‘eat-SE apples’ (Nishida 1994, Zagona 1996, De Miguel 1999, De Miguel and Fernández Lagunilla 2000, Basilico 2010, but see also MacDonald 2004, 2008, 2017, Armstrong 2013; see also Vivanco [this volume] and García-Pardo [this volume] for the analysis of the so-called aspectual se with different types of verbs). In several studies, occurrences with transitive verbs (comerse ‘eat-SE’, leerse ‘read-SE’) and with intransitive verbs (irse ‘go-SE’, morirse ‘die-SE’, dormirse ‘sleep-SE’) are not considered separately. For example, De Miguel (1999, 2986, 2995–2996) claims that se with transitive verbs names the final endpoint of the event (1), whereas with intransitive verbs (including DM verbs) the presence of the clitic names the initial

The Role of SE and NE in Romance Verbs of Directed Motion

267

endpoint of the event: thus, the interpretation in (2) is that there is an implicit or explicit mention of the beginning of the event, a mention that delimits the event:1,2 (1) Andrés se comió la manzana Andrés REFL.3SG eat.PST.3SG the apple ‘Andrés ate the apple up’ (2) Sara {se fue de casa/ Sara REFL.3SG go.PST.3SG from house/ se bajó del bus} REFL.3SG get.down.PST.3SG of.the bus ‘Sara {left the house/got out of the bus}’ All these occurrences have been subsumed under the label aspectual se, but García Fernández (2011) explicitly criticizes this joint analysis and argues that important differences tease different verbal classes apart. We agree with his point and, from now on, will concentrate on the particular case of a subclass of intransitive verbs, that of DM verbs. In general Spanish, se can appear with some DM verbs: ir ‘go’, salir ‘go out’, venir ‘come’, bajar ‘go down’, subir ‘go up’, volver ‘go back’, escapar ‘escape’, marchar ‘leave’. However, there is important dialectal variation, as already pointed out by authors such as Martín Zorraquino (1979, 280–281), who claimed that both in American Spanish and in the colloquial speech of Spain reflexive pronouns with intransitive verbs are used more frequently than in standard language –as we will see below, entrar ‘go in’ and llegar ‘arrive’, among others, can also take se in several Spanish varieties. The semantic contribution authors have attributed to se with DM verbs is related to the initial endpoint of the movement event, as we have already shown for De Miguel’s (1999) analysis. Masullo (1992, 246) also argues that with these verbs se stands for an incorporated source argument. A similar idea is defended by Schroten (1972, 89), who claims that DM verbs with and without se entail “a difference in

1

The clitic se has traditionally been labeled a reflexive marker, hence we use the label REFL (followed by specifications for person and number, e.g. REFL.3SG) for glossing purposes. Later on, we will also use ABL (‘ablative’) for the locative source clitic that appears together with se in some varieties. The other abbreviations used in glosses are the following ones: APPL ‘applicative’, ASP ‘aspect’, DAT ‘dative’, FUT ‘future’, FV ‘final vowel’, GEN ‘genitive’, IMP ‘imperative’, INF ‘infinitive’, IPFV ‘imperfective’, LOC ‘locative’, M ‘masculine’, PL ‘plural’, PRS ‘present’, PST ‘past’, PTCP ‘participle’, SBJV ‘subjunctive’, SG ‘singular’, SCD ‘subject concord class’. 2 De Miguel (1999) adds that verbs already bounded in their initial endpoint (salir ‘go out’ and venir ‘come’) accept se denoting the point of departure but also the destination: (i)

Sara se salió {de la habitación/al balcón} Sara REFL.3SG go.out.PST.3SG of the room /to.the balcony ‘Sara went out of the room/on the balcony’

268

A. Pineda

semantic interpretation”, although he admits that it is “not clear what this difference should be taken to be”. He claims that “the most plausible suggestion is that the reflexive ‘variants’ imply movement from some place (that need not be specified explicitly), whereas the non-reflexive ‘variants’ do not imply such movement from some place”. Along the same lines, Sánchez López (2002, 118) argues that DM verbs with se add a meaning of leaving behind the place of departure: Me voy de aquí para no volver ‘I am leaving this place to never come back’, Me vine de Alemania para siempre ‘I came from Germany to stay forever’. Other authors, such as Llorente Maldonado (1980) and Cuervo (2014, 62), have pointed out that neither of these approaches can account for several other occurrences where se does not seem to have anything to do with the initial endpoint or the source of the event, e.g. subirse ‘go up’, possible in general Spanish. JiménezFernández and Tubino (2014, 2019) add that the same happens with Southern Peninsular Spanish occurrences such as entrarse ‘go in’ (3) and llegarse ‘arrive’, where the initial endpoint (i.e. the place of departure) is left unspecified, and the final endpoint of the movement (a la casa, a Madrid) is focused instead; and they conclude that se can co-occur with DM verbs not only delimiting events at the beginning but also at the end:3 (3) Juan se entró a la casa Juan REFL.3SG go.in.PST.3SG to the house ‘Juan went in the house’ (Jiménez-Fernández and Tubino 2019, 186) Moreover, following Gallardo (2008, 2010), Jiménez-Fernández & Tubino claim that, when it comes to the pronominal use of DM verbs, what is relevant is the notion of permanence in a position, either in the place that is left behind or in the new location. Specifically, Jiménez-Fernández & Tubino (2019, 205) claim that there is a clear difference between María entró en el camarote ‘Maria went in the cabin’ and María se entró en el camarote ‘Maria went in the cabin to stay’: “While both sentences mean that Maria goes in the cabin, only the option involving se denotes that Maria stops being outside to go and stay inside the cabin, that is, permanence first in the original and then in the final location”. In Gallardo’s words, ‘no one leaves a place if they have not stayed there a minimal amount of time or they are not going to stay in the new destination’ [our translation]. Jiménez-Fernández & Tubino further exemplify this contrast showing that with se a continuation denying the permanence in the final position would be pragmatically odd—they give parallel examples with salirse ‘go out’, bajarse ‘go down’, subirse ‘go up’:

3

As shown in fn. 2, a similar point was made by De Miguel (1999), although restricted to the particular case of salir ‘go out’ and venir ‘come’, verbs already bounded in their initial endpoint (one always goes out of a place, and comes from a place).

The Role of SE and NE in Romance Verbs of Directed Motion

269

(4) Juan se entró en su casa #pero no Juan REFL.3SG go.in.PST.3SG in his house but no se quedó REFL.3SG stay.PST.3SG ‘Juan went into his house but he didn’t stay there’ (Jiménez-Fernández and Tubino 2014, 18) From this overview we can conclude that se with Spanish DM verbs is claimed to be an aspectual marker of bounded events, contributing a meaning of permanence in the initial location and/or in the final location. But, still, we think that an account for the role of se based only on Spanish data is not sufficient, and a broader crosslinguistic perspective is needed.

2.2

The Clitic Cluster se + ne

In order to determine the contribution of se in Romance DM verbs, we must look at the role of the ablative locative clitic that appears in Catalan, Italian and Aragonese varieties together with se. In Catalan, anar ‘go’ can appear with the clitic cluster se + ne, acquiring either the meaning of departing from some place in order to go somewhere else (5) or simply the meaning of leaving (6). Similarly, tornar ‘go back’ can also take se’n with the particular meaning of departing from some place in order to return to the original location (7): (5) Se

n’ anirà a Cambridge a estudiar go.FUT.3SG to Cambridge to study.INF ‘(S)he will go to Cambridge to study’ (6) La Maria ja se n’ ha anat The Maria already REFL.3SG ABL have.PRS.3SG go.PTCP ‘Maria is already gone’ (7) Ara se’ n tornarà a casa Now REFL.3SG ABL go.back.FUT.3SG to home ‘Now (s)he will go back home’ REFL.3SG ABL

Aside from these two verbs, the pronominal use of DM verbs is no longer productive in general Catalan. In some Catalan dialects though, the use of se’n is possible with virtually all DM verbs: this has been noticed for Valencian (Sanchis Guarner 1950; Todolí 1990, 2002) and some North-Western varieties spoken in La Franja, in the administrative region of Aragon (Giralt 1995). Thus, in addition to (5)–(7), in these dialects we find venir-se’n ‘come’, baixar-se’n ‘go down’, pujar-se’n ‘go up’, entrarse’n ‘go in’, eixir-se’n go out’, envolar-se’n ‘fly’, fugir-se’n ‘run away’. This more

270

A. Pineda

generalized use of se’n with DM verbs was also present in Old Catalan4, but nowadays, aside from the varieties just mentioned, it is no longer productive –if anything, some remainders are found in folk storytelling, proverbs and religious prayers (see Pineda 2018, § 3.1).5 The appearance of a locative clitic toghether with se is not unique of Catalan varieties. For Italian, Rohlfs (1954, 188) mentions the “reflexive” use of several intransitive verbs (andarsi ‘go, leave’, uscirsi ‘go out’, venirsi ‘come’, fuggirsi ‘run away’), giving also examples that include the ablative clitic, such as andarsene (Rohlfs 1954, 166). In fact, quite interestingly, his translator S. Persichino clarifies in a footnote that the language prefers this construction when the reflexive is linked to clitics such as ne, or even only admits the option with the clitic cluster: in his words, ‘no one would say, reflexively, ci andiamo, me venivo or vatti, while they are in common use ce n’andiamo ‘we leave’, me ne venivo ‘I came’, vattene ‘go!, leave’, and so on’ [our translation] (Rohlfs 1954, 188, fn. 1 [N. de T.]). Similary, Cennamo (1999, 141) points out that “residues of unaccusative SIBI verbs are relatively rare in contemporary standard Italian, cropping up however in one of the classes of verbs SIBI occurred with in Late Latin, i.e., unaccusatives denoting inherently directed change of location”, as in (8) and (9). (8) Mario se ne andò Mario REFL.3SG ABL go.PST.3SG ‘Mario left’ (9) Mario se ne scappò Mario REFL.3SG ABL run.away.PST.3SG ‘Mario ran away (Cennamo 1999, 141) As pointed out for Catalan, in the varieties of Italian there are also some dialectal differences concerning the extension of pronominal DM verbs. According to Cennamo (1999, 141), these forms are especially common in central/southern dialects and in the regional variety of Italian spoken in these areas, “clearly reflecting the dialect”. Luigi Andriani (p.c.) confirms this for Barese and regional Italian from Bari, as does Silvio Cruschina (p.c.) for Sicilian (e.g. acchianarisinni ‘go.up.REFL. ABL’, scinnirisinni ‘go.down.REFL.ABL’) and regional Italian from Sicily. Likewise, 4

Just as an illustration, see the verbs in boldface of the following passage from the Decameron: “se
n partí [‘he left’], havent enperò ben considerat la disposició del loch; e esperant la nit, e de aquella lexada passar bona part, llà se
n tornà [‘he went back’], e arrapinyant-se en loch hon no
s foren affarats moscarts, en lo jardí se n’entrà [‘he went in’], en lo qual trobà una anteneta, ab lo qual en la finestra de la donzella se
n pujà [‘he went up’], e per aquella per semblant laugerament se
n podia devallar [‘he was able to go down’]” (Decameró, fifteenth century, Corpus informatitzat del català antic: www.cica.cat) 5 An anonymous reviewer points out that a Romance language with a behaviour similar to general Catalan is contemporary French, where only s’en aller (‘go’), s’en retourner (‘go back’) and se’n venir (‘come’) exist, and where the clitic en has been morphologically incorporated in the case of ‘run away’, s’enfuir (and the same seems to be occurring in some dialects for s’en aller, where something like Jean s’est en allé is replacing Jean s’en est allé).

The Role of SE and NE in Romance Verbs of Directed Motion

271

Neapolitan (10)–(12) and Calabrian (13) exhibit several pronominal DM verbs with se and the ablative clitic ne (Ledgeway 2009, 313, 350), as shown in (10)–(12). (10) Po0 me ne tornaie ccà then REFL.1SG ABL come.back.PST.3SG here’ ‘Then I came back here’ (11) Tu te ne viene, Federi’? you REFL.2SG ABL come.PRS.2SG Federi’ ‘Do you come (with us), Federi’?’ (Ledgeway 2009, 350) (12) Me ne saglio REFL.1SG ABL go.up.PRS.1SG to. . . ‘I’m going up’ (Adam Ledgeway, p.c.) (13) Si nn’ è partuta REFL.3SG ABL be.PRS.3SG leave.PTCP.F.SG ‘She has left’(Adam Ledgeway, p.c.) Aragonese DM verbs may feature the clitic cluster as well. Alvar (1953, 298) claims that this usage abounds since the old stages of the language and is present almost all over Aragonese-speaking territory (except for Southern varieties). Again, the existence of this pattern is confirmed in fine-grained dialectal descriptions. In Eastern Aragonese there is a systematic usage of the pronoun ne with verbs of movement that are reflexive, thus verbs such as íse ‘go’, veníse ‘come’, subíse ‘go up’, bajáse ‘go down’, salíse ‘go out’, etc., have become ísene, venísene, subísene, bajásene, salísene: (14) Ha dicho tío que te ‘n venises have.PRS.3SG say.PTCP uncle that REFL.2SG ABL come.SBJV.PST.2PL ‘The uncle has said that you came’ (15) Súbitene t’ astirriba go.up.IMP.2SG-REFL-ABL to up.there ‘Go up to the top’ (16) La mullé ya se ‘n ha torna the woman already REFL.3SG ABL have.PRS.3SG come.back.PTCP.F.SG uta casa to home ‘The woman has already come back home’ (Arnal Purroy 1998, 317) The same pattern is frequently found in Central Aragonese: (17) Cuando quieras í-te ne, t’ en bas When want.SBJV.2SG go-REFL.2SG ABL REFL.2SG ABL go.PRS.2SG ‘When you want to go/leave, you go/leave’ (18) Baxa te ne Come.down.IMP.2SG REFL.2SG ABL ‘Come down’

272

A. Pineda

(19) M’

en puyo ta ras cumbres más altas go.up.PRS.1sg to the peaks most high ‘I climb the highest peaks’ (20) Nos ne tornaremos REFL.1PL ABL come.back.FUT.1PL ‘We will come back’ (Nagore Lain 1986, 108–109) REFL.1SG ABL

Finally, for Western Aragonese, examples with the clitic cluster (21) are found together with examples with only the reflexive, aspectual clitic se (22)—the latter thus patterning with Spanish: (21) Ellos sen son tornaus They REFL.3SG-ABL be.PRS.3PL come.back.PTCP.M.PL ‘They REFL ABL have come back’ (22) Te yeras tornado REFL.2SG be.IPFV.3PL come.back.PTCP.M.SG ‘You had come back’ (Alvar 1953, 292) With these Aragonese data we have completed our overview on the availability of pronominal DM verbs in several Romance varieties. Let us turn to assess the role of the two different clitics that appear in the linguistic spectrum just described.

3 On the Role of se and ne with Verbs of Directed Motion Using Spanish data, Jiménez-Fernández and Tubino (2019) establish a distinction we think is intuitively right: DM verbs without se “name only the culmination point or denote just the achievement of a particular position”, whereas with se they entail a resultant state, related to the notion of remaining in the new position achieved, after having left behind the original position. Let us show now that this generalisation also holds for the other Romance languages under study where a locative clitic appears in addition to se. This is clearly illustrated by the sentences (23) and (24), featuring a non-pronominal and a pronominal verb of motion, respectively: (23) El roder fugia de la Guàrdia Civil the bandit run.away.IPFV.3SG from the Guard Civil ‘The bandit ran away from the Civil Guard’ (24) El roder, com que li feien la vida impossible the bandit as that him.DAT do.IPFV.3PL the life impossible se ’n va fugir a la muntanya REFL.3SG ABL go.PRS.3SG run.away.INF to the mountain ‘The bandit, as they were making his life impossible, ran away to the mountains’

The Role of SE and NE in Romance Verbs of Directed Motion

273

Both sentences were produced by a Valencian Catalan speaker, who was telling a story of a bandit (recall that Valencian is one of the dialects of Catalan where pronominal DM verbs are used extensively, as mentioned in Sect. 2.2). The crucial point here is the contrast between the choice of a non-pronominal verb in (23), and a pronominal verb in (24), emphasizing that the latter implies that the bandit ran away and stayed in the mountains for years. The existence of such a resultant state/location in (24) but not in (23) is backed up by contrasts regarding the licensing of the interpretation that some adverbs (such as quasi ‘almost’, bé ‘well’ and ràpid ‘quickly’) have when they do not refer to the whole event but attach to the resultant state/location (Armstrong 2011; Cuervo 2014). Crucially, the specific interpretation that these adverbs have when attaching to the resultant state/location is only available for (24): for example, the adverb bé (se’n va fugir bé a la muntanya) has the reading of ‘in good condition’, whereas if attaching to the whole event the meaning would be ‘correctly’— the latter interpretation is the only one available for the non-pronominal DM verb in (23) (va fugir bé a la muntanya). In Sect. 5, however, we will refer to these tests in more detail, taking into account the differences that emerge between Romance varieties with pronominal DM verbs (such as Valencian Catalan) and Romance varieties where such pronominal forms are not productive (such as general Catalan). The distribution in (23)–(24) reflects a pattern identical to that observed by Jiménez-Fernández and Tubino in the case of Spanish, where a pronominal verb of motion followed by a phrase that denies the permanence of the subject in the position reached (‘but he didn’t stay’) is pragmatically odd (as we saw in (4) above). Along these lines, in (24) a continuation such as ‘but he didn’t stay’ would be pragmatically odd, whereas if the verb was the non-pronominal form va fugir ‘he ran away’ such a continuation would be absolutely compatible, meaning that the bandit had run away to the mountains (in that direction) but then he ended somewhere else. Therefore, agreeing with Jiménez-Fernández and Tubino’s view, we assume that pronominal DM verbs have a complex event structure that includes a resultant state or, more specifically, a resultant location, as exemplified in (25), where said location is denoted by the small clause el hombre en casa ‘the man at home’. Also, in the spirit of the Distributed Morphology neoconstructionist approach, we assume that verbs are formed in the syntax by combination of a root and a verbalizing head v (Marantz 1997); and in the particular case of pronominal uses of DM verbs, the root combines with a verbalizing head vCAUSE (25). The claim that the subject of pronominal DM verbs has a higher degree of implication in the event (i.e. the subject is the causer of the event) has been made for Spanish by Otero (1999, 1471), García Fernández (2011, 63–64) 6 and also Jiménez-Fernández & Tubino (2014, 32), albeit with differences in terminology.

6

For García Fernández (2011, 63–64), se signals the degree of participation of the subject in the event in which it is involved, in the sense that se can co-appear with agents La gente se salía del cine porque la película era aburrida (‘People REFL were coming out of the cinema because the movie was boring’) as well as experiencer subjects, as in dialectal Catalan L’aigua se n’eixia (‘The water REFL ABL went out’), where something causes the water to undergo a process of spilling over, or in Spanish El cadaver se salió del ataúd (‘The corpse SE came out of the coffin’). In contrast, se is not compatible with patients (#La estatua se salió del museo ‘The statue REFL went out of the museum’).

274

A. Pineda

(25) Sp. El hombre se entra en casa ‘The man goes into the house’

Thus, pronominal DM verbs are assumed to have a complex event structure formed by a causative subevent and a subevent denoting the resultant state.7 In (25) the subject of entrarse is the causer of a change of state (consisting of a change of location), and it is also the undergoer of this change.8 In such a configuration, Jiménez-Fernández & Tubino (2019, 206–207) take se to be “the realization of vCAUSE as an indication that, even though the same referent is associated with both resultant and causing subevents, this must be interpreted as two separate entities”; as they state, this is “the only way the structure may entail a change of state/location rather than a pure [i.e. simple] achievement”. In this regard, the authors point out that the most salient semantic contrast between a complex and a simple achievement “has to do with the interpretation of the event as no longer denoting a mere culmination but rather a full change of state or position, depending on the semantic denotation of the verb” (Jiménez-Fernández and Tubino 2019, 186). Although the analysis sketched in (25) seems correct to us, our proposal does not directly hang on the particular formalization of se combining into the structure: the only relevant thing to us is the role attributed to it. Thus, we could also adopt Cuervo’s (2014) view on se with break-type verbs and adapt it to our verbs by claiming that the clitic spells out, as agreement on vCAUSE, the φ-features of the DP subject, thus signaling its dual role, namely the causer of the change of location and the holder of the new state of ‘being located in some place’ (i.e. undergoer of the change of location).9 7 We assume that for contextual reasons it may be the case that the locative PP remains covert, as in Emilio, después de pasar horas sentado delanto de su casa, al final se entró (‘Emilio, after spending several hours sitting in front of his house, finally SE went in’). In such cases, the implication of a resultant state/location is still there, and so is the presence of a SC complement in the structure. 8 In this approach, it is assumed that causers are not necessarily [-human] (such as natural forces), but they can also be [+human], following Folli and Harley (2005). 9 See Cuervo (2014, 53) for the original proposal on verbs like romperse ‘break’, which however differ from DM verbs in many respects –and the analysis of such differences falls beyond the scope of this paper. It is however worth pointing out that the view of anticausatives (such as romperse ‘break’) as underlyingly reflexive has been argued by Koontz-Garboden (2009). A similar analysis

The Role of SE and NE in Romance Verbs of Directed Motion

275

Having made the above observations, we think the analysis in (25) should be refined in the light of the data offered by our cross-Romance perspective (recall Sect. 2.2). In our view, Catalan, Italian and Aragonese varieties are crucial for improving the existing proposals on the role of se with DM verbs. In particular, these languages and their varieties make us reconsider the idea that se stands for an incorporated source argument or names the initial endpoint of the movement described by the event, as argued by the authors that based their analyses only on Spanish data (recall Sect. 2.1). Thus, we do not assume the before-mentioned aspectual role attributed to Spanish se, that is, we do not believe the aspectual contribution of boundedness (the notion of leaving behind a place) relies on se. Instead, we argue that there is a specific syntactic head that contributes this source-related meaning. (In Sect. 4, we suggest that an analysis of such a locative component in terms of an applicative projection could be feasible.) Recall that, as we saw in Sect. 2.1, the most salient interpretation of pronominal DM verbs entailed leaving behind a place (generally after having remained there, and in order to stay in a new place). (26) Sp. El hombre se entra en casa ‘The man goes into the house’

is proposed in Wood (2014, 1402–1406) for some Icelandic verbs with the reflexive affix –st/sja, which is taken to function as an expletive pointing to the Theme (or “figure”, in the author’s account) being correferent with the Agent, or, in other words, the subject having two θ-roles.

276

A. Pineda

As can be seen in (26), the locative projection remains covert in Spanish (Ø), whereas it can be phonologically overt in the other Romance languages under study, crucially spelled out by a locative clitic (en, ne).10 We think the ablative nature of this clitic reasonably supports our view that there is a specific syntactic head responsible for the source meaning entailed by pronominal DM verbs.11 As we saw in Sect. 2.1, the most salient interpretation of pronominal DM verbs entailed leaving behind a place (generally after having remained there, and in order to stay in a new place). By adopting a cross-Romance perspective we have been able to refine several existing proposals based only on Spanish data, whose claim was that the reference to the initial endpoint or source of the movement was contributed by se. Finally, in the face of obvious contrasts such as Spanish entrarse and Catalan entrar-se’n, it remains to be said that, under the Distributed Morphology approach we adopt, the morphophonological outcome of the locative head (en vs. Ø) is regulated by different rules across languages and varieties. An exhaustive account for interlinguistic and intralinguistic differences will be provided in Sect. 5. 12

10

Crucially, as will be shown in Sect. 4 devoted to the status of ne, this grammaticalized use of ne that co-occurs with se is to be distinguished from the very productive use of the locative clitic ne to replace actual source PPs, just like another locative clitic, hi, is productively used to replace goal and location PPs, as shown by Catalan examples below. See also fn. 16.

(i)

11

a. La Maria torna del mercat. ! La Maria en torna. ‘Mary goes back from the market. ! She goes back from there.’ b. La Maria va al mercat ! La Maria hi va. ‘Mary goes to the market. ! She goes there.’ c. La Maria és al mercat ! La Maria hi és. ‘Mary is in the market. ! She goes is there.’

As an anonymous reviewer notices, entrar-se’n (with both the aspectual se and the locative clitic) is attested in Valencian Catalan but not in general Catalan nor in Aragonese or Italian dialects, which seems to be indicating that that denoting a movement away from some location are accepted everywhere, but not with verbs denoting a movement (from some location) into some location. The notion of source, then, is clearly salient, which supports, we believe, our analysis. 12 It could also be accounted for as an instance of silent variation (Kayne 2005). This locative clitic is actually a sort of agreement in the sense that it replicates a directional component already present in the verbal root. In this sense, the optional coding of a grammatical function is not that unexpected (Comrie 1989). Actually, we could also argue that the non-realization of this head is the most economical solution, whereas what needs to be accounted for is its effective realization. This is Sigurðsson’s (2004, 245) view: “Given the Silence Principle, it is lexicalization that is last resort, requiring some licensing or justification [...], whereas non-lexicalization is the unmarked or the minimal strategy, applied whenever possible”. There are several reasons to keep a category silent: it is more economical for the speaker and the hearer, since the pronunciation of dispensable information is nothing but unnecessary and antieconomical noise (Merchant (2001, 1), Sigurðsson (2004, 245 i 254, fn. 25)). Obviously, this Silence Principle is constrained by a more general one, the Full Interpretation Principle (Chomsky 1986) which ensures the interpretability of syntactic derivations. We leave the exploration of such alternatives for future research.

The Role of SE and NE in Romance Verbs of Directed Motion

277

4 On the Status of ne The locative meaning contributed by ne presents interesting similarities with a particular type of applicative construction. Applicatives are a well-known distinctive aspect of the syntax of Bantu, Austronesian, Salish, Mayan and Uto-Aztecan language families (Polinsky 2013). They are functional heads responsible for the introduction of oblique, peripheral complements into the core argument structure of a predicate. Take for example the goal/recipient argument introduced by a preposition in (27a), which by means of an applicative becomes part of the core argument structure, that is, becomes a central, direct argument of the verb in (27b): (27) a. Mbidzi zi-na-perek-a msampha kwa nkhandwe zebras they-past-hand-ASP trap to fox ‘The zebras handed the trap to the fox’ b. Mbidzi zi-na-perek-er-a nkhandwe msampha zebras they-past-hand-APPL-ASP fox trap ‘The zebras handed the fox the trap’ (Chicheŵa) (Baker 1988) In other words, an applicative is a valency-increasing strategy, since in (27b) the verb can take an “extra” object. As can be seen from the English translations above, a parallelism has been argued to exist between the Bantu pair in (27) and the so-called English dative alternation, where a prepositional paraphrase (John gives a book to Mary) alternates with the so-called double object construction (John gives Mary a book), the latter arguably being an instance of applicative construction (see Marantz 1993, Pylkkänen 2002/2008). In the case of English, there is no verbal affix corresponding to the applicative head, which would be phonologically covert. The existence of applicative constructions in the realm of ditransitives has also been extended to the Romance area (see Cuervo (2003) for Spanish, Diaconescu and Rivero (2007) for Rumanian, Torres Morais and Salles (2010) for Portuguese, Fournier (2010) for French, and Pineda (2013, 2016, 2020) for Catalan). In the case of Romance languages, some authors argue that what signals the presence of an applicative head is the presence of a dative doubling clitic (e.g. Sp. Juan le dio el libro a María), whereas others argue that the applicative head may also remain covert, as in English (for extensive discussion, see Pineda (2013, 2016, 2020)). The latter view, according to which the applicative head may be overt or remain covert in Romance languages, is the one we will also pursue here in regard to locative applicatives. In fact, we have already shown this double possibility (en/Ø) in (26). To our knowledge, the identification of applicative(like) patterns outside the language families that have traditionally been considered applicative has been limited to the realm of dative-like constructions, with goal/recipient/beneficiary arguments. However, in those languages, e.g. in the Bantu family, applicatives can introduce very different θ-roles, including instrumentals (28) and locatives (29):

278

A. Pineda

(28) Umwaana a-ra-ri-iish-a ikanya child he-PRS-eat-APPL-ASP fork ‘The child is eating with the fork’ (29) Umuhuungu a-r-iig-ir-a-ho ishuuri imibare boy he-PRS-study-APPL-ASP-APPL school Maths ‘The boy is studying Maths at school’ (Kinyarwanda) (McGinnis 2008) Our aim here is to tentatively explore the possibility of an analysis of ne as an instantiation of a locative applicative.13 Two main properties would define this Romance locative applicative: it does not imply a major change in the valency of the verb, and it emphasizes a meaning component. Actually, these two properties can also be found in the Bantu cases. It is true that most examples of Bantu applicatives imply a change in valency, as we saw in (27) and can see in (30) and (31), where (a) examples contain a locative marker (ni14 and ta respectively) but (b) examples do not: (30) a. Salma a-li-ka-a kiti-ni Salma SCD1-PST-sit-FV chair-LOC ‘Salma was sitting on a chair’ b. Salma a-li-kal-i-a kiti cha uvivu Salma SCD1-PAST-sit-APPL-FV chair GEN laziness ‘Salma was slouching/sitting in a comfortable chair’ (Swahili) (Marten 2003, 214) (31) a. Poro cise ta horari big house in live ‘He lives in a big house’ b. Poro cise e-horari big house APPL-live ‘He lives in a big house’ (Ainu) (Shibatani 1996, 159) In (30a) and (31a) the locative argument is oblique (introduced by a postposition), whereas in (30b) and (31b) the locative argument appears as a direct object of the verb, together with concomitant applicative morphology on the verb. Interestingly, though, examples where the applicative does not imply any change in the valency of the verb (i.e. the postposition is not suppressed) can also be found –more examples are provided by Bresnan and Moshi (1990, 148–149), Ngonyani (1995, 8) and Bentley (1998, 193):

13

As Peterson (2007, 203 and 229–230) points out in a comparative survey of 50 languages, for a language to have a locative applicative construction, “it appears that it must also have either a benefactive [. . .] or an instrumental applicative construction”. Given that benefactive (dative-like) applicatives have been posited for Romance languagues, the postulation of a locative applicativelike element seems feasible. 14 On the postpositional nature of the suffix -ni, see Bentley (1998, 188).

The Role of SE and NE in Romance Verbs of Directed Motion

279

(32) a. mpishi a-li-pik-a jiko-ni cook SCD1-PST-cook-FV kitchen-LOC ‘The cook was cooking in the kitchen’ b. mpishi a-li-pik-i-a jiko-ni cook SCD1-PST-cook-APPL-FV kitchen-LOC ‘The cook was cooking in the kitchen’ (habitually) (Swahili) (Marten 2003, 216) As Marten (2003, 213, 217) reports, here the applicative “conveys a different interpretation of the predicate”, and its function “can be analysed as concept strengthening or predicate emphasis”. This is exactly what we argue for the Romance locative projection ne. In particular, we claim it emphasizes a directionality meaning (movement from a place). In fact, applicatives that specifically emphasize the direction of the movement are also reported to exist in Bantu languages with DM verbs: (33) a. waziri a-li-anguk-a chini minister SCD1-PAST-fall-FV down ‘The minister fell down’ b. waziri a-li-anguk-i-a chini minister SCD1-PAST-fall-APPL-FV down ‘The minister fell down’ (with an implied meaning of directionality) (Swahili) (Abdulaziz 1996, 32) In our view, this kind of applicative found in Bantu languages, non-valency augmenting and emphasizer of a directionality meaning already existent in DM verbs, shows striking similarities with the locative source projection that co-occurs with Romance pronominal DM verbs, instantiated by ne.15 Importantly, such a clitic does not replace anything; it does not stand for any actual source argument.16 This is

15

Yet another point in support of our proposal is made by Campanini & Schäfer’s (2011, 30): in an appendix of their work, they briefly mention the case of pronominal verbs in Spanish (with se) and French and Italian (with se and ne) and state that “[t]he locative semantics of the relevant verbs are consistent with the prepositional nature of ne/en: the clitic appears to express a (locative) PP which is inherently encoded by the lexical/root meaning of the verb”. 16 Actually, pronominal DM verbs that are source-oriented, such as ‘go out’, can combine with a source argument, as in the following Valencian Catalan example: (i)

L’ home se n’ eixí de casa The man REFL.3SG ABL go.out.PST.3SG of home ‘The man left home’

It is thus important to keep in mind that our ne is different from its full-fledgded counterpart, which in the languages under study is used to replace a source PP with the meaning ‘from some place’, as shown by the following Catalan example:

280

A. Pineda

why it is considered a grammaticalized, inherent element. In other words, in the above-mentioned uses ne has lost its pronominal function and just co-occurs with the verb, as noted for Catalan by several descriptive grammarians (Fabra 1956, Lacreu 1990). Likewise, for Aragonese Arnal Purroy (1998, 316–317) shows that the extensive use of ne with DM verbs entailed the loss of meaning so ‘it became just an index required by verbs of movement in their reflexive form’ [our translation]. Similar appreciations hold for Southern Italian dialects such as Calabrian (Alessandra Lombardi, p.c.). Summing up, the clitic ne in these occurrences does not anaphorically refer to any location.17 This is reflected in the dictionary definitions given for general Catalan forms anar-se’n and tornar-se’n, which are said to imply leaving a place (DIEC2 s.v. anar, tornar); similarly, for the particular case of Valencian Catalan, where, as we pointed out, a greater set of pronominal DM verbs exists, Todolí (2002, 1426–1427) reports that ne delimits the movement denoted by the verb, implying a meaning of ‘going away’ that the equivalent non-pronominal verbs do not have. This description fits into our suggestion that ne could be analysed as an instance of a locative (high) applicative head (located above the vP), whose contribution is to emphasize a directional meaning that is already an intrinsic component of the verbal

(ii)

L’

Anna ha

the

Anna have.PRS.3SG come.back.PTCP from Paris? yes,

tornat

de

París? Sí,

ja n’ ha tornat already ABL have.PRS.3SG come.back.PTCP ‘Is Anna back from Paris? Yes, she ABL is back.’

At the same time, it is true that in several of the varieties under study the actual preservation (in the spoken language) of ne as an pronoun standing for source locative complements (2) is weak: this is the case of Valencian Catalan, as well as Eastern Aragonese (Arnal Purroy 1998, 316–317), Central Aragonese (Nagore Lain 1986, 106), Southern Italian varieties (Ledgeway 2009, 350; Adam Ledgeway and Alessandra Lombardi, p.c.). It is our view that such a decline is due to independent reasons, but of course it constitutes a matter of future research. Another issue of further research has to do with the deictic interpretation (in particular, speaker-oriented) that pronominal DM verbs seem to have in some varieties such as Calabrian (Alessandra Lombardi, p.c.). 17 Very interestingly, Burzio (1986) and Tortora (1998, 2001) report that the Piedmontese ye and the Borgomanerese gghi, which are goal-indicating locative clitics, can co-occur with DM verbs while being anaphorically non-referential, just like the source-indicating locative clitics we are dealing with. These Northern Italian goal clitics are not referring to any location, they can combine with explicit goal-PP (i), just as we saw for ne in fn. 16: (i)

N gh CL

LOC.CL

è riva-gghi tre matai a la is arrive.PTCP-LOC.CL three boys at the

staziôn station ‘Three boys arrived at the station’

The Role of SE and NE in Romance Verbs of Directed Motion

281

root. In conclusion, the locative projection we put forth, spelled out by ne in some Romance languages (and phonologically covert in others18), emphasizes a directional component of the meaning of the verbal root: the notion of leaving behind the initial location (after having remained there). This syntactic head is a part of a complex event structure that also includes a resultant state: a new position reached by the subject, in which it also remains for some time. Furthermore, as shown in (26) above, the LocativeSOURCE Appl does not license an argument in its specifier position: only case-marked individuals can occupy such position, but Romance languages have not preserved locative/ablative case marking on DPs, only on clitics. This situation is similar to the one described by Roberge and Troberg (2009) with regard to the dativus commodi/incommodi in languages such as French or Italian: unlike Spanish, whose clitic-doubled a-DPs are dative-marked DPs (Cuervo 2003), in non-doubling languages such as French or Italian there are no dative-marked DPs (a/à-DPs are PPs), such that “the only candidate in the lexicon that can satisfy the syntactic selectional requirement of Appl in these languages is an operator” (Roberge and Troberg 2009, 281), as shown in (34): (34)

ApplP Op[iD]

Appl’ Appl[Case, θ , uD] clDAT[iφ, uCase, uθ]

vP … DP

(Roberge and Troberg 2009, 281) Our view is that, similarly, Romance languages have no locative/ablative casemarking available for DPs, and thus what merges in Spec,Appl in (26) is a null, expletive operator, which “satisfies the selectional requirement of the Appl head, whose case and theta properties have been absorbed by the clitic” (Roberge and Troberg 2009, 250). We thus have an ablative case-marked clitic that contributes a source meaning/nuance. The fact that the analysis of this clitic partially resembles that of dativus commodi/incommodi is not as surprising as it might seem: in both cases, we are dealing with Romance oblique clitics (datives, locatives) corresponding to a high applicative head which contributes a particular semantic nuance or strengthening but which do not introduce any additional argument. More importantly, the operator analysis allows Roberge and Troberg (2009) to ensure that

18

A similar line of reasoning, albeit with differences, is found in Tortora (1998, 2001): she claims that the goal locative clitic gghi that coappears with DM verbs in Borgomanerese (see fn. 17) constitutes evidence for claiming that a phonologically null goal projection is part of the structure of such verbs in general Italian.

282

A. Pineda

such an applicative head in French or Italian necessarily combines with transitive verbs. (The operator has the structural requirement that a referential DP must merge within the VP, thus accounting for the infelicitousness of dativus commodi/ incommodi with intransitive verbs in these languages.) Building on this, we argue that the operator analysis may also account for the distributional properties of the LocativeSOURCEAppl: the operator has the structural requirement that a DP undergoing a complex event of change of location must merge under its scope. As argued, such a complex event whereby an individual is caused to undergo a change of location is structurally represented by a DP embedded under a SC, when there is a causing vP composing on top of the verbal root. Recall that the presence of this causing vP “is responsible for the theme DP’s interpretation as changing location to the location named by the goal PP” (Jiménez-Fernández and Tubino 2019, 206). Therefore, for ne to co-occur, the presence of vCAUSE is required. In addition, such a requirement rules out the combination of a LocativeSOURCEAppl with non-pronominal DM verbs that denote just the achievement of a particular position without entailing any actual change of position, i.e. the root does not combine with vCAUSE but with vDO (Jiménez-Fernández and Tubino 2019, 202, 206). In other words, the need of emphasizing/strengthening the reference to the initial endpoint of the movement described by a verb of motion arises only when the subject is undergoing an actual change of position, i.e. when we have the structure in (26), where the focus is on the fact that the individual is leaving a place or going somewhere else after having remained in a place for a while, at the initial endpoint. Put differently, given that leaving a place behind implies a result, our analysis can semantically capture the intuition that the causer is related to the existence of a resulting state: without the resulting state that follows from leaving a location behind (or, by metaphorical extension, a property) the notion of ‘source’ is not computable, and such a source is the one that determines that there is a component of causation, if we understand causation, in a broad sense, as the origin of the changing process.

5 Accounting for Inter- and Intralinguistic Variation Some words need to be said regarding the huge amount of variation we find across and within languages when it comes to the realization of these pronominal verbs. In particular, we should account for the difference across and within languages concerning which DM verbs can be used pronominally. Some of these differences are summarized in Table 1 below. We assume that all varieties do indeed have the structure in (26), although differences arise when it comes to the lexicalization of the structure, in nanosyntactic terms. Under the view that syntactic structure is formed by very small terminal nodes

The Role of SE and NE in Romance Verbs of Directed Motion

283

Table 1 Cross-linguistic differences between directed motion verbs Spanish Catalan Aragonese

General Dialectal General Dialectal Western Central

subirse, bajarse, entrarØ, salirØ. . . subirse, bajarse, entrarse, salirse. . . entrarØ, pujarØ, baixarØ. . . entrar-se’n, pujar-se’n, baixar-se’n tornáse tornásene

(hence the term nano-), the lexicalization of the structure is a post-syntactic operation and different strategies, such as phrasal spell-out, let lexical items spell out more than a single node (i.e. non-terminal nodes) (Pantcheva 2009, 17–18). This allows us to explain why different varieties and languages, despite sharing the same structure, provide different outcomes. 19 (35) entrar (general Spanish)

19

A similar solution we could resort to is Fusion, an operation invoked within the Distributed Morphology approach (Halle and Marantz 1993) by virtue of which two nodes are fused into a single terminal node and then spelled out as one.

284

(36) entrarse (dialectal Spanish)

(37) entrar (general Catalan)

A. Pineda

The Role of SE and NE in Romance Verbs of Directed Motion

285

(38) entrar-se’n (dialectal Catalan, e.g. Valencian Catalan)

Note that the realization of this head is conditioned to the presence of se. Recall that the operator requirement ensured the presence of vCAUSE in its domain, but it remains to be explained why it is the case that vCAUSE has to be lexicalized by a dedicated lexical item se, or otherwise (if phrasal spell out applies) ne cannot be overt (cf. Val. Cat. entrar-se’n vs. *entrar-ne). It seems that the answer is to be found in the diachronic evolution from Latin to Romance, which shows that the “reflexive” clitic was already present in Latin (and more widespread in Vulgar Latin), and the locative clitic grammaticalized later.20 Therefore, our proposal is that speakers have at their disposal the structure in (26) regardless of whether they actually pronounce the clitics, that is, regardless of

20

Rohlfs (1954, 188) points out that the pronominal use of DM verbs (in Italian) was a further development of a possibility already existent strategy in Latin. Bassols de Climent (1948, 32) and Bobes (1974, 112–113) explain that the use of se with Romance DM verbs is historically related to the ethical reflexive in Latin that occurred rather often with these verbs: Ambulavimus nobis per heremum, Vade tibi in propiam ecclesiam. Bastardas Parera (1953, 112) shows that ethical datives with intransitive verbs appeared as an analogy with transitive verbs (e.g. separare ‘separate’). Such verbs, when combined with SIBI, acquired a meaning rather similar to that of DM verbs (e.g separare + SIBI ‘leave’). In her exhaustive account of the use of SIBI and SE in Late Latin, Cennamo (1999, 140) shows that by the eight–ninth centuries, “one can also find SE with verbs that in earlier centuries only occurred with SIBI (i.e., also with unaccusatives) (e.g. se ire ‘go’, se turnare, se reverti ‘go back’)”. As for Romance diachrony, a corpus survey on Old Catalan confirms to us that those verbs appear with se in the earlier texts, and with both se and ne later on (Corpus informatitzat del catala antic: www.cica.cat). An exhaustive diachronic study of such verbs falls however beyond the scope of this paper. See also fn. 4.

286

A. Pineda

whether they opt for (35), (36), (37) or (38). This conclusion is additionally backed up by tests of adverbial modification. For example, we have conducted some experiments with Catalan speakers from different dialectal origin, using adverbs that can be interpreted in more than one way (Armstrong 2011; Cuervo 2014). In particular, we have used adverbs that have a specific interpretation when attaching to the resultant state: quasi (‘almost’) can refer to the whole event or to the resultant state only (the final state/location); bé (‘well’) can have a manner meaning (‘correctly’), or it can be a degree modifier of the resultant state; and ràpid (‘quickly, fast’) can signal the manner interpretation (‘at high speed’), or express a temporal frame in the sense that the final location/resultant state is achieved within a short time. Due to space constraints, we will sum up the results of the experiments conducted. It is crucial that the resultantstate-related interpretation of the above mentioned adverbs is always available, not only for speakers of Valencian Catalan (a variety where the clitics are systematically overt), but also for speakers from other areas where the clitics generally remain covert (recall the geographical distribution presented in Sect. 2.2). We obtain the same result with Spanish speakers, no matter which dialect they speak. The case of entrar ‘go in’ and its opposite salir ‘go out’ is of particular interest. Recall that, as shown in Sect. 2.1, in general Spanish, only the latter can take se. De Miguel and Fernández Lagunilla (2000, 29–31) account for this by claiming that the two verbs have a different event structure: only salir corresponds to an achievement followed by a resultant state (and thus can take se), whereas entrar is a simple achievement. However, as García Fernández (2011, 50–51) points out, this approach is quite stipulative, since it seems that the verbs are assigned an event structure depending exclusively on the possibility of co-occuring with se, which yields counterintuitive results (entrar and salir having different event structures). He convincingly shows that an adverbial expression can target a resultant state within the structure of both salir(se) and entrar, and therefore both of these verbs can be conceived by speakers of general Spanish as a complex event, regardless of the presence of se. Putting it in the terms of our analysis presented above, speakers have at their diposal a complex event structure for both verbs, regardless of the lexicalization pattern they opt for. Although García Fernández does not mention it, his conclusion is reinforced by the observation that entrarse is actually possible in some Spanish dialects, as we saw in the examples provided in the previous sections. The variation between entrar (general Spanish) and entrarse (dialectal Spanish), both of which can have a complex event structure, fits perfectly into our account based on different lexicalization patterns available for one structure.21,22

21

Actually, we think the approach here presented, where variation boils down to different lexicalization patterns, can be extended to other cases concerning se. Take the example, pointed out by Martín Zorraquino (1979, 297–298), of some north-western Spanish varieties where the clitic of irse ‘leave’ is omitted even when implicitly or explicitly there is a complement indicating source: Va de aquí, ¿Ya va? ‘(S)he is leaving from here. Is (s)he already leaving?’, that being completely out in any other Spanish variety, where Se va de aquí, ¿Ya se va? is the only grammatical option. The same could be said for differences across and within languages such as Catalan marxar ‘leave’, general Spanish marcharse and northern Peninsular Spanish marchar (Sánchez López 2002, 116). 22 As an anonymous reviewer points out, as a complement of single vs. phrasal spell out, we need to assume that there is some mechanism that tells vocabulary insertion to insert se in the presence of a

The Role of SE and NE in Romance Verbs of Directed Motion

287

We thus conclude that, regardless of whether Romance speakers actually pronounce the clitic se or the clitic cluster se + ne, it is possible for them to conceive the different DM verbs here studied as denoting complex events.

6 Conclusions In this paper we have analyzed the use of se with DM verbs in Romance languages. Building on some observations that have been made for Spanish, we have adopted a broader cross-linguistic perspective, bringing into discussion an element that has, until now, gone generally unnoticed (aside from descriptive works): the ablative locative clitic that appears, together with se, in Catalan, Italian and Aragonese varieties. Our data from different Romance languages and dialects have allowed us to refine the settings of the connection between pronominal DM verbs and the existence of a source component. In particular, we have posited the existence of a locative head (here analysed as an applicative), which can be spelled out by an ablative locative clitic. We have also argued that DM verbs can be conceived of by Romance speakers as simple, punctual events denoting the achievement of a particular position, but also as denoting a complex event that consists of a causing subevent and a resultant state (which is connected to achieving a new position and remaining there for some time, after having left behind the original location). In the latter case, the DM verbs can surface in their pronominal form, even if it does not happen always. As we have proven, there is cross-linguistic and cross-dialectal variation regarding the availability of pronominal forms for these verbs, due to different lexicalization patterns. Acknowledgments Parts of this work were presented at the Seminari de Lingüística Teòrica at the Universitat Autònoma de Barcelona in December 2015, the IKER Mintegia at the Centre National de la Recherche Scientifique in Bayonne (France) in January 2016, the Linguistics Symposium on Romance Languages (LSRL46) in Stony Brook (New York) in March 2016, the Workshop on Romance SE/SI constructions in Madison (Wisconsin) in April 2016, and the Cambridge Comparative Syntax 5 (CamCoS5) in Cambridge in May 2016. I would like to thank the participants, as well as two anonymous reviewers, for their valuable comments and suggestions that helped to improve this paper. All errors remain mine This work has been supported by the postdoctoral research fellowship Beatriu de Pinós 2014 BP_A 00165 (Secretaria d’Universitats i Recerca del Departament d’Economia i Coneixement de la Generalitat de Catalunya) and the research project FFI2014-56968-C4-1-P (Ministerio de Economía y Competitividad of the Spanish Government).

list of some roots, which varies between languages and dialects. A similar mechanism, depending on the former one, would be responsible for the insertion of ne in the presence of se in some languages and varieties too.

288

A. Pineda

References Abdulaziz, Mohamed. 1996. Transitivity in swahili. Köln: Köppe. Alvar, Manuel. 1953. El dialecto aragonés. Madrid: Gredos. Armstrong, Grant. 2011. Two classes of transitive verbs: Evidence from Spanish. PhD thesis, Georgetown University. ———. 2013. Agentive reflexive clitics and transitive se constructions in Spanish. Borealis. An international journal of Hispanic linguistics 2 (2): 81–128. Arnal Purroy, María Luisa. 1998. El habla de la Baja Ribagorza occidental: Aspectos fónicos y gramaticales. Zaragoza: Institución “Fernando el Católico”. Baker, Mark C. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Basilico, David. 2010. The se clitic and its relationship to paths. Probus 22: 271–302. Bassols de Climent, Marià. 1948. Sintaxis histórica de la lengua latina. Barcelona: CSIC. Bastardas Parera, Joan. 1953. Particularidades sintácticas del latín medieval. Barcelona: CSIC. Bentley, Mayrene. 1998. The marking of grammatical relations in Swahili. Studies in African Linguistics 27: 177–197. Bobes, María del Carmen. 1974. Construcciones castellanas con “SE”: Análisis transformacional continuación. Revista española de lingüística, year 4, fasc. 1, 87–128. Bresnan, Joan, and Lioba Moshi. 1990. Object asymmetries in comparative bantu syntax. Linguistic Inquiry 21: 147–185. Burzio, Luigi. 1986. Italian syntax. Dordrecth: Foris. Campanini, Cinzia, and Florian Schäfer. 2011. Optional Se-constructions in romance: Syntactic encoding of conceptual information. In Ms, Generative Linguistics in the Old World 34. Wien: Universität. Cennamo, Michela. 1999. Late Latin pleonastic relexives and the unaccusative hypothesis. Transactions of the Philological Society 97 (1): 103–150. Chomsky, Noam. 1986. Knowledge of language. New York: Praeger Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago: The University of Chicago Press. Cuervo, M. Cristina. 2003. Datives at large. PhD diss., MIT. ———. 2014. Alternating unaccusatives and the distribution of roots. Lingua 141: 48–70. De Miguel, Elena. 1999. El aspecto léxico. In Gramática descriptiva de la lengua española, ed. I. Bosque and V. Demonte, vol. 2, 2977–3060. Madrid: Espasa-Calpe. De Miguel, Elena, and Marina Fernández Lagunilla. 2000. El operador aspectual 'se'. REL 30: 13–44. Diaconescu, Constanţa Rodica, and María Luisa Rivero. 2007. An applicative analysis of double object constructions in Romanian. Probus 19: 209–233. DIEC2. n.d. Diccionari de la llengua catalana. de l’Institut d’Estudis Catalans. http://dlc.iec.cat. Fabra, Pompeu. 1956. Gramàtica catalana. Barcelona: Teide. Folli, Raffaella and Heidi Harley. 2005. Flavours of v: Consuming Results in Italian and English. In Aspectual Inquiries, ed. by Paula Kempchinsky & Roumyana Slabakova, 95–120. Dordrecht: Springer. Fournier, David H. 2010. La structure du prédicat verbal: une étude de la construction à double objet en français. PhD diss., University of Toronto. Gallardo, Elías. 2008. Estrategias de inacusatividad en inglés, japonés y lenguas románicas. In Actas del VIII congreso de lingüística general, 747–762. Madrid: UAM. ———. 2010. A spatial-deictic explanation for the use of the inchoative reflexive pronoun. In Space in language, ed. G. Marotta, A. Lenci, L. Meini, and F. Rovai, 461–476. Piza: Edizioni ETS. García Fernández, Luis. 2011. Algunas observaciones sobre el 'se' aspectual. In Estudios sobre perífrasis y aspecto, ed. J. Cuartero Otal, L. García Fernández, and C. Sinner, 43–71. München: Peniope.

The Role of SE and NE in Romance Verbs of Directed Motion

289

Giralt, Javier. 1995. Valors i funcions d’IBI i INDE als parlars lliterans. Alazet 7: 57–74. Halle, Morris, and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In The view from building 20, ed. K. Hale and S.J. Keyser, 111–176. Cambridge, MA: MIT Press. Jiménez-Fernández, Ángel, and Mercedes Tubino. 2014. Variación sintáctica en la causativización léxica. Revista Española de Lingüística 44 (1): 7–37. ———. 2019. Causativity in southern peninsular Spanish. In The syntactic variation of Spanish dialects, ed. Á. Gallego, 181–212. Oxford: OUP. Kayne, Richard. 2005. Movement and silence. Oxford, UK: Oxford University Press. Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27: 77–138. Lacreu, Josep. 1990. Manual d’ús de I’estandard oral. Valencia: Universitat de Valencia. Ledgeway, Adam. 2009. Grammatica diacronica del napoletano. Tübingen: Max Niemeyer Verlag. Llorente Maldonado, Antonio. 1980. Consideraciones sobre el español actual. Anuario de Letras XVIII: 5–61. MacDonald, Jonathan E. 2004. Spanish reflexive pronouns: A null preposition hypothesis. In Proceedings of WCCFL 23, ed. G. Garding and M. Tsujimura, 528–540. Somerville, MA: Cascadilla Press. ———. 2008. The syntactic nature of inner aspect: A minimalst perspective. Amsterdam/Philadelphia: John Benjamins. ———. 2017. Spanish aspectual se as an indirect object reflexive: Evidence from (a)telicity, bare nouns and leísta PCC repairs. Probus 29 (1): 73–117. Marantz, Alec. 1993. Implications of asymmetries in double object constructions. In Theoretical aspects of Bantu grammar, ed. Sam Mchombo, 113–150. Standford: CSLI Publications. Marantz, Alex. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. U. Penn Working Papers in Linguistics 4 (2): 201–225. Marten, Lutz. 2003. The dynamics of bantu applied verbs: An analysis at the syntax-pragmatics interface. In Actes du 3e Congrès Mondial de Linguistique Africaine Lomé 2000, ed. Kézié K. Lébikaza, 207–221. Köppe: Köln. Martín Zorraquino, M. Antonia. 1979. Las construcciones pronominales en español. Madrid: Gredos. Masullo, Pascual José. 1992. Incorporation and case theory in Spanish. A cross-linguistic perspective. PhD diss., University of Washington. McGinnis, Martha J. 2008. Applicatives. Language and Linguistics Compass 2 (6): 1225–1245. Merchant, Jason. 2001. The syntax of silence. Sluicing, Islands, and the theory of ellipsis. New York: Oxford University Press. Nagore Lain, Francho. 1986. El aragonés de Panticosa. Gramática. Huesca: Diputación Provincial. Ngonyani, Deo. 1995. Towards a typology of applicatives in Bantu. Ms, 26th annual conference on african linguistics, University of Los Angeles. Nishida, Chiyo. 1994. The Spanish reflexive clitic se as an aspectual class marker. Linguistics 23: 425–458. Otero, Carlos Peregrín. 1999. Pronombres reflexivos y recíprocos. In Gramática descriptiva de la lengua española, ed. I. Bosque and V. Demonte, 1427–1517. Madrid: Espasa Calpe. Pantcheva, Marina. 2009. Directional expressions cross-linguistically: Nanosyntax and lexicalization. In Nordlyd, Tromsø Working Papers on Language and Linguistics: 36, Special issue on Nanosyntax, ed. P. Svenonius, G. Ramchand, M. Starke, and T. Taraldsen, 7–39. Tromsø: University of Tromsø. Peterson, David A. 2007. Applicative constructions. Oxford: OUP. Pineda, Anna. 2013. Double object constructions in Spanish (and Catalan) revisited. In Romance languages and linguistic theory 2011, ed. S. Baauw, F. Drijkoningen, L. Meroni, and M. Pinto, 193–216. Amsterdam/Philadelphia: John Benjamins. ———. 2016. Les fronteres de la (in)transitivitat. Estudi dels aplicatius en llengües romàniques i basc, 6. Barcelona: Institut d’Estudis Món Juïc, Col
lecció Cum Laude.

290

A. Pineda

———. 2018. Causativization of verbs of directed motion in romance languages. In Romance languages and linguistic theory. Selected proceedings of the LSRL, ed. P. Ordóñez and L. Repetti, vol. 46, 245–262. Amsterdam/Philadelphia: John Benjamins. ———. 2020. Double object constructions in romance: The common denominator. Syntax 23: 203–240. https://doi.org/10.1111/synt.12193. Polinsky, Maria. 2013. Applicative constructions. In The world atlas of language structures online, ed. M.S. Dryer and M. Haspelmath. Leipzig: Max Planck Institute for Evolutionary Anthropology. Pylkkänen, Liina. 2002. Introducing arguments. Tesi doctoral, MIT. [Also Pylkkänen, Liina. 2008. Introducing arguments. Cambridge: The MIT Press.] Roberge, Yves, and Michelle Troberg. 2009. The high applicative syntax of the dativus commodi/ incommodi in romance. Probus 21: 249–289. Rohlfs, Gerhard. 1954. Grammatica storica della lingua italiana e dei suoi dialetti. Vol. 1. Morfologia. Torino: Einaudi. Sánchez López, Cristina. 2002. Las construcciones con se. Estado de la cuestión. In Las construcciones con se, ed. C. Sánchez López, 13–163. Madrid: Visor Libros. Sanchis Guarner, Manuel. 1950. Gramàtica valenciana. València: Editorial Torre. Schroten. 1972. Concerning the deep structure of Spanish reflexive sentences. The Hague: Janua Linguarum. Shibatani, Masayoshi. 1996. Applicatives and benefactives: A cognitive account. In Grammatical constructions, ed. M. Shibatani and S. Thompson, 157–194. Oxford: Clarendon Press. Sigurðsson, Halldór Ármann. 2004. Meaningful silence, meaningless sounds. Linguistic Variation Yearbook 4: 235–259. Todolí, Júlia. 1990. Les funcions pronominals en el català del País Valencià. In Miscel
lània Joan Bastardas, vol. 4, 253–261. Barcelona: Abadia de Montserrat. ———. 2002. Els pronoms. In Gramàtica del català contemporani, ed. Solà et al., vol. 2, 1337–1433. Barcelona: Empúries. Torres Morais, Maria Aparecida, and Heloisa Salles. 2010. Parametric change in the grammatical encoding of indirect objects in Brazilian Portuguese. Probus 22: 181–209. Tortora, Christina. 1998. The post-verbal subject position of Italian unaccusative verbs of inherently directed motion. In Semantic issues in romance syntax, ed. E. Treviño and J. Lema, 283–298. Amsterdam/Philadelphia: John Benjamins. ———. 2001. Evidence for a null locative in Italian. In Current studies in Italian syntax: Essays offered to Lorenzo Renzi, ed. G. Cinque and G. Salvi, 313–326. Amsterdam/London: Elsevier. Wood, Jim. 2014. Reflexive -st verbs in Icelandic. Natural Language & Linguistic Theory 32 (4): 1387–1425. Zagona, Karen. 1996. Compositionality of aspect: Evidence from Spanish aspectual se. In Aspects of romance linguistics, ed. C. Parodi, 475–488. Washington, DC: Georgetown University Press.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited Margot Vivanco

Abstract This chapter addresses the co-existence of two morphological mechanisms to express the causative-unaccusative alternation in Spanish: se and Ø. It has been observed that the choice between these patterns seems to be related to aspect in Romance. However, a detailed study of Spanish data will allow us to refine this hypothesis by claiming that aspect itself is determined by scale structure. Se and Ø will be analyzed as two lexical items competing to spell out the same head, v[BECOME], related to unaccusativity and located above AspP. Therefore, scale structure and, eventually, aspect will be the relevant grammatical factors determining the competition to spell out v[BECOME] in Spanish. Keywords Anticausative alternation · Labile alternation · Unaccusativity · Aspect · Scale structure · Degree achievements

1 Introduction Verbs undergoing the causative-unaccusative alternation in Spanish are divided into three classes depending on the morphological mechanism they choose to express the alternation. Firstly, most Spanish alternating verbs undergo the anticausative alternation, i.e. they obligatorily mark the Unaccusative Variant (UV) using the clitic se. Adopting Schäfer’s (2008) terminology, I will refer to this group as “class A”: (1) a. b.

Andrea abrió el balcón. Andrea opened the balcony El balcón se abrió. the balcony SE opened

M. Vivanco (*) Universidad de Castilla La-Mancha, Ciudad Real, Spain e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_12

291

292

M. Vivanco

Secondly, a small group of verbs (between 15 and 20),1 undergoes the labile alternation, where neither variant is morphologically marked, i.e., where the UV is obligatorily unmarked. This group will be henceforth called “class B”: (2)

a. b.

El accidente cambió su vida. the accident changed his life Su vida (*se) cambió. his life SE changed

Finally, the approximately 10 verbs2 found in “class C” can undergo both the anticausative and the labile alternation, i.e., the presence of se is apparently optional. (3)

a. b.

El trueno despertó a Vicky. the thunder woke.up DOM Vicky Vicky (se) despertó. Vicky SE woke.up

Thus, the clitic se is the default pattern used to express the alternation in Spanish and, consequently, it has received much attention in the previous literature. The topic of this paper will be, however, why se is absent—either obligatorily or optionally— with such a small number of verbs (around 30 between classes B and C): the labile alternation is in fact the marked pattern, the one most restrictively used, and so it must be explained by a fine-grained generalization. Crucially the isolation of those factors underlying Ø-marking is expected to eventually shed light on the role of semarking itself. The co-existence of two (or more) mechanisms within one language has been one of the main concerns in the study of the causative-unaccusative alternation, especially with regard to whether there are any syntactic or semantic differences between them (Labelle 1992; Folli 2001; Alexiadou and Anagnostopoulou 2004; Alexiadou, Anagnostopoulou and Schäfer 2006; Schäfer 2008; Labelle and Doron 2010; Letuchyi 2010; Kailuweit 2012; Heidinger 2014, 2015). Assuming that lexical items are inserted post-syntactically (Late Insertion), my hypothesis is that se and Ø are two exponents competing to spell out the same head, v[BECOME] (see Sect. 4). This is a case of context-dependent spell out (Halle and Marantz 1993), so that the spell out conditions of v[BECOME] are sensitive to the properties of AspP, the head immediately below v, but, more specifically, it is scale structure—on DegP—that determines the value of aspect. Therefore, this analysis states that the only semantic and syntactic differences between se and Ø are those derived from the property scale that every Change Of State (COS) verb lexicalizes.

1

There is significant dialectal variation with regard to classes B and C in Spanish. Furthermore, even if class B is usually smaller than class A in Romance, the difference in number found in Spanish is striking. 2 The verbs that I include in this class are the ones that in my own dialect (Peninsular Spanish, Madrid) can optionally take se.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

293

Scales are known to play a major role in the interaction among argument structure, event structure and syntax (Tenny 1994; Krifka 1992; Hay et al. 1999; Ramchand 2008; Rappaport-Hovav 2008), determining the internal development of the event. A key characteristic of scales is that they are composed of an iteration of achievements—degrees—(cf. Hay et al. 1999): thus, in the case of property scales, each achievement leads to a comparative result state (Kearns 2007). If the iteration is unbounded (open multi-point scale), an activity reading arises; however, this is not a truly atelic event since it is not homogeneous but involves consecutive achievements. As we will see, this is one of the contexts where we find Ø in the UV. On the other hand, if the iteration is bounded (closed multi-point scale), an accomplishment reading arises. Crucially, all these verbs belong to class A in Spanish, i.e. their UV is always se-marked. Finally, all deadjectival verbs have an achievement reading available (cf. Kearns 2007), targeting a single COS event. In Sect. 3.2. I will argue that there are simple (Ø-marked) and complex (se-marked)—two-point scale—COS achievements.

2 The Aspectual Hypothesis Revisited It has long been observed that optionally marked UVs seem to choose se or Ø depending on aspect in some Romance languages. Descriptively speaking, Ø-UVs can be either telic or atelic, while se-UVs must be telic (Zribi-Hertz 1987; Labelle 1992; Folli 2001; Labelle and Doron 2010; Legendre and Smolensky 2010). French and Italian examples with class C verbs (4–5) show that, unlike Ø, se/si is incompatible with for X time, and the same holds in Spanish (6): (4)

(5)

(6)

Le ciment a / *s’est durci pendent trois heures. the cement has/SE-is hardened for three hours (Labelle 1992, 398) Il cioccolato è fuso / * si è fuso per pochi secondi. the chocolate has melted/ SE is melted for few seconds (Folli 2001, 128) a. El paciente mejoró en / durante una semana. the patient got.better in / for a week b. El paciente se mejoró en /*durante una semana. the patient SE.got.better in/ for a week

Zribi-Hertz (1987), Labelle (1992), Folli (2001), Labelle and Doron (2010) and Legendre and Smolensky (2010) observe this contrast also in classes A and B. Folli (2001) claims that class B verbs are always atelic, while Basilico (2010) argues that se selects a (bounded) complement that measures the event, the property denoted by the root. Labelle and Doron (2010) and Cuervo (2003, 2014) reinterpret these and other data claiming that there is a difference in event structure: for Labelle & Doron, se focuses on the result while Ø focuses on the process; for Cuervo, se-unaccusatives involve a result subevent, which is missing in the Ø-marked ones.

294

M. Vivanco

The hypothesis of a difference in event structure is related to, but different from, the aspectual hypothesis; however, it is beyond the scope of this study. Following Schäfer (2008), Martin and Schäfer (2014) and Vivanco (2016), I assume that both marked and unmarked unaccusatives have a complex event structure, involving a dynamic and a result subevent.3 The behavior of Romance class C in (4–6) might be due to diachronic reasons since, according to Cennamo et al. (2015), aspect was at the heart of the division between the anticausative and the labile alternation in Latin: class A involved only inherently telic verbs, while variable aspectual behavior verbs belonged to class B. The labile alternation is documented very early in Latin texts (cf. Gianollo 2014) and, together with se, eventually replaced the r- passive as the expression of the alternation. Nevertheless, its use had already decreased significantly in Latin because of the spread of se, and so class B is smaller than class A across Romance.4 However, aspect alone is not enough to account for the data in modern Spanish: although it seems related to se in (4–6), it is irrelevant with regard to Ø. The first problem is that se became the default pattern and spread to variable aspectual behavior verbs already in Latin: such verbs can be found in all three classes across Romance, as well as inherently telic verbs. Authors like Schäfer (2008), Legendre and Smolensky (2010) and Martin and Schäfer (2014) argue that this a serious problem for the aspectual hypothesis, since the contrasts found in class C (4–6) do not hold in classes A and B: thus the question is to what extent se truly rejects atelicity. As we see in (7), Spanish speaker’s judgments vary as to whether class A verbs allow for X time modifiers (see Sect. 3.1.): (7)

?El té se enfrió durante diez minutos. the tea SE cooled for ten minutes

Secondly, Ø-UVs can also be telic (6a, 8); so the question remains: what exactly is Ø related to? Moreover, how can the relation between se and telicity be special if we can get the latter without the former? (8)

Jesús resucitó en tres días. Jesus resurrected in three days

This problem sharpens if we look at the distribution of se and Ø in class C achievements: since these are inherently telic, what difference does the choice make? (9)

Vicky (se) despertó en cinco minutos. Vicky SE woke.up in five minutes

Notice, for instance, that both (6a) and (6b) allow durante (‘for’) in a ‘duration of the result state’ reading (The patient was better for a week). This and other diagnosis are presented in detail by the aforementioned authors, together with explanations for some apparent counterexamples. 4 In fact, Heidinger (2014) treats French class B as a fossil in terms of “diachronic persistence”. 3

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

295

The aspectual hypothesis’ lack of systematicity has led Schäfer (2008) and Alexiadou et al. (2015) to reject any analysis of this kind. They claim that German sich and Romance se/si are exclusively related to the absence of an External Argument (EA) and so they are expletive elements in VoiceP, the locus of diathesis changes (see Sect. 4.2). The aim of this study is, nevertheless, to cope with this apparent lack of systematicity in the aspectual contrasts by providing a more accurate explanation in terms of scale structure, arguing that the relevant distinction is not the telic/atelic one, but the open/closed scale one—with regard to durative events—and also the difference between simple and complex COS achievements. Folli (2001), Cuervo (2003, 2014) and Basilico (2010) place se in v in order to configurationally explain its twofold function: its relationship with the absence of an EA and with aspect/event structure. The analysis presented here builds on these and provides a new perspective on the aspectual and scalar properties of alternating verbs.

3 The Data Section 3.1 is devoted to durative verbs (multi-point scales) showing variable aspectual behavior, such as enfriarse (class A, ‘cool’), aumentar (class B, ‘increase’) and ennegrecer(se) (class C, ‘blacken’). Durative telic verbs, such as secarse (‘dry’), are not problematic because they all belong to class A; nevertheless, the fact that they are the only ones confined to just one class will be relevant in the discussion. Section 3.2 addresses COS achievements, such as encenderse (class A, ‘light’), resucitar (class B, ‘resurrect’) and desperar(se) (class C, ‘wake up’).

3.1

Variable Aspectual Behavior Verbs

First, we will deal with variable aspectual behavior verbs (derived from open scale adjectives) in atelic contexts. We have already observed that there is a strong contrast in class C, since the se variant disallows for X time modifiers (4–6). The following examples add two tests, the imperfective paradox (10) and the construction poner algo a (‘to put something to’) (12). This construction focuses on the beginning of the event and means that the subject creates the necessary conditions for it to happen, without controlling its end (11a); as we see in (11b–c), the construction selects atelic events (Vivanco 2019, 2020)—activities or an interrupted reading with accomplishments—and rejects se in class C (12). (10)

a. b.

El anillo está ennegreciendo ¼> ha ennegrecido the ring is blackening has blackened El anillo se está ennegreciendo ≠> ha ennegrecido the ring SE is blackening has blackened

296

M. Vivanco

(11) a.

b.

c. (12)

Pusimos a los niños a limpiar la cocina/ver la tele y the kitchen/watch the TV and Put.1PL. DOM the kids to clean fuimos al bar. went.1PL. to.the pub Pusimos a los niños a limpiar la cocina Put.1PL. DOM the kids to clean the kitchen durante / *en una hora. for / in an hour Pusimos a los niños a ver la tele durante/*en una hora. Put.1PL. DOM the kids to watch the TV for / in an hour Diana puso la cebolla a caramelizar(*se). Diana put the onion to caramelize.SE

The relevant question is whether these contrasts hold in classes A and B. The latter, as expected, passes atelicty tests: (13) a. b. c.

Luisa engordó durante un mes. Luisa got.fat for a month Luisa está engordando ¼> ha engordado Luisa is getting.fat has got.fat Puse a las gallinas a engordar Put.1SG. DOM the hens to get.fat

With class A verbs, se is disallowed in the poner algo a periphrasis (14). This means that, although se is supposedly obligatory, it can be removed in certain contexts (see also Zribi-Hertz 1987; Labelle 1992; Martin and Schäfer 2014). (14) Diana puso la comida a calentar(*se) / enfriar (*se). Diana put the food to warm.SE / cool.SE There are, however, two apparent counterexamples, since class A verbs derived from open-scale adjectives seem to pass the other atelicity tests: (15) a. b.

?El té se enfrió durante diez minutos. the tea SE cooled for ten minutes El té se está enfriando ¼> se ha enfriado the tea SE is cooling SE has cooled

Why does se not produce the imperfective paradox in (15b)? Remember that deadjectival verbs always entail “some past”, even in the progressive, because they involve an iteration of achievements. If the scale is inherently closed, it entails the last achievement but if it is open, the paradox works thanks to the presence of other achievements in the iteration. Therefore, I conclude that this is not a reliable test when dealing with scalar verbs, and so (15b) must not be considered a counterexample. With regard to (15a), speakers’ judgments vary as to whether the combination of se and for is acceptable, and I myself am among the ones who think it is not. Nevertheless, even if we claimed that (15a) is grammatical, it cannot be considered a

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

297

counterexample either, because telicity can be canceled by a modifier, thus producing an ‘interrupted’ reading with accomplishments (16): (16)

Nerea escribió la tesis durante dos horas y salió a por un café. Nerea wrote the thesis for two hours and went.out to for a coffee

Now we will focus on telicity with variable aspectual behavior verbs in classes B (b examples) and C (a examples) in order to refine the claim that it is compatible with Ø: Ø is allowed in telic contexts, but only under an achievement reading, not under an accomplishment reading. As is well known (cf. Filip 1999; Kearns 2007), modifiers like in X time or the periphrasis take (Spanish ‘tardar’) X time may measure the time lapse between the beginning and the end of an accomplishment, getting the so-called Whole Event Reading (WER), but they may also measure the interval previous to the beginning of the event—Delay Reading (DR). The latter is the only interpretation available with activities, since they lack an endpoint, and achievements, since they lack duration: (17)

a.

Luisa tardó una hora en llegar / bailar. Luisa took an hour in arrive / dance. At the end of an hour, Luisa arrived / danced ! DR/#WER

b.

Sol tardó media hora en regar las plantas Sol took half hour in water the plants 1. At the end of half an hour, Sol watered the plants ! DR 2. Sol watered the plants and the whole plant-watering event had a duration of half an hour ! WER

In (18–19) we see that Ø allows these modifiers; however, the only interpretation available is that the event started after a week (delay reading), not that it lasted a week (whole event reading). Another classical accomplishment diagnosis, terminar de (‘to finish of’) (20)—which may only target endpoints—leads to ungrammaticality. (18)

a.

b.

(19)

a.

b.

(20)

a. b.

El tomate (se) enmoheció en una semana. the tomato SE went.moldy in 1 week DR/#WER El paro aumentó en una semana. the unemployement increased in 1 week DR/#WER El tomate tardó una semana en enmohecer. the tomato took one week in go.moldy DR/#WER El paro tardó una semana en aumentar. the unemployment took one week in increase DR/#WER *El tomate terminó de enmohecer. the tomato finished of go.moldy *El paro terminó de aumentar. the unemployment finished of increase

298

M. Vivanco

Finally, the Scalar Reading (SR) of casi (‘almost’) is not available (22) with Ø5 (see García Pardo [this volume] for similar examples). Almost always has a CounterFactual Reading (CFR)—the event was about to happen, but it did not–, while the scalar reading—the event began, but did not finish—only appears with durative telic events, i.e. with accomplishments (21) (see Dowty 1979; Hay et al. 1999; Rapp and von Stechow 1999).6 (21)

a.

b.

(22)

a.

a’.

b.

b’.

Luisa casi llega / baila. Luisa almost arrives / dances ‘Luisa almost arrives / dances, but she did not’!CFR Luisa casi lee La Metamorfosis. Luisa almost reads the metamorphosis CFR: ‘Luisa almost starts reading The Metamorphosis, but she did not’. SR: ‘Luisa read some pages, but did not finish the book’. El tomate casi enmohece, pero nos lo comimos antes. the tomato almost goes.moldy but us it ate.1PL. before CFR: ‘The tomato was about to go moldy, but we ate it up before this happened’. ?El tomate casi enmohece, pero no lo hizo completamente. the tomato almost goes.moldy but not it did completely #SR: ‘The tomato started going moldy but did not do it completely’. El paro casi aumenta, pero el gobierno the unemployment almost increases but the government intervino a tiempo de evitarlo. intervened in time of prevent.it CFR: ‘Unemployment was about to increase, but the government intervened just in time to prevent this from happening’ ?El paro casi aumenta, pero no lo hizo the unemployment almost increases but not it do completamente. completely #SR: ‘Unemployment started increasing, but did do it completely’.

5 Unlike English, Spanish uses the present tense in the almost-constructions with a past meaning, which does not affect the validity of these examples as telicity tests. The past tense is ungrammatical: *Antonio casi murió (‘Antonio almost died’). 6 The actual truth conditions of each reading are controversial (see Xu 2016 for recent discussion); thus for the purposes of this paper it will suffice to say that the CFR targets the initiation of the event—preventing it from starting– while the SR targets the end of the event—preventing it from finishing.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

299

Therefore, we can interpret the Ø-UVs above as either activities or achievements, but not as accomplishments, which means that Ø allows telicity, but not any kind of telicity. Since all variable aspectual behavior COS verbs involve a multi-point scale—an iteration of achievements–, they all have an achievement reading available. Therefore, the discussion so far suggests that Ø selects achievements and open multi-point scales, while se selects closed multi-point scales. In order to obtain an accomplishment reading with Ø, an explicit measure complement is needed (23), which is unnecessary with se (24): (23)

a.

b. (24)

a.

b.

El tomate casi enmohece hasta parecer musgo. the tomato almost goes.moldy until look.like moss ‘The tomato almost went moldy to the extent of looking like moss’. CFR/SR El paro casi aumenta un 20%. the unemployment almost increases a 20% CFR/SR El tomate casi se enmohece. the tomato almost SE goes.moldy CFR: The tomato was about to get mouldy, but it did not. SR: The tomato got a little bit mouldy, but not to the point of being uneatable. El té casi se enfría. the tea almost SE cools CFR: The tea was about to get cold, but it did not. SR: The tea cooled a bit, but not to the point of being undrinkable.

Finally, another difference between se and Ø is that the latter does not entail the completion of the event with the periphrasis dejar (‘let’) + infinitive. Moreover, class A verbs can leave se out under this periphrasis, although the clitic is supposedly obligatory. In (25) we have a class C verb without se (25a), a class B verb (25b) and a class A verb (25c), also without se: all three sentences mean that the event starts but Antonio does not control whether it ends or not, he just allows its initiation. In (26), on the other hand, we have the corresponding se-versions of the class C (26a) and A (26b) verbs: now both sentences mean that Antonio controls the event until it finishes—i.e. he did not only allow the event to start, but he also allowed it to finish. This shows, again, that se can disappear in atelic contexts, even with class A verbs. (25)

a. b. c.

Antonio dejó enmohecer el tomate. Antonio let go.moldy the tomato Antonio dejó aumentar las quejas en su departamento. Antonio let increase the complaints in his department Antonio dejó enfriar el té. Antonio let cool the tea ! Antonio let all three processes begin

300

(26)

M. Vivanco

a. b.

3.2

Antonio dejó enmohecerse el tomate. Antonio let go.moldy.SE the tomato Antonio dejó enfriarse el té. Antonio let cool.SE the tea ! Antonio let both processes finish

Preliminary Conclusions

The data discussed so far show that class C variable aspectual behavior verbs choose between se and Ø depending on aspect. In this group, where the clitic is optional, the contrast is robust. Ø-UVs—classes B and C—denote either activities or achievements, but cannot denote accomplishments by themselves—in order to do so, they require a complement introducing the telos. In class A se is supposed to be the obligatory marker but it can disappear in atelic contexts (14, 25c). This supports the hypothesis defended here that se does not depend exclusively on the verb but also on the syntactic context. Se-UVs can denote accomplishments by themselves and, what is more, all verbs involving inherently closed multi-point scales belong to class A in Spanish. The apparent compatibility between se and atelicity (15) can be independently explained and does not represent a counterexample for the generalizations established in this section.

3.3

Achievements

As mentioned earlier, the mere opposition telic/atelic is pointless when applied to seand Ø-marked COS achievements; nevertheless, the following contrasts will help us to find a fine-grained explanation to this problem. The first test is the progressive, which is typically odd with achievements (27a) unless it produces a Delay Reading (DR) (27b) or a Slow-Motion Reading (SMR) (26c) (Dini and Bertinetto 1995; Filip 1999): (27)

a. b. c.

#Cristina está marcando un gol. Cristina is scoring a goal El tren está llegando. ! DR The train is arriving ¡Mira, el mago está desapareciendo! !SMR Look the magician is disappearing

If we consider alternating verbs, we observe that only the se-marked ones may produce the SMR. The contrast is easier to notice in class C (28), but holds when comparing classes A (30) and B (29).

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

(28)

a.

b.

301

#Vicky está despertando. Vicky is waking.up ?DR: ‘Vicky is about to wake up’ #SMR: ‘Vicky is slowly opening his eyes and starting to move’ Vicky se está despertando. Vicky SE is waking.up DR: ‘Vicky is about to wake up’ SMR: ‘Vicky is opening his eyes slowly and starting to move’

(29)

#Jesús está resucitando. Jesus is resurrecting ?DR: ‘Jesus is about to resurrect’ #SMR: ‘Jesus is resurrecting slowly, his body is coming back to live little by little’.

(30)

La lámpara se está encendiendo. the lamp SE is lighting.up #DR: ‘The lamp is about to light up’. SMR: ‘The lamp is lighting up slowly, illuminating the room little by little’

Secondly achievements are not expected to produce the scalar reading with casi because they are punctual, and this is what we find with class B verbs (32) and the seless version of class C verbs (31a). However, with the se-version of class C achievements (31b), the scalar reading becomes possible. This contrast is stronger in class C than in class A (33), because the clitic is optional, probably because se is the default pattern. (31)

a.

b.

(32)

(33)

Vicky casi despierta a causa del trueno. Vicky almost wakes.up a cause of.the thunder CFR: ‘Vicky was about to wake up because of the thunder, but he did not.’ #SR: ‘Because of the thunder, Vicky woke up a bit, but not completely, and then continued sleeping’ Vicky casi se despierta a causa del trueno. Vicky almost SE wakes.up to cause of.the thunder CFR: ‘Vicky was about to wake up because of the thunder, but he did not.’ SR: ‘Because of the thunder, Vicky woke up a bit, but not completely, and then continued sleeping.’

Jesús casi resucita. Jesus almost resurrects CFR: ‘Jesus was about to resurrect, but he did not.’ #SR: ‘Jesus resurrected a bit, but not completely.’ La lámpara casi se enciende. the lamp almost SE light.up CFR: ‘The lamp was about to light up, but it did not’.

302

M. Vivanco

?SR: ‘The lamp lit up a bit, but not completely.’ Thirdly we observe similar contrasts with tardar X time (‘take X time’) and terminar (‘finish’) de. If grammatical, Ø-marked UVs of achievements only allow the delay reading, as expected (34, 36). Nevertheless, we notice that se-UVs allow the whole event reading, that is, an accomplishment interpretation. Again, this contrast is clearer in class C (34–35), and holds also in class A (37). Notice that only (35a), but not (34a) may mean that Vicky was half sleepy for a while and when she finally felt completely awake was capable of making herself a cup of tea; although the difference between a delay reading and a whole event reading in (35a) is subtle, the crucial fact is that (34a) disallows both: if we try to lengthen the duration of a COS achievement by using modifiers, we need se. (34)

a.

b.

(35)

a.

b.

(36)

a.

b.

(37)

a.

b.

*Cuando Vicky terminó de despertar, se preparó un té. when Vicky finished of wake.up REFL. prepared a tea #DR/#WER Vicky tardó diez minutos en despertar. Vicky took ten minutes in wake.up DR/#WER Cuando Vicky terminó de despertarse, se preparó un té. when Vicky finished of wake.up.SE REFL. prepared a tea DR/WER Vicky tardó diez minutos en despertarse. Vickytook ten minutes in wake.up.SE DR/WER *Cuando Jesús terminó de resucitar, la gente aplaudió. when Jesus finished of resurrect the people applauded #DR/#WER Jesús tardó tres días en resucitar. Jesus took three days in resurrect DR/#WER Cuando la luz terminó de encenderse, Antonio vio al asesino. when the light finished of turn.on.SE Antonio saw the killer DR/WER La luz tardó diez segundos en encenderse. the light took ten seconds in turn.on.SE DR/WER

What do these contrasts mean? At first sight it seems that se “lengthens” the duration of an achievement, allowing a slow motion reading (28b, 30) or even an accomplishment reading (31b, 35, 37). The initial and end points of an achievement are supposedly simultaneous, which is why these periphrases and the adverb casi can normally target only the beginning of the event; the puzzle is that se seems to make

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

303

the end of an achievement accessible for modification. I will account for these contrasts by using a distinction between simple and complex COS achievements. The former involves a transition to a result state, while the latter describes a transition between the end of a previous state and the beginning of another. The data in this section are very subtle, and are best perceived in class C, where se is optional. In order to reach a better understanding of what is going on, the next section will address other unaccusative contexts with an optional se.

3.3.1

Aspectual se

Some Spanish non-alternating unaccusative verbs can take se (38); furthermore, when combined with certain verbs like ir (‘go’) se creates a distinct unaccusative predicate (‘to leave’, 39c) (see Pineda, [this volume]): (38) Antonio (se) ha muerto. Antonio SE has died (39) a. b. c.

Antonio (se) fue al cine. Antonio SE went to.the cinema Antonio (se) fue de su casa al cine. Antonio SE went from his house to.the cinema Antonio *(se) fue del cine. Antonio SE went of.the cinema ‘Antonio left the cinema’

This clitic has been called “aspectual se”, as opposed to “anticausative se”, because these verbs lack a causative variant and also because it is said to have some aspectual implications whose exact nature is still a matter of debate in Spanish linguistics (see Bull 1950; De Mello 1997; De Miguel and Fernández Lagunilla 2000; Cuervo 2003, 2014; García Fernández 2011; Fernández Jiménez and Tubino Blanco 2014; de Benito 2016). Many of these verbs are achievements, so telicity alone might be, once again, an insufficient explanation for the distribution of se. Therefore we expect that the comparison between alternating (despertar(se), ‘wake up’) and non-alternating (morir (se), ‘die’) unaccusatives with an optional se might shed some light on the problem. In fact, authors like De Miguel and Fernández Lagunilla (2000) and Cuervo (2014) have explored the hypothesis that both instances of se are one and the same. The term aspectual se is also used to refer to the one that may combine with some transitive verbs7 (Nishida 1994; Rigau 1994; Zagona 1996; Sanz and Laka 2002; Basilico 2010; Teomiro 2013; Armstrong 2013; MacDonald 2017). In this case, the

7

Se can appear with other verbs, like estar (‘be’)—Pepe se estuvo callado toda la tarde (‘Pepe remained quiet during the whole afternoon’)–, where Sánchez López (2002) sees a connection with agentivity while De Miguel and Fernández Lagunilla (2000, 29) see a COS event. 8 Although most studies agree about the aspectual nature of se in transitive contexts, some authors have argued that the clitic might be an affected argument (see Rigau 1994; Teomiro 2013; MacDonald 2017).

304

M. Vivanco

aspectual contribution seems clearer, since se appears when the incremental theme bounds the event, creating an accomplishment8: (40)

a. b.

Cora (se) comió tres donuts. Cora SE ate three donuts Cora (*se) comió donuts. Cora SE ate donuts

Building on Sanz (2000) and Folli and Harley (2005), Basilico (2010) argues that se is connected to scalarity in transitive as well as in anticausative constructions and unifies both by placing se on an eventive head selecting a bounded path complement that measures out the event. These hypotheses share important aspects with the one presented here: the clitic could be sensitive to scales in general. However, for reasons of space we must leave aside the study of comer(se)-type verbs, which clearly differ from alternating and non-alternating unaccusatives because of the presence of an EA. Building on ideas already present in authors like De Miguel and Fernández Lagunilla (2000), Cuervo (2003, 2014) and García Fernández (2011), the following discussion will show that the aspectual se found in unaccusative contexts contrasts two states, the one previous to the change and the resultative one, which is exactly what anticausative se does. More specifically, se selects complex COS achievements, while Ø selects simple COS achievements. Coming back to the examples in (39), se is optional in (39a), where ir (‘go’) combines with a goal complement, and in (39b), where there is also an origin complement. However, in (39c) the clitic becomes obligatory if only the origin is mentioned (Masullo 1992; Sánchez López 2002; see Pineda [this volume] for an account of the appearance of the locative clitic ne in Catalan and other Romance varieties). García Fernández (2011) points out that, although se seems related to the origin with ir (‘go’), such an explanation does not account for its role with a verb like salir (‘go out’) which implies an origin by itself and can also take se in certain contexts (41), but not in others (42): (41)

Cora (se) salió de la habitación / a fumar al balcón. Cora SE went.out from the room / to smoke to.the balcony (García Fernández 2011, 63)

(42)

a. b.

A las 12 en punto, la Macarena salió de la basílica. at the 12 o’clock the Macarena went.out from the basilica ?A las 12 en punto, la Macarena se salió de la basílica. at the 12 o’clock the Macarena SE went.out from the basilica

During a procession, the interpretation of (42a) is that the sculpture representing the Macarena Virgin leaves the church, carried by some people. García Fernández 8

Although most studies agree about the aspectual nature of se in transitive contexts, some authors have argued that the clitic might be an affected argument (see Rigau 1994; Teomiro 2013; MacDonald 2017).

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

305

(2011) thinks that (42b) is odd because of agentivity reasons, since it implies that the Macarena is going out on her feet. However, other contrasts show that agentivity is not the key factor. The following pair presents the opposite situation: (43a) is odd unless the corpse is leaving the coffin on its feet (unlike the Macarena), while (43b) is interpreted as a fall. (43)

a. b.

?Debido al accidente, el cadáver salió del ataúd. due to.the accident the corpse went. out from.the coffin Debido al accidente, el cadáver se salió del due to.the accident the corpse SE went.out from.the ataúd. coffin (García Fernández 2011, 63)

A similar puzzle is presented in (44–45). In (44a–b) we see the Ø-variants of two sentences, and only one of them (b) is acceptable. In (45a–b) we see the corresponding se-variants, but only (a) is acceptable. These pairs work in opposite ways: depending on what goes out, a breast (se) or a pimple (Ø), se is necessary (a) or not (b). (44)

a.

b.

(45)

a.

b.

?A la bailarina le salió un pecho. to the dancer CL. went.out a breast ‘The dancer grew a breast’ (a third one??) A la bailarina le salió un grano. to the dancer CL. went.out a pimple ‘The dancer got a pimple’ (García Fernández 2011, 62; cf. also Cuervo 2003) A la bailarina se le salió un pecho. to the dancer SE CL. went.out a breast ‘The dancer’s breast slipped out’ (of the maillot) ?A la bailarina se le salió un grano. to the dancer SE CL. went.out a pimple ‘The dancer’s pimple slipped / came out’ (out of where??)

García Fernández (2011) claims that se is disallowed if it is impossible to refer to the origin, like in (45b), while Cuervo (2003, 2014) argues that se selects a result subevent, which is missing in the unmarked variant. In her analysis, non-alternating unaccusatives with se have the same bi-eventive structure as anticausatives, while se-less unaccusatives are simple achievements that involve only a dynamic event. Although I think there is always a result subevent,9 I do agree that the structure with se is more complex somehow, though not in terms of event structure, but in terms of scale structure.

9

(Se) salió del cine durante cinco minutos (‘He left the cinema for 5 min’) expresses the duration of the result state with and without se.

306

M. Vivanco

De Miguel and Fernández Lagunilla (2000) establish a difference between (1) simple achievements, which are truly punctual, like llegar (‘arrive’); (2) events whose culmination takes place at the beginning and are followed by a process, like hervir (‘boil’); (3) events whose culmination takes place at the beginning and are followed by a state, like marearse (‘get dizzy’); and (4) complex achievements that express the transition between two states, like morirse (‘die’) and irse (‘leave’). In their analysis, only the last two groups allow the presence of se because the clitic is an aspectual operator that focuses on a phase of the event and requires a culmination followed by a COS. In a similar vein, my claim is that the Ø-constructions denote a simple transition to a result state, while the ones with se denote a complex transition involving the end of the previous state (the “origin”) and the beginning of the new one. Se is preferable or even obligatory if we need to contrast the result state and the previous one, especially if the change is unexpected. On the contrary, Ø is used if we cannot or do not need to identify the previous state. Therefore in (45a) the breast is not expected to leave the maillot, and so se is needed, while in (45b) there is no clear origin for the pimple, which simply comes into existence, and so se is odd. Likewise, the corpse is not expected to leave its coffin in (43), provided that it is truly dead; thus we use se if it comes out accidentally, like the breast, and Ø if it is a zombie waking up. Notice that when talking about a zombie se is not ungrammatical, but just optional, like (41), depending on the emphasis we put on the fact that he is leaving a place to go to another. This is why (42b) is odd: if there is nothing special about leaving a place, se is optional, a matter of emphasis (41, 43), and so in (42b) the default reading would be that the Macarena is going out the church just like Cora is going out to smoke. Since the virgin is, as expected and not accidentally, carried out, using se is the odd option. To finish, notice that non-alternating unaccusatives behave like the achievements in the previous section with regard to the almost test (compare 46–31): the se-less version only allows the counter-factual reading, while the se-version legitimates the scalar reading, making the accomplishment interpretation available: (46)

a.

b.

Vicky y Hugo casi salen de paseo. Vicky and Hugo almost go.out for walk CFR / #SR Vicky and Hugo almost left for a walk, but they did not Casi se salen dos clavos. almost se go.out two nails ‘Two nails almost come off’ CFR/SR (Cuervo 2003, 138)

In conclusion, the so-called aspectual se contrasts two states, the one previous to the change and the result state, and so does anticausative se. I formalize this as a complex achievement (47b) with a two-point scale—each state is a point–, which allows an accomplishment interpretation, thus creating the impression that the achievement is “lengthened”. On the other hand, Ø is chosen with simple

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

307

achievements (47a). These involve a result state, but no previous state, which means that they lack a scale and denote only a transition (dynamic subevent) to a result.10 (47)

3.4

a. b

Achievement ├ state ! Ø state ┤Achievement ├ state ! se

Preliminary Conclusions Revised

Our discussion has shown that it is scale structure which determines the choice between se and Ø in Spanish. The claim that Ø-UVs can be either telic or atelic has proven to be an insufficient description of the facts. The data demonstrate that Ø selects, more specifically, open scales and simple COS achievements, and that it is unable to denote accomplishments by itself. What open scales and simple COS achievements have in common is that there is something missing: the former lack an inherent end point, so they involve a number of comparative result states but not a final result state; the latter, on the other hand, lack a scale and, consequently, any reference to the state previous to the change. In class C, se is optional and is strongly related to telicity (closed scales) and complex COS achievements. In class A, on the other hand, se is obligatory, which weakens some of the contrasts, but, crucially, it can still be left out if the right scale boundary is missing. This is how the study of the conditions under which se is absent has helped us to understand the role it plays when present. The aspectual hypothesis fails to account for COS achievements, which are found with and without se and need a more fine-grained explanation. Therefore any analysis of Spanish se/Ø must take into account the fact that they are related both to the absence of an EA and to scale structure. The next section presents a proposal where both units compete to spell out the same head, v, above AspP.

4 Analysis 4.1

AspP and DegP

In this section MacDonald’s (2008) theory of aspect will be adapted to formalize how scale structure (encoded on AP and DegP) reflects on AspP. According to MacDonald, “initiation” ({ie}) and “final” ({fe}) are the grammaticalized phases of events—interpretable features, not subevents. AspP

10

See Caudal and Nicolas (2005) for discussion about the existence and relevance of two-point scales. By distinguishing between simple and complex COS achievements, I am distinguishing between non-scalar and scalar COS achievements (see Sect. 4).

308

M. Vivanco

Table 1 Diagram key ├ ┤ k ╢

A S S+ ST

Initiation point (beginning of the event/state) Final point (end of the event/state). COS verbs involve an iteration of achievements, and so several initiation and final points. Telos (term used in the broad, traditional sense; the telos may be introduced by a complement and does not need to be followed by a result state) Terminative end: This symbol, as opposed to ┤, is used to signal a point after which no further iteration of achievements will follow (typical of closed scales) and, unlike k, this end point is necessarily followed by a result state. Achievement State Comparative state: each achievement in the iteration leads to a comparative state. Terminative end state: this is the last state in the property scale, nothing will follow after it.

always introduces an {ie} feature, while {fe} may be introduced in different positions, producing different davidsonian events: a) If {fe} is absent, we get an activity; b) if {fe} is introduced by another syntactic element c-commanded by AspP—namely an incremental theme—a time interval interpretation arises, and so we get an accomplishment; c) finally, if {fe} is introduced also on AspP, its temporal reference coincides with that of the {ie} feature—punctual event–, so we get an achievement. In this latter case, it is {ie} that projects (48), the visible one, which is why achievements get delay readings with temporal modifiers. (48) presents MacDonald’s syntactic structure for achivements and the adaptation that will be used in this chapeter: (48)

[AspP{ie} [Asp [{i}] [{f}]]] ! (adaptation) [AspP{i} [Asp{i}, {f}]]

In the case of COS verbs, the features on AspP depend on the scalar properties of the embedded PredP, which contains a DegP on top of the adjectival root. Most studies on deadjectival verbs11 agree that they encode a scale, which I formalize as a DegP following standard analysis of gradable adjectives (cf. Kennedy 1999). The Deg head hosts features such as {durative}—multi-point scales—and {f}—open/closed scales.12 Moreover, I label two-point scales—complex COS achievements—as Degx2 and assume that simple COS achievements, on the other hand, lack a DegP.13 The diagrams in the following paragraphs describe scalar COS events and will be formalized later using this syntactic model.

I use the term ‘deadjectival’ in a broad sense: not all COS verbs derive directly from adjectives, but they all denote a change along a scale. 12 These features cannot be simply placed on the root because this would lead to overly strong predictions about the aspectual properties of deadjectival nouns, for instance, and would prevent deadjectival verbs from showing variable aspectual behavior. 13 Again, if this is the difference between despertar (‘wake up’) and despertarse, this information cannot belong to the root. 11

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

309

1. Simple COS achievement (Ø-marked):14 a simple achievement like despertar (‘wake up’) only involves a transition—A(chievement)—that means the beginning (├ ) of a terminative result state—ST in (49), AP in (50)–which will not be followed by any more transitions. There is no DegP in (50) because there is no scale. Despertar (‘wake up’). (49) (50)

A{i}├ ST (DESPIERTO/‘AWAKE’) [vP [v] [AsP{i} [Asp{i}, {f}] [PredP DP [Pred] [AP [A] [√]]]]]

2. Unbounded iteration of achievements (Ø-marked): a verb like engordar (‘gain weight’)15 in (51) encodes the end (┤) of a state that is followed by the beginning (├) of another, but there is no telos (neither k nor ╢) nor ST. This can be formalized by switching the features: it is {fe} that projects on AspP, pointing to the end of the previous state. The iteration of achievements comes from the scale, formalized as a DegP{+durative} (52). If a complement like three kilos is added (53), it introduces a boundary (k), but there is no subsequent result state. This introduces an extra telos c-commanded by AspP and so a time lapse interpretation between {ie} and {fe} that produces an accomplishment reading. Engordar (‘gain weight’). (51) (52)

A{i}├S +FAT ┤{f}A{i}├S +FAT ┤{f}A{i}├ S+FAT [vP [v] [AspP{f} [Asp{i}, {f}] [PredP DP [Pred] [DegP [Deg{+durative}] [AP [A] [√]]]]]]

Engordar 3 kilos (‘gain 3 kilos’). (53) (54)

A{i}├ S+1KG ┤{f}A{i}├ S+1KG ┤{f}A{i}├ S+1KG ┤{f}k [vP [v] [AspP{f} [Asp{i},{f}] [PredP DP [Pred] [DegP DP 3 kilos [Deg{+durative}] [AP [A][√]]]]]]

3. Bounded iteration of achievements (se-marked): With a verb like secarse (‘to dry’), the iteration will not continue after a certain point (┤╢), which is followed by a ST—the positive degree of the adjective (seco (‘dry’)). Now, given that accomplishment readings come from the presence of a boundary in a head c-commanded by Asp—in order to obtain the time lapse interpretation–in this case that head would be Deg. More specifically, this boundary cannot be only {f} (┤), which is already present in every achievement within the iteration, but {F} (╢), which ends the iteration as a whole. Thus in this context the feature {F} on

14

Achievements which do not express a COS event do not have a result state (score two points). Notice that in Spanish, unlike English, this verb derives from the open scale adjective gordo (‘fat’).

15

310

M. Vivanco

AspP means that an iteration has taken place but is over, and it comes from Deg, with a {F} feature (closed scale). Secarse (‘dry’). (55) (56)

A{i}├ S+DRY ┤{f}A{i}├ S+DRY ┤{f}A{i}├ S+DRY {f} ╢A{i}├ ST DRY [vP [v] [AspP{F} [Asp{i},{f}] [PredP DP [Pred] [DegP [Deg{+dur.},{f}] [AP [A] [√]]]]]]

4. Complex COS achievement (se-marked): According to our previous discussion, se-marked achievements like despertarse (‘wake up’) are complex, so a state (S) ends (╢) with the change (A), and another one (ST) begins (├). A two-point scale (Degx2) does not involve an iteration strictu sensu, but refers to a previous state that ends and means that no iteration will follow (╢, {F}). Therefore, {F} stops the iteration from taking place in this context, while in the previous context (closed scales) it stops the iteration from going on. Despertarse (‘wake up’). (57) SASLEEP ╢{F}A{i}├ ST (AWAKE) (58) [vP [v] [AspP{F} [Asp{i}, {f}] [PredP DP [Pred] [DegP [Deg{x2}] [AP [A] [√]]]]]] These diagrams and syntactic structures explain the puzzles found in the previous sections. Firstly, the construction poner algo a (‘put something to’) selects atelic infinitives, and forces the terminative telos and the terminate state to be removed: (59)

a. b.

Poner la ropa a secar(*se). put the clothes to dry.SE [Poner a [A ├ S+SECO ┤ A ├ S+SECO ┤ A ├ S+SECO ╢A ├ ST (SECO)]]

Secondly, temporal modifiers can only access the beginning of the event if {i} is the feature that projects on AspP (60), producing a delay. However, if {F} is the projected one, these modifiers can access the end of the previous state, producing this apparent “lengthening” of the event (61). (60)

a. b.

(61)

a. b.

Vicky tardó diez minutos en despertar. Vicky took 10 minutes in wake.up Ten minutes ! A├ awake Vicky tardó diez minutos en despertarse. Vicky took ten minutes in wake.up.SE Ten minutes ! ╢A├ awake

In conclusion, se selects {F} (╢) on AspP or, more specifically, requires a configuration of the type [S ╢{F}A{i}├ S]. In turn, Ø selects either {i} or {f} on AspP. These aspectual properties depend on the scale structure introduced by DegP. Section 4.3. presents an analysis where these scalar properties are the spell out

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

311

conditions of v. Before presenting this analysis, in Sect. 4.2, I will briefly reflect on the main theoretical and empirical arguments in favor of treating se/Ø as the phonetic expressions of v rather than Voice.

4.2

Se/Ø as Unaccusativity Markers: Voice or v?

Once the scalar/aspectual behavior of alternating COS verbs has been established, this section addresses the syntactic position of Ø and se as unaccusativity markers. First, I assume that both marked and unmarked UVs are bi-eventive structures, since they denote COS events: a dynamic subevent—the change—embeds a result state.16 As said before, it is beyond the aims of this chapter to refute the hypothesis that the difference in marking reflects a difference in event structure. Second, I assume a difference between a verbalizing v head, which introduces the event, and a Voice head, which introduces the EA (Kratzer 1996; Cuervo 2003; Pylkkänen 2008; Harley 2013; Legate 2014). Thus a problem arises with regard to the syntactic locus of unaccusativity: are se and Ø to be placed on v or on Voice? According to Folli (2001), Cuervo (2003), Folli and Harley (2005) and Pylkkänen (2008) the features ‘cause’/‘become’ on v are responsible for the eventual insertion of Voice. As mentioned earlier, the analysis presented here follows those developed by these authors17 and provides a configurational account for the double function of se—its relation to aspect and to unaccusativity—by placing it on little v. The main problem for the v-analysis is the nature of v itself. According to Key (2013), purely verbalizing morphology can co-appear with cause/become morphology in Turkish, something unexpected if both are placed on the same head. Similar examples are easily observed in other languages, like Spanish, where se co-appears with a light verb in the UV of alternating complex predicates (62). (See García Pardo [this volume] for an analysis of the co-occurrence of se with light verbs.) (62)

Cristina se puso nerviosa. Cristina SEbecome putlight verb nervousadjective ‘Cristina got nervous’

This means that v is supposed to be carrying too much information, which Key (2013) solves with a fission operation on little v. However, fission is meant to account for those cases where one feature has two or more exponents (Noyer, 1992) while in (62) there are two features—‘verbalizing’ and ‘become’–two exponents and, maybe, two heads.18 For reasons of space I cannot deal with this problem

16

There is no need to use a specific head with causative semantics, the embedding relation between both subevents is enough (Hale and Keyser 1993; Cuervo 2003; Schäfer 2008). 17 See also Basilico (2010) for a different implementation of the little v analysis of se, where this head introduces the EA. 18 See Chung (2007), Caha (2009) and Pantcheva (2011), among others, for discussion about the theoretical adequacy of post-syntactic operations in Distributed Morphology.

312

M. Vivanco

here and I refer the reader to Vivanco (2016) for further discussion and an alternative proposal where the cause/become features are hosted by a different head between v and AspP. In this chapter, nevertheless, I will adhere to a more traditional version of v, leaving the problem of (62) open for future research. On the other hand Alexiadou et al. (2015) propose a complex system of features on Voice that produces different diathesis, like the passive voice or the causativeunaccusative alternation. They claim that Ø-UVs have a true unaccusative syntax with no Voice projection, while se-UVs have a transitive syntax, so that Romance se/ si and German sich are expletive elements in Voice. Nevertheless many of the empirical arguments provided are from German rather than Romance. The transitive syntax of German sich-UVs is mainly supported by the positional freedom of sich and the selection of the auxiliary haben (‘have’) instead of sein (‘be’) but, as is well known, Romance se/si is a clitic, and the constructions where it appears select être/ essere (‘be’) in French and Italian. An additional argument provided by Schäfer (2008) is that German sich-UVs, unlike Ø-UVs, block the ‘unintentional causer reading’ of a dative19 due to the fact that sich-UVs have a transitive syntax that prevents the argument on the high applicative— above Voice–from being interpreted as the causer of the event. According to this author, Ø-UVs allow this reading because they lack a Voice projection. However, this does not hold in Romance, as we see in (63), where the situation is the opposite: seUVs allow unintentional dative causers while Ø-UVs reject them. Thus, in (63a) the dative Cristina can be interpreted both as an affected argument—the phone broke on her—or as the unintentional causer—she accidentally broke the phone (see Fernández Soriano 1999 and García Pardo [this volume] and Basilico [this volume] for further examples)–while in (63b) the dative can only get an affected reading. (63)

a. b.

A Cristina se le ha roto el teléfono. to.DAT Cristina SE her.DAT has broken the phone A Cristina le aumentaron los problemas. to.DAT Cristina her.DAT increased the problems

Finally, this kind of analysis of Voice is not meant to account for the relationship between se and aspect/scale structure; moreover, it presents two different structures for marked and unmarked UVs, with and without Voice, something contradictory with the fact that both show the same behavior with regard to those diagnoses that target the presence of this projection (see Vivanco 2016 for discussion). Therefore, I conclude that se- and Ø-UVs must receive a unified analysis as unaccusative constructions—without Voice—that accounts for the role played by scale structure. In the next section I will argue that the distribution of se and Ø can be better understood in terms of context-dependent spell-out.

19

See Cuervo (2003) for an in-depth study of datives, focused on Spanish.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

4.3

313

Spell Out Conditions on v

This chapter has dealt with the co-existence of the anticausative and the labile alternation in Spanish; however it is well known that there are three other mechanisms in the world’s languages (cf. Haspelmath 1993): the causative alternation, where the causative variant is the marked one (duys/a-duy-ebs; Georgian, ‘cook’); the equipollent alternation, where each variant in the alternation gets a different morphological marker (atum-aru/atum-eru; Japanese, ‘gather’), and the suppletive alternation, where each variant is expressed using a different root (to die/to kill). Since the causative-unaccusative alternation is thought to be a universal phenomenon, I believe that the crosslinguistic differences in marking, as well as the co-existence of more than one mechanism within one language, are to be understood as spell out issues. From a late insertion perspective, languages may differ with respect to the (number of) lexical items they have to express the alternation, i.e. to spell out the head which acts as the locus of the alternation, and they are also expected to differ with respect to the specific spell out conditions driving the choice between those lexical items. In the case of Spanish, I have treated se and Ø as one and the same thing: they have the same function—marking UVs in the alternation–so I assume they spell out the same head, v, in the same kind of structure—an unaccusative structure which lacks Voice and involves a result subevent. The only difference between them has to do with the property scale that measures out the COS event: Ø-ACs simple COS achievement (despertar, ‘wake up’) (64) [vP [vBECOME Ø] [AspP{i} [Asp{i}, {f}] [PredP DP [Pred] [AP [A] [√]]]]] Ø-ACs open scale (engordar, ‘gain weight’) (65) [vP[vBECOME Ø][AspP{f}[Asp{i}, {f}][PredP P[Pred][DegP[Deg{+dur.}][AP[A [√]]]]]] Se-ACs complex COS achievement (despertarse, ‘wake up’) (66) [vP[vBECOME se][AspP{F}[Asp{i}, {f}][PredP P[Pred][DegP[Deg{x2}][AP[A][√]]]]]] Se-ACs closed multi-point scale (secarse, ‘dry’) (67) [vP[vBECOME se][AspP{F}[Asp{i},{f}][PredP[Pred][DegP[Deg{+dur.},{+bounded}] [AP[A[√]]]]]] Distributed Morphology argues for a one-to-one correspondence between syntactic nodes and phonological exponents, whose competition is determined by the Subset Principle (Halle 1997)—the chosen item is the one containing all or a subset

314

M. Vivanco

of the features present in the derivation—and eventually conditioned by postsyntactic readjustment operations such as fusion, fission and impoverishment. On the other hand, Nanosyntax avoids the use of such operations by claiming that one lexical item can spell out not only a terminal node, but a bigger chunk of syntactic structure—“phrasal spell out” (Caha 2009; Starke 2011). This theory claims that a lexical item can contain more features than the structure it spells out (Superset Principle), and so the “biggest” exponent will be the winner in a competition. However, the choice between se and Ø is not a matter of size, since they only spell out one head,20 but a matter of grammatical context: the information on DegP determines the feature projection on AspP which determines, in turn, the choice of se or Ø on v. Therefore this is an instance of context-dependent spell out (Halle and Marantz 1993) where the v is spelled out differently depending on the grammatical properties of the complement it selects, as stated in (68): (68)

Spell out conditions of vBECOME in Spanish: a. vBECOME will be spelled out as se if it selects a complement AspP that projects a {F} feature. b. vBECOME will be spelled out as Ø if it selects a complement AspP that lacks a {F} feature.

This analysis is consistent with the hypothesis that spell out is cyclic and applies to phases (cf. Uriagereka 1999; Epstein 1999; Chomsky 2000, 2001), since little v constitutes a phase boundary and the relevant information for spell out is contained within its local domain.

5 Conclusions Any analysis of Spanish alternating verbs must capture the fact that anticausative morphology is simultaneously related to aspect/scale structure and to the lack of an EA. This chapter has shown that the problem of the distribution of se/Ø goes beyond telicity and is directly related to scalarity. The lack of se is due to the lack of a very specific event configuration where either a previous state or a terminative one are missing. Therefore the change of state meaning has been defined in syntactic terms,

20

One might think that there could be a difference in size if se was the phrasal spell out of [v [AspF]] as opposed to Ø, which would spell out only [v]. However, there seems to be other idiosyncratic factors involved in the choice between se and Ø that depend on how speakers conceptualize COS event as more or less likely to happen spontaneously (see Haspelmath 1993; Schäfer 2008; Letuchyi 2010; Haspelmath et al. 2014; Heidinger 2015; Vivanco 2017). Because of this and also because aspect is actually determined by DegP—i.e. it is not the lowest head involved in this problem–I opt for an analysis where Spanish se/and Ø do not differ in size, but in their grammatical and idiosyncratic conditions on spell out. See Kempchinsky (2004) for an analysis of se directly in AspP.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

315

making explicit reference to the scalar nature of the change, thus providing a deeper understanding of the aspectual properties of these verbs. From a late insertion perspective, the cross-linguistic morphological variation found in alternating verbs can be understood as a matter of spell out. Languages differ with regard to the specific vocabulary items available to spell out the head responsible for the diathesis change as well as with regard to the specific conditions on spell out. Acknowledgments Earlier versions of this study were presented at the Going Romance 2014, in Lisbon; at the Workshop on Romance SE/SI constructions, held in Madison (Wisconsin) in April 2016, and at the Linguistics Seminar of the University of Toronto, in july 2016. I thank the participants for their feedback, as well as the anonymous reviewers of this book. I would also like to acknowledge the outstanding work the editors of this volume have done. Finally, my special thanks are due to Cristina Sánchez López, Antonio Fábregas, Ignacio Bosque, Florian Schäfer, Jaume Mateu, Elena de Miguel, Víctor Acedo-Matellán, María Jesús Fernández Leborans and Luis García Fernández for their helpful and valuable comments. All errors remain my own.

References Alexiadou, Artemis, and Elena Anagnostopoulou. 2004. Voice morphology in the causativeinchoative alternation: Evidence for a non-unified structural analysis of unaccusatives. In The unaccusativity puzzle: Explorations of the syntax-lexicon interface, ed. Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, 114–136. Oxford: Oxofrd University Press. Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2006. The properties of anticausatives cross-linguistically. In Phases of interpretation, ed. M. Frascarelli, 187–211. Berlin: Mouton. ———. 2015. External arguments in transitivity alternations. A layering approach. Oxford: Oxford University Press. Armstrong, Grant. 2013. Agentive reflexive clitics and transitive se constructions in Spanish. Borealis. An International Journal of Hispanic Linguistics 2 (2): 81–128. https://doi.org/10. 7557/1.2.2.2526. Basilico, David. 2010. The se clitic and its relation to paths. Probus 22: 271–302. https://doi.org/10. 1515/prbs.2010.010. ———. n.d.. Spanish se as a high and low verbalizer. Bull, William E. 1950. The intransitive reflexive ir and irse. Modern Language Journal 26: 382–389. Caha, Pavel. 2009. The nanosyntax of case. PhD diss., Tromsø University. Caudal, Patrick, and David Nicolas. 2005. Types of degrees and types of event structures. In Event arguments foundations and applications, ed. Claudia Maienborn and Angelika Wöllstein, 277–300. Tübingen: Niemeyer. Cennamo, Michela, Thórhallur Eythórsson, and Jóhanna Barðdal. 2015. Semantic and (morpho)syntactic constraints on anticausativization: Evidence from Latin and Old Norse-Icelandic. Linguistics 53 (4): 677–730. https://doi.org/10.1515/ling-2015-0015. Chomsky, Noam. 2000. Minimalist inquiries. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. Roger Martin, David Michaels, and Juan Uriagereka, 89–155. Cambridge: MIT Press. ———. 2001. Derivation by phase. In Ken Hale: A life in language, ed. Michael Kenstowicz, 1–52. Cambridge: MIT Press.

316

M. Vivanco

Chung, Inkie. 2007. Suppletive negation in Korean and Distribuited Morphology. Lingua 117 (1): 95–148. https://doi.org/10.1016/j.lingua.2005.10.002. Cuervo, María Cristina. 2003. Datives at large. PhD diss., MIT. ———. 2014. Alternating unaccusatives and the distribution of roots. Lingua 141: 48–70. https:// doi.org/10.1016/j.lingua.2013.12.001. de Benito, Carlota. 2016. Las construcciones con 'se' desde una perspectiva variacionista y dialectal. PhD Thesis, Universidad Autónoma de Madrid. De Mello, George. 1997. Morir vs. Morirse. In Lengua escrita y habla culta en América y España: Diez casos, ed. Josse de Kock and George De Mello, 99–117. Salamanca: Universidad de Salamanca. De Miguel, Elena, and Maria Fernández Lagunilla. 2000. El operador aspectual se. Revista Española de Lingüística 30 (1): 13–43. Dini, Luca, and Pier Marco Bertinetto. 1995. Punctual verbs and the ontology of events. In Quaderni del Laboratorio di Linguistica della SNS. Pisa: Scuola Normale Superiore. Dowty, David. 1979. Word, meaning and Montague grammar. In The semantics of verbs and time in generative semantics and Montague’s PTQ. Dordrecht: Reidel. Epstein, Samuel D. 1999. Un-principled syntax: The derivation of syntactic relations. In Working minimalism, ed. Samuel Epstein and Norbert Hornstein, 317–345. Cambridge: MIT Press. Fernández Jiménez, Ángel, and Mercedes Tubino Blanco. 2014. Variación sintáctica en la causativización léxica. Revista Española de Lingüística 44 (1): 7–37. Fernández Soriano, Olga. 1999. Datives in constructions with unaccusative se. Catalan Working Papers in Linguistics 7: 89–106. Filip, Hana. 1999. Aspect, eventuality types and nominal reference. New York: Garland. Folli, Raffaella. 2001. Constructing telicity in English and Italian. PhD diss., Oxford University. Folli, Raffaella, and Heidi Harley. 2005. Flavors of v: Consuming results in Italian and English. In Aspectual inquiries, ed. Roumanya Slabakova and Paula Kempchinsky, 95–120. Dordrecht: Kluwer. García Fernández, Luis. 2011. Algunas observaciones sobre se aspectual. In Estudios sobre perífrasis y aspecto, ed. Juan Cuartero Otal, Luis García Fernández, and Carsten Sinner, 43–71. München: Peniope. García Pardo Alfredo. n.d. Light verbs and the syntactic configurations of se. Gianollo, Chiara. 2014. Labile verbs in late Latin. Linguistics 52 (4): 945–1002. https://doi.org/10. 1515/ling-2014-0013. Hale, Ken, and Samuel J. Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In The view from building 20, ed. Ken Hale and Samuel J. Keyser, 53–109. Cambridge: MIT Press. Halle, Morris. 1997. Distributed morphology: Impoverishment and fission. In MIT working papers on linguistics (30): Papers at the interface, ed. Benjamin Bruening, Yoonjung Kang, and Martha McGinnis, 425–449. Cambridge: MIT Press. Halle, Morris, and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In The view from building 20, ed. Ken Hale and Samuel J. Keyser, 111–176. Cambridge: MIT Press. Harley, Heidi. 2013. External arguments and the mirror principle: On the distinctness between Voice and v. Lingua 125: 34–57. Haspelmath, Martin. 1993. More on the typology of inchoative/causative alternations. In Causatives and transitivity, ed. B. Comrie and M. Polinsky, 87–120. Amsterdam: John Benjamins. Haspelmath, Martin, Andreea Calude, Michael Spagnol, Heiko Narrog, and Elif Bamyaci. 2014. Coding causal–noncausal verb alternations: A form–frequency correspondence explanation. Journal of Linguistics 50 (3): 587–625. https://doi.org/10.1017/S0022226714000255. Hay, Jennifer, Christopher Kennedy, and Beth Levin. 1999. Scalar structure underlies telicity in ‘degree achievements’. In Proceedings of SALT 9, ed. T. Mathews and D. Strolovitch, 127–144. Ithaca: Cornell Linguistics Circle Publications. Heidinger, Steffen. 2014. The persistence of labile verbs in the French causative-anticausative alternation. Linguistics 52 (4): 1003–1024. https://doi.org/10.1515/ling-2014-0014.

Scalar Constraints on Anticausative SE: The Aspectual Hypothesis Revisited

317

———. 2015. Causalness and the encoding of the causative-anticausative alternation in French and Spanish. Journal of Linguistics. 51: 562–594. https://doi.org/10.1017/S0022226714000607. Kailuweit, Rolf. 2012. Construcciones anticausativas: El español comparado con el francés. In Aspectualidad—transitividad—referencialidad. Las lenguas románicas en contraste, ed. Valeriano Bellosta von Colbe and Marco García, 133–158. Frankfurt A. M.: Peter Lang. Kearns, Kate. 2007. Telic senses of deadjectival verbs. Lingua 117 (1): 26–66. https://doi.org/10. 1016/j.lingua.2005.09.002. Kempchinsky, Paula. 2004. Romance se as an aspectual element. In Contemporary approaches to romance linguistics, ed. J. Auger et al., 239–256. Amsterdam: John Benjamins. Kennedy, Christopher. 1999. Projecting the adjective: The syntax and semantics of gradability and comparison. New York: Garland. Key, Greg. 2013. The morphosyntax of the Turkish causative construction. PhD diss., University of Arizona. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase structure and the lexicon, ed. Johan Rooryck and Laurie Zaring, 109–137. Dordrecht: Kluwer. Krifka, Manfred. 1992. Thematic relations and links between nominal reference and temporal constitution. In Lexical matters, ed. Ivan A. Sag and Anna Szabolcsi, 29–53. Stanford, CA: CSLI Publications. Labelle, Marie. 1992. Change of state and valency. Journal of Linguistics 28: 375–414. Labelle, Marie, and Edit Doron. 2010. Anticausative derivations (and other valency alternations) in French. Probus 22 (2): 303–316. Legate, Julie Anne. 2014. Voice and v. lessons from acehnese. Cambridge, MA: MIT Press. Legendre, Geraldine, and Paul Smolensky. 2010. French inchoatives and the unaccusativity hypothesis. In Hypothesis A/Hypothesis B: Linguistic explorations in honor of David M. Perlmutter, ed. Donna Gerdts, John Moore, and Maria Polinsky, 229–246. Cambridge, MA: MIT Press. Letuchyi, Alexander. 2010. Lability and spontaneity. In Transitivity. Form, meaning, acquisition and processing, ed. Patrick Brand and Marco García García, 237–256. Amsterdam: John Benjamins. López, Luis. 2012. Indefinite objects. Scrambling, choice functions and differential marking. In Linguistic Inquiry Monograph 63. Cambridge: MIT Press. MacDonald, Jonathan E. 2008. The syntactic nature of inner aspect: A minimalist perspective. Amsterdam: John Benjamins. ———. 2017. Spanish aspectual se as an indirect object reflexive: The import of atelicity, bare nouns and leísta PCC repairs. Probus 29 (1): 73–118. https://doi.org/10.1515/probus-20150009. Martin, Fabienne, and Florian Schäfer. 2014. Anticausatives compete but do not differ in meaning: A French case study. In Proceedings of the 4e Congrès Mondial de Linguistique Française, ed. F. Neveu, L. Hriba, J. Meinschaefer, P. Blumenthal, A. Gerstenberg, and S. Prevost, 2485–2500. New York: Curran Associates, Inc. Masullo, Pascual José. 1992. Antipassive constructions in Spanish. In Romance languages and modern linguistic theory, ed. P. Hirschbuhler and K. Koerner, 175–194. Amsterdam: John Benjamins. Nishida, Chiyo. 1994. The Spanish reflexive se as an aspectual class marker. Linguistics 32 (3): 425–458. Noyer, R. R. 1992. Features, positions and affixes in autonomous morpholical structure. PhD Thesis, Massachusetts Institute of Technology. Pantcheva, Marina. 2011. Decomposing Path. The nanosyntax of directional expressions. PhD diss., Tromsø University. Pineda, Anna. n.d. The role of SE and NE in romance verbs of directed motion. Evidence from Catalan, Italian, Aragonese and Spanish varieties. Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, MA: MIT Press.

318

M. Vivanco

Ramchand, Gillian. 2008. Verb meaning and the lexicon: A first phase syntax. Cambridge: Cambridge University Press. Rapp, Irene, and Arnim von Stechow. 1999. Fast ‘almost’ and the visibility parameter for functional adverbs. Journal of Semantics 16 (2): 149–204. Rappaport-Hovav, Malka. 2008. Lexicalized meaning and the internal temporal structure of events. In Theoretical and crosslinguistic approaches to aspect, ed. Susan Rothstein, 13–42. Amsterdam: John Benjamins. https://doi.org/10.1075/la.110.03hov. Rigau, Gemma. 1994. Les propietats dels verbs pronominals. Els Marges 50: 29–39. Sánchez López, Cristina. 2002. Las construcciones con se. Estado de la cuestión. In Las construcciones con ‘se’, ed. Crisitina Sánchez López, 13–163. Madrid: Visor Libros. Sanz, Monserrat. 2000. Events and predication: A new approach to syntactic processing in English and Spanish. Amsterdam/Philadelphia: John Benjamins Publishing. Sanz, Monserrat, and Itziar Laka. 2002. Oraciones transitivas con se: el modo de acción en la sintaxis. In Las construcciones con se, ed. Crisitina Sánchez López, 309–338. Madrid: Visor. Schäfer, Florian. 2008. The syntax of (anti)causatives. External arguments in change of state contexts. Amsterdam: John Benjamins. Starke, Michal. 2011. Towards elegant parameters: Variation reduces to the size of lexically stored trees. Talk given in Barcelona workshop on linguistic variation in the minimalist framework. Tenny, Carol. 1994. Aspectual roles and the syntax semantics interface. Dordrecht: Kluwer. Teomiro, Iván. 2013. Low applicatives and optional se in Spanish non-anticausative intransitive verbs. Revista de Lingüística y Lenguas Aplicadas 8: 140–153. Uriagereka, Juan. 1999. Multiple spell-out. In Working minimalism, ed. Samuel Epstein and Norbert Hornstein, 251–282. Cambridge, MA: MIT Press. Vivanco, Margot. 2016. Causatividad y cambio de estado en español. La alternancia causativoinacusativa. PhD diss., Universidad Complutense de Madrid. https://eprints.ucm.es/37123/1/ T37038.pdf ———. 2017. La conceptualización de los eventos de cambio de estado y la alternancia lábil en español. Estudios de Lingüística de la Universidad de Alicante 31: 327–347. https://doi.org/10. 14198/ELUA2017.31.17. ———. 2019. To be or not to be an auxiliary verb: the case of Spanish poner(se) a + infinitive. Borealis: An International Journal of Hispanic Linguistics 8 (1): 35–54. https://doi.org/10. 7557/1.8.1.4655. ———. 2020 ¿Qué hay entre el control y la reestructuración? Sobre la construcción . Revista Española de Lingüística 50 (2): 2297–258. Xu, Ting. 2016. Almost again: On the semantics and acquisition of decomposition adverbs. PhD diss., University of Connecticut. Zagona, Karen. 1996. Compositionality of aspect: Evidence from Spanish aspectual se. In Aspects of romance linguistics. Selected from the SLSR XXIV, ed. Claudia Parodi et al., 475–488. Georgetown: Georgetown University Press. Zribi-Hertz, Anne. 1987. La réfrexivité ergative en français moderne. Le Français Moderne 55 (1): 23–52.

Part IV

A Unifying Perspective

Spanish se as a High and Low Verbalizer David Basilico

Abstract The clitic se in Spanish appears in a number of different constructions, including the reflexive, anticausative and antipassive. It even appears with certain unergative verbs. So far, a unified explanation for the polyfunctionality of se has remained elusive. In this paper, I propose that se functions as a verbalizer (see also Kallulli D, (Non-)canonical passives and reflexives. In Alexiadou A, Schäfer F. Non-canonical passives. Amsterdam: John Benjamins, 337–358, 2013) in two different domains. When it merges low, it functions to verbalize a nominal element and introduces an external argument. In this instance we see the clitic se in an unergative structure. In the second domain, the clitic merges higher and takes as its complement a Predicate Phrase (PredP) that contains in its specifier the derived position of the internal argument. This PredP semantically does not denote a predicate of events but a predicate of individuals. In this higher position, the clitic (re)verbalizes the structure, taking the predicate of individuals and creating a predicate of events. By creating a predicate of events, the verb can interact with tense and other verbal functional elements in the clause. The polyfunctionality of the clitic se results because of its verbalizing function. Keywords Reflexive · Anticausative · Antipassive · Unergatives · Verbalizer

1 Introduction A well-known property of Spanish and many other Romance languages is the polyfunctionality or syncretism of the clitic se. Most analyses take as their basis of comparison the reflexive use of se, and thus this morpheme is often termed a reflexive clitic.

D. Basilico (*) University of Alabama at Birmingham, Birmingham, AL, USA e-mail: [email protected] © Springer Nature Switzerland AG 2021 G. Armstrong, J. E. MacDonald (eds.), Unraveling the complexity of SE, Studies in Natural Language and Linguistic Theory 99, https://doi.org/10.1007/978-3-030-57004-0_13

321

322

D. Basilico

(1) a.

b.

Juan lava el coche Juan washes the car ‘Juan washes the car.’ Juan se lava. Juan se wash ‘Juan washes himself.’

transitive

reflexive

However, se is also found in anticausative constructions as well as in a construction which Masullo (1992) likens to an antipassive, since the construction turns from transitive to intransitive in the presence of the se morpheme while maintaining the external argument, rather than the internal argument as in the anticausative. (2) a.

b.

(3) a.

b.

Juan rompió el Juan broke the ‘Juan broke the vase.’ El jarrón se the vase se ‘The vase broke’.

jarrón. vase

transitive

rompió. broke

anticausative

Juan confiesa sus pecados. John confesses his sins ‘John confesses his sins.’ Juan se confiesa (de sus John se confesses (of his ‘John confesses (his sins).’

transitive

pecados). sins)

antipassive

This syncretism of the ‘reflexive’ morpheme is cross-linguistically robust, occurring outside the Romance family, as in Russian (4)–(6) and Icelandic (7)–(8), and even outside the Indo-European language family as in Chukchi (9)–(11). Russian: (4) a.

b.

(5) a.

b.

Medsestra breet pacienta. nurse shaves patient ‘The nurse is shaving the patient.’ Pacient breet-sja. patient shaves-sja. ‘The patient shaves himself’. Anton otkryl dver’. Anton opened door ‘Anton opened the door.’ Dver’ otkryl-s’. door opened-sja ‘The door opened’

transitive

reflexive

transitive

anticausative

Spanish se as a High and Low Verbalizer

(6) a.

Sobaka kusaet mal’čik-a. dog bites boy-acc ‘The dog bites a/the boy.’ Sobaka kusaet-sja. dog bites-sja ‘The dog bites (is a biter).’

b.

323

transitive

antipassive

Icelandic. (7)

Jón dulbjóst sem prestur. John.nom disguised-st as priest. ‘John disguised himself as a priest’.

(8) a. Þær opnu-ðu dyrnar. They open-3pl.pst door.the ‘They opened the door.’ b. Dynar opnu-ðu-st. door.the opened-3pl.pst-st ‘The door opened.’

reflexive

transitive

anticausative

Chukchi. (9)

a.

b.

(10) a.

b.

(11) a.

b.

atewla-nen shake-3sg:3sg/aor ‘He shook it off.’ tewla-tko-γɂe shake-tku-3sg/aor ‘He shook himself.’ ejpə-nin close-3sg:3sg/aor ‘He closed it.’ ejpə-tku-ɣʔi close-tku-3sg/aor ‘It closed.’ ǝtlǝg-e keyŋ-ǝn penrǝ-nen father-erg bear-abs attack-3sg:3sg/aor ‘The father attacked the bear.’ ǝtlǝg-en penrǝ-tko-gɁǝ keyŋ-etǝ father-abs attack-tku-3sg/aor bear-dat ‘The father rushed at the bear.’

transitive

reflexive

transitive

anticausative

transitive

antipassive

Though extensively studied, an analysis of this syncretism remains elusive. (For some recent ideas, see Embick 2004; Schäfer 2008; Alexiadou et al. 2015; Wood 2012.) One particular issue concerns whether or not the reflexive clitic itself is nominal or verbal (see the references above, Armstrong and MacDonald [this volume] as well as Harley and Folli 2005; Basilico 2010; Armstrong 2011; Cuervo

324

D. Basilico

2003, 2014 and others). In this paper, I argue that this clitic is a verbalizer. (For a similar claim, see Kallulli 2013.) However, there are two places where this verbalizer can merge. One position is quite low and creates a verb from the root itself. The second position is much higher, with the clitic merging with the phrase that contains a derived position for the internal argument. By positing two positions, we can also account for some differences in the morphosyntax of the se clitic between its appearance in reflexive/anticausative/antipassive constructions on the one hand and in certain unergative verbs on the other, while still maintaining that across all instances this clitic is a verbalizer.

2 Assumptions About Clause Structure The idea that the external argument is ‘severed’ from the verb both syntactically and semantically has become the standard approach, with the internal argument specified by the verb. In Kratzer’s approach, there is a special Voice head which contains a thematic role predicate that introduces the external argument; this thematic role predicate is integrated semantically through a special rule of Event Identification. Here, instead of VoiceP, I consider that the thematic role predicate is introduced through a v head. (12)

λe[V(e) &agent(e, NP)]

vP NPsbj

λxλe [V(e)

v´

v ag

VP V

agent(e, x)]

λe[V(e, NP)] NPobj

More recently, however, some researchers have argued that even the internal argument is severed from the verb (Borer 2005, 2013; Ramchand 2008; Bowers 2010; Lohndahl 2012, 2014; Alexiadou 2014; Acedo-Matellán and Mateu 2015). A more nuanced approach to the syntax of the internal argument considers that there are two (base) positions for the internal argument (Basilico 1998; Hale and Keyser 2002; Svenonius 2002; Ramchand 2008; Alexiadou and Schäfer 2011; Cuervo 2003, 2014). These positions are often understood as ‘subject of a small clause’ vs. ‘complement of V’ or ‘specifier of VP’ vs. ‘complement of V’ or ‘higher in the VP’ vs ‘lower in the VP’. In this paper, I follow this latter line of thinking and consider that there are two positions for the internal argument: one is a ‘severed’ position within a separate vP outside of the VP, headed by a thematic role predicate, the other is a position within the VP, with the internal argument specified by the verb.

Spanish se as a High and Low Verbalizer

325

(13) external to the VP; the verb is a predicate of events; the undergoer is integrated through event identification. λe [V(e) & und(e, NP)]

vP NP

λxλe [V(e) & und(e, x)]

v´ v und

VP V

λe [V(e)]

(14) internal to the VP; the verb is a relation between an event and an entity. λe [V(e, NP)]

VP λxλe [V(e, x)]

V

NP

In the case where the internal argument has been ‘severed’, the verb is a predicate of events only; an undergoer thematic role predicate heads a vP that takes the VP projection as a complement. Like the agent predicate, this undergoer predicate introduces the (internal) argument and is semantically integrated through Event Identification. When it is the verb that introduces the internal argument, the verb is a relation between an event and an entity and the internal argument is introduced within the VP itself. I also consider that there is a functional projection above the base position for the internal argument but below the position of the transitive external argument (Johnson 1991; Basilico 1998; Travis 2010; Borer 2013 and others). This functional projection, which I notate as PredP (for Predicate Phrase), hosts the direct object in transitive clauses. The internal argument generated in vP moves to PredP, where it is assigned case by the transitive v. (15) λe [V(e) & und(e, NPDO & ag(e,NPSBJ)]

vP NPSBJ

λyλe [V(e) & und(e, NPDO ) & ag(e,y)]

v´

λx∃e [V(e) & und(e, x)]NPDO

v[trans] PredP NPDO

λx∃e [V(e) & und(e, x)]

Pred´

Pred

λe[V(e) & und(e, NP)]

vP NPDO v und

v´

λxλe[V(e) & und(e, x)] VP V

λe[V(e)]

326

D. Basilico

There is a difference in interpretation: VP/vP is an event description, while PredP is a predication over individuals (Kuroda 1972; Kluender 1992; Basilico 1998). Pred existentially closes the event argument; movement of the internal argument to PredP introduces a lambda operator that binds a variable in the argument position of the thematic role predicate, generating a predicate of individuals. (16)

Event description

λe[P(e,X ..Y. . .)]

Predication λx∃e[P(e, x . . .Y. . .)] (z)

The introduction of v[trans], which introduces the external argument, existentially discloses the event argument, (Dekker 1993). Syntactically, the structure is ‘reverbalized’. Semantically the predicate of individuals that appears at the PredP level becomes again a predicate of events that can interact with tense and other event operators. Given this background, in the next section I discuss a ‘high’ use of the clitic, in which se appears in a high verbalizing head associated with the external argument.

3 The Anticausative In this analysis, the derivation of the anticausative is similar to that of Schäfer (2008) and Alexiadou et al. (2015), in that the se clitic relates to the head that introduces the external argument, but this reasoning is couched differently. In the present work, se is a v head that is in complementary distribution with the v head that introduces the external argument and assigns case to the internal argument, as with the above authors. Thus, the transitive and anticausative clauses have the same structure but a different v head which merges with PredP. Here, though, I consider that this higher v head is responsible for ‘reverbalization’. Both heads disclose the event argument, but only the transitive v head introduces a thematic role predicate and assigns case. (17) a.

b.

(18)

Juan rompió el Juan broke the ‘John broke the vase.’ El jarrón se the vase se ‘The vase broke.’ transitive structure

jarrón. vase rompió. broke

Spanish se as a High and Low Verbalizer

327

λe [break(e) & ag(e, John) & und(e, vase)]

vP NP Juan

λxλe[break(e) & ag(e, x) & und(e, vase)]

v´ v trans]

λx∃e[break(e) & und(e,x)] (vase)

PredP

NP Pred´ el jarron Pred vP

λe[break(e) & und(e, vase)]

NP el jarron

λxλe[break(e) & und(e, x)]

v´ v und

VP V rompió

λe[break(e)]

(19) intransitive structure λe [break(e) & und e, vase)]

vP v

λx ∃e[break(e) & und(e,x)] (vase)

PredP NP el jarron

Pred´ Pred

λe [break(e) & und(e, vase)]

vP

λxλe [break(e) & und(e, x)]

NP el jarron VP und V rompió

λe [break(e)]

Since the transitive and intransitive verbalizer occupy the same position, se will be ungrammatical in a transitive structure. (20) *Juan se rompió el Juan se break the ‘Juan broke the vase.’

jarrón vase

328

D. Basilico

4 The Reflexive The challenge of giving a unitary account of se is that although it is associated with intransitivity, the ‘suppressed’ argument appears to be different between the anticausative and the reflexive; in the former it is the internal argument that has a theme/patient thematic role with the external argument suppressed, while in the latter it is the external argument that has an agent thematic role and the internal argument is suppressed. The proposal that I make here is that while there is a different ‘missing’ argument between the reflexive and the anticausative, the se clitic in both constructions share the same syntactic position. In the reflexive, there is a null reflexive morpheme that takes a verb that is a predicate of events and turns it into a verb that is a relation between an event and an entity. Importantly, the clitic is not this reflexive morpheme. In this case, the internal argument is introduced within the VP, not in a separate head. However, the entity argument that is introduced by the special reflexive morpheme is a variable that must be bound; syntactically in most instances, this bound variable is PRO.1 (21) a. b.

[V refl V] [VP [V refl V ] PRO]

λxλe[V(e, X)] where X is a bound variable λxλe[V(e, PRO)]

Now, in this case, we can generate the agent predicate within a v head that merges with VP. In this way, the external argument will be generated low, below PredP and then move to the specifier of PredP. Movement of the external argument to PredP introduces a lambda expression that binds PRO, giving the reflexive reading. The clitic is added for the same reason it is added in the anticausative: it reverbalizes the PredP structure. I give example derivations in the following.

1

A reviewer reminds me that it is possible for the argument position to contain a reflexive pronoun along with the clitic, as in the following French sentence from Labelle (2008). 1. Le ministre se copie lui-même. the deputy se imitate-pres-3s himself The deputy imitates himself. The reviewer suggests that this may pose a problem for the analysis advocated here. However, in Labelle’s (2008) analysis, this overt reflexive pronoun does not saturate the argument position of the verb and considers the possibility that this pronoun “translates as a formula containing a variable” (847). If this pronoun does translate as a variable, then it is another way for the internal argument of the [V refl V ] complex to have its internal argument saturated and meet the bound variable requirement of the internal argument imposed by refl.

Spanish se as a High and Low Verbalizer

(22)

lavar

329

λe [wash(e) &und(e, the car)]

vP NP el coche

λx λe [wash(e) & und(e, x)]

v´

v und

VP λe [wash(e)]

V lava lavar+refl

λe [wash(e, PRO)]

VP

λxλe [wash(e, X)]

V

NP PRO V ava

refl

In the latter structure, it will be possible to add vP[ag] above VP, since the internal argument is introduced within the VP. (23)

λe wash[(e, PRO) & ag(e, John)]

vP NP Juan

λxλe [wash(e, PRO) & ag

v´

v ag λxλe [wash(e, X)] V refl

λe[wash(e, PRO)]

VP NP PRO V

Note that both structures (19a) and (20) are similar, with a vP above VP. The difference between the two lies in what thematic role predicate is within vP. In the transitive case, there is an undergoer predicate in v, while in the reflexive case it is the agent predicate. In both cases, the Pred head is added next, with the movement of the noun phrase in vP to the specifier of PredP. This movement to PredP creates a predication; a lambda operator is introduced. In the reflexive, this operator also binds PRO, generating the reflexive interpretation.

330

D. Basilico

(24) lavar

λx∃e[wash(e) & und(e, x)](car)

PredP NP el coche

Pred´ Pred

λe [wash(e) & und(e, car)]

vP

NP el coche

λxλe[wash(e) & und(e, x)]

v´

v und

VP V lavar

λe [wash(e)]

(25) lavar+refl λx∃e[wash(e, x) & ag(e, x)](John)

PredP NP Juan

Pred´ Pred

λe[wash(e, PRO) ag(e, John)]

vP

λe wash(e, PRO) & ag(e, x)]

NP Juan v [ag]

VP V

refl

λe[wash(e, PRO)] NP PRO

V lava

With the next step, the derivations differ again. In the transitive, v[trans] merges, discloses the event argument and assigns case to el coche as well as introduces an external theta role.

Spanish se as a High and Low Verbalizer

(26)

331

λe[wash(e) & ag(e, John) & und(e, car)]

vP NP Juan

v´ v[trans]

λx∃e[wash(e) & und(e, x)](car)

PredP

NP Pred´ el coche Pred vP

λe [wash(e) & und(e, car)]

NP el coche

λe [wash(e) & und(e, x)]

v´

v und

VP V lava

λe [wash(e)]

With the reflexive, v[se] is added; it discloses the event argument but does not assign an additional theta role nor does it assign case to the single NP agent, which then gets its case from Tense (and there is likely additional movement beyond vP [se]). (27)

λe [wash(e, John) & ag(e, John)]

vP v se

λx∃e[wash(e, x) & ag(e, x)](John)

PredP NP Juan

Pred´ Pred

λe [wash(e) & ag(e, John)]

vP NP Juan

λe [wash(e, PRO) & ag(e, x)]

v´ v ag

VP V

refl

NP PRO V lava

In this analysis of the reflexive, there is a null refl element that takes a verb which is a predicate of events and creates a relation between an event and an individual. However, in languages such as Russian, we see the clitic only appears with certain verbs; many of these verbs that have a reflexive meaning are those which are ‘naturally reflexive’ such as ‘wash’, ‘shave’, etc. though there are a few others. Thus, a verb such as ‘see’ is out in the –sja construction (but fine in the corresponding transitive construction with reflexive pronoun).

332

D. Basilico

(28) a.

b.

*On vidit-sja v zerkale. he.nom sees-sja in mirror ‘He saw himself in the mirror.’ On vidit sebja v zerakale. he.nom sees self.acc in mirror ‘He sees himself in the mirror.’

Also, Say (2005) states that these reflexive verbs have a “conventionalised meaning that often goes beyond the compositional sum of the meaning of the base verb and the indication of coreference of its two arguments” (260). Consider the ungrammaticality of the following sentence. (29) *On slučajno zastrelil-sja He.nom unintentionally shot-sja ‘He unintentionally shot himself’ (e.g. shivered while holding a gun in his hands). The reflexive verb doesn’t just mean ‘to shoot oneself’ but “to commit suicide by way of shooting oneself” (260). This evidence points to the lack of a general reflexivization operation as in Spanish. In Russian, certain verbs have at least two lexical representations. In one, the verb is a predicate of events. In the other, the verb is a relation between an event and an entity, where the entity must be a bound variable. This argument structure would be limited to naturally reflexive verbs. (30) breet1 breet2

λe [shave(e)] λXλe [shave(e, X)] where X is a bound variable

The derivation for the reflexive will be the same as above. In these cases, we can consider that these verbs are lexically relations between an event and an entity, with the entity argument being restricted to a bound variable. There is no refl element that creates reflexive verbs and thus it appears that the ‘reflexive’ interpretation when the clitic appears is lexically restricted.2 So, in the structure in which we get a reflexive interpretation with the clitic, the clause starts with breet2. Note again that there is no null refl element to create the relation between an event and an entity because the verb is lexically specified as a relation between an event and an entity. Because the internal argument has already been introduced in a ‘low’ position, v[ag] is merged into the structure, not v[und], and thus the external argument is introduced. After v[ag] is introduced, Pred is merged, with the external argument noun phrase moving to the specifier of Pred.

Kempchinsky (2004) notes that ‘verbs of grooming’ in Spanish behave differently with respect to clitic doubling and omission of se under causative verbs such as hacer; thus it may be the case that these verbs, too, in Spanish also have a different lexical specification. 2

Spanish se as a High and Low Verbalizer

333

Now we have created a predicate of individuals, which then needs to get reverbalized. It is at this point that sja merges to reverbalize the structure. (31)

λe[wash(e, patient) & ag(e, patient)]

vP v sja

λx∃e[wash(e, x) & ag(e, x)](patient)

PredP NP pacient

Pred´ Pred

λ

vP

NP pacient

ag(e, patient)]

λe [wash(e, PRO) & ag(e, x)]

v´ v ag

wash(e)

VP V breet

NP PRO

If breet1 is the verb that is chosen, the derivation will proceed by merging v[und] instead of v[ag], since the verb itself does not introduce an argument. In this case, to derive the reflexive reading, the overt reflexive sebja pronoun must merge into the structure. This pronoun merges in the specifier of v[und] where it is assigned a thematic role and moves to the specifier of PredP. The v[trans] head merges next, reverbalizing the structure and introducing the external argument. This external argument noun phrase binds the reflexive in the specifier of PredP, giving the reflexive reading. (32)

λe[shave(e) & ag(e, patienti) & und(e, selfi)]

vP NP pacient

v´ vtrans

PredP λx∃e[shave(e) & und(e, x)](self)

NP sebja

Pred´ Pred

λe [shave(e) & und(e, self)]

vP

NP sebja

v´ v [und]

λe [shave(e) & und(e, x)] VP V breet

λe [shave(e)]

334

4.1

D. Basilico

Ditransitive Reflexives

In Spanish, it is possible for a reflexive clitic to appear in a ditransitive clause, as in the following. (33) Juan se compró un libro. John se bought a book ‘John bought himself a book.’ In this case, we have a low applicative, in the terminology of Pylkkänen (2008). With a low applicative, the applicative head Appl first merges with the direct object noun phrase, then the indirect object noun phrase, creating an Applicative Phrase (ApplP). The verb and the ApplP merge together next. Such a structure has been extended to Spanish by Cuervo (2003). Now, in this case, the Appl head introduces the arguments, not the verb and not a v[und]. I consider that Appl acts like refl in that it merges with a verb to create the argument structure but unlike refl it introduces more than one argument. In order to highlight the parallel between argument introducing refl and Appl, the structure I adopt is different than that of Pylkkänen (2008). In the ditransitive structure below, the Appl element, like the refl, first merges with the verb and introduces both the direct and indirect object positions. The complex V + appl head then merges first with the direct object and then with the indirect object to create a VP. I borrow the denotation of the Appl head from Pylkkänen (2008), though with important differences. In Pylkkänen (2008), the ApplP merges with a verb that has a direct object argument. Here, though, the verb lacks an argument and it is the Appl head that introduces the direct object argument. Second, the Appl head semantically combines first with the verb. (34)

appl: λfλxλyλe[f(e) & und(e, x) & to-the-possession(x,y)] VP

λe[V(e) & und(e, DO) & to-poss.(DO, IO)]

V λxλyλe[V(e) & und(e, x) appl & to-the-possession(x,y)]

λyλe [V(e) & und(e, DO) & to-the-possession(DO,y)]

V¢

IO

DO V

Spanish se as a High and Low Verbalizer

335

The indirect object position contributed by Appl can be saturated by PRO. This PRO must be bound, and in this case we add v[ag] above VP. This external argument noun phrase then moves to PredP, binding the PRO in indirect object position. The v[se] head merges next, reverbalizing the structure. (35)

λe[buy(e) & und(e, book) & to-the-possession(book, John) & ag(e, John)] λ ∃ und to-the-possession(book, x) & ag(e,x)](John)

vP

se NP Juan

Pred¢ Pred

vP NP Juan

v¢ v[ag] PRO

λxλyλe [buy(e) & und(e, x) & to-the-possession(x,y)]

λe [buy(e) & und(e, book) & to-possession(book, PRO)]

VP V¢ V

appl

NP un libro V compró

We might expect the reflexive clitic to appear with all such ditransitive structures, since the indirect object is in a ‘low’ position. However, if we take the position of Basilico (1998) and consider that the indirect object participates in an ‘inner’ categorical predication, then the indirect object moves to PredP. In the reflexive case, since the PRO indirect object is bound by the external argument, and the external argument participates in the inner categorial predication, PRO itself need not be in PredP. Note that I follow Cuervo (2003) and treat the ‘a + NP’ as a noun phrase.3 (36) Pablo (le) mandó un diccionario Pablo cl.dat sent a dictionary ‘Pablo sent Gabi a dictionary.’

3

a Gabi. Gabi.dat

The optional dative clitic which agrees with the indirect object appears in the head of Appl, following Cuervo (2003, 2014). I propose that this clitic does not appear in the reflexive case because PRO lacks features for agreement.

336

D. Basilico

vP NP Pablo

v¢ v[trans] PredP NP a Gabi

Pred¢ Pred

VP

NP a Gabi

V¢ V

appl

NP un diccionario V

5 Antipassive The antipassive cases are lexically restricted.4 As Armstrong (2011) notes, there are only a few such verbs. Other verbs which can appear in both a transitive and antipassive construction are lamentar ‘to regret’, rehusar to ‘refuse’, compadecer ‘to sympathize with’. Armstrong notes, too, that not all verbs have the same meaning in the ‘antipassive’ construction, as seen with verbs such as acordar ‘to agree’/ acordarse de ‘to remember’, despedir ‘to fire’/despedirse de ‘to say goodbye’, and deshacer ‘to take apart’/deshacerse de ‘to get rid of’.5

4

Although this construction is often termed the antipassive, this construction is not completely parallel to canonical antipassive structures seen in ergative languages such as Chukchi or Inuit. (For a recent discussion of the antipassive in Inuit, see Spreng 2012.) Like the canonical antipassive, we see an alternation where, with special morphology on the verb, the external argument remains while the internal argument appears as an oblique or is missing entirely. However, unlike these ergative languages, where the antipassive construction is productive, this construction in Spanish is limited to a few lexical items, as noted by Armstrong (2011). 5 A reviewer notes that the reflexive/non-reflexive forms are more related in meaning than what the translations above suggest. The verb despedir means ‘fire’ in the sense of ‘dismiss from a job’. If you dismiss someone, you send them away, which relates to the sense of despedirse de in that when you say goodbye to someone, you are sending them away. We can see the same relatedness with the verb deshacer ‘to take apart’ and its reflexive counterpart. When something is taken apart, each elements is getting away from another element. In the reflexive deshacerse de, the elements are getting away from you or your belongings, which relates to the sense of ‘get rid of’. Even though these two verbs have related senses, they do not have identical senses in every case. We expect as the meanings drift we would get two separate lexical items.

Spanish se as a High and Low Verbalizer

(37) a. b.

337

Juan confiesa sus pecados. ‘John confesses his sins’ Juan se confiesa (de sus John se confesses (of his ‘John confesses (his sins).’

transitive pecados). sins)

antipassive

Given these facts, I propose that there are two lexical items, one which is a predicate of events and the other which is a relation between an event and an individual. In this case, though, the entity argument need not be a bound variable. (38) a. b.

λe [confess(e)] λxλe [confess(e, x)]

confesar1 confesar2

Just like the transitive/reflexive cases, we get two different clause structures. With the transitive structure, the internal argument is added within a v[und]P phrase and moves to the specifier of PredP, where it gets case from v[trans]. In this case, we merge confesar1. (39)

vP λe [confess(e) & ag(e, John) & und(e, his sins)] NP Juan

v´ v[trans]

PredP λx∃e[confess(e) &und(e, x)](his sins)

NP sus pecados

Pred´ vP λe [confess(e) & und(e, his sins)] λe [confess(e) & und(e, x)] sus pecados v und

VP V λe [confess(e)] confiesa

In the intransitive case, since the verb itself introduces the internal argument, there is no need for a v[und] head. In this case, like the reflexive, the v head above VP introduces the external argument. Unlike the transitive, it is the external argument, not the internal argument, that moves to the specifier of PredP.6

6 A reviewer worries that this analysis is merely descriptively adequate rather than explanatorily adequate, in that the two argument structures are simply listed. However, this ‘antipassive’ construction is lexically limited, so there must be something special with respect to these verbs. My goal here is to explain the appearance of the se clitic and link the appearance of the clitic in this construction with its appearance in other constructions. The explanatory link between the antipassive and the reflexive is that in both cases, the internal argument is introduced in a low position and in a different position than that of the transitive.

338

D. Basilico

(40) intransitive λe [confess(e,his sins) & ag(e, John)]

vP v se

λx∃e[confess(e, his sins) &ag(e, x)](John

PredP NP Juan

Pred´ Pred

vP NP Juan

λe confess[(e, his sins)

ag(e, John )]

v´ λe [confess(e, his sins)

ag(e, x)]

v ag

VP V confiesa

PP de sus pecados

It is important to note that the v head above VP does not assign case, whether there is the external argument thematic role predicate introduced within this v head or the internal argument thematic role predicate (either v[ag] or v[und]). As a result, when the head is v[ag] and the verb introduces an argument, the P de steps in to assign case to the NP sus pecados that is introduced by the verb. This analysis is similar to that given in Armstrong (2011), who considers that in these cases, the external argument is introduced by a special v head that does not assign case. The se clitic heads this special v head. Here, I have allowed this v head that does not assign case to introduce both an internal as well as an external argument thematic role predicate (the agent or undergoer predicate). Also, this v head does not host se; se appears higher up and alternates with a head that assigns only an external argument thematic role and accusative case. In this way, we can unify the appearance of the clitic with its appearance in reflexive and anticausative constructions.7

7

The Russian construction in (6) above has been likened to the antipassive, in that the agent argument remains while the internal argument is missing when–sja is present. This construction is also termed the ‘absolutive’, and only appears with a limited set of verbs (such as bodat’(sja) ‘to butt’, ljagat’(sja) ‘to kick’). We see the same use of the tko/tku suffix in Chukchi (Kozinsky, Nedjalkov and Polinskaja 1988). (1)

ɁǝttɁ-e juu-nin ‘the dog bit him.’ ⟶ ɁǝttɁ-ǝn nǝ-jγu-tku-qin ‘the dog bites.’

In this case, we can consider that there are two verbs ‘bite’, one which is just a predicate of events and one which is a relation between an event and an entity. However, unlike in the reflexive, the entity argument is not a bound variable. Syntactically, it will be represented as arbitrary PRO. (2)

kusaet1 kusaet2

λe [bite(e)] λx λe [bite(e, xARB)]

Interestingly, this construction does not occur in Spanish.

Spanish se as a High and Low Verbalizer

339

6 Extending the Analysis: Unaccusative Doublets Cuervo (2003, 2014) notes the existence in Spanish of ‘unaccusative doublets’, in which the same verb can appear with or without the se clitic. (41) a. Cayeron tres hojas. fall.pl three leaves ‘Three leaves fell.’ b. Se cayeron tres vasos. se fall.pl three glasses ‘Three glasses fell (down).’

se less unaccusative

se unaccusative

Cuervo posits two distinct structures. With the se less unaccusative, the verbal root acts as a manner component and is adjoined to a v[go], and the single argument of the verb appears as a complement. With the se unaccusative, the verbal root acts as the complement of the verb and the single argument of the verb appears in the specifier of this v[be] head; here the root is interpreted as a state. This v[be] structure is the complement of a v[go] verb which is headed by se; the se clitic is coindexed with the single argument of v[be]P. In this way, the argument of the change event given by v[go] and the argument of the state event given by v[be] are the same argument. (42) se less variant

se variant [go]

[go] [go]

[go] se

[be]

[go] +

Here, I propose that the difference between these two variants results from a different lexical specification of the verbs. The se unaccusative contains a verb that is a predicate of events only; the single argument is generated within v[und]P. In this case, the structure is the same as the anticausative, with the merger of PredP and then the se verbalizer. The difference here is that these verbs do not occur with an external argument. With the se less unaccusatives, the verb is specified as a relation between an event and entity. Since this verb also does not allow an external argument, there is no v head above VP. In this case, I propose that the VP combines directly with Tense. Thus, there is no PredP and thus no need for the se verbalizer.

340

D. Basilico

(43) se unaccusative λe [fall down(e)

vP v

λx∃e[fall down(e) und(e, x)](three glasses)

PredP NP tres vasos

und(e, glasses)]

Pred´ Pred

λe [fall down(e)

vP NP

und(e, three glasses)

λe [fall down(e) & und(e, x)]

v´ v und

VP V λe [fall down(e)] cayeron

(44) se less unaccusative TP T

VP

λe [fall (e, three leaves)]

V NP cayeron tres hojas

λxλe[fall(e,x)]

In this way, we assimilate the se less unaccusative to ‘true’ unaccusative verbs such as llegar ‘arrive’. These verbs, too, would be lexically specified as relations between an event and an entity. (45)

TP T

λxλe[arrive(e,x)]

VP V llegan

λe [arrive(e, trains)] NP trenes

Russian gives an interesting confirmation of this particular account. Like Spanish, we see ‘unaccusative doublets’ (Tatevosov 2013), one which shows the presence of ‘reflexive’ marking and one without. Interestingly, unlike in Spanish, the verbs are different between the variants. As can be seen below, with the clitic –sja, the verb ‘dry’ appears as suš and without the clitic the verb is sox. Other verbs which have these doublets include ‘freeze’ (zamerz-nu-t’ and zamoroz-i-t’-sja), ‘get wet’ ((na)mok-nut’ and (na)moč-i-t’-sja), ‘melt’ (rastaj-a-t’ and rastop-i-t’-sja) and others.8 8

A similar situation occurs in Polish (Szczesniak 2006).

Spanish se as a High and Low Verbalizer

(46) a.

b.

341

Bel’je suš-it-sja. linen dry-prs.3sg-refl ‘The linen is drying.’ Bel’je sox-n-et. linen dry.ipvf-cl-prs.3sg ‘The linen is drying.’

Importantly, the form of the transitive verb is the same as the sja unaccusative (anticausative) and not the sja less unaccusative (unaccusative). (47) Vasja suš-it Basil dry.ipvf-prs.3sg ‘Basil is drying the linen.’

bel’je. linen.acc

This is exactly what you would expect given the argument structure and syntactic differences discussed above for Spanish. Both the transitive and the anticausative would be based on the same verb, which would be a predicate of events only. Thus, the internal argument would be generated within v[und] outside of VP. In this case, PredP is merged, creating a predication. The difference would be the verbalizer added after the merger of PredP. The anticausative would involve the merger of an intransitive v[sja], while the transitive would involve the merger of v[trans] that adds the external argument and assigns accusative case to the direct object in the specifier of PredP. (48)

suš λe [dry(e)] [vP v[trans] or v[sja] [PredP [NP bel’je] [Pred´ Pred [vP [NP bel’je] [v´ v[und] [VP suš ]]]]]]

The unaccusative variant would be based on the verb form that is a relation between an event and an entity. In this case, like the se less unaccusative above, there is no vP projection above and also no PredP. We cannot merge v[se] because there is no PredP; we also cannot merge v[trans] and thus there is no transitive variant for this form. (49)

6.1

sox λxλe [dry(e,x)] [TP T [VP soz [NP bel’je]]]

Low Applicatives

Cuervo (2014) also notes that there is a difference in the kinds of applicatives allowed with the different unaccusatives. With the se less intransitive, the applied argument is interpreted as a relation between two individuals; here it is a ‘low

342

D. Basilico

applicative’ in the sense of Pylkkänen (2008). This relation is, according to Cuervo, ‘possession-related’; the dative argument can be interpreted as a recipient, as a source or as a static possessor relative to the nominative argument. On the other hand, with se intransitives, the dative applied argument cannot be understood to be in a relation with the nominative argument; it is understood as an ‘affected’ argument; the dative is somehow affected by the change of state and is not interpreted in relation to the single argument. (50)

(51)

Applicative with se less unaccusative a. A Gabi le cayeron visitas de Gabi.dat cl.dat fell.pl visitors from ‘Gabi got guests from London.’ b. A todos los árboles les salieron All the trees.dat cl.dat came.out.pl ‘All the trees got fungi.’

Londres. London hongos. fungi

Applicative with se unaccusative a. A Carolina se le salieron dos clavos (de la pared). Carolina.dat refl cl.dat came.out.pl two nails (from the wall) ‘Two nails came out (from the wall) on Carolina.’ b. A Gabi se le cayó la biblioteca (encima de la comida). Gabi.dat refl cl.dat fell the bookcase (on.top of the food) ‘The bookcase fell (on top of the food) on Gabi.’

Cuervo notes that the generalization appears to be that a low applicative can only relate to a direct object that is a complement and not to one that is in a specifier position. She adopts Pylkkänen’s (2008) analysis of the low applicative. Recall, though, in the above analysis of ditransitives, I treated the low applicative head as merging with a verb and introducing two arguments, the undergoer argument and the possessor argument. With these particular verbs, the verb itself also introduces its own argument. Semantically we need for the undergoer argument of the v[appl] head and the single argument of the verb to be ‘caught’ by the same lambda operator. I propose that the v[appl] head and its theme argument are integrated into the event not just through the event argument but through the entity argument within VP. That is, the single argument of the verb and the undergoer argument of the applicative head are identified as the same argument through a process of Argument Identification, similar to the process of Event Identification (see Alexiadou et al. 2013; Wood 2012). Here, the verb salir introduces an entity argument that is identified as the same argument as the undergoer of the applicative. I follow Cuervo (2003, 2014) and treat the dative clitic as the head of v[appl].

Spanish se as a High and Low Verbalizer

(52)

343

λe [fall(e, guests) &und(e, guests) & to-the-possession(guests, Gabi)] NPdat V´ λyλe [fall(e, guests) & und(e, guests a Gabi. to-the-possession(guests,y)] λxλyλe [fall(e, x) & und(e, x) V NP & to-the-possession(x ,y)] visitas appl V λzλe[fall(e,z)] cayeron VP

With the se unaccusative, a low applicative is not available, since the undergoer argument is in a specifier, being introduced by v[und]. However, a high v[appl] head that semantically integrates through the event argument is able to merge with this verb. (53)

vP v se

PredP λxλe [come out(e) &und(e, x) & aff(e, Carolina)](two nails) NP Pred´ dos clavos Pred vP λe [come out(e) & und(e, two nails) & aff(e, Carolina)] NPdat v´ λyλe [come out(e) &und e, two nails) & aff(e,y)] A Carolina v[appl] vP λe [come out(e) & und(e, two nails)] NP dos clavos v und

v´ λe [come out(e) & und(e, x)] VP V salieron

λe [come out(e)]

This analysis of the low applicative fits well with an analysis of the English benefactive given in Basilico (2008). Levin (1993) shows that it is limited to verbs of creation, verbs of obtaining and verbs of preparation. Basilico (2008) argues that the benefactive head introduces an argument that is identified with the argument of the verb. In this case, the verbs which participate in the benefactive alternation are those verbs which are relations between an event and an entity. Change of state verbs cannot add an additional benefactive argument because they are only predicates of events. (54) (55)

The chef baked a cake/The chef baked me a cake. ‘The verb ‘bake’ is lexically a relation: λxλe [bake(e, x)]’ *The chef broke me a piece of chocolate. ‘The verb ‘break’ is lexically a predicate of events: λe [break(e)]’

344

D. Basilico

7 The Unergative: Se as a Low Verbalizer Finally, we need to account for unergative intransitives which lack se marking, as well as those which do show se. (56) a. Los niños bailaron the children dance.pl ‘The children danced.’ b. Los niños jugaban en el jardin. The children play.pl in the garden ‘The children were playing in the garden.’ BUT (57) Los niños se rieron. the children se laugh.pl ‘The children laughed.’ We might suppose that the external argument is introduced by vP[ag] since both lack a direct object, but this analysis would predict the appearance of the se clitic on all instances of unergative verbs. So far, I have considered that se occupies a high position in the structure, in a v head above PredP, and is responsible for ‘reverbalizing’ the structure, taking a predicate of individuals created by PredP and giving back a predicate of events. Interestingly, cross-linguistically, we see this element that marks anticausatives and reflexives (as well as antipassives for those languages which have them) is also a denominal verbalizer. Thus, in Chukchi, the –tko/tku suffix combines with nouns to create instrumental verbs. (58) Chuckchi –tku/tko (Kozinsky et al. 1988; Dunn 1999) a. Kozinsky et al. (1988) write that -tku/tko “forms denominal verbs, eg. wǝlpǝ ‘spade’⟶wǝlpǝ-tko ‘to dig with a spade”. b. ŋote-nqac ta-ɣʔe ewən ənpənacɣ-ə-n jʔilɣ-ə-kin there-side pass-th ints old.man-e-3sg.abs moon-e-rel3sg.abs orw-ə-taraŋ-rajwacə n-ə-ɣatɣa-tko-qen sled-e-build.house-lee.side hab-e-adze-util-3sg Similarly for Icelandic, the –st suffix that appears on some anticausatives and reflexives also appears as a denominal verbalizer. (59) Icelandic

st (Wood 2012)

Ég væri til að gítarast með I would.be up in guitar-st with ‘I would be up for guitaring with y’all.’

yllur. you.pl

Spanish se as a High and Low Verbalizer

345

Even in Romance, we see a denominal use of the reflexive clitic. Kallulli (2013) points out with respect to Italian (and German) inherently reflexive and pseudoreflexive verbs that “one crucial characteristic [of these verbs] is their denominal and/or deadjectival nature”. For example, the verb vergognar-si ‘to be ashamed’ in Italian is derived from the nominal root vergogna ‘shame’. (60) Quando Dio *(si) vergogna degli uomini e gli uomini *(si) when God si ashamed of-the men and the men si vergognano di Diio ashamed of God. ‘When God is ashamed of men and men are ashamed of God.’ Note that without extra morphology these inherent reflexives do not occur in a transitive frame. Thus, we cannot consider them to be anticausatives. (61) a. *Gli uomini hanno vergognato the men have shamed ‘Men have put God to shame.’ b. Gli uomini hanno svergognato The men have shamed ‘Men have put God to shame.’

il Dio. the God il Dio. the God

Let us suppose, then, that in those unergative cases where we see the clitic, we have a denominal structure; it takes a nominal root and creates a verb. In this case, the clitic is low in the structure, but with the same function as when it occurs higher in the structure when it merges with PredP: here, the clitic is a verbalizer. Unlike the high verbalizer, however, it introduces the external argument as well. (62)

reír:

λe [do(e, y, laugh)] TP

λe[do(e, the children, laugh)] λxλe[do(e, x, laugh)]

T

VP

NP los niños

V´ V se

ri

The difference between a verb such as ‘laugh’ with the clitic and a verb such as ‘dance’ without the verbalizer is the verbalizing element. With a verb such as ‘dance’, the verbalizer is null. This difference is the consequence of the idiosyncratic morphological requirements of the different roots.

346

D. Basilico

(63) bailar: λxλe[do(e, x, dance)] TP T

λe[do(e, the children, dance)]

VP

NP los niños

λxλe[do(e, x, dance)]

V´ V do

bail

However, in both cases, the DO verbalizer itself introduces the single argument. If the external argument were added by a separate v[ag] head, we would expect a clitic in the ‘dance’ case, as discussed above. Furthermore, since there is no v element above VP that introduces an argument, there is no PredP; the unergative verb, like ‘true’ unaccusatives, combines directly with Tense.

8 Conclusion This paper gives a unified analysis of se as an intransitive verbalizer, which takes predicates of entities and turns them into predicates of events. This verbalizer occurs in two positions: high up in the structure in a position where it is in complementary distribution with a verbalizer that introduces an external argument, and low down in the structure where it acts as a denominal verbalizer and introduces the external argument. This analysis exploits two theoretical innovations: (1) that the undergoer (internal) argument, as well as the agent/experiencer (external) argument, can be severed from the verb; and (2) that the internal argument can appear in two different positions (within VP or vP[und]), reflecting whether the verb is a predicate of events or a relation between an event and an entity. The verbalizing nature of the clitic is supported by its use as a denominal verbalizer, both within the Romance language family and outside of it. The work here is also part of the conversation about the nature of argument structure and whether or not it is lexically or syntactically specified. Ever since Kratzer (1996), building on Marantz (1984), proposed to ‘sever’ the external argument from the verb, a lingering question has been whether or not verbs introduce any of their arguments, even their internal argument. An alternative is that all arguments are introduced syntactically and licensed by a special head in the syntax (Borer 2005, 2013; Ramchand 2008; Bowers 2010; Lohndahl 2012, 2014; Alexiadou 2014; Acedo-Matellán and Mateu 2015). This work supports the claim that while some verbs do not introduce any of their arguments, there are some verbs that actually do specify their internal argument. This analysis also extends the theory by allowing even the external argument to be generated in different positions (within vP[trans] above PredP, or vP[ag]) (see

Spanish se as a High and Low Verbalizer

347

also Wiltschko 2006). In this analysis, an agent/experiencer and undergoer argument can appear in structurally the same position (within a vP above VP but below PredP), thus complicating our understanding of the notion of unaccusativity as well as the relationship between syntactic structure and thematic relations, as with Baker’s (1988) UTAH. Acknowledgments I thank the audience at the Workshop on se/si for their questions and suggestions as well as the two anonymous reviewers for further input on this paper. I especially thank Grant Armstrong and Jonathan MacDonald for organizing and hosting the conference as well as for their editorial input. Of course, all errors are my responsibility.

References Acedo-Matellán, Victor, and Jaume Mateu. 2015. From syntax to roots: A syntactic approach to root interpretation. In The roots of syntax, the syntax of roots, ed. Artemis Alexiadou, Hagit Borer, and Florian Schaefer, 14–32. Oxford: Oxford University Press. Alexiadou, Artemis. 2014. Roots don’t take complements. Theoretical Linguistics 40: 287–297. Alexiadou, Artemis and Florian Schäfer. 2011. There-insertion: an unaccusative mismatch at the syntax-semantics interface. Online Proceedings of WCCFL 28. https://sites.google.com/site/ wccfl28pro/alexiadou-schaefer. Alexiadou, Artemis, Florian Schäfer, and Giorgos Spathas. 2013. Delimiting voice in Germanic: On object drop and naturally reflexive verbs. In Proceedings of 44th meeting of the North East linguistic society, ed. Jyoti Iyer and Leland Kusmer, 1–14. Amherst, MA: Graduate Linguistics Students Association, University of Massachusetts, Amherst. Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer. 2015. Exernal arguments in transitivity alternations. Oxford: Oxford University Press. Armstrong, Grant. 2011. Two classes of transitive verbs: Evidence from Spanish. PhD diss., Georgetown University. Baker, Mark. 1988. Incorporation. Chicago: University of Chicago Press. Basilico, David. 1998. Object positions and predication forms. Natural Language and Linguistic Theory 16: 541–595. ———. 2008. Particle verbs and benefactive double object constructions: High and low attachments. Natural Language and Linguistic Theory 26: 731–773. ———. 2010. The se clitic and its relationship to paths. Probus 22: 271–302. Borer, Hagit. 2005. Structuring sense Vol II: The normal course of events. Oxford: Oxford University Press. ———. 2013. Structuring sense Vol III: Taking form. Oxford: Oxford University Press. Bowers, John. 2010. Arguments as relations. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press. Cuervo, Maria Cristina. 2003. Datives at large. PhD diss., MIT. ———. 2014. Alternation unaccusatives and the distribution of roots. Lingua 141: 48–70. Dekker, Paul. 1993. Existential Disclosure. Linguistics and Philosophy 16: 561–587. Embick, David. 2004. Unaccusative syntax and verbal alternations. In The Unaccusativity puzzle, ed. A. Alexiadou, E. Anagnostopolou, and M. Everaert, 137–158. Oxford: Oxford University Press. Hale, Kay, and Samuel Jay Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge, MA: MIT Press.

348

D. Basilico

Harley, Heidi, and Rafaella Folli. 2005. Flavors of V: Consuming results in Italian and English. In Aspectual inquiries, ed. Roumyana Slabakova and Paula Kempchinsky, 95–120. Dordrecht: Kluwer. Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9: 577–636. Kallulli, Dalina. 2013. (Non-)canonical passives and reflexives. In Non-canonical passives, ed. Artemis Alexiadou and Florian Schäfer, 337–358. Amsterdam: John Benjamins. Kempchinsky, Paula. 2004. Romance SE as an aspectual element. In Contemporary approaches to romance linguistics, ed. J. Auger et al., 239–256. Amsterdam/Philadelphia: John Benjamins. Kluender, Robert. 1992. Deriving Island constraints from the principles of predication. In Island constraints: Theory, acquisition and processing, ed. Helen Goodluck and Michael Rochemont, 223–258. Dordrecht: Kluwer. Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27: 77–138. Kozinsky, Isaac, Vladimir P. Nedjalkov, and Marija S. Polinskaja. 1988. Antipassive in Chukchee: Oblique object, object incorporation, zero object. In Passive and voice, ed. Masayoshi Shibatani, 651–707. Amsterdam: John Benjamins. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase structure and the lexicon, ed. Johan Rooryck and Laurie Zaring, 109–137. Dordrecht: Kluwer. Kuroda, S.-Y. 1972. The categorical and the thetic judgement. Foundations of Language 9: 153–185. Labelle, Marie. 2008. The French reflexive and reciprocal se. Natural Language and Linguistic Theory 26: 833–876. Levin, Beth. 1993. English verb classes and alternations. Chicago: University of Chicago Press. Lohndahl Terje. 2012. Without specifiers: Phrase structure and events. PhD diss., University of Maryland. Lohndahl, Terje. 2014. Phrase structure and argument structure: A case study of the syntaxsemantics interface. Oxford: Oxford University Press. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge, MA: MIT Press. Masullo, Pascual José. 1992. Antipassive constructions in Spanish. In Romance languages and modern linguistic theory, ed. Paul Hirschbühler and E.F.K. Koerner, 175–194. Amsterdam: John Benjamins. Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, MA: MIT Press. Ramchand, Gillian. 2008. Verb meaning and the lexicon: A first phase syntax. Cambridge: Cambridge University Press. Say, Sergei. 2005. Antipassive –sja verbs in Russian: Between inflection and derivation. In Morphology and its demarcations, ed. Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfeiffer, and Franz Rainer, 253–276. Amsterdam: John Benjamins. Schäfer, Florian. 2008. The syntax of (anti-)causatives. Amsterdam: John Benjamins. Spreng, Bettina. 2012. Viewpoint aspect in inuktitut. PhD diss., University of Toronto. Svenonius, Peter. 2002. Icelandic case and the structure of events. Journal of Comparative Germanic Linguistics 5: 197–225. Tatevosov, Sergei. 2013. Event structure of the anticausative and unaccusative in Russian. Paper presented at IATL 29. Travis, Lisa de Mena. 2010. Inner aspect: The articulation of VP. Dordrecht: Springer. Wiltschko, Martina. 2006. On ‘Ergativity’ in Halkomelem Salish. In Ergativity: Emerging issues, ed. Alana Johns, Diane Massam, and Juvenal Ndayiragije, 197–228. Dordrecht: Kluwer. Wood, Jim. 2012. Icelandic morphosyntax and argument structure. PhD diss., NYU.