Romance Languages and Linguistic Theory: Selected papers from 'Going Romance' Amsterdam 2007

Romance Languages and Linguistic Theory

Romance Languages and Linguistic Theory (RLLT) The yearly ‘Going Romance’ meetings feature research in formal linguistics of Romance languages, mainly in the domains of morphology, syntax, and semantics, and, to a certain extent, phonology. Each volume brings together a peer-reviewed selection of papers that were presented at one of the meetings, aiming to provide a representation of the spread of topics at that conference, and of the variety of research carried out nowadays on Romance languages within theoretical linguistics.

Volume 1 Romance Languages and Linguistic Theory. Selected papers from ‘Going Romance’ Amsterdam 2007 Edited by Enoch O. Aboh, Elisabeth van der Linden, Josep Quer and Petra Sleeman

Romance Languages and Linguistic Theory Selected papers from ‘Going Romance’ Amsterdam 2007 Edited by

Enoch O. Aboh Elisabeth van der Linden University of Amsterdam

Josep Quer Pompeu Fabra University

Petra Sleeman University of Amsterdam

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

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

Library of Congress Cataloging-in-Publication Data Romance languages and linguistic theory : selected papers from ‘Going romance,’ Amsterdam, 2007 / edited by Enoch O. Aboh ... [et al.]. p. cm. (Romance Languages and Linguistic Theory, issn 1574-552X ; v. 1) Includes bibliographical references and index. 1. Romance languages--Congresses. I. Aboh, Enoch Oladé. PC11.R637   2003 440--dc22 2009038954 isbn 978 90 272 0381 6 (Hb ; alk. paper) isbn 978 90 272 8870 7 (Eb)

© 2009 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents Foreword

vii

Tense domains in BP and EP – vP, CP and phases   Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

1

Variable-behavior Ps and the location of PATH in Old French   Heather Burnett & Mireille Tremblay

25

Hebrew and Arabic children going Romance: On the acquisition of word order in Semitic and Romance   João Costa & Naama Friedmann Adjectives and deleted nominals in Spanish   Luis Eguren On the nature of covert operations: The case of focus in Spanish pseudoclefts   Olga Fernández-Soriano Ellipsis and Restructuring in European Portuguese   Anabela Gonçalves & Gabriela Matos The early steps of modal and negation interactions: Evidence from child Italian   Andrea Gualmini & Vincenzo Moscati

51 67

87 109

131

Structural patterns blocking plural in Romance nominalizations   Gianina Iordăchioaia & Elena Soare

145

On the distribution of adjectives in Romanian: The cel construction   Mihaela Marchis & Artemis Alexiadou

161

Subject doubling in European Portuguese dialects: The role of impersonal se   Ana Maria Martins On the Quebec French interrogative particle tu   Annick Morin

179 201

 Romance Languages and Linguistic Theory 2007

Autonomous typological prosodic evolution versus the Germanic superstrate in diachronic French phonology   Roland Noske

223

Dummy prepositions and the licensing of null subjects in Brazilian Portuguese   Jairo Nunes

243

OV sequences in early child Catalan and English   Susagna Tubau

267

Index

287

Foreword From early on Going Romance became the reference conference in the area of formal linguistic research on Romance languages, both at the European and international level, and it has kept this status to date. Organized every year at a different university in the Netherlands, its continuity and quality standards are guaranteed by a steering committee formed by Romanists from all the participating institutions. This yearly meeting features research mainly in the domain of morphology, syntax and semantics and, to a certain extent, phonology too and it attracts speakers and attendees not only from Europe but also from the rest of the world (North and South America mostly). This volume assembles a significant number of selected papers that were presented at the 21st edition of Going Romance, which was organized by the Chair of Romance Linguistics of the University of Amsterdam and hosted in December 2007 by the Istituto Italiano di Cultura per i Paesi Bassi. The main two-day programme had no specific theme and it included 16 selected presentations, 3 alternate papers, as well as two invited addresses by Jairo Nunes (Universidade de São Paulo) and Dominique Sportiche (University of California, Los Angeles). The third day of the conference was devoted to a workshop with the topic ‘L1/L2 acquisition of word order and information structure in Romance’, with 6 selected papers, one alternate presentation and an invited talk by Marina Nespor (Università di Ferrara). The volume you have before you features 14 papers: 10 papers and one of the invited addresses from the main session, as well as 3 papers from the thematic workshop. As is common practice in the publication of volumes from Going Romance, all the submissions were thoroughly peer-reviewed by external referees and on the basis of those judgements the editorial team decided which ones to publish and asked for revisions where required. The result is quite representative of the spread of topics at the conference itself, but also of the variety of research carried out nowadays on Romance languages within theoretical Linguistics. Over the last three decades, the contribution of Romance to the range of empirical phenomena that underlie the research agenda has grown exponentially. Moving forward beyond the initial interest of syntacticians in pro-drop, subject inversion or clitics, for instance, linguistic theorizing has incorporated Romance ‘problems’ into virtually every single topic addressed in theoretical approaches to phonology, morphology, syntax, semantics, information structure and their multiple interfaces, as well as in the study of diachronic development, (monolingual and bilingual) acquisition, creoles, language contact, language attrition or language deficits. In many cases existing descriptions from

 Romance Languages and Linguistic Theory 2007

more traditional analyses have been revitalized, reinterpreted and incorporated into discussions of a very different nature from the original ones. In parallel, as a consequence of the methodology used, new data have been incorporated into the picture which shed new light on the problems and they are often put in connection with other sets of data that were previously considered to be unrelated. The interest for the differences within the established ‘languages’ of the same family has progressively drawn more and more attention to the dialects and spoken varieties of those languages as a window into microvariation. The range of languages (both standard and non-standard varieties) analyzed in this volume is also significant: Catalan, French, Italian, European and Brazilian Portuguese, Romanian and Spanish. Several contributions deal with rather different phenomena related to the syntax of subjects. Annick Morin discusses interrogative force in Quebec French. While standard French mainly expresses interrogative force by means of intonational patterns, Quebec French involves a question particle tu. At first sight, this particle could be regarded as a typical question particle that occurs sentence-finally. However, it exhibits a distribution similar to that of clitic doubling, and it appears to require definite subjects. The author proposes that the question particle tu and the pronominal subject clitic are endowed with the feature [Definite]. The two elements differ, however, in that the pronominal clitic tu is marked as [+definite; +interpretable], while the question particle tu expresses [+ definite; –interpretable]. The distribution of these two elements and the interaction with clitic doubling is further accounted for by assuming that the [Definite] feature on the pronominal clitic must agree with a [Definite] feature on the subject NP, in a Spec-head relationship at LF. Ana Maria Martins offers a ‘big-DP analysis’ for the impersonal se-construction in dialects of European Portuguese, whereby the clitic se is doubled by a strong pronoun or a DP. In her account, the clitic ends up in T and the pronoun or DP lands in a subject position. The doubling of the subject by se is possible because this element is ϕ-incomplete. Jairo Nunes provides further evidence for the status of referential null subjects in Bazilian Portuguese as traces of A-movement and not as null pronouns on the basis of their distribution in different types of noun complement clauses. The dummy preposition de ‘of ’ is argued to be the realization of inherent Case assigned to the embedded clause by the subcategorizing head. The basic logic of the analysis is extended to object control structures and to impersonal constructions involving an infinitival clause. Still dealing with syntactic varieties of Portuguese at the clausal level, Manuela Ambar, Esmeralda Negrão, Rita Veloso, and Luís Graça discuss variations in Brazilian Portuguese (BP) and European Portuguese (EP) in structures that involve Aux-to-Comp movement, Complementizer Deletion, verbal complex



Foreword 

gerunds, and adjunct gerunds licensing. The paper investigates to what extent the variations observed between BP and EP involve properties of the tense system, and how these properties correlate with other aspects of the two grammars in consideration. In this regard, the authors show that the variations are related to the presence versus absence of V raising to the CP field: BP has no V-to-C. It is shown that the verb does not move out of the first phase (i.e., vP) in this language. In EP, however, T has different specifications and the verb can move out of vP and subsequently target C. Three papers deal with ellipsis at the clausal level and in DPs. Anabela Gonçalves and Gabriela Matos derive the complementary distribution of Null Complement Anaphora (NCA) and Restructuring in European Portuguese from the selectional properties of the main predicates licensing them: verbs allowing for NCA select for a sentential complement with an active T domain, whereas Restructuring verbs select a defective T. Mihaela Marchis and Artemis Alexiadou also contribute to the discussion on ellipsis. They examine nominal ellipsis in Romanian with cel (cel frumos ‘the beautiful one’) and another cel construction, in which cel follows a lexically expressed noun (băiatul cel frumos ‘the boy cel nice’). For both cases they present an analysis of cel as involving an appositive specification clause, building on De Vries (2002). In line with Corver & van Koppen’s (2006) claim that nominal ellipsis is licensed by contrastive focus, Luis Eguren proposes that the restricted distribution of adjectives in Spanish elliptical DPs can be reduced to a contrastive focus condition on nominal ellipsis, whereby only adjectives that create Roothian sets of alternatives, giving rise to a partitive set-subset relation, can combine with an elided noun. Still within the domain of nominal syntax, Gianina Iordăchioaia and Elena Soare contribute with their paper to the ongoing discussion on nominalizations, see, e.g., Roodenburg (2006), Sleeman & Brito (in progress), and Alexiadou (2009). They show that the Romanian infinitive contradicts the generalization in Grimshaw (1990) according to which Complex Event Nominals do not pluralize, although the Romanian supine and the Spanish nominal infinitive obey this generalization. The authors explain the availability and the blocking of pluralization in relation to the internal structure of the Complex Event Nominal, more specifically in relation to the (un)boundedness of aspect. Besides Eguren’s paper there is another contribution that deals with focus, but this time at the clausal level. Olga Fernández-Soriano analyzes a specific type of pseudoclefts in spoken Spanish as an instance of monoclausal structure where the focus of the construction propagates after covert reconstruction of the clefted element into the relative CP, and she finds additional support for the proposed analysis in the so-called split interrogatives in the same language.



Romance Languages and Linguistic Theory 2007

Roland Noske’s paper is the only phonology paper in this volume. He argues that the traditional idea that in Early French the Frankish superstrate had a major influence on the phonology is mistaken. Noske proposes to consider the evolution of the entire prosodic system of the language to account for the Early French data. Heather Burnett and Mireille Tremblay investigate the interaction between lexical semantics and syntactic structure in the interpretation of prepositional phrases, concentrating on the expression of directionality in prepositions and particles in Old French (OF). This language has a series of locative and directional prepositions that may be interpreted to encode direction or aspect when used intransitively. The authors show that this semantic variation can be accounted for in a principled manner by assuming that the distinct meanings stem from the syntactic configurations in which the prepositions occur. The discussion here reminds us of the ongoing debate on the structure of prepositional phrases (Svenonius to appear) which itself has to be anchored in the context of the general discussion of argument structure (e.g., Burzio 1986; Hale & Keyser 1993). Acquisition of word order and phenomena related to information structure was the topic of the thematic workshop and three of those contributions are included in this volume. Two of them concentrate on word order acquisition and the third one on the interpretation of scope bearing elements. João Costa and Naama Friedmann studied the acquisition of word order in two Semitic languages (Palestinian Arabic and Hebrew) and two Romance languages (Spanish and European Portuguese). In all four languages, SV as well as VS word order is possible in the adult variety, but these options are not used to the same amount in all languages by the children studied: in the first stage, Hebrew and European-Portuguese children use both SV and VS orders with unaccusative verbs, but only SV with unergative and transitive verbs. In Spanish and Palestinian-Arabic, on the other hand, children prefer VS order with unaccusative, unergative, and transitive verbs. The authors argue that children at this stage cannot move the subject outside of VP yet, and the cross-linguistic difference stems from the identification of Spell-Out Domains (Fox & Pesetsky 2004). Spanish and Palestinian-Arabic allow the verb to appear before the subject, whereas Hebrew and European-Portuguese do not allow the verb to move to I until the Spell-Out domain widens beyond VP, which takes place after subject movement to Spec,IP is acquired. Susagna Tubau also tackled word order, comparing OV Sequences in Early Child Catalan and English. OV sequences are rare in both VO languages but they do occur between approximately 19 months and 2 years of age. Tubau claims that these OV sequences emerge in child English and Catalan after Telicity Phrase and Aspect Phrase have been acquired but are still underspecified, thus not forcing movement of objects and verbs to Tel0 and Asp0 respectively. While V-to-Asp0



Foreword 

without object-to-TelP yields VO word order, the opposite yields OV. For child Catalan, Tubau distinguishes ‘true’ OV sequences, which occur with non-finite and non-adult inflected verbs, from Focus fronted OV combinations, which are attested with adult-like inflected verbs and whose syntax is different from ‘true’ OVs. In their paper, Andrea Gualmini and Vincenzo Moscati report on the early steps of modal and negation interactions in child Italian. In their experimental research, they looked at children’s interpretation of sentences containing an ambiguity between different scope-bearing elements, namely negation and a modal verb in Italian, as in Every horse didn’t jump over the fence. The results of their study suggest that Italian-speaking children differ from Italian-speaking adults in that they choose strong inverse scope interpretations rather than weak surface scope interpretations. This brief rendering of the contents of the contributions published here proves the wealth of topics, theoretical approaches, analytical solutions and empirical coverage that Going Romance stands for, as well as the vitality of the research carried out on Romance languages within theoretical Linguistics. The editors of this volume would like to thank here Margot Rozendaal, Mauro Scorretti and Fred Weerman, who were members of the local organizing committee too, and Mike Olson for proofreading and editing the final manuscript of this book. We are also specially grateful to Mr. S. Marchetti, director of the Istituto Italiano di Cultura per i Paesi Bassi in Amsterdam, who generously offered us the venue and the infrastructure for the conference. Financial support from ACLC (Amsterdam Centre for Language and Communication, University of Amsterdam) and KNAW (Royal Dutch Academy of Sciences) made the event possible and we are grateful for that too.

References Alexiadou, Artemis. 2009. “On the Role of Syntactic Locality in Morphological Processes: The case of (Greek) derived nominals”. Quantification, Definiteness and Nominalization ed. by Anastasia Giannakidou & Monika Rathert, 253-280. Oxford: Oxford University Press.  Burzio, Luigi. 1986. Italian Syntax, A Government and Binding Approach. Dordrecht: Reidel. Corver, Norbert & Marjo van Koppen. 2006. “Let’s Focus on Noun Ellipsis”. Glow Newsletter 57. Fox, Daniel & David Pesetsky. 2004. Cyclic Linearization and the Typology of Movement. Ms., MIT. Hale, Kenneth & Samuel Jay Keyser. 1993. “On Argument Structure and the Lexical Expression of Syntactic Relations”. The View from Building 20 ed. by Kenneth Hale & Samuel Jay Keyser, 53–109. Cambridge, Mass.: MIT Press. Roodenburg, Jasper. 2006. “The Role of Number within Nominal Arguments”. Paper presented at the 36th Linguistic Symposium on Romance Languages. Rutgers University, 31 March-2 April, 2006.

 Romance Languages and Linguistic Theory 2007 Sleeman, Petra & Ana Maria Brito. In progress. “Nominalization, Event, Aspect, and Argument Structure: a Syntactic approach”, To appear in Argument Structure from a Crosslinguistic Perspective, ed. by Maia Duguine, Susana Huidobro & Nerea Madariaga. Amsterdam & Philadelphia: John Benjamins. Svenonius, Peter. To appear. “Spatial P in English”. To appear in Cartography of Syntactic Structures, vol. 6 ed. by Guglielmo Cinque & Luigi Rizzi. Oxford: Oxford University Press. Vries, Mark de. 2002. The Syntax of Relativization. Ph.D. dissertation, University of Amsterdam.

Tense domains in BP and EP – vP, CP and phases Manuela Ambar1, Esmeralda Negrão2, Rita Veloso3 & Luís Graça4 1CLUL/2USP/3CLU/4CLUL

In this article we will deal with intriguing oppositions in Brazilian and European Portuguese in structures involving Aux-to-Comp, Complementizer Deletion and verbal complex gerunds versus adjunct gerunds licensing. The objectives of this paper are twofold: (i) to investigate to what extent those oppositions involve properties of the tense system and how these properties correlate; (ii) to understand whether the variation observed is or is not co-variation, identifying the properties responsible for it. We will posit that the oppositions described are the effect of the presence versus absence of V raising to the CP field in EP versus BP. Going back to work by Den Besten (1977) and following Ambar (2005, 2007), we will assume that the understanding of this phenomenon is to be seen in properties of Tense, conspiring for the following working hypothesis: BP has no V-to-C because V doesn’t move out of vP, EP has V-to-C because V can move out of vP.

1.  Introduction Brazilian and European Portuguese (henceforth BP and EP) exhibit intriguing oppositions in structures involving the following phenomena: Aux-to-Comp, Complementizer Deletion (C-Deletion), and verbal complex gerunds vs. adjunct gerunds licensing. These oppositions are exemplified under (i)–(iii) below, respectively: (i) Although both in BP and EP inflected infinitival structures may occur as complements to certain matrix predicates, only in EP are they fully acceptable in epistemic or declarative contexts:1 (1) Penso terem os deputados votado a proposta. I.think to.have.3pl the deputies voted the proposal “I believe the deputies to have voted for the proposal.”

(ok EP; *? BP)

1.  In BP, although few speakers could accept the infinitive form with realized lexical subjects, the incontestable acceptable form of complement clauses to epistemic matrix predicates is a tensed sentence introduced by the complementizer que.



Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

(ii) Although in EP, as in Italian, C-deletion is limited to embedded clauses selected by a restricted set of verbs, in BP no cases of C-deletion are allowed: (2) a.

Solicito me seja concedida uma bolsa.ˉ I.request to.me be-subj-3sg given a subsidy “I request to be given a subsidy.”

b. Credo Gianni abbia telefonato. I.believe Gianni have-subj phoned “I believe Gianni to have phoned.”

(okˉEP;ˉ*ˉBP)

(Cocchi & Poletto 2002)

(iii) although in BP the gerund is productive in verbal complexes such as (3) and (4), adjunct gerundives of the type (5) are not; inversely, in EP the gerund is productive in the later structures, whereas in standard EP prep+infinitive is preferred to the gerund in the constructions illustrated in (3) and (4): (3) a.

O Marcelo está estudando Medicina.  Marcelo is studying Medicine “Marcelo is studying Medicine.”

b. O Marcelo está a estudar Medicina.  Marcelo is to study Medicine. “Marcelo is studying Medicine.” (4) A Ana vai estar vindo na USP a semana inteira.  Ana goes to.be coming to USP the week whole “Ana will be coming to USP everyday for the whole week.” (5) Sabendo ele da história, ficou calado.  knowing he of.the story, kept quiet “Knowing about the story, he didn’t say anything.”

(*?EP; ok BP)

(ok EP ; * BP)

(*EP, ok BP)

(ok EP; *? BP)2

Table 1 below summarizes the contrasts described above: Table 1. Phenomena

EP

BP

Infl. Infinitives under epistemic/declarative verbs C-Deletion Gerund in adjunct clauses

Yes

No

Yes Yes

No No

The objectives of this paper are twofold: (i) to investigate to what extent those oppositions involve properties of the tense system and how these properties correlate and (ii) to understand whether the variation observed is or is not co-variation, identifying the property or properties responsible for it. We will posit that the 2.  We adopted the convention of marking with *? the sentences that are less productive in the language in question.



Tense domains in BP and EP – vP, CP and phases

variation above, actually co-variation, is the effect of the presence versus absence of V raising to the CP field (V-to-C) in EP versus BP. The next goal is to discover what underlying property allows or excludes V-to-C across languages. Going back to work by Den Besten (1977) and following Ambar (2005, 2007), we will assume that the understanding of this phenomenon is to be seen in properties of Tense, conspiring for the following working hypothesis: BP has no V-to-C because V doesn’t move out of vP, EP has V-to-C because V can move out of vP. The paper is organized as follows: Section 2 deals with the Aux-to-C phenomenon, Section 3 with C-Deletion, Section 4 with gerundives licensing. In Section 5 we outline a view positing that vP, in its relation with Tense, seems to be the locus of the parameterization between the two languages, in Section 6 we make some brief concluding remarks.

2.  Aux-to-Comp As largely described in the literature (Rouveret 1980; Negrão 1986; Raposo 1987; Ambar 1988, 1992; Figueiredo Silva 1996; Galves 2001), inflected infinitival structures in EP occur in sentential complements to different classes of verbs (factive, causative, perception, epistemic and declarative predicates, though not volitional predicates). However, its occurrence in sentential complements to epistemic and declarative verbs is submitted to the condition of subject verb inversion, interpreted by Raposo (1987) as the result of Infl raising to C pied-piping the verb: (6) a. *Penso os deputados terem votado a proposta. think the deputies have-3pl voted the proposal “I believe that the deputies have voted for the proposal.” b. Penso terem os deputados votado a proposta. think have-3pl the deputies voted the proposal “I believe that the deputies have voted for the proposal.”

Rizzi (1982) identified this type of strategy as the Aux-to-Comp phenomenon, available in Italian structures like (7a–c) (Rizzi 1982: 79–80): (7) a.

Suppongo non esser la situazione suscettibile di ulterior suppose not to.be the situation susceptible of further



miglioramenti. improvements



“I believe that the situation is not susceptible of further improvements.”

b. Mario afferma non esser lui in grado di affrontare la situazione. Mario asserts not to.be he/ him able to face the situation “Mario asserts not to be able to face the situation.”





Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

c.

Questa commissione ritiene [aver loro sempre this commission believes to.have they/them always



ottemperato agli obblighi previsti dalla legge] obeyed to.the obligations anticipated by.the law



“This commission believes that they have always fulfilled the legal duties.”

The deepest understanding of this movement is far from being clear. Both in Rizzi’s and in Raposo’s analyses, Aux-to-Comp is motivated by Case. In Rizzi (1982), Aux in Comp is able to assign Case to the subject whereas it is unable to do so when it remains in situ. In Raposo (1987), Infl (=Agr), pied piping the verb, moves to Comp, because Agr, being nominal, has to be assigned Case by the matrix predicate, the only Case assigner available given his assumption that infinitival structures don’t have Tense. The first opposition between BP and EP we will deal with concerns this phenomenon. The acceptability of sentence (1) in EP contrasts with the ungrammaticality of its equivalent in BP (Negrão 1986; Figueiredo Silva 1996). Below we repeat (1) as (8b): (8) a. Penso os deputados terem votado a proposta. (*EP, *? BP) b. Penso terem os deputados votado a proposta. (ok EP, *? BP)

Assuming Raposo’s analysis, one could think of the locus of parameterization between BP and EP as being Agr. In fact, linguists working with the syntax of BP (Galves 1993; Duarte 1995; Figueiredo Silva 1996; Negrão 1999; and Negrão & Viotti 2001, a.o.) have shown that this language is going through a parametric change in its Agr specifications;3 thus, it would be plausible to assume that the requirement on Case marking for Agr proposed by Raposo does not operate in BP (a non null subject language); if this were the case Agr raising to C would be unnecessary, therefore forbidden in minimalist terms. In this view, absence of Aux-to-Comp in BP would be related to its status of not being a null subject language.4 We wouldn’t however clarify the ungrammaticality of (8a) in BP: being insensitive to Case, Agr in situ should be legitimate in sentence (8a).5 Moreover, this hypothesis wouldn’t say

3.  Some agreement marks have disappeared from the BP inflection paradigm, e.g. loss of 2nd person singular. 4.  One of the reviewers asks for ‘a list of phenomena related to […] the parametric change in the Agr specification’ in BP. In fact, BP exhibits the cluster of properties typical of non null subject languages, e.g. lack of pro-drop, weakness of inflection, loss of subject inversion, including the so-called ‘free inversion’. 5.  The same reviewer “believes this distinction between BP and EP [the possibility of inverting the subject] does not play a role in these sentences” and suggests that “one might hypothesize



Tense domains in BP and EP – vP, CP and phases

anything on the other two phenomena we will concentrate on, namely the availability versus unavailability of C-Deletion in EP versus BP, respectively and the different behaviours of gerundive constructions in each language. Regarding the former, Agr seems not to be the core issue involved in C-Deletion: on the one hand C-Deletion is also available in non null subject languages, such as English; on the other hand what causes C-Deletion seems to be somehow associated with properties of C in connection with properties of Tense/Mood, rather than with Agr. This is why in languages of the EP or Italian type, C-Deletion requires the presence of the subjunctive, whereas in other languages it doesn’t, e.g. English, Florentine, (Cocchi & Poletto 2002). As for the second, in standard Portuguese gerundives have no Agr specifications and the relevant restrictions seem to concern Tense, Aspect or Aktionsart as well, as will become clearer in what follows (Lobo 2001, 2003, 2007). Thinking of the general ban on V-to-C (T-to-C) in other Brazilian Portuguese constructions, e.g. wh-questions, focus, etc. (Ambar 1996b, 2001; Ambar & Veloso 1999; Ambar & Pollock 2002; Mioto 2004; Mioto & Kato 2006; Kato & Raposo 2007), it is tempting to keep the idea that V-to-C (I-to-C) is at stake in Aux-to-Comp, as argued by Rizzi and Raposo. If this were the case, the absence of inflected infinitives in complements to epistemic predicates in BP would be nothing but another manifestation of the absence of V-to-C (T-to-C) in this language, in spite of their apparent disconnection. We will then shift the focus from Agr to Tense, assuming that, in inflected infinitival clausal complements to epistemic verbs, the embedded verb moves to C, with restrictions recalling observations on the boundaries of V movement in other contexts (Emonds 1976; Pollock 1989; Enç 1991; Haegeman 1996; Ambar 1988, 1994, 1998, a.o.).6 In our view Aux-to-Comp is the reflex of universal requirements on tense valuation, rather than a language specific operation to solve licensing of

that sentences such as (1) are impossible in BP because of inversion and sentences like (8) are impossible because the Agr feature in inflected infinitives is located in C, so the subject in (8) has no Case”. First, in the spirit of Chomsky (2001), we believe that Case is Tense-related, rather than Agr-related and that Agr doesn’t exist as an independent category. We also assume that, in the spirit of Chomsky (2001, 2005), Case is valued when Tense is φ-complete (cf. Chomsky 2001: “α must have a complete set of φ features (it must be φ-complete) to delete uninterpretable features of the paired matching element β”). Second, our main point is to discover the underlying property of the phenomenon ‘inversion’ in relation to tense properties which oppose this class of verbs to others selecting inflected infinitival clauses where however inversion is not required, e.g. factive verbs, rather than disconnecting the phenomena. 6.  Rizzi (1982: Fn.10) already notes a threshold of acceptability for verb raising to C. Ambar (1988, 1994, 1998) considered that what causes the ungrammaticality is the habitual vs. non





Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

Case in a particular context, which is why, throughout the paper, we use the labels Aux-to-Comp or V-to-C to mean ‘verb movement’ (with or without restrictions on the verbs that raise to C). Let us come back to sentences in (8). Ambar (1988) observed that inflected infinitival clauses with SV order under epistemic and declarative verbs are possible in EP if the subject is focused, as illustrated by (9) below:7 (9) a.

Disseram-me [só eles terem visto esse filme] told-3pl-me only they to-have-3pl seen that movie “I’ve been told that only they saw that movie.”

b. Penso ELES terem votado a proposta.8 think they to-have-3pl voted the proposal “I believe they voted for the proposal.”

At first glance it seems that in (9) the focus strategy saves the structure dispensing with inversion. However, as pointed out by Raposo (l994: 40) “the possibility of material occurring before the inflected infinitive is much more general” and “the phenomenon is not restricted to subjects.” Some of his examples: (10) Disseram-me [muita gente ter visto esse filme] told-3pl-me many people to-have seen that movie “I’ve been told that many people saw that movie.” (11) a.

Disseram-me [nada terem esses turistas visitado] told-3pl-me nothing to.have-3pl those tourists visited “I’ve been told that those tourists didn’t visit anything.”

b. Disseram-me [só essa cidade terem os turistas visitado] told-3pl-me only that city to.have-3pl the tourists visited “I’ve been told that the tourists have only visited that town.”

Notice that inversion is required, when the focused phrase is different from the subject as in (11), the equivalents of these sentences without inversion being bad. We conclude that inversion is dissociated from focus in these structures, i.e. inversion is plausibly needed to value the structure, say Tense and Case, in a way which will become clearer in what follows, and Focus-movement is necessary to value

habitual reading, with eventive and stative verbs, rather than the auxiliary-non-auxiliary status of the verb, as also independently observed by Enç (1991) for English. 7.  We thank a reviewer for having suggested that we develop this part of the paper involving the left periphery, which in the first version was nothing but a short footnote. We also thank her/him for other very good comments. 8.  The capital letters translate a special intonation associated with the focus interpretation resulting from movement of the focused phrase to Spec,FocusP (cf. Ambar 1996b for the description of this interpretation).



Tense domains in BP and EP – vP, CP and phases

Focus. In the line of work by Rizzi (1997), Ambar (1996b) a.o., we will assume that the focused element in sentences (9)–(11) moves to Spec,FocusP, a projection located above IP (TP) and below CP. But then we face a problem: on the assumption that FocusP is located above IP (TP) and below CP, if the verb in the inflected infinitive raises to C and the focused phrase moves to Spec,FocusP, as we have been assuming for paradigm (9)–(11), the desired word order will be not derived. Throughout we have been calling CP, the projection where the verb moves to, the locus of Tense and complementizers. Obviously, considering a finer structure of the left periphery, CP has to be redefined.9 We will roughly assume that Tense is valued in a position of the type Fin(iteness) in Rizzi’s (1997) left periphery proposal, i.e. a position located above TP and below FocusP: (12) …..[FocusP só elesi [ Fin teremk [TP ti tk visto o filme ]]]

Clearly, something else has to be said with respect to overt complementizers, which appear in different positions across structures and languages.10 Further evidence for the hypothesis illustrated by structure (12) comes from BP. As expected, if we are right, the BP sentences equivalent to (9)–(11) should be out. And they are. Given that BP has focus structures like: (13) Só os deputados votaram a proposta. only the deputies voted the proposal “Only the deputies have voted for the proposal.”

The ungrammaticality of (9)–(11) is not derivable from Focus. They are excluded because the verb in the inflected infinitive cannot move to Finite to license Tense and Case, due to the ban against V-to-C in BP. A reviewer asks whether the subject could “occupy a lower position than T”. Barbosa (2001) used Ambar and Raposo’s observations to argue in favour of her hypothesis according to which the subject is inside VP in EP, focused phrases are

9.  For the structure of the left periphery dealing with phenomena like Focus, Evaluative structures, wh-exclamatives, wh-questions and Adverbs in Portuguese compared to other languages, see (Ambar 1996b; Ambar & Veloso 1999; Ambar et al. 2004). Of course we cannot go through these phenomena here, but they also provide evidence for this analysis. 10.  See complementizers occurring in wh-exclamatives in languages of the EP type, or in wh-questions, either preceding the wh-phrase (as is the case of embedded wh-questions in Hungarian) or following it (as is the case of wh-questions in BP). Further evidence for the need of different layers for complementizers is found in the ‘recomplementation’ or Double C phenomenon, discussed in (Uriagereka l995) and in (Raposo 1994, 1996; Cocchi & Poletto 2002; Manzini & Savoia 2003, a.o.), where a given element, plausibly a topic, may be sandwiched between two complementizers.





Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

in Spec,IP, an A-position, and CLLD phrases are adjoined to IP, an A’-position11 – (14a–b) below are taken from Barbosa (2001: 60): (14) a. Pensava [IP [só os deputados]k [I´ terem tk aprovado essa thought only the deputies to-have-3pl approved that “I thought that only the deputies approved that proposal.” b. ?? Pensavaˉ[IPˉ[IP os deputadosˉ[IP teremˉpro aprovado esse thought the deputies to.have-3pl approved that “I thought that the deputies approved that movie.”

proposta]] proposal filme]] movie

She explains: (14b) “violates the ban against adjunction to an argument, so it has a dubious status; (14a) doesn´t involve adjunction, so it is fine”. In her argumentation there is a crucial assumption: “Portuguese doesn’t have Focus movement of definite DPs (or of sentential adverbs) unlike the other Romance languages”. Although we cannot go through this discussion here, let us say that, on the contrary, Raposo (1994), Ambar (1996b) have argued for movement of operator phrases like those in (9a) and (10–11) to a left periphery position. As for definite DPs or sentential adverbs in EP, Ambar (1996b) has also argued for movement to Spec,FocusP. On the other hand, discussing what she considers to be a problem for her analysis – absence of inversion in clausal complements to factive verbs – Barbosa assumes that in these structures CLLD phrases different from the subject can adjoin to the embedded clause; thus, absence of inversion in this context follows. However the author doesn’t say why factive verbs behave differently from epistemic ones with respect to this phenomenon. Note that evidence drawn from “the position occupied by adverbs”, as asked by the reviewers, is hard to test, first because both in EP and BP those positions are quite free (apparently), second because if we don’t assume that in the unmarked case adverbs are adjoined to VP, and we rather adopt a view of the type of Cinque (1999), Ambar (1989) adverbs are generated in different positions according to the heads they are related to. Anyway, there are some adverbs whose positions seem to be more restricted, as is the case of bem (well), in its ‘manner’ interpretation: (15) a. *Penso terem bem eles lido o livro. think to.have-3pl well they read the book b. *Penso terem eles bem lido o livro. think to.have-3pl they well read the book c.

Penso terem eles lido bem o livro. think to.have-3pl they read well the book “I believe them to have read the book well.”

11.  For relevant discussion on the topic see (Duarte 1987 and Costa 1996).



Tense domains in BP and EP – vP, CP and phases

Suppose ‘manner’ adverbs are adjoined to VP, as generally assumed, and that the past participle has moved out of VP, plausibly to I (T), as (15c) suggests. If the subject is in Spec,IP and terem has raised to C (Finite), only (15c) is derived, as desired. Take now Barbosa’s representation in (16): (16) [ IP [ I’ V [ VP subject t ]]]

where terem is under inflection, bem is adjoined to VP and eles inside VP: (15c) is never derived, neither with PP raising nor without it. Let us turn to the mechanism underlying V-to-C (Fin). What follows is taken from different works by Ambar on Tense valuation and sentence structure, crucially Ambar (2005).12 Recall that the requirement on inversion in epistemic and declarative13 contexts is not present in other classes of matrix predicates, e.g. factives. However, the peculiarity of this behaviour is only apparent. In fact, other phenomena opposing epistemic verbs to others conspire for the conclusion that the opposition ‘inversion’ correlates with other differences in a principled way. Table 2 summarizes the description of those phenomena presented in Ambar (1998, 2005): Table 2.

Independent binding domain in embedded clausal complements Selected category Morphological tense independency Selected mood Clitic Climbing Obviation Referential DPs selection Inflected Infinitives selection Definite determiner + inflected infinitive

Epistemic

Factives

Volitional

Yes

Yes (restrictions)

No

CP Yes

DP_CP No (restrictions)

CP No

Indicative No No No Yes No

Subjunctive No No (restrictions) No Yes Yes

Subjunctive Yes Yes Yes No No

Ambar (2005) claims that the cluster of phenomena above provides empirical evidence for the hypothesis that in clausal complements to epistemic verbs

12.  Ambar (2005, 2007) reinterprets in the probe-goal system of the minimalist framework Ambar’s (1988, 1992, 1994, 1996a, 1998) proposals, where sentence structure was already seen as the effect of Tense (and Agreement) valuation, through tense features in C. 13.  Henceforth, for economy, we will refer to epistemics as covering also declaratives.



 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

C (whatever C is) bears a valued t-feature (tense feature) and an unvalued tev-feature (ev=Event),14 assuming Chomsky’s conception of bundles of features. The verb has to raise to value that unvalued tev-feature. The verb in (inflected) infinitives has no valued t-feature - infinitives have no tense morphology; the auxiliary can however value the unvalued tev-feature in C, it has event properties, plausibly inherited from vP. In factive constructions, there is no need of inversion because these verbs select DP, which in turn selects CP. In factive contexts D plays over the role of verb raising in epistemic ones, i.e. D has a valued tev-feature that values the unvalued tev-feature of C, everything else being equal. As for clausal complements to volitionals predicates, the author proposes that, contrary to epistemics, C has a valued tev-feature and an unvalued t-feature. Verb raising however does not value the t-feature (it values the tev-feature, see epistemics above). Therefore, the t-feature can only be valued by the matrix tense, deriving the dependence of tense and all the transparency effects, such as clitic climbing, which led to the restructuring hypothesis, a process that converts a bi-clausal structure in a monoclausal one, first put forth by Rizzi 1976.15 Ambar (2005, 2007) concludes that CPs and DPs embedded to epistemic and to factive verbs are t- and φ-complete, i.e. they are Phases. CPs under volitonals are not Phases. Suppose this is so, abstracting from evidence the author presents drawn from other structures, e.g. control, ECM, raising, subjunctive clauses.16 Returning to the oppositions between BP and EP, observe that the unacceptability of (8a–b) above in BP is rescued if the embedded subject is wh-moved as in (17b): (17) a. b.

*Que meninos você acredita sonharem? which boys you believe to.dream-3pl “Which boys do you believe to dream of?” Que meninos você acredita terem sonhado? which boys you believe to.have-3pl dreamt “Which boys do you believe to have dreamt of?”

14.  Event features cover properties of Tense object (TobjP in Ambar 1996a, 1998 and Pesetsky & Torrego 2001, 2004), Aktionsart, object determination and other vP properties, which in part have been called Aspect (see those works for discussion). 15.  After Rizzi (1976, 1982) different proposals have been made, (a.o. Kayne 1989; Bok-Bennema & Kampers-Manhe 1994; Rooryck 1994; Cinque 2004; Bošković 1997; Ambar cf, Fn.12). Although restructuring is a crucial issue for our motivation on tense valuation and clause structure, also covering Aux-to-Comp, we cannot deal with it here. 16.  See for instance the rich array of facts described in (Manzini 2000) for the subjunctive.



Tense domains in BP and EP – vP, CP and phases 

Equivalent contrasts have been discussed in the literature for French: (18) a. b.

*Pierre croit Marie avoir acheté des fraises. Peter believes Mary to.have bought strawberries “Peter believes Mary to have bought strawberries.” Qui Pierre croit-il avoir acheté des fraises ? who Peter believes-he to.have bought strawberries “Who does Peter believe to have bought strawberries?”

Kayne (1984) makes the proposal that in (18b) the wh-phrase is assigned Case in an A’-position while undergoing wh-movement. We will reinterpret Kayne’s assumption in terms of tense valuation, combining it with Pesetsky & Torrego’s (2004) proposal that Case is T on the DP: while undergoing wh-movement, the wh-phrase gets its t-feature valued in the higher domain; the t-feature in the embedded C (Fin) gets valued through the trace of the wh-phrase. As for the contrast in (17) in BP, due to the presence versus absence of the auxiliary, note that: (i) it is not different from the ones resulting from restrictions on verb raising, which have led to the Aux-to-Comp phenomenon, (ii) it also shows up in French. We adopt Ambar’s analyses on the issue:17 auxiliaries and stative verbs can move to C; eventive verbs can move to C if their single event reading (nonhabitual) shifts into a generic/habitual one. Ambar’s key to solve the problem is found in considering that Event, Tense object, Atktionsart, object determination, i.e. properties codified in the lower portion of the clause – vP and TobjP (or the so-called AspP) – play an important role. The restrictions result from the relation TP establishes with those properties in vP. Summing up, as described above, we will roughly assume here Ambar’s (2005) proposal: C has a tev-feature in need of being valued, verb raising is needed to value that feature. After verb movement to C all features, φ-features included, are valued in the embedded domain, which is a phase. Being a phase, no transparency

17.  Déprez (1989) relates the phenomenon to Aux-to-Comp; observing that Aux-to-Comp is not otherwise attested in French, Bošković proposes that the restrictions result from Enç’s (1991) claim: eventive, unlike stative predicates, have an event variable which needs to be bound; tense, adverbs of quantification and aspectual auxiliaries are anchoring elements for that variable, combined with his own proposal: I is [–tense] in this context. But having assumed that infinitival clauses are IPs he has to stipulate that, exceptionally in this context, infinitival clauses are CPs. He also stipulates that for cases where in French the nonhabitual reading is possible, I is [+tense]. Ambar’s key to solve the problem is found in vP or in the so called Aspect (cf. Fn 12). The reader will find the argumentation in the quoted works.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

effects will show up, e.g. clitic climbing will not occur, the embedded tense will be not dependent on the matrix one.18 As already observed, V-to-C movement is productive in EP in different contexts whereas in BP it is not (see references under (8)). If there is no V-to-C in BP the ungrammaticality of (8)–(11) in BP is expected: in (8a), (9) and (11) there is an unvalued tev-feature in C, which has not been valued; in (8b) and in (11) an improper movement of V-to-C (Fin). The following step would be to understand why V-to-C movement is or is not available across languages. Discussing Vikner’s (1997) condition on V-to-I movement, according to which a language has V-to-I movement if and only if person morphology is found in all tenses, i.e. if INFL is strong, Aboh (2000) convincingly argues that Gbe languages provide evidence for the claim that “verb movement is driven by the strong features of INFL which need not be overtly realised as inflectional or agreement affixes in the language”. Furthermore, interesting data lead Aboh to conclude that Gbe-languages involve two groups: one allowing V-to-Asp movement, another allowing V-to-T movement, which provides a straightforward account of word order variation in these languages. Aboh’s work clearly shows that, across languages, the lower area of the clause structure (AspP) may be more active than the higher area of the clause (TP) and vice versa. Although with poorer empirical evidence, Ambar (1998) has argued for a similar view. Turning to BP, it is tempting to propose that only short V movement is available, just like on the V-to-Asp type languages described by Aboh. Implementing the proposal in the minimalist probe-goal system, we will hypothesize that in

18.  Updating Ambar (1998), sentential complements to volitional predicates are not phases, because they are not tense-complete, predicting that clitic climbing and dependent tense may occur. Assuming, however, that Stowell (1982) is right about the independent temporal denotation of infinitives under volitionals, which led to the assumption that these clauses are specified for [+tense]. Ambar (2005) proposes that the key to solve the paradox resulting from the need of specifying these domains as [+tense] to cover Stowell’s observation, and [–tense ] to cover morphological tense dependence, clitic climbing and other transparency effects must be seen in properties of event, Aktionsart, TobjectP, or the so-called Aspect. C under volitional predicates would have a cluster of features (a bundle of features): some valued (event features), others unvalued (tense features- morphological tense features). This predicts that the embedded domain of volitional predicates are not phases, as well as Stowell’s observation: event-features being valued in the embedded C would be responsible for the independent ‘unrealized’ temporal interpretation described by Stowell, the unvalued (morphological) tense-feature being valued in the higher domain would account for the dependent Tense of these structures.



Tense domains in BP and EP – vP, CP and phases 

BP the verb doesn’t move out of vP, where all features are valued. Different phenomena conspire for this working hypothesis.19

3.  C-Deletion C-Deletion apparently poses a problem to the minimalist theory, since the phenomenon consists in the apparent optional choice between the presence versus absence of a lexically realized complementizer.20 The phenomenon occurs in different languages and in different forms, e.g. English, unlike EP, has C-Deletion in the context of Indicative. Pesetsky and Torrego (2004) offer the hypothesis illustrated in (19) below for C-Deletion in English: (19) a. Mary thinks Sue will buy the book b. Mary thinks [CP [Sue ----] [C, -----] [IP _ [Tns iT +fut] buy the book]]

In (19), C contains the uninterpretable, unvalued tense-feature with which it was endowed in the Lexicon. This feature can be valued in two ways: through agreement with T, which raises under the form of that, or through T on the subject, which raises to CP as in (19b).21 In languages of the EP type, however, C-Deletion is restricted to subjunctive clauses, under certain predicates: (20) a.

A Maria pensa que a Joana comprará o livro. Mary thinks that Joan will.buy the book “Mary thinks that Joan will buy the book.”

b. *A Maria pensa a Joana comprará o livro. Mary thinks Joan will.buy the book “Mary thinks Joan will buy the book.”

19.  See for instance the last section of this paper. A reviewer asks for evidence drawn from the behaviour of adverbs. On this, see discussion on (15) in the text. Anyway, there is a tendency: where, for given adverbs, BP prefers preverbal positions, EP prefers post-verbal ones (see Ambar, Negrão & Gonzaga 2004). 20.  Note that only that-type complementizers (e.g. que in BP, EP, French, che in Italian, that in English, etc.) can undergo C-deletion, interrogative complementizers cannot. 21.  Duarte et al. (2005) adopt Pesetsky & Torrego’s analysis for EP.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

(21) a.

Solicito que seja concedida uma bolsa a este aluno. ask that be-subj given a grant to this student “I ask this student be given a grant.”

b. Solicito seja concedida uma bolsa a este aluno. Ask be-subj given a grant to this student “I ask this student be given a grant.” (22) a.

Quero que seja concedida uma bolsa a este aluno. want that be-subj given a grant to this student “I want a grant to be given to this student.”

b. *Quero seja concedida uma bolsa a este aluno. want be-subj given a grant to this student “I want a grant to be given to this student.”

As shown in sentences (20)–(22) C-Deletion is ruled out from indicative contexts (20b) and allowed in subjunctive ones, under predicates of the solicitar type (21b), though not of the querer type (22b), for example. In BP C-Deletion is banned from all contexts; neither subjunctive, nor indicative allow for it. Thus, sentence (21b) is out in BP. Benincà (1994), Poletto (1995), Giorgi & Pianesi (2000), Cocchi & Poletto (2002) bring to light different aspects of C-Deletion observed in standard modern Italian, in the history of Italian and on Italian dialects. Cocchi & Poletto (2002, 2007), concentrating on two types of C-Deletion available in the Italian dialect Florentine, propose an analysis based on two assumptions: Italian has a split left periphery and different elements may be ‘alternative checkers’ of the same feature located in the CP area, as proposed by Obenauer (2004). These ‘alternative checkers’ may be either an inflected verb with modal morphology, or a clitic, or negation, or an infinitival verb or a relative operator. Standard Italian only allows the former type, which is available for all the varieties of C-Deletion, as in (23): (23) a.

Credo che abbia già parlato con te. think that have-subj already spoken with you “I think that he has already spoken to you.”

b. Credo abbia già parlato con te. think have-subj already spoken with you “I think he has already spoken to you.”

As observed by the authors, in standard Italian C-Deletion is restricted to contexts where the main verb is a bridge verb and, as in EP, C-Deletion does not occur with the indicative (25), contrary to English: (24) a.

Credo lo faccia. think it do-subj “I think it does it.”



Tense domains in BP and EP – vP, CP and phases 

(25) a. *Ha detto viene. has said comes “He said he is coming.”

C-Deletion in Florentine is less restricted: the selecting verbs don’t have to be bridge verbs and clitics and negation are ‘alternative checkers’. Cocchi & Poletto conclude that the two C-Deletion processes (CD 1 and CD 2) in Florentine crucially differ: CD1 and CD2 target different features of the C0 layer, respectively Finiteness and Force; the alternative checkers also differ: a [–realis] verb can check finiteness in CD1 (both in standard Italian and in Florentine) while clitics, preverbal negation or auxiliaries can check Force in CD2 (available in Florentine). The ideas put forth by Cocchi & Poletto are also interesting for what we are defending: having recourse to a split left periphery to account for the different processes of C-deletion, their analysis excludes at once the possibility of C-deletion in a language which has no V-to-C, the case of BP. Moreover the authors establish puzzling correlations with other phenomena, such as Double-C and V2, noting that in Old Italian no C-deletion at all was observed, contrary to V2 and Double-C. The incompatibility between V2 and C-Deletion results, in their view, from the fact that V2 was the result of V movement to a low C position (identified as Focus), which would block any alternative checking. Leaving these questions for future research and adopting Cocchi & Poletto’s analysis, we conclude that in the absence of V-to-C, BP has no way to license C-Deletion.

4.  Gerundive licensing Lobo (2001) presents a complete description of constructions involving gerunds in EP,22 namely verbal complex gerunds, predicative gerunds, adjunct gerundives and gerunds inside DPs. For the purpose of the comparative study developed in this paper, we present some properties of adjunct gerundives as opposed to verbal complex gerunds. In addition, we discuss some syntactic properties of what the author calls sentence adjunct gerundives and we contrast them with predicate adjunct gerundives. Sentences (26a) and (26b) are

22.  See also Fernandez-Lagunilla (1999) for Spanish, Lonzi (1988) for Italian.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

examples of sentence adjunct gerundives and predicate adjunct gerundives in EP, respectively: (26) a.

Arrombando / *A arrombar a porta com um maçarico, opening /  prep to.open the door with a blowtorch,



os ladrões conseguiram entrar.23 the thieves managed to.enter



“Opening the door with a blowtorch, the thieves broke in.”

b. Os ladrões entraram em casa arrombando / *a arrombar a porta the thieves entered in house opening /  Prep to.open the door

com um maçarico. with a blowtorch

“The thieves managed to break in the house, opening the door with a blowtorch.”

The author gives a set of properties in order to distinguish the two classes of gerundives: their semantic values, unmarked position in the sentence, and behaviour in regard to constructions involving focus, possibility of realization of a lexical subject in post-verbal position with nominative case, among others. In regard to their semantic values, sentence adjunct gerundives may denote cause, concession, time and condition (27), whereas predicate adjunct gerundives may denote condition, time and means (28): (27) a.

Cause



Chegando atrasado, o Zé não arranjou lugar sentado. arriving late, Joe not found place seated “Having arrived late, Joe didn’t find a place to sit.”

b. Concession

Estudando pouco, o Paulo consegue boas notas. studying little, Paul gets good grades “Although he doesn’t study much, Paul gets good grades.”

c.

Time



Atravessando o banco, o assaltante dirigiu-se ao caixa. crossing the bank, the robber went to the teller “Crossing the bank, the robber headed to the teller.”

d. Condition

Saindo de casa às oito e meia, conseguirás chegar a horas. leaving of home at 8 and half, manage arrive on time “Leaving home at half past eight, you’ll manage to arrive on time.”

23.  The data presented in this section was extracted from Lobo’s and Móia & Viotti’s papers. Móia & Viotti (2004) offer an account of the contrast in the distribution of gerund forms in EP and BP.



Tense domains in BP and EP – vP, CP and phases 

(28) a.

Condition/Means



Os atletas teriam melhores resultados treinando mais horas the athletes would have better results training more hours



por dia. per day



“The athletes would have better results training more hours per day.”

b. Time

O Zé encontrou a solução para o problema passeando. Joe found the solution to the problem walking



pela cidade through the city



“Joe found the solution to the problem walking through the city”

c.

Means



Os ladrões arrombaram a porta usando um maçarico. the thieves opened the door using a torch “The thieves opened the door using a torch.”

The initial is the unmarked position in the sentence for clause adjunct gerundives in contrast with the predicate adjunct gerundives, realized in final position as shown in examples (26) and (27). As the following examples show, sensitivity to clefting appears another good diagnosis for distinguishing between sentence adjunct gerundives and predicate adjunct gerundives: (29) a. *Foi chegando atrasado que o Zé não arranjou lugar sentado. was arriving late that Joe didn’t find place seated “It was arriving late that made Joe not find a place to sit.” b. Seria treinando mais horas por dia que os atletas would.be training more hours per day that the athletes

teriam melhores resultados would.have better results

“It would be by training more hours a day that the athletes would have better results”

Among the internal properties that distinguish the two classes of gerundives, she points out that sentence adjunct gerundives allow the realization of a lexical subject in post-verbal position with nominative case as in: (30) a.

Sentence adjunct gerundive

Tendo este aluno desistido, poderemos abrir mais uma vaga. having this student givenˉup, will.can open more one vacancy “With this student having given up, we will be able to open another vacancy”

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

b. Predicate adjunct gerundive

*Os chimpanzés constroem os ninhos juntando as fêmeas the chimps build the nests gathering the females



pequenos ramos. small branches



“Chimps build their nests by gathering the females’ small branches.”

Those differences lead the author to posit that those two kinds of adjunct gerundive sentences occupy different structural positions in the sentence: sentence adjunct gerundives occupy positions in the left periphery of the sentence whereas predicate adjunct gerundives are VP adjuncts. Lobo´s conclusions in regard to the distribution of constructions involving the gerund in EP need to be extended to BP. We have already pointed out that only the gerund form of the main verb make the sentences grammatical in BP verbal complexes. In standard EP, however, prep+infinitive is preferred to the gerund in these constructions. Another contrast appears relevant to this discussion. Although predicate adjunct gerundives are fully productive in both languages, sentence adjunct gerundives, on the other hand, are only fully productive in EP. In BP, sentence adjunct gerundives sound European Portuguese and their semantic relationship with the matrix clause is very hard to recover. In BP, the sentences in (26) are acceptable if the semantic relationship is overtly expressed by adverbs and prepositions as in: (31) a.

Cause



Por ter chegado atrasado, o Zé não arranjou lugar sentado. for have arrived late, Joe not found place seated “For having arrived late, Joe didn’t find a seat.”

b. Concession

Mesmo estudando pouco, o Paulo consegue boas notas. even studying little, Paul gets good grades “Even hardly studying, Paul still gets good grades.”

c.

Time



Depois de atravessar o banco, o assaltante se dirigiu ao caixa. after of cross the bank, the robber walked to.the teller “After crossing the bank, the robber approached the teller.”

d. Condition

Se você sair de casa às oito e meia, vai conseguir chegar if you leave of home at eight and half, will manage to.arrive



a tempo. on time



“If you leave home at half past eight, you’ll manage to arrive on time.”



Tense domains in BP and EP – vP, CP and phases 

Besides the specific semantic relationship between the adjunct and the matrix clause in the gerundive constructions, there is also a temporal relationship established between the two sentences involved in these constructions, as also argued by Lobo (2007). In as much as in BP there is no movement of the verb to C,24 as proposed by Lobo (2007) for EP, the temporal values in C cannot be interpreted. This explains the strangeness of the sentence adjunct gerundives in BP.

5.  Tense as the locus of variation between BP and EP We have shown in Ambar, Gonzaga & Negrão (2004) two crucial differences between BP and EP: (32) (i) EP allows a confirmative reading for a pre-verbal position of the adverb sempre, either with present or with past tense, whereas BP doesn’t; (ii) BP allows a pre-verbal temporal position for sempre with present tense whereas EP doesn’t.

The following paradigm briefly illustrates those contrasts: (33) O João sempre vai a Paris a. “John ‘really/indeed’ goes to Paris” (EP, *BP) b. “John ‘always’ goes to Paris” (*EP, BP)

The first contrast, related to the confirmative reading, was derived from a more general property: absence versus presence of V movement outside IP in BP and in EP respectively. As for the second contrast, related to the temporal reading, we proposed that inside IP no V movement to AgrS is available in BP, whereas in EP it is. The reason underlying this opposition was attributed to the weakening of the identification role played by the inflectional markers of BP as often claimed in the literature. However, EP and BP vary in regard to tense. In EP, present morphology receives a temporal future denotation associated with episodic single event readings through the action of punctual adverbs (Ambar in preparation). In BP, on the other hand, present is opened for temporal values. Recall that for Chomsky (2006: 14), either “Tense is a property of C, and is inherited by T” or “Tense is a property of T, but receives only some residual interpretation unless selected by C.” Therefore, under the latter option, it is reasonable to consider that T has some semantic content, overcoming the problem of null categories which don’t have

24.  Note that (30), where the lexical subject is in post-verbal, shows there is V-to-C.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça

any feature in the lexicon, such as AGR.25 As Chomsky also notes the agreement “features of T are determined by its context, so it should enter the lexicon without values for these features. T bears these features if and only if it is selected by C, hence it should inherit these from C.” Finally, “T may or may not raise to C, but V must raise to v*”. Although, as observed by Chomsky, “whether T has semantic content remains open” and “a broader range of languages should, obviously, be considered seriously before conclusions can be drawn”, on the basis of empirical evidence observed till now let us hypothesize that the reduction of the inflectional tense paradigm in BP (cf. the loss of both different forms for past and present first person plural and the reduction on the marking of person and number agreement on the verb already pointed out, which in turn resulted in the loss of φ-features specification) plausibly renders T vulnerable to be valued by other features of C, namely, modality, and focus.26 On the other hand, in BP, differently than in EP, arguments can be realized not only as bare plurals but also as bare singulars. Müller & Oliveira (2004) claim that bares in this language are indefinites that get bound unselected, either by existential closure or by generic operators. Those operators, in our analysis are generated under v0. This means that, having all its features validated in the v*phase, the verb does not move to Tense in BP. Consequently, the TSubj layer is inert in BP. 6.  Concluding remarks Our main concern in this paper was to bring to light the oppositions between BP and EP referred to in the Introduction, involving three apparently disconnected phenomena: Aux-to-Comp, C-Deletion and Gerunds. We have hypothesized that the contrasts observed in inflected infinitival complements to epistemic and declarative verbs, in C-deletion and in adjunct gerundive clauses plausibly reduce to a property: absence versus presence of V-to-C movement. Limitations of space precluded going through detailed argumentation on all those phenomena here. The working hypothesis that there is no V-to-C movement in BP because V doesn’t move out of vP, first put forth by Ambar (2005), is to be seen as a step in the explanation of this phenomenon across languages. 25.  As noted by Chomsky (2006: 14): “One advantage of the latter option is that T will then have at least some feature in the lexicon, and it is not clear what would be the status of an LI with no features (one of the problems with postulating AGR or other null elements).” 26.  Considering Chomsky’s second option, what features of T differentiate EP from BP remains open. Work in progress starts showing that different structures in BP acquire modality values not available in their EP equivalents.



Tense domains in BP and EP – vP, CP and phases 

The observations made on C-Deletion and adjunct gerundives opened the way for a deeper comparative research on these phenomena, due to the insightful analyses which have given support to ours own, respectively Cocchi and Poletto (2002, 2007) and Lobo (2001, 2003, 2007).

References Aboh, Enoch. 2000. “Object Shift and Verb Movement in Gbe”. Generative Grammar in Geneva 2. 1–13. Ambar, Manuela. 1988. Para uma Sintaxe da Inversão Sujeito Verbo em Português. Ph.D. Dissertation, Lisbon University. Ambar, Manuela. 1989. “Sobre a Posição do Sujeito - Movimento do Verbo e Estrutura da Frase”. Actas do 5.º Encontro da Associação Portuguesa de Linguística, 369–399. Lisboa: APL. Ambar, Manuela. 1992. “Temps et Structure de la Phrase”. Structure de la Phrase et Théorie du Liage ed. by Hans-Georg Obenauer & Anne Zribi-Hertz, 29–49. Paris: Presses Universitaires de Vincennes. Ambar, Manuela. 1994. “Aux-to-Comp and Lexical Restrictions on Verb Movement”. Paths Towards Universal Grammar: Studies in Honour of Richard Kayne ed. by Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaella Zanuttini, 1–23. Washington D.C.: Georgetown University Press. Ambar, Manuela. 1996a. “Infinitives vs. Participles”. Semantic Issues in Romance Syntax: Proceedings of the Linguistic Symposium on Romance Languages, 1996 ed. by Esthela Trevino & José Lema, 1–20. Amsterdam & Philadelphia: John Benjamins. Ambar, Manuela. 1996b. “Aspects of the Syntax of Focus in Portuguese”. Paper presented at the International Workshop on Focus, Paris, 1996.. Ambar, Manuela. 1998. “Inflected Infinitives Revisited: Genericity and single event.” Canadian Journal of Linguistics 43.5–36. Ambar, Manuela. 2001. “Wh-Asymmetries”. Asymmetry in Grammar ed. by Anna-Maria Di Sciullo, 209–250. Amsterdam: John Benjamins. Ambar, Manuela. 2005. “Inflected Infinitives in Romance. Tense domains and Phases”. Ms., University of Lisbon. Ambar, Manuela. 2007. “Verb Movement and Tense – EPP and t-Completeness”. Proceedings of the XXXII Incontro di Grammatica Generativa ed. by M. Cecilia Picchi & Alan Pona, 1–20. Firenze: Edizioni dell’Orso. Ambar, Manuela. & Rita Veloso. 1999. “On the Nature of Wh-phrases – Word Order and Wh-in-Situ: Evidence from Portuguese, French, Hungarian and Tetum”. Romance Languages and Linguistic Theory 1999 ed. by Yves D’Hulst, Johan Rooryck & Jan Schroten, 1–37. Amsterdam: John Benjamins. Ambar, Manuela & Jean-Yves Pollock. 2002. “Topic vs. Comment in some Subject Inversion Sentences in French and Portuguese”. Journal of Portuguese Linguistics 1:1. 119–138. Ambar, Manuela, Manuela Gonzaga & Esmeralda Negrão. 2004. “Tense, Quantification and Clause Structure in EP and BP. Evidence from a comparative study on sempre”. Romance Languages and Linguistic Theory 2002 ed. by Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe & Petra Sleeman. Amsterdam: John Benjamins.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça Barbosa, Pilar. 2001. “Clitics: A window into the null subject property”. Portuguese Syntax: New Comparative Studies ed. by Joao Costa, 31–93. New York: Oxford University Press. Benincà, Paola. 1994. La Variazione Sintattica. Studi di dialettologia romanza. Bologna: Il Mulino. Besten, Hans den. 1977. “On the Interaction of Root Transformations and Lexical Deletive Rules”. Groninger Arbeiten zur Germanistischen Linguistik 20. 1–78. Bok-Bennema, Reineke & Brigitte Kampers-Manhe. 1994. “Transparency effects in the Romance Languages”. Issues and Theory in Romance Linguistics: Selected Papers from the Linguistic Symposium on Romance Languages XXIII ed. by Michael Mazzola, 199–217. Washington, DC: Georgetown University Press. Bošković, Željko. 1997. The Syntax of Nonfinite Complementation: An economy approach. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1999. Adverbs and Functional Heads – A cross linguistic perspective. New York: Oxford. Cinque, G. 2004. “Restructuring and Functional Structure”. Ms., University of Venice. Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale. A Life in Language ed. by Michael Kenstowicz, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2005. “On phases”. Ms., MIT. Chomsky, Noam. 2006. “Approaching UG from Below”. Ms., MIT. Cocchi, Gloria & Cecilia Poletto. 2002. “C-deletion in Florentine: The interaction between merge and move”. Ms., University of Padua. Cocchi, Gloria & Cecilia Poletto. 2007. “Complemetizer Deletion and Double Complementizers”. Proceedings of the XXXII Incontro di Grammatica Generativa ed. by M. Cecilia Picchi & Alan Pona, 49–62. Firenze: Edizioni dell’ Orso. Costa, João. l996. “Positions for Subjects in European Portuguese”. Proceedings of WCCFLXV. Stanford, Calif.: CSLI Publications. Déprez, Viviane. 1989. On the Typology of Syntactic Positions and the Nature of Chains. Ph.D. dissertation, MIT. Duarte, Maria Eugênia L. 1995. A Perda do Princípio ‘Evite Pronome’ no Português Brasileiro. Ph.D. Dissertation, University of Campinas, São Paulo. Duarte, Inês. l987. A Construção de Topicalização na Gramática do Português: Regência, Ligação e Condições sobre Movimento. Ph.D. Dissertation, UL, Lisboa. Duarte, Inês, Anabela Gonçalves & Matilde Miguel. 2005. “Propriedades de C em Frases Completivas”. Actas do XX Encontro Nacional da Associação Portuguesa de Linguística ed. by Inês Duarte & Isabel Leiria, 549–562. Lisboa: APL/Colibri. Emonds, Joseph. 1976. “The Verbal Complex V’-V in French”. Linguistic Inquiry 9. 151–175. Enç, Mürvet. 1991. “On the Absence of the Present Tense Morpheme in English”. Ms., University of Wisconsin, Madison. Fernandez-Lagunilla, Marina. 1999. “Las Construcciones de Gerundio”. Gramática Descriptiva de la Lengua Española. Vol. 2 ed. by Ignacio Bosque & Violeta Demonte. Madrid: Espasa. Figueiredo Silva, Maria Cristina. 1996. A Posição do Sujeito no Português Brasileiro: Frases finitas e infinitivas. Campinas, UNICAMP. Galves, Charlotte. 1993. “O Enfraquecimento da Concordância no Português Brasileiro”. Português Brasileiro: uma Viagem Diacrônica ed. by Ian Roberts & Mary A. Kato, 387–425. Campinas: Editora da UNICAMP. Galves, Charlotte. 2001. “Infinitivo Flexionado e a Legitimação de Agr”. Ensaios sobre as Gramáticas do Português ed. by Charlotte Galves, 181–194. Campinas: Editora da UNICAMP.



Tense domains in BP and EP – vP, CP and phases 

Giorgi, Alessandra & Fabio Pianesi. 2000. “Complementizer Deletion”. Ms., University of Ca’Foscari, Venice & ITC/IRST, Povo, Trento. Haegeman. Liliane. 1996. “Verb Second, the Split CP and Null Subjects in Early Dutch Finite Clauses”. GenGenP 4. 133–175. Kato, Mary A. & Eduardo Raposo. 2007. “Topicalization in European and Brazilian Portuguese”. Romance Linguistics: Selected papers from the 36th Linguistic Symposium on Romance Languages ed. by Maria Jose Cabrera, Jose Camacho, Viviane Déprez, Nydia Flores & Liliana Sánchez, 213–226. Amsterdam: John Benjamins. Kayne, Richard. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Kayne, Richard. 1989. “Facets of Romance Past Participle Agreement”. Dialect Variation and the Theory of Grammar ed. by Paola Benincà, 85–103. Dordrecht: Foris. Lobo, M. 2001. “Aspectos da Sintaxe das Orações Gerundivas Adjuntas do Português”. Actas do 17.º Encontro Nacional da Associação de Linguística, 247–265. Lisboa: APL. Lobo, Maria. 2003. Aspectos da Sintaxe das Orações Subordinadas Adverbiais do Português. Ph.D. Dissertation, Universidade Nova de Lisboa. Lobo, Maria. 2007. “Dependências Temporais: A sintaxe das orações subordinadas gerundivas do português”. Veredas 10. Lonzi, Lidia. 1988. “Tipi di Gerundio”. Rivista di Grammatica Generativa 13. 59–80. Manzini, Rita 2000. “Sentential Complementation: The subjunctive”. Lexical Specification and Insertion ed. by Peter Coopmans, Martin Everaert & Jane Grimshaw, 241- 267. Amsterdam: John Benjamins. Manzini, Rita & Leonardo Savoia. 2003. “The Nature of Complementizers”. Rivista di Grammatica Generativa 28. 87–110. Mioto, Carlos. 2004. “Focalização e Quantificação”. Revista Letras 61. 169–189. Mioto, Carlos & Mary Kato. 2006. “As Interrogativas Q do Português Europeu e do Português Brasileiro Atuais”. Revista da ABRALIN 4. 171–196. Móia, Telmo & Evani Viotti. 2004. “Differences and Similarities between European and Brazilian Portuguese in the Use of the 〈〈Gerúndio〉〉“. Journal of Portuguese Linguistics 3. 111–139. Müller, Ana & Fátima Oliveira. 2004. “Bare Nominals and Number in Brazilian and European Portuguese”. Journal of Portuguese Linguistics 3:1. 9–36. Negrão. E.V. 1986. Anaphora in Brazilian Complement Structures. Ph.D. Dissertation, University of Wisconsin, Madison. Negrão, Esmeralda. 1999. O Português Brasileiro: Uma língua voltada para o discurso. Universidade de São Paulo, Livre-Docência Thesis. Negrão, Esmeralda & Evani Viotti. 2001. “Brazilian Portuguese as a Discourse-oriented Language”. Brazilian Portuguese and the Null Subject Parameter ed. by Mary Kato & Esmeralda Negrão, 97–116. Frankfurt: Editorial Vervuert/Iberoamericana. Obenauer, Hans-Georg. 2004. “Nonstandard Wh-questions and Alternative Checkers in Pagotto”. Syntax and Semantics of the Left Periphery ed. by Horst Lohnstein & Susanne Trissler, 343–383. Berlin: Mouton de Gruyter. Pesetsky, David & Esther Torrego. 2001. “T-to-C Movement: Causes and consequences”. Ken Hale: A life in Language ed. by Michael Kenstowicz. Cambridge, Mass.: MIT Press. Pesetsky, David & Esther Torrego. 2004. “Tense, Case, and the Nature of Syntactic Categories”. The Syntax of Time ed. by Jacqueline Guéron & Jacqueline Lecarme. Cambridge, Mass.: MIT Press. Poletto, Cecilia. 1995. “Complementizer Deletion and Verb Movement in Italian”. Working Papers in Linguistics 5:2. 1–15.

 Manuela Ambar, Esmeralda Negrão, Rita Veloso & Luís Graça Pollock, Jean-Yves. 1989. “Verb Movement, Universal Grammar, and the Structure of IP”. Linguistic Inquiry 20. 356–424. Raposo, Eduardo. 1987. “Case Theory and Infl-to-Comp: the Inflected infinitive in European Portuguese”. Linguistic Inquiry 18. 85–109. Raposo, Eduardo. 1994. “Affective Operators and Clausal Structure in European Portuguese and European Spanish.” Ms., University of California at Santa Barbara. Raposo, Eduardo. l996. “Towards a Unification of Topic Constructions.” Ms., University of California at Santa Barbara. Rizzi, Luigi. 1976. “Ristrutturazione”. Rivista di Grammatica Generativa 1. 1–54. Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. Elements of Grammar: Handbook of Generative Syntax ed. by Liliane Haegeman, 281–337. Dordrecht : Kluwer. Rooryck, Johan. 1994. “Against Optional Movement for Clitic Climbing”. Issues and Theory in Romance Linguistics: Selected Papers from the Linguistic Symposium on Romance Languages XXIII ed. by Michael Mazzola, 417–443. Washington, DC: Georgetown University Press. Rouveret, A. 1980. “Sur la Notion de Proposition Finie: Gouvernement et inversion”. Langages 60. 75–107. Stowell, Tim. 1982. “The Tense of Infinitives”. Linguistic Inquiry 13. 561–570. Uriagereka, Juan. l995. “An F Position in Western Romance”. Discourse Configurational Languages ed. by Katalin Kiss, 153–175. Oxford: Oxford University Press. Vikner, Sten. 1997. “V-to-I Movement and Inflection for Person in all Tenses”. The New Comparative Syntax ed. by Liliane Haegeman. London: Longman.

Variable-behavior Ps and the location of PATH in Old French* Heather Burnett1 & Mireille Tremblay2 1UCLA/2Université

de Montréal

This paper investigates the interaction between lexical semantics and syntactic structure in the interpretation of prepositional phrases through a study of the lexical encoding of directionality in prepositions and particles in Old French (OF). OF had a series of locative and directional prepositions that could be used intransitively, where they were interpreted directionally or aspectually. We argue that the different interpretations available to these elements are a result of the syntactic configurations into which they are placed, not a systematic homophony in the lexicon. We first show that, for each element in the class under consideration, its interpretation can be predicted based on its transitivity properties and the type of verb with which it is paired. We then present our analysis of the OF prepositional system, and show how the lexical semantics of individual particles and general rules of composition conspire to create the semantic variation found in our data.

1.  Introduction This paper investigates the interaction between lexical semantics and syntactic structure in the interpretation of prepositional phrases through a study of the lexical encoding of directionality in prepositions and particles in Old French. In the literature on the syntax and semantics of locative expressions, prepositions are often divided into distinct classes depending on the type of relation they express *This study is based on the Base de français médiéval elaborated by Christiane Marchello-Nizia (ENS-Lyon, France). We wish to thank the participants of Going Romance 2007 for their comments and suggestions. Previous versions of this work have been presented at the Bilingual Workshop in Theoretical Linguistics (York U. 2004), the 2005 annual meeting of the Canadian Linguistic Association (U.of Western Ontario) and LSRL 36 (Rutgers). We would like to thank the audiences of these events for their constructive remarks. This research has been partly supported by the Social Sciences and Humanities research Council of Canada (grants # 410-05-1518 (M. Tremblay) and # 412-2003-1003 (A.-M. Di Sciullo) and graduate fellowships to H. Burnett).

 Heather Burnett & Mireille Tremblay

between an eventuality and an object (the ground). Primarily, it is common to distinguish between locative and directional prepositions (Examples from English shown in (1)). (1) Locative prepositions: in, on, at, behind, under, in front of, etc. Directional prepositions: into, onto, to, towards, out of, from, etc

It is often claimed that the elements of these two classes differ in their lexical semantics: locative prepositions combine with DPs to denote regions (Nam 1995), sets of regions (Kracht 2002) or vector spaces (Zwarts & Winter 2000); whereas, directional PPs denote sets of paths (Bierwisch 1988; Verkuyl & Zwarts 1992; Nam 1995; Zwarts 2005). However, in many cases, a single lexical item can appear to be part of both classes. For example, in English, certain locative PPs, such as under the table (2a), can also be interpreted directionally (2b). (2) a.

The mouse is under the table “The mouse is located in the area under the table”

b. The mouse ran under the table “The mouse ran to the area under the table”

Locative PP Directional PP (Jackendoff 1983: 163)

Given that locative and directional prepositions have different lexical semantics, (2b) would seem to suggest that the under in this sentence is not the same lexical item as in (2a). Thus, one way of accounting for directional uses of locative prepositions would be to postulate the existence of two separate lexical items with identical phonological content: one having locative semantics, and the other one being lexically specified as directional. This style of analysis, which we will henceforth refer to as the Lexical analysis, is explicitly adopted by Jackendoff (1983) and is implicit in the works of authors (such as those cited above) that assume that directional and locative PPs refer to different semantic objects that bear no direct relation to each other. More recently, however, there have been attempts to solve the problem of prepositional semantic variability not through systematic homophony, but through picking one interpretation as exemplifying the true lexical semantics of the preposition, and showing how its alternate can be derived from the syntactic and semantic contexts in which the preposition occurs. For example, Gerhke (2007, 2008) argues that variable prepositions in English such as under are inherently locative, but contribute to creating a directional interpretation when they are paired with a certain class of verbs that denote sets of complex events. In this paper, we also argue that the lexical analysis is insufficient to account for the patterns of semantic variability observed in the Old French (OF) particle and prepositional system, and we show that the different interpretations of particles and prepositions (P-elements) in OF are derivable from the syntactic



Variable-behavior Ps and the location of PATH in Old French 

structures into which these elements are placed. Thus, this paper serves as an additional argument that the distinction between directional and locative P-elements is derivational. The paper is organized as follows: Section 2 contrasts the basic properties of the particle-preposition alternation in OF with that of other Romance languages and presents the basic data concerning the interpretation of OF P-elements. Section 3 provides arguments against a lexical analysis for the semantic alternations described in Section 2. Section 4 presents our analysis of the OF system and shows how the interpretation of a P-element contributes to the ways in which it combines with other elements in the sentence. Finally, Section 5 summarizes our argumentation and presents the implications of our analysis for the cross-linguistic realization of particles and the category P. 2.  Variable-behavior Ps 2.1  Particles in Romance Particles are often considered to be a Germanic feature absent in Romance (see for example, Talmy’s (1985) distinction between satellite-frame vs. verb-frame languages). However, an increasing number of studies have challenged this claim, and evidence of verb-particle constructions has been provided for a number of Modern Romance languages such as Italian (3), Catalan (4) and Spanish (5). (3) venire giù portare via mettere sotto

“to come down, to descend” “to take away” “to put under, to run over” (from Iacobini & Masini 2007:(1))

(4)

anar tirar tirar tornar

avall fora amunt enrere

“to go down” “to throw out” “to throw up” “to turn back”

(from Mateu & Rigau 2007:(8))

(5)

ir echar echar volver

abajo fuera arriba atrás

“to go down” “to throw out” “to throw up” “to turn back”

(from Mateu & Rigau 2007:(8))

Moreover, according to Masini (2005, 2006), verb-particle constructions were also attested in late Latin (ire via) and Old Italian (6). (6)

Andare giú “go down” buttare fuori “throw out” tirare su “throw up” tornare in dietro “turn back” (from Masini 2006, cited in Mateu & Rigau 2007)

 Heather Burnett & Mireille Tremblay

In this context, the existence of productive particle system in OF (Buridant 2000; Dufresne, Dupuis & Tremblay 2003; Burnett & Tremblay in press) comes as no surprise and seems to support a view of particles as a proto-Romance feature. 2.1.1  The OF particle system The OF particle system consists of 10 elements: ariere “back”; avant “forward”; sus “up”; jus “down”; ens “in”; hors/fors “out”; amont “upwards”; aval “downwards”; contremont “upwards”; contreval “downwards”. As shown in the examples below, the OF particles all have a directional interpretation. This is consistent with many analyses of English particles, such as Jackendoff (1983) and Svenonius (2006), where these P-elements are proposed to refer directly to paths. (7) a.

…et pour ce se trait il ariere …and for this himself draw he back “…and, because of this, he left”

Trispr, p. 263

b. Lors saut avant Girflez et dist a la reïne:ˉ…ˉ Artu,ˉp.ˉ319 Then jumps forward Girflet and says to the queen: … “Then Girflet comes forward and says to the queen:…” c.

Et toutevoies recort il sus au serpent … And however re-run he up to.the snake … “However, he pursues the snake again…”

QGraal, p. 94

d. Ains descent tout maintenant jus de son ceval Trispr,ˉp.ˉ127 So descends all now down from his horse “He then dismounts from his horse” e.

et lors entrent enz li plus riche baron and then entered in the most rich barons “and then the richest barons came in”

f.

A ces criz et a ces noises issi At these cries and at these noises, came



hors Messires Gauvains de son ostel out Sir Gawain.nom from his house

Artu, p. 213

Artu,ˉp.ˉ130

“When he heard the cries and the noise, Sir Gawain came out of his house” g. si tost come il fu amont venuz en la roche… Qgraal,ˉp.ˉ94 so soon as he was upwards came on the rock “As soon as he came up on the rock” h. il descendent del pals et viennent en They descended from-the. palaces and came into

la cort aval the courtyard down



“They descended from the palaces and came down into the courtyard”

Artu,ˉp.ˉ194



Variable-behavior Ps and the location of PATH in Old French 

i.

Et il vient as degrez et monte contremont, …  Graal, p. 139 And he comes to the.stairs and climb up… “And he goes to the stairs and goes up…”

j. 

Si se leva et devala comme ainz pot contreval. “He got up and went down as fast as he could” SagesP, 21,12, (cited in Buridant 2000: 439)

Like other Romance particles, OF particles form a semantic unit with the verb they appear with. However, unlike its Romance counterparts (Iacobini & Masini 2007; Rigau p.c.), but similar to Germanic verb-particle constructions, this verb-particle construction is not syntactically atomic. This can be seen in the examples in (8), where the verb is separated from the particle1 by the direct object (8a), and an adverb (8b). (8) a. le mers reportoit le nef ariere. the sea re.bring.imp the ship back “the sea pushed the ship back”  Clari, p. 74, (cited in Dufresne, Dupuis & Tremblay 2003:Example (26a)) b. Je lour courui esranment sus … I them.dat run.pst immediately up “I pursued them immediately…”  Trispr, p. 62, (cited in Dufresne, Dupuis & Tremblay 2003:Example (73))

The OF particles differ from English-style particles in that they can appear before infinitives (9) and past participles (10), i.e. in sentences which do not involve verb movement.2 (9) a.

Voldrent le faire sus lever /ˉMais il ne pot sor wanted it make.inf up raise /ˉbut he neg can on



piez ester feet be.inf

“They wanted to make him stand up, but he could not stay on his feet.”  VieSGrég1, ms. A1, 2537–38, (cited in Buridant 2000: §440)

1.  As shown in (7a) and (7c), the verb could also be separated from the particle by the subject, as a result of movement of the verb in second position. 2.  Although the preposing of particles appears more frequent in verse than prose, the existence of such structures in Modern Acadian French seems to indicate that the phenomenon is neither marginal, nor a remnant of the OV word order (see Tremblay 2005 for a comparison between OF ariere and Acadian back and King (2000) for an alternative analysis of back in Acadian French). (i)

elle est back sortie she is back exited “she went back out”

Roy corpus, (cited in Perrot 1994: 242)

 Heather Burnett & Mireille Tremblay

b. …si Dex me laist de cest champ issir vif/ Et sain et …if God me let from this field leave alive/ and sound and

sauf arriere revertir safe back return.



“If God lets me leave from this field alive, and return safe and sound.” Amiami, p. 46

(10) a. 

Li sëaus est jus avalez The pail is down descended “The pail is brought down.” RenartR, II, 3641, (cited in Buridant 2000: §440)

b. Garde le Keep it

bien, tant que tu soies well, until that you are



Cornuaille. Cornwall



“Keep it well until you return to Cornwall.”

ariere revenus en back returned in

Trispr, p. 237

2.1.2  The productivity of OF particles The verb-particle combinations can be divided into two classes: idiomatic and transparent. Idiomatic verb-particle constructions are those that have become lexicalized, such that the meaning of the entire construction is not predictable from the meanings of their separate parts. For example, the meanings of combinations such as metre sus “to accuse” and corre sus “to attack” in (11) are unpredictable and, therefore, must be listed together in the lexicon. (11) a.

la reïne te pardonra ce meffait que tu li the queen you.dat forgive.fut this mischief that you her.dat



as mis sus… aux put on

“The queen will forgive you for this mischief that you accused her of.”  Artu, p. 106 b. Lors li corent li autre sus de parole et dient : Dame, Then him.dat run.3PP the one on of words and say.3P : Lady,

vostre escondires ne vos vaut neantˉ; your refusal neg you.dat worth nothing.

“Then, the others attack her with words and say : Lady, your refusal will not bring you anything.”

Additionally, the symmetry of the particle system has given rise to lexicalized expressions of the form P et P, where two contrasting particles are coordinated to create the meaning ‘everywhere’:



Variable-behavior Ps and the location of PATH in Old French 

(12) a.

Tant l’ai quisse ariere et avant que jou ne sais So much CL.searched back and forward that I neg know



u la guerre where the war

“So much did I look for her everywhere, that I didn’t know where the war was” Escoufle, p. 207 b. …et l’endemain cerchierent amont et aval et trouverent une ..and the next.day looked.for up and down and found an

vielle old

fame qui… woman who…

“and the next day, they looked everywhere and found an old woman who….”

Qgraal, p. 75

c. …et vont sus et jus por savoir s’il porroient ceus …and go up and down to know if they can them

trouver qu’il queroient. find that they looked.for.

“…and they go everywhere to see if they can find the ones they were looking for.”

Artu, p. 128

However, the meaning of the majority of verb-particle combinations is fully compositional. Semantically transparent combinations such as metre avant “put forward” and aller ariere “go back” are not listed in the lexicon, but formed productively in the syntax. (13) a.

…qu’il est menterres de metre tel chose avant. …that it is dishonest to put such things forward.

b. je m’en irai ariere a mon signour I CL.CL go.fut back to my lord “I will return to my lord”

Artu, p. 37

Artu, p. 143

2.1.3  The semantic variability of particles While most OF transparent particles only have a directional interpretation, the particles ariere “back”, avant “forward”, and hors/fors “out” appear to also have an aspectual reading. In (14), the particles ariere, avant and hors are interpreted as directional; in (15), the same particles yield iterative, inchoative,3 and perfective interpretations respectively.4

3.  According to Buridant (2000: §442), the aspectual interpretation of avant is that of a durative. We believe that examples such as (15b) provide empirical evidence against this interpretation. 4.  Other Romance particles have also been claimed to have an aspectual interpretation. For example, Iacobini & Masini (2007) distinguish between tendentially telic particles and tendentially

 Heather Burnett & Mireille Tremblay

– Directional (14) a.

il se traient adont un poi ariere They CL moved therefore a bit back “Therefore, they moved back a bit”

Trispr, p. 39

b. et mesire Gauvains vient avant et li deslace son hiaume. And Sir Gawain comes forward and CL unties his helmet “And Sir Gawain comes forward and undoes his helmet” Artu, p. 106 c. 

Is t’an fors contre moi go you.from.there out against me “Get out of there and fight me” Saisnes, L 3637, (cited in Buridant 2000: 54)

– Aspectual (15) a.

arriere les voit consillier again them see counsel “(he) sees them conversing again” 

Belinagr, p. 85

b. il s’ entrecommencent a regarder et semont li uns l’autre they CL.ENTRE.begin to look and ask the one theother

de parler avant. to talk orward.



“They begin to look at each other and ask each other to start talking.” Artu, p. 13

c.

boire fors/ manger fors/ paiier fors drink up/ eat up / pay up (examples from Buridant 2000: 544)

Furthermore, as shown in the examples in (16), OF particles may occasionally have a locative interpretation.

atelic particles, depending on whether their meanings are bounded or unbounded. They identify a subclass of particles which may function both as a direction (or Path) marker and as a telos indicator, and propose that such particles inherently refer to a specific spatial endpoint and thus contribute to the overall telic meaning of the verb particle construction. (i)

tirare andare sbattere saltare

“to pull” “to go” “to dash/throw” “to jump”

tirare fuori andare via sbattere fuori saltare giù

“to take out” “to go away” “to throw out” “to jump down”

While some OF directional particles also appear to have a telic contribution, OF appears to be unique in allowing some of its particles to have purely aspectual (i.e. non-directional) uses.



Variable-behavior Ps and the location of PATH in Old French 

(16) a.



…et quant Mordrés, qui hors étoit entre lui et …and when Mordred, who outside was between him and sa compaignie, …. his retinue…

“ ..and when Mordred, who was outside between him and his retinue, ….”  Artu, p. 178 b. Et si tost come il est amont, si troeve le lyon qui… “And as soon as he is up, he finds the lion that…”  Qgraal, p. 104

In summary, while some particles in OF may have directional, aspectual, and locative meanings, others only appear to have directional and locative readings, or only a directional interpretation. A lexical analysis of this data would postulate ambiguity in the lexicon: some directional particles could have aspectual and locative homophones. 2.2  The prepositional use of particles The semantic variability of OF particles is further increased by the fact that these elements also exhibit syntactic variability. As noted by Buridant (2000); Dufresne, Dupuis & Tremblay (2003); and Burnett & Tremblay (in press), the most striking difference between particles in medieval French and those found in other Romance languages is the fact that most OF particles can also be used transitively.5 (17) a.

il laissa tous ses compaignons ariere soi “he left all his companions behind him”

Tristpr, p. 204

b. De deduit d’oiseax et de bois/ ne savoit nus hom “No man knew how to be delighted so much

avant lui. before him”

Dolem, p.ˉ1

c. Sus sa poitrine tenoit ses mains croisant On his chest kept his hands crossed. “He was keeping his hands crossed on his chest.”  Aliscans, 827, (cited in Buridant 2000:§388) d. Et la mist ju del mullet afeutré And it put down of the mule saddled “And he put it at the bottom of the saddled mule”  La Geste de Loherins, (cited in Greimas (2001: 327))

5.  As exemplified in (17d) and (17f), some particles are sometimes followed by the preposition de “of ”. This property does not seem immediately relevant to the present discussion since this usage is not restricted to particles used transitively, but is also found with complex prepositions such as au dessus (de), and à côté (de).

 Heather Burnett & Mireille Tremblay

e. 

Ens cele canbre, biaus dous nies, enterés. In this room, beautiful sweet nephew, enter.will “You will enter this room, my sweet and beautiful nephew” H. de Bord., (cited in Greimas 2001: 209)

f.

Quant il furent hors de Kamaalot… When they were out of Camelot “When they were outside Camelot…”

g. Amont Sainne naja. Up Seine navigate “He sailed up the Seine.”

Artu, p. 6–7

Rou, III, 422, (cited in Buridant 2000: 486)

h. si s’en vint aval la roche “Thus he came down the rock”

Qgraal, p. 95

i.

Quant Perceval voit ceste aventure, si cort au plus tost When Perceval sees this event, thus runs to.the more soon



qu’il puet contremont la montagne. that he can up the mountain.

“When Perceval sees this, he runs as soon as possible up the mountain.”  Qgraal, p. 94 j.

estoit as fenestres de la sale, et regardoit was at.the windows of the room, and looked



contreval la riviere, down the river



“He was at the windows of the room, and he looked down the river” Artu, p. 87

We know that the P-elements in sentences such as (17) form a constituent with the DP they precede, as they are treated as a single unit by syntactic movement rules such as PP fronting. For example, in (17c), sus sa poitrine appears in first position followed by the tensed verb (see also 17e, 17g, and 23a). The transitive variants of the class of P-elements under discussion can be divided into two semantic classes. As shown in (17), elements from Class 1, consisting of ariere, avant, sus, jus, ens and hors/fors, combine with a direct object to form a locative prepositional phrase. On the other hand, elements from Class 2, made up of amont, aval, contremont, and contreval, have a directional interpretation when paired with a DP: the PP in (18) may only mean ‘down (on) her face’, not ‘at the bottom of her face’. (18) Ele amena ses mains contreval sa face She brought her hands downwards her face “She ran her hands down her face” SagesP, 5,3, (cited in Buridant 2000: 541)

Just as in the case of the semantic variability of intransitive P-elements, we may investigate the question of how to treat the semantic and syntactic variability of



Variable-behavior Ps and the location of PATH in Old French 

these elements. The lexical analysis would postulate that each P-element is ambiguous between (at least) one lexical entry for the intransitive P-element (the particle), and a separate lexical entry for the transitive one (the preposition) with locative semantics (for Class 1 prepositions), or directional semantics (for Class 2 prepositions). However, in Section 3, we will argue that the transitive/intransitive alternation could also be derived structurally. 2.3  Summary OF P-elements can be used both intransitively, where they form a complex predicate with the verb, and transitively, where they form a PP with a direct object. When intransitive, P-elements can have up to three interpretations: directional, aspectual, and locative. When used transitively, P-elements only have one interpretation: locative (Class 1) or directional (Class 2). Thus, a lexical treatment of this four-way syntactic and semantic ambiguity would have to propose that the lexicon contains at least the items found in Table 1; whereas, a fully structural analysis would be able to collapse all four elements into a single lexical entry. Table 1.  Fragment of the OF lexicon under a lexical analysis Lexical item

Phonological content

Syntax

Semantics

1 2 3 4

Ariere Ariere Ariere Ariere

–transitive –transitive –transitive + transitive

directional aspectual locative locative

In the next section, we provide arguments that, in addition to being less elegant, the lexical analysis of the OF particle and prepositional system is empirically inadequate.

3.  Arguments against the lexical analysis Although a lexical analysis easily solves the meta-linguistic problem of how to write the proper semantic definitions and subcategorization frames for members of the class of OF P-elements, we argue that it does not accurately capture the data it is meant to describe. Such an analysis misses the following empirical fact about the particle and prepositional system of medieval French: (19) The semantic and syntactic variability observed with this class is predictable from other elements in the sentence.

 Heather Burnett & Mireille Tremblay

An analysis that proposed that the different interpretations and argument structures reflected a systematic, yet accidental, homophony in the lexicon would predict the opposite state of affairs: that there should be environments in which more than one variant is possible and attested. We therefore conclude that the different meanings and syntactic characteristics attributed to a single phonological representation are derivable from a single lexical item. We first show that the semantic variability of intransitive P-elements is conditioned by the type of verb they are paired with. Then, we argue that the syntactic variability of this class of elements exhibits a systematicity that is unexpected under a lexical analysis. 3.1  Semantic variability Interestingly, each interpretation available to an intransitive P-element corresponds to a distinct environment, namely, the type of verb with which it forms a complex predicate. In other words, directional, aspectual and locative particles seem to be in complementary distribution, and this distribution is determined by the interaction between the semantics of the verb and the particle when they are merged together in the syntax. We first consider the locative interpretation. A detailed study of our 13th century corpus shows that the locative interpretation of particles is extremely restricted. It is limited to what we will call predicational environments: sentences where the particle is either the complement of the verb estre “to be” (20a) or is semantically predicated of a verbal argument, usually the direct object (20b). (20) a.

Et si tost come il est enz, si voit parmi les rues “And as soon as he is inside, he sees among the streets



tant de puceles que… so many young girls that… ”

Qgraal, p. 49

b. Quant Boorz le vit jus, il le regarde et dist: When Bors him sees down, he him looks and says:

‘Traïtres, desloiax…’ ‘Traitor, disloyal…’

“When Bors sees him on the ground, he looks at him and says, ‘Traitor, disloyal…’ ” Artu, p. 255

An analysis that would propose the existence of particles with locative meanings would predict more meanings for verb-particle combinations than actually exist. There is nothing about the semantics of ‘locative-ness’ that requires a predicational structure, as shown by the fact that verbs and events can be modified by locative PPs. In principle, particles with a locative interpretation could be used to specify the location of an event, just like locative PPs do. Thus, there would be no reason



Variable-behavior Ps and the location of PATH in Old French 

to rule out the existence of verb-particle combinations like parler ens, meaning ‘to talk inside’. We therefore conclude that the locative alternation of intransitive P-elements is structural. Having identified the contexts in which particles receive a locative interpretation, we now turn to the remaining contexts in which particles are interpreted directionally or in the cases of ariere, avant, and hors/fors, aspectually. Again we see that for the particles that have attested aspectual uses, these uses are in complementary distribution with the directional ones. As argued in Burnett, Petrik and Tremblay (2005), the conditioning factor seems to be the semantics of other elements in the sentence. For example, we show in (21a) the list of verbs which, when used with ariere yield a directional interpretation. Likewise, in (21b), we list the verbs that combine with this particle to yield an iterative interpretation. (22a) gives the list of verbs used with directional avant, while (22b) gives the list of verbs used with aspectual avant. The verbs marked with an asterisk appear more than twice in the corpus. (21) a. revenir*, traire*, aller*, retorner*, torner*, venir, remettre*, mettre*, enchacier*, realler*, repairier*, reporter*, mener(se)*, retraire, courir, suivre, assambler, porter, remener, verser, revertir. b. consiller*, chier*, bouter, mander, remander, refuser, rebouter, s’abattre, armer (22) a. venir*, mettre*, aller*, sauter/saillir*, traire*, amainer*, lanchir*, envoyer*, porter*, departir, fere, garder, passer b. dire, parler

The data in (21a) and (22a) seem to indicate that, when ariere and avant are used with verbs that take a locative argument (primarily motion verbs), henceforth known as locative contexts, the verb-particle combination has a directional interpretation. Otherwise, these particles are aspectual. Table 2.  Distribution of non-locative ariere and avant in the Old French corpus (prose and verse6) (data from Burnett, Petrik & Tremblay 2005) Particle

Interpretation

Number of Occurrences

Locative or Motion verbs

Ariere

Directional Iterative Directional Inchoative

199(94%) 12(6%) 141(98.6%) 2 (1.4%)

100%

Avant

100%

Other Verbs

100% 100%

6.  Our Old French corpus includes two series of texts: the texts in verse come from the 12th century, and the texts in prose from the 13th century.

 Heather Burnett & Mireille Tremblay

The fact that we can only directly observe the directional/aspectual alternation with ariere and avant raises the question of whether the association of directionality with locative contexts is a robust pattern or a lexical quirk of these two particular lexical items. Unfortunately, in our corpus, with the exception of the aforementioned two particles, all the verb-particle combinations that we have identified are formed with locative or motion verbs. However, a result that is consistent with the generalization we have proposed is that, in our corpus, all uses of particles in non-predicational environments are directional. We therefore suppose that the absence of tokens of aspectual particles other than ariere and avant in our corpus simply reflects the low frequency of these uses to begin with (for example, although we have found 258 total instances of avant and 120 total instances of ariere, there are only 19 tokens of contreval and 20 tokens of contremont in our prose corpus (cf. Burnett & Tremblay in press). The case of the particle hors illustrates this point. Although aspectual uses of the particle hors/fors are attested elsewhere in the literature (cf. Buridant 2000: 544), we have found no verb-particle combinations with hors formed from a non-locative verb in our corpus. As predicted by our generalization, we find no occurrences of aspectual hors. Furthermore, the examples of aspectual hors given by Buridant are exclusively with verbs that do not take locative arguments (manger “to eat”, boire “to drink”, and paiier “to pay”), so it seems that the semantic variability of hors conforms to the same pattern as avant and ariere. In sum, the different interpretations of particles in Old French are conditioned by the contexts in which they appear: when they are used in predicational contexts, they acquire a locative meaning; when they are used in locative contexts, they acquire a directional meaning; and, when they are used with other verbs, they acquire an aspectual meaning. This precise distribution is unexplained by an analysis that postulates independent homophones in the lexicon. However, this kind of complementary distribution is expected under an analysis that postulates only one entry for each intransitive P-element and sees semantic variability as an epiphenomenon. 3.2  Syntactic variability Having established the need for a unified analysis of intransitive P-elements, we turn to the relationship between intransitive P-elements and their transitive counterparts. Are there two homophonous arieres in the lexicon, one that is a locative preposition, and one that is an intransitive particle (with a meaning that has yet to be determined)? Or is there only one P-element, either transitive or intransitive, that acquires a locative meaning when it is merged with a direct object? In this section, we argue for the second analysis. Our principal argument for the



Variable-behavior Ps and the location of PATH in Old French 

identity between arierePART and arierePREP comes from the organization of the P-element system. Although, looking at each particle individually, it may seem reasonable to postulate the existence of a transitive doublet with locative (Class 1) or directional (Class 2) semantics, the regularity of these doublets, i.e. the fact that, for every particle, there exists a corresponding preposition, suggests that the +transitive variants are related to the –transitive ones by syntactic or morphological rules. Furthermore, we claim that it is significant that transitive P-elements may have only one of two meanings: locative or directional, never aspectual. Under a lexical analysis, there is no reason to expect such systematicity. This approach has nothing to say about both the completeness of the system and the limited range of meanings available to transitive P-elements. Because these patterns seem robust, we conclude that there is a derivational relationship between transitive and intransitive P-elements in Old French. 3.3  Summary In this section, we argued that the different interpretations of particles in OF should be reducible to a single basic interpretation, and that they should arise as a result of the particular syntactic structures that the particle is inserted into. We have argued, furthermore, that a similar analysis should apply to the transitive/intransitive alternation. Our argument is based on the observation that the semantic interpretation of P-elements is predictable from the context in which they appear. A summary of the different environments and the corresponding interpretations is shown below. Table 3.  The interpretations of P-elements in Old French Class

Class 1 (ariere, avant, sus, jus, ens, hors/fors) Class 2 (amont, aval, contremont, contreval)

Intransitive

Transitive

Locative Contexts

Predicational Contexts

Other Contexts

Directional

Locative

Aspectual

Locative

Directional

Locative

Unattested

Directional

The distribution of meanings shown in Table 3 is unexpected under a lexical analysis, but can be incorporated into a structural analysis. However, a structural analysis of this system is not immediately obvious. For example, a challenge for this approach is to explain the existence of the Class 1/Class 2 distinction. If, as

 Heather Burnett & Mireille Tremblay

we claim, the transitive/intransitive alternation is a result of a general syntactic compositional rule, it is unexpected to find that two intransitive elements with identical semantics (say sus and amont) have different meanings when input into the same rule. However, as we will see in the next section, the semantic differences between Class 1 and Class 2 prepositions are not arbitrary, but arise from the fact that Class 2 elements are morphologically complex, and Class 1 elements are not. We can therefore derive the difference in interpretation between Class 1 prepositions and Class 2 prepositions from semantic contribution of the morphemes that make up the complex P-elements. 4.  A structural account of variable behavior Ps in Old French In this section, we provide our analysis of how both the directional/aspectual/ locative alternations and the transitive/intransitive alternations arise despite the existence of a single lexical entry for each P-element. Since we propose that prepositions and particles derive from the same lexical entry, we must also account for why prepositional uses of P-elements have the meanings that they do, and what relation these meanings bear to the meanings available to particle-uses of these elements. We will therefore first provide an analysis of the semantic behavior of Class 1 P-elements, whose prepositional uses are locative, and then we analyze the elements of Class 2, which are directional when used transitively. 4.1  Class 1 P-elements Class 1 is made up of six P-elements: ariere “back”, avant “forward”, sus “up”, jus “down”, ens “in”, hors “out”. We claim that these forms are monomorphemic in 13th century French, which seems uncontroversial in the cases of sus, jus, ens, and hors. However, that avant and ariere should not be analyzed into a root -vant and -riere combined with a prefix a- requires some argumentation since, historically, avant derives from the Latin ab+antes “from + in front”, and ariere comes from the Latin ad+retro “to+behind”. Furthermore, in Very Old French, riere itself could be used both as a locative preposition (23a), and as a directional particle (23b). (23) a. Rier lui regarde, et vit maint chevalier behind him look, and saw many knights “He looks behind him and he sees many knights”  Chevalerie Ogier, (cited in Greimas (2001: 528)) b. Dist Ysorés: ‘Jantil duch de Clermont, Torné vos rier!’ Said Ysoré: ‘Jantil duke of Clermont, turn you back “Ysoré said, ‘Jantil, Duke of Clermont, turn back!’ ”  L’entrée d’Espagne 6328, (cited in Tobler-Lommatzch)



Variable-behavior Ps and the location of PATH in Old French 

However, rier had stopped being able to appear independently by the 13th century: there are no examples of it in our corpus or in the literature on this time period. Additionally, vant was never an independent root in the history of French, so it is likely that these two particles were treated as having the same morphological complexity as the other P-elements in Class 1. 4.1.1  The locative interpretation of Class 1 P-elements Class 1 particles are locative in transitive and predicational contexts, directional in locative contexts (with motion verbs), and (in some cases) aspectual when merged with other eventive verbs. Since Class 1 P-elements are locative in more environments than any other interpretation, we propose that the underlying semantics of these elements is locative. We propose that, when Class 1 particles are merged with DPs, they form PPs denoting places.7 For example, a PP like ariere lui “behind him” denotes the area behind the location of the referent of the pronoun. In conceptual semantic theories, such as Jackendoff (1983, 1990), one would assign the conceptual structure in (24a) to ariere. Model-theoretically, the denotation of ariere can be viewed as a function that takes an individual argument to yield the area that is behind the location of that individual (24b) (24) a. [Place BEHIND [Thing ] i] b. {<x, area behind x>, , …}

The interpretation of a sentence like Li chevaliers est ariere le chastel “The knight is behind the castle”, can now be given in exactly the same way as in other works. For example, Zwarts & Winter (2000) employ a type-shifting function loc' that maps places to the set of individuals located in those places. So loc'([[ariere le chastel]]) is the set of individuals located behind the castle, and the entire sentence is true only if li chevaliers is in that set. 4.1.2  The directional interpretation of Class 1 P-elements We must now explain why, given that Class 1 P-elements are semantically locative, they appear to be directional and aspectual in certain contexts. Recall that OF particles are directional when they combine with, primarily, motion verbs. Following, among others, Jackendoff (1983), Zwarts (2005, 2006), and Kracht (2002), we assume that a significant distinction between motion verbs and other verbs is that the former make reference to paths in their lexical semantics. For example, 7.  As mentioned in the introduction, there are many ways of conceiving of the denotations of locative PPs, so we will use the term place as a way of being agnostic about the nature of these set theoretic objects. Since our proposal can be easily executed in models where locative PPs denote either sets of points (Nam 1995); families of sets of points (Kracht 2002), or vector spaces (Zwarts & Winter 2000), this terminology is justifiably general.

 Heather Burnett & Mireille Tremblay

Jackendoff (1983, 1990: 45) represents the conceptual structure of a motion verb as in (25). (25) [Event GO ([Thing ]i [Path ]j)]

(25) indicates that a motion verb denotes a set of events made up of a theme and a path upon which the theme travels over the course of the event. However, the PATH argument of a motion verb is not of the same order as its THING argument. Generally, individuals are treated as primitive elements of the model; however, most analyses of paths view them as constructed from other, simpler objects (see Zwarts 2005 for arguments for constructed paths). It is very often assumed that paths can be decomposed into a [Path [Place]] structure (Jackendoff 1983; Koopman 2000; Svenonius 2006; Gehrke 2008), forming the following structure: (26) [Event GO ([Thing ]i [Path TO [Place [Thing ]j]] ])]

In this approach, the sentence John ran into the house would be represented as in (27). (27) [Event GO ([Thing John ]i [Path TO [Place IN [Thing The house ]j]] ])]

As shown above, the directional PP into the house saturates the entire complex PATH argument of run. This sentence is true when an event occurs with John running, which starts outside the house and then ends inside the house. To account for the directional reading of Class 1 particles, we propose that when it is combined with a motion verb to form a complex predicate, the particle becomes a locative modifier of the PATH argument of the motion verb. In particular, we propose that merging a P-element directly with a motion verb restricts the set of events denoted by the VP to only those whose paths end within the place denoted by the P-element. Since, in complex predicates, P-elements are merged with the verb before their THING argument is saturated, in order for the final structure to be interpretable, the semantic argument of the particle must be saturated by other means. We propose that when this predicate is merged with the verb, its open argument position is co-indexed with the open theme position in the motion verb. In other words, merging the particle and the verb restricts the set of individuals that can fill the THING position in verb by the set of individuals that can fill the THING argument of the particle. These two restricting operations are done by a single compositional rule that we call Complex Predicate Formation. (28)





Complex Predicate Formation Verb-Particle Combination [Event GO ([Thing ]i [Path TO [Place α [Thing ]i]] ])]

Motion Verb [Event GO ([Thing ]i [Path TO [Place [Thing ]j]] ])]

Particle [Place α [Thing ]k]



Variable-behavior Ps and the location of PATH in Old French 

The derivation of the complex predicate aller ariere “to go back” is shown below. (29)

aller ariere [Event GO ([Thing ]i [Path TO [Place BEHIND [Thing ]i]] ])]

aller [Event GO ([Thing ]i [Path TO [Place [Thing ]j]] ])]

ariere [Place BEHIND [Thing ]k]

We can view ariere as semantically transitive, in the sense that it requires an individual argument to be semantically saturated; however, it is not syntactically transitive, since this argument does not need to be syntactically realized. It can be saturated through the composition of the P-element with a motion verb followed by the binding of this argument position with the first argument of the verb. 4.1.3  The aspectual interpretation of Class 1 P-elements Finally, we discuss how the locative class 1 particles might acquire aspectual interpretations. Since we only have aspectual meanings attested with half of this class of particles, the mechanisms that derive these interpretations remain somewhat unclear. However, we can make some conjectures on the subject based on the distribution patterns that we find in our corpus. We saw in the previous section that aspectual verb-particle combinations arise with eventive verbs that do not subcategorize for a PATH argument. More specifically, we saw that we get aspectual readings with verbs like parler “to talk” and mander “to ask” that do not seem to take any arguments that are not of the category THING (individuals). Therefore, in these cases, we propose that, instead of modifying part of one of the complex arguments of the verb, the particles modify the set of events denoted by the verb itself. Consider the aspectual use of ariere “back”. When used with mander, it means roughly “again”. (30) Se li manderent ariere que ore le desfioient il Thus him asked back that …. “Thus they asked him again that”

Clari, p. 62

We propose that ariere combines pointwise with the elements of the denotation of the VP. So the VP li manderent ariere will now denote a set of ordered <event of asking, area (temporally) behind the event of asking> pairs. We may then again employ Zwarts & Winter’s loc' to transform the areas behind the events into the sets of events located in those temporal areas. In other words, loc'(area behind eventx) is the set of events that occur before eventx. The sentence will be true iff the set of ordered pairs denoted by li manderent ariere is non-empty, i.e. if there is an

 Heather Burnett & Mireille Tremblay

event e such that e is an event of them asking him, and there is at least one event e’ such that e’ has the same properties as e,8 and e’ occurred before e. Intuitively, these are the truth conditions that we are looking for, and in fact, they correspond exactly to those given for sentences with the English iterative adverb again by Beck & Johnson (2004). Although we can give a straightforward account of the semantic composition of sentences with aspectual ariere as a modifier of the set of events denoted by the VP, it is unclear whether this is possible in a principled way for all particles. Unfortunately, the exact meaning of aspectual avant is not agreed upon by all scholars: Buridant (2000) claims that it is a durative; whereas Santucci (1991) glosses it as an inchoative. Furthermore, as discussed in Section 2, since we have no occurrences of aspectual hors in our corpus, it is impossible for us to formulate an analysis of its semantics derived from the contexts that we observe it in. Because of these problems inherent to the study of historical data, we will leave the question of a structural account of the aspectual reading here. 4.2  Class 2 P-elements Class 2 is made up of amont “upwards”, aval “downwards”, contremont “upwards”, and contreval “downwards”. As mentioned earlier, we claim that these particles maintain their directionality when used transitively because of the interaction between their two main components: a prefix (a- or contre-), and a root (mont “mountain” and val “valley”). Unlike the class 1 P-elements ariere and avant, we argue that Class 2 P-elements are bi-morphemic. While -riere and -vant are not found independently in 13th century French, mont and val both exist as nouns, as well as verbs (as in (a)monter “to go up”, and (de)valer/avaler “to go down”). We propose that, like they do for the paths in the lexical semantics of monter and devaler, mont and val serve to specify the direction of the paths denoted by the P-element. We further propose that, in the case of the P-element, these paths are introduced by the prefixes a- and contre-. The OF prefix a- was studied in great detail by Dufresne, Dupuis & Longtin (2001), and these authors showed that the addition of a- onto a verb has the effect of either adding a locative ‘endpoint’ argument to a motion verb (i.e. adding a PATH argument to the verb) or making the realization of the verb’s PATH argument necessarily explicit. For example, in their OF corpus, they found that 81% of the uses of the verb aporter “to bring”

8.  Technically speaking, nothing in our proposal ensures that e and e’ be the same type of event; however, we can either complicate the way in which ariere is added to the VP to specify this, or we can assume that the fact that the two events must have most of the same properties is pragmatically required to make the truth of the sentence not trivial.



Variable-behavior Ps and the location of PATH in Old French 

were accompanied with a locative or dative “to” argument, compared to 38% of the uses of the unprefixed verb porter “to carry”. Furthermore, unlike its modern French homophone, OF contre- could also have a truly directional meaning in Old French, as illustrated in (31). (31) a.  b. 

Contre le tré Sébile a conduit ses conrois. Towards the tent Sebile aux lead his detachment. “He led his detachment towards Sebile’s tent” Saisnes, L, 1964, (from Buridant 2000: 497) Quant voit venir Charlon, contre lui est alés When sees come Charles, towards him aux went. “When he sees Charles, he goes towards him.” Saisnes, L, 2371, (from Buridant 2000: 497)

We therefore propose that the conceptual structures of a- and contre- are simply bare, unanalyzable PATH structures. (32) [Path ]

Model-theoretically, we claim that these prefixes denote the set of all paths. Merging a- or contre- with a nominal root restricts this set to just those that have the direction specified by the root. Thus, we propose that amont simply denotes the set of paths whose end points are higher than their start points, and has the conceptual structure shown below. (33) [Path UP]

We now consider why merging amont and a direct object results in a directional PP, not a locative one, as in the case of Class 1 prepositions. Since, under our proposal, Class 2 P-elements are inherently directional, that they remain directional when used transitively is hardly unsurprising. However, we propose that what distinguishes these P-elements from those in Class 1 on the one hand and other directional prepositions on the other hand is that Class 2 P-elements are semantically (and syntactically) intransitive, and because of this, the DPs they are paired with in PPs are actually locative adjuncts. Our argument for this analysis comes from a comparison between transitive Class 2 P-elements and other spatial expressions in OF that are at the same time both transitive and directional. For example, the preposition vers “towards” had these properties. (34) Vers dulce France chevauchet l’emperere Towards sweet France rides the.emperor “The emperor rides towards sweet France”  Roland 706, (cited in Buridant 2000: 496)

 Heather Burnett & Mireille Tremblay

In many theories of the semantics of directional prepositions (e.g. Kracht 2002; Zwarts 2005), the PP vers dulce France denotes the set of paths leading towards sweet France. In this example, the ground, dulce France, provides an endpoint to the paths denoted by the PP, a place with respect to which these paths are oriented. Thus Vers dulce France “towards sweet France” exemplifies the standard Path-Place structure, in which the DP dulce France is the argument of the preposition vers. (35) [Path TO [Place AT [Thing dulce France]]]

However, the relationship between amont and les degrez in the PP amont les degrez “up (on) the stairs” is different. Les degrez does not provide an endpoint to the paths that amont refers to (the PP would mean either “up to the stairs”, or “at the top of the stairs”, were this the case); rather, this expression provides a location for the paths themselves. We therefore propose that les degrez in the PP amont les degrez has the same function as in the pool in the sentence I swam in the pool: just as in the pool is an adjunct that provides information about the location of the swimming event, les degrez provides information about the location of the paths denoted by amont, namely that they are located on the stairs. Compositionally, merging amont and les degrez involves intersecting the union of the set of places that make up each path with the place occupied by les degrez. The conceptual structure of this PP is therefore as in (36), not as in (35). (36) [Path UP [Place LOCATION [Thing les degrez]]]

Since intersective semantics is a characteristic property of adjuncts (Heim & Kratzer 1998; Zwarts & Winter 2000, among others), we conclude that the relationship between ‘prepositional’ uses of Class 2 particles is best modeled as adjunction (37a), rather than complementation, as with the true preposition vers (37b). (37)

a.

PP P′





b. DP les degrez

PP P vers

DP dulce France

P Amont

4.3  The locative interpretation of particles With respect to the locative interpretation of particles with predicational contexts, we propose that the semantics of expressions exemplifying these contexts is identical to the semantics of copular sentences with locative PPs (like Li chevaliers est arriere le chastel “The knight is behind the castle”). The only difference is that sentences with locative Class 1 P-elements contain null pronouns that get their



Variable-behavior Ps and the location of PATH in Old French 

referents from either the context or the co-text, as proposed by Zribi-Hertz (1983) for Modern French.9 For example, consider (20a), repeated here in an expanded version as (38). (38)

il prent les cles et entre ou chastel. “He takes the keys and enters the castle”



Et si tost come il est enz, si voit “And as soon as he is inside (the castle), he sees,

parmi les ruesˉ tant de puceles que… among the streets, so many young.girls that…”

Qgraal, p. 49

Since the object of the preposition enz is recoverable from the previous sentence, it may be left implicit. 4.4  Summary In this section, we have presented a structural analysis of the syntactic and semantic variability of P-elements in Old French. We have argued that these lexical items should be divided into two classes based on their morphological composition and that this composition has semantic consequences. We first proposed that the simplex Class 1 P-elements ariere, avant, sus, jus, ens and hors have the underlying semantics of locative prepositions; however, since their complement does not need to be syntactically realized, they may be merged by themselves with a motion verb, at which point they acquire a directional interpretation. We then proposed that the complex Class 2 P-elements amont, aval, contremont, and contreval are underlyingly intransitive directional particles denoting sets of paths. Since they are semantically and syntactically intransitive, merging one of these particles with a DP is only possible if this DP is construed as a locative adjunct. This analysis makes two predictions. First, while the present analysis does not rule out adjunction structures for Class 1 P-elements, given the structural difference between Class 1 and Class 2 P-elements, and the fact that Class 1 prepositions may license implicit arguments, structural differences between the two types of adjunction structures are expected. This prediction is borne out: Class 1 particles require the presence of an additional prepositional element en “in”, as exemplified in (39), giving rise to an adjunction structure such as (40).10

9.  For a discussion of implicit prepositional objects in Old French, see Tremblay, Dupuis & Dufresne (2003). 10.  As discussed earlier, we know that prepositional strings such as enz enl fou and sus el paleis in (39) form a constituent since they can be fronted as a single unit.

 Heather Burnett & Mireille Tremblay

(39) a. Enz en l fou la getterent com arde tost In in. the fire her throw as burn soon “They threw her in the fire so that she burn quickly.”  Eulalie, 19, (cited in Buridant 2000: 488) b. Sus el paleis est li marchis venuz Up in.the palace is the marquis come “The marquis came up in the big room.”  Aliscans 2963, (cited in Buridant 2000: 488) (40) [PP [P’ [P enz ] pro ] [PP enl fou ]]]

A second prediction is also made by the proposed analysis. Recall that, unlike Class 2 P-elements which are inherently directional, intransitive Class 1 P-elements are inherently locative but get a directional interpretation when combined with a motion verb. In principle, nothing precludes these elements from also shifting in a transitive context and thus we expect to find at least a subset of Class 1 P-elements which are ambiguous between a locative and a directional reading. This prediction is also borne out. For example, the preposition fors “out” gets a locative reading in a predicative context (41a), but a directional reading in the context of a motion verb (41b). (41) a.

… et quant il fu hors del bois, … … and when he was out of.the wood… “…And when he got outside the wood…”

Artu, p. 75

b. Si cort meintenant a une espee qui ert seur un lit, si If run now to a sword which was on a bed, thus

la trest hors del fuerreˉ it-draw out of-the sheath

“He now runs towards a sword which was on a bed and unsheathes it”  Artu, p. 109

However, the opposite is not true and Class 2 P-elements cannot get a locative reading since they are inherently directional. 5.  Conclusion This paper argues that most, if not all, semantic and syntactic ambiguities in the Old French particle system should be treated structurally, not lexically. We argued that an analysis that postulates an independent lexical item for each difference in interpretation or transitivity has no account for both the systematicity of the particle/preposition alternation and the predictability of a particle’s interpretation from the context into which it is inserted. We then gave both a conceptual



Variable-behavior Ps and the location of PATH in Old French 

and a model-theoretic analysis of the semantics of these elements showing exactly how their different interpretations are computed from the interaction between the semantics of other lexical items and general compositional rules. We therefore conclude, along with Gerhke (2008), that the approach to ambiguity in the category P that derives variability from structure and mode(s) of composition is a fruitful research project, and should be continued with studies of other languages. Our analysis also has implications for the standard view of the syntax and semantics of particles. Very often, particles are analyzed as straightforward instantiations of PATH (e.g. Svenonius 2006). However, we have seen that, in Old French, an entire class of particles are underlyingly locative, i.e. PLACE. So it seems that the possible semantic values for syntactically intransitive P-elements are richer than it is generally thought. Nevertheless, whether or not PATH/PLACE divisions such as these are common in the cross-linguistic realization of the category P is a question left to future research.

References Beck, Sigrid & Kyle Johnson. 2004. “Double Objects Again”. Linguistic Inquiry 35: 1. 97–123. Bierwisch, Manfred. 1988. “On the Grammar of Local Prepositions”. Syntax, Semantik, und Lexicon, Studia Grammatica XXIX ed. by Manfred Bierwisch, Wolfgang Motsch & Ilse Zimmerman. 1–65. Berlin:Akademie Verlag. Buridant, Claude. 2000. Grammaire Nouvelle de l’Ancien Français. Paris: Sedes. Burnett, Heather, Katrina Petrik & Mireille Tremblay. 2005. “La Grammaire des Particules en Ancien Français: Sémantisme, distribution et perte de productivité”. Proceedings of the 2005 Canadian Linguistics Association Annual Conference ed. by Claire Gurski. http://ling.uwo. ca/publications/CLA-ACL/CLA-ACL2005.htm Burnett, Heather & Mireille Tremblay. In press. “Directionalité et Aspect en Ancien Français: L’apport du système prépositionnel”. Approches de la Variation Linguistique Gallo-romane ed. by Mario Barra-Jover, Guylaine Brun-Trigaud, Jean-Philippe Dalbera, Patrick Sauzet, Tobias Scheer & Philippe Ségéral. Saint-Denis: Les Presses universitaires de Vincennes. Dufresne, Monique, Fernande Dupuis & Mireille Tremblay. 2003. “Preverbs and Particles in Old French”. Yearbook of Morphology 2003 ed. By Geert Booij & Ans van Kemenade, 30–60. Great Britain: Kluwer Academic Publishers. Dufresne, Monique, Fernande Dupuis & Catherine-Marie Longtin. 2001. “Un Changement dans la Diachronie du Français: La perte de la préfixation aspectuelle en a-”. Revue québécois de linguistique 29: 2. 33–54. Gehrke, Berit. 2007. “On Directional Readings of Locative Prepositions”. Proceedings of Console XIV ed. by Sylvia Blaho, Luis Vicente & Erik Schoorlemmer, 99–120. Leiden: Student Organisation of Linguistics in Europe. Gehrke, Berit. 2008. Ps in Motion: On the Semantics and Syntax of P Elements and Motion Events. Ph.D. Thesis, Utrecht University. LOT Dissertation Series 184. Greimas, Algirdas Julien. 2001. Dictionnaire de l’Ancien Français, 3rd edition. Paris: Larousse-Bordas. Heim, Irene & Angelika Kratzer. 1998. Semantics in Generative Grammar. Malden & Oxford: Blackwell.

 Heather Burnett & Mireille Tremblay Iacobini, Claudio & Francesca Masini. 2007. “Verb-particle Constructions and Prefixed Verbs in Italian: Typology, Diachrony and Semantics”. On-line Proceedings of the Fifth Mediterranean Morphology Meeting (MMM5) Fréjus, 15–18 September 2005 ed. by Geert Booij et al. http://mmm.lingue.unibo.it/ Jackendoff, Ray. 1983. Semantics and Cognition. Cambridge, Mass.: MIT Press. Jackendoff, Ray. 1990. Semantic Structures. Cambridge, Mass.: MIT Press. King, Ruth. 2000. The Lexical Basis of Grammatical Borrowing: A Prince Edward Island French case study. Philadelphia: Benjamins. Koopman, Hilda. 2000. “Prepositions, postpositions, Circumpositions, and Particles”. The Syntax of Specifiers and Heads ed. by Hilda Koopman, 204–260. London: Routledge. Kracht, Marcus. 2002. “On the Semantics of Locatives”. Linguistics and Philosophy 25. 157–232. Masini, Francesca. 2005. “Multi-Word Expressions between Syntax and the Lexicon: The Case of Italian Verb-Particle Constructions”. SKY Journal of Linguistics 18. 145–173. Masini, Francesca. 2006. “Diacronia dei Verbi Sintagmatici in Italiano”. Archivio Glottologico Italiano XCI:1. 67–105. Mateu, Jaume & Gemma Rigau. 2007. “Romance Paths as Cognate Complements: A lexicalsyntactic account”. Ms., Universitat autònoma de Barcelona. Nam, Seungho. 1995. The Semantics of Locative Prepositional Phrases in English. Ph.D. thesis, UCLA. Perrot, Marie-Éve. 1994. “Le chiac ou… whatever: Le vernaculaire des jeunes d’une école secondaire francophone de Moncton”. Études Canadiennes 37. 237–46. Santucci, Monique. 1991. La Mort du Roi Arthur. Translated from Jean Frappier’s edition. Paris: Champion. Svenonius, Peter. 2006. “The Emergence of Axial Parts”. Nordlyd: Tromso Working Papers in Linguistics 33. 49–77. Talmy, Leonard. 1985. “Lexicalization Patterns Semantic Structure in Lexical Forms”. Language Typology and Syntactic Description III: Grammatical Categories and the Lexicon ed. by T. Shopen, 57–149. Cambridge: Cambridge University Press. Tobler, Adolf, Erhard Lommatzsch & Hans Helmut Christmann. 2002. Tobler-Lommatzsch: Altfranzösisches Wörterbuch, XI. Electronic edition by Peter Blumenthal & Achim Stein. Stuttgart: Franz-Steiner-Verlag. Tremblay, Mireille. 2005. “Back en Français Acadien: Archaïsme ou innovation?”. Français d’Amérique: Approches morphosyntaxiques: Actes du colloque Grammaire comparée des variétés des français d’Amérique, Université d’Avignon, ed. by Patrice Brasseur & Anika Falkert, 263–73. Paris: L’Harmattan Tremblay, Mireille, Fernande Dupuis & Monique Dufresne. 2003. “Les Prépositions dans l’Histoire du Français: Transitivité, grammaticalisation et lexicalisation”. Verbum 25. 549–62. Verkuyl, Henk & Joost Zwarts. 1992. “Time and Space in Conceptual and Logical Semantics: The notion of path”. Linguistic 30. 483–511. Zribi-Hertz, Anne. 1983. “Prépositions Orphelines et Pronoms Nuls”. Recherches Linguistiques 12. 46–91. Zwarts, Joost. 2005. “Prepositional Aspect and the Algebra of Path”. Linguistics & Philosophy 28. 739–79. Zwarts, Joost. 2006. “Event Shape: Paths in the Semantics of Verbs”. Ms., Nijmegen & Utrecht Universities. Zwarts, Joost & Yoad Winter. 2000. “Vector Space Semantics”. Journal of Logic, Language and Information 9. 171–213.

Hebrew and Arabic children going Romance On the acquisition of word order in Semitic and Romance* João Costa1 & Naama Friedmann2 1Universidade

Nova de Lisboa/2Tel-Aviv University

In Hebrew, European-Portuguese, Spanish, and Palestinian-Arabic both SV and VS orders are possible. In the early stages of sentence construction, however, children acquiring these languages do not use the whole array of word orders in their language. Their word order preference differs in the different languages: in the first stage, Hebrew and European-Portuguese children use both SV and VS orders with unaccusative verbs, but only SV with unergative and transitive verbs. In Spanish and Palestinian-Arabic, on the other hand, children prefer VS order with unaccusative, unergative, and transitive verbs. We propose an account according to which children at this stage cannot move the subject outside of VP yet, and the cross-linguistic difference stems from the identification of Spell-Out Domains (Fox & Pesetsky 2004a). Spanish and Palestinian-Arabic allow the verb to appear before the subject, whereas Hebrew and European-Portuguese do not allow the verb to move to I until the Spell-Out domain widens beyond VP, which takes place after subject movement to Spec,IP is acquired.

1.  Introduction In recent literature on the acquisition of word order, attention has been paid to the development of word orders related to different movement types. However, most results focus on determining whether specific movement operations have been acquired at a given stage of development, and there are very few comparative

*We are grateful to the research students with whom we collaborated in these experiments. Nivin Shahin and Ula Huri from Tel-Aviv University in the experiments in Palestinian Arabic, Ivana Nachman-Katz from Tel-Aviv University in the experiments in Spanish, Maria do Mar Adragão from Universidade Nova de Lisboa in European Portuguese, and Lilach Shvimer, Noa Kerbel, and Julia Reznick from Tel-Aviv University in the Hebrew experiments. This research was supported by the Lieselotte Adler Laboratory for Research on Child Development, and the writing of the manuscript was supported by the COST A33 European initiative.

 João Costa & Naama Friedmann

results, based on similar methodologies, to assess whether there are any important crosslinguistic differences in the acquisition of word order. In this paper, we present the summary of several experiments eliciting SV and VS orders in four different languages: Hebrew, European-Portuguese, Spanish, and PalestinianArabic. Our aim is twofold: first, we show that children do not arbitrarily produce any word order. The emergence of a specific word order depends on the stage of acquisition of specific movement types, together with characteristics of the language being acquired and on specific verb classes. Second, we show that the four languages cluster in two groups: European Portuguese and Hebrew behave alike, and Spanish and Palestinian Arabic also behave alike. We offer an account of the witnessed variation within and between languages which ties the acquisition of various movements and the setting of Spell-Out domains for linearization, following proposals by Fox and Pesetsky (2004a, 2004b), taking into account independent properties of each language set. 2.  Word order in the four languages under consideration Hebrew, European-Portuguese, Spanish, and Palestinian-Arabic are very similar in what regards the available word orders for subject and verb in adult grammar. In fact, as shown in (1)–(4), the four languages display SV/VS variation both with unergatives and with unaccusative verbs.

(1) Hebrew:

a.

Unaccusative verbs



Ha-ale nafal. / Nafal ha-ale. the-leaf fell fell the-leaf “The leaf fell”

b. Unergative verbs

Etmol ha-yeled rakad. / Etmol rakad ha-yeled. yesterday the-boy danced   yesterday danced the-boy “Yesterday the boy danced.”

(2) European Portuguese:

a.

Unaccusative verbs



Um rapaz chegou. / Chegou um rapaz. a boy arrived arrived a boy “A boy arrived.”

b. Unergative verbs

O rapaz dançou. / Dançou o rapaz. the boy danced danced the boy “The boy danced.”





Hebrew and Arabic children going Romance 

(3) Spanish:

a.

Unaccusative verbs



Un chico llegó. /Llegó un chico. a boy arrived arrived a boy “A boy arrived.”

b. Unergative verbs

Un chico habló. /Habló un chico. a boy spoke spoke a boy “A boy spoke.”

(4) Palestinian Arabic:

a.

Unaccusative verbs



Al-balon infajar. / Infajar al-balon. the-balloon popped popped the-balloon “The balloon popped.”

b. Unergative verbs

Al-walad biki. / Biki al-walad. the-boy cried cried the-boy “The boy cried.”

For the four languages we assume that in the adult grammar preverbal subjects are base-generated within the VP and then moved to Spec,IP in the SV orders. The VS order in European Portuguese, Spanish, and Palestinian Arabic has been analyzed by several authors as a case of in-situ subject: with unergatives, we assume that the subject is in Spec,VP and the verb has moved to I. With unaccusatives, the verb moves to I, and the argument is either in Spec,VP or in its base-generated complement position (cf. Ouhalla 1994; Shlonsky 1997; Costa 1998; Mohammad 2000, among many others). As for Hebrew, we follow that analysis according to which VS orders with unergatives and transitives are a consequence of I-to-C movement (cf. Shlonsky 1997). The similarity and differences between the four languages in terms of word order patterns raises the following questions for language acquisition: 1. Do children produce all word orders for all verb classes? 2. Are there crosslinguistic differences in the acquisition of subject-verb order? In what follows, we try to answer these two questions. 3.  The acquisition of subject-verb order: Results In order to assess the children’s ability to produce sentences with unaccusative and unergative verbs, we used three methods: two structured tasks, sentence repetition

 João Costa & Naama Friedmann

and story retelling, as well as the analysis of spontaneous speech. For details regarding the structure of the tests, we refer the reader to Friedmann and Costa (2007). In the analysis of spontaneous speech data, only utterances containing both subject and verb were considered. In the repetition tasks, the sentences were read by a native speaker of the relevant language, and the children were asked to repeat the sentences as accurately as they could. For each sentence they repeated (regardless of their success) they were awarded with a building block which they used to build a building-block tower ‘way up to the sky’. Whenever the child requested, the experimenter repeated the sentence, as many times as the child needed. In the story retelling tasks, the children were looking at an illustrated story, which was read to them enthusiastically by the experimenter, with matching facial expressions and gestures, in order to keep the child as attentive and focused as possible. After reading some of the sentences, the child was requested to tell the story s/he has just been told. The Figures present the percentage of correct responses in each condition. Let us consider the overall results for each of the languages separately. For space reasons, we only present the general results; for a more detailed presentation of the analyses and results obtained for each task, see Friedmann and Costa (2007). 3.1  Palestinian Arabic For Palestinian Arabic, two repetition tests were used. One included repetition of SV and VS sentences with unaccusative and unergative verbs by 20 children aged 1;10–3;0. The second test included repetition of SVO and VSO sentences with transitive verbs by 27 children aged 2;0–3;6. The results of both experiments were very consistent, as seen in Figure 1. Both revealed a strong preference at the early stage of word order acquisition for VS word order: Before the age of 2;5, children correctly repeated VS and VSO sentences at a rate above 50%, but were very poor at repeating SV sentences. They correctly repeated only 20% of the SV sentences with both unaccusative and unergatives, and were 27% correct on SVO sentences. No significant difference was found between the repetition of unaccusative and unergative verbs. 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

SV VS

1;10–2;4

2;6–3;0

90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

SVO VSO

2;0–2;5

2;6–3;0

3;1–3;6

Figure 1.  Repetition of SV and VS order in Palestinian Arabic. Left: repetition of unergative and unaccusative verbs, right: repetition of transitive verbs



Hebrew and Arabic children going Romance 

3.2  Spanish The data in Spanish are taken from the analysis of spontaneous speech in Iberian Spanish, and from two structured tests in Argentinean Spanish, repetition and story retelling. For the spontaneous speech, we analyzed 29 samples of a Spanish child, containing 11,715 utterances between the ages 1;7–2;7 (from CHILDES, Aguirre database, MacWhinney, 2000), which we classified for verb class and word order. The story retelling test included 20 Argentinean-Spanish speaking children aged 2;0–4;0, and tested the production of SV and VS sentences with unaccusative verbs. The repetition task included 17 Argentinean-Spanish speaking children aged 2;8–4;0, and tested SV and VS sentences with unaccusative and unergative verbs. Very similarly to the observations from Palestinian Arabic, the Spanish results reveal a preference for VS for both unaccusative and unergative verbs. As seen in Figure 2, the analysis of spontaneous speech yielded a clear preference for VS order: the child produced 35% of his sentences that included an unergative verb in SV order, compared to 65% in VS order, and 38% of the sentences with unaccusatives in SV order, compared with 62% in VS order. The story retelling task showed that the younger children (aged between 2 and 3) could not produce any sentence in SV order in unaccusative contexts, but still produced 17% VS orders. The children kept showing a clear difference between SV and VS until they were 4 years old. In a very similar manner, the scores on the repetition test showed the same preference (albeit with better overall performance) for production of VS in the younger group (from 2;8 to 3;6). Like in Palestinian Arabic, no significant differences were found between unaccusative and unergative verbs in either SV or VS orders.

70% 60% 50% 40% 30% 20% 10% 0% 1;7–2;7

80% 70% 60% SV 50% VS 40% 30% 20% 10% 0%

100% SV VS

90%

SV VS

80% 70% 60%

2:0–3:0

3:1–3:6

3:7–4:0

50%

2:8–3:6

3:7–4:0

Figure 2.  SV and VS order in Spanish. From left to right: unergatives and unaccusatives in spontaneous speech; story retelling with unaccusative verbs; repetition of unergative and unaccusative verbs

3.3  European Portuguese For European Portuguese, we used a repetition task and spontaneous speech data. The repetition task consisted of sentences with unaccusative or unergative verbs

 João Costa & Naama Friedmann

in SV and VS order, as well as Wh-questions involving I-to-C movement, and sentences with unambiguous V-to-I movement, that is, with the order SVAdvO. The participants were 21 children aged 2;1 to 3;0. The picture that emerged from European Portuguese was crucially different from what we reported above for Palestinian Arabic and Spanish. In European Portuguese, the results reveal a strong preference for SV order in unergative contexts (93% correct repetition of SV vs. only 42% for VS, see Figure 3). On the contrary, in the unaccusative context, there was no significant difference between SV and VS, and the repetition of VS sentences was significantly better with unaccusative than with unergative verbs. Children had difficulties in the repetition of I-to-C (12% correct) and V-to-I (17% correct). The analysis of spontaneous speech of a Portuguese child, when he was between 2;7 and 3;7 (reported in Adragão 2001 and Adragão & Costa 2004), yielded similar results. Whereas in unaccusative contexts both SV and VS orders were used, there was an overwhelming preference for SV in unergative and transitive contexts. In contexts in which information structure required VS, VS was appropriately used with unaccusative verbs (since children can use the VS order with this verb class, there is no reason for them to use a pragmatic infelicitous word order). Crucially, with unergative verbs VS order was hardly used, even in these VS obligatory contexts. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

SV VS

unergative

unaccusative

80% 70% 60% 50% 40% 30% 20% 10% 0%

SV VS

unergative

unaccusative

Figure 3.  SV and VS order in European Portuguese. Left: repetition of unergative and unaccusative verbs in SV and VS orders; Right: spontaneous speech data- sentences produced when the context required VS sentences with unaccusative or unergative verbs

3.4  Hebrew For Hebrew, we used data from repetition, story retelling, and spontaneous speech analysis. For repetition, in Experiment 1, reported in Friedmann and Lavi (2006), we tested 60 children aged 2;2–3;10, who were asked to repeat SV and VS sentences with unaccusative, unergative, and transitive verbs. In Experiment  2, a similar task was run on 18 children aged 2;2–3;10 for the repetition of unaccusative and transitive verbs in the orders SV(O) and VS(O) (Experiments 2, 3, 5, and



Hebrew and Arabic children going Romance 

the first spontaneous speech analysis are presented in detail in Friedmann 2007). In Experiment 3, 20 children aged 2;0–4;0 repeated sentences with transitive and unaccusative verbs with a sentence-initial temporal adverb, which may serve as a trigger for subject verb inversion in Hebrew. The fourth repetition experiment included 40 sentences in SV and VS order – half declarative, half object and adjunct Wh-questions, with 22 children aged 3;7 to 4;10. In the story retelling task (Experiment 5), 17 children aged 1;9–2;0 retold a story containing unaccusative verbs in SV and VS orders, and unergatives in SV order. In addition to the structured tests, we also used spontaneous speech in three different types of analyses. We first analyzed the spontaneous speech of 21 children aged 1;6–2;11 for the number of occurrences of unaccusative and unergative verbs in SV and VS orders (these transcripts were taken from Berman & Dromi 1984; Berman 1985; MacWhinney 2000). We then analyzed 56 samples including a total of 6400 utterances of children between the ages 1;6 and 6;1 (MacWhinney 2000; Davidson 2002; Bibi 2003) looking for each individual whether or not there were occurrences of sentences including SV and VS orders with various verb types. Finally, a longitudinal analysis of the spontaneous speech of one child between ages 1;7–2;11 including 2269 utterances was analyzed for the occurrence of sentences including SV and VS orders with various verb types in each monthly sample. All experiments revealed that SV order is strongly preferred for unergative and transitive verbs (with the rates of SV production ranging from 73% to 100%, and the rates of VS production ranging from 5% to 34% in the younger groups), whereas for unaccusative verbs both word orders are used (there was no significant difference between the rates of production for SV and VS, and both word orders were successfully produced by the youngest children). When the children were as young as 1;9, they could already produce both SV and VS with unaccusatives in repetition, story retelling, and spontaneous speech. When they were as old as 4;10, they still could not repeat or produce in story retelling VS sentences with unergative and transitive verbs (see also Zuckerman 2001). Similarly, the samples of 56 children indicated that only 10 of the 56 children ever produced VS sentences with unergatives or transitives, and the spontaneous speech of the longitudinal sample indicated that out of 2269 utterances, only a single utterance included V-to-C, and this occurred at age 2;6. The highlights of these results are presented in Figure 4. This pattern is similar to our findings for European Portuguese, with one exception – in both Hebrew and European Portuguese, the production of VS in transitive and unergative contexts is delayed, but in Hebrew it appears in production even later than in European Portuguese: VS was found only after the production of Wh movement (See Friedmann & Lavi 2006 for the acquisition of V-C later than Wh-movement). Taken together with the need for a specifier in CP to trigger VS in Hebrew, but not in European Portuguese, these differences in time table

 João Costa & Naama Friedmann

might relate to the fact that VS in Hebrew involves I-to-C movement, whereas in European Portuguese the verb can move to I in VS sentences. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

unergative/ transitive SV unergative/ transitive VS 2:2–2:9

2:10–3:2

3:3–3:10

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

unaccusative SV unaccusative VS transitive SVO transitive VSO 2;3–2;8

100% 90% 80% Transitive VSO 70% 60% Wh question SV 50% Wh question VS 40% 30% 20% 10% 0%

2;11–3;4

3;6–3;10

Transitive SVO

3;7–4;10

unergative SV unergative VS unaccusative SVO unaccusative VSO

1;6–2;11

Figure 4.  Hebrew data. From left to right, then bottom: Repetition of SV and VS with unergative and transitive verbs (Experiment 1); Repetition of unaccusative and transitive verbs in SV and VS order (Experiment 2); repetition of declarative and interrogative sentences in VS and SV orders (Experiment 4); Spontaneous speech of 21 children

3.5  Summary According to the results presented above for the four languages, two very clear and very different patterns emerge at the early stage of acquisition: Hebrew and European Portuguese pattern together, showing preference for SV order in sentences with unergative and transitive verbs, and allowing both SV and VS orders with unaccusative verbs. Spanish and Palestinian Arabic also pattern together, but form a different pattern: both prefer VS for all verb types in the early stage (see Table 1). Table 1.  Summary of the results from the four languages – production of SV and VS orders in the early stage of word order acquisition

Hebrew European Portuguese Spanish Palestinian Arabic

Unergative/ transitive SV(O)

Unergative/ transitive VS(O)

Unaccusative

Unaccusative

SV

VS

  – –

– –  

  – –

   



Hebrew and Arabic children going Romance 

4.  Discussion Summarizing the word order patterns in the four languages, children acquiring Spanish and Palestinian Arabic prefer to use VS over SV both with transitive and unergative and with unaccusative verbs, whereas children acquiring Hebrew and European Portuguese use both SV and VS with unaccusative verbs, but only SV with unergative verbs. How can these patterns be accounted for? Firstly, the results from Hebrew and European Portuguese indicate that from very early on children distinguish between unaccusative and unergative verbs, in line with many previous studies (cf. Friedmann 2007 and Guasti 2002, among others). Based on the results just presented, the following questions must be answered: 1. Why is there no distinction for verb classes in Spanish and Palestinian Arabic and VS is preferred in all contexts? 2. Why is SV ruled out in Spanish and Palestinian Arabic? 3. Why is VS dispreferred in European Portuguese and Hebrew only with unergatives and transitives? 4. What underlies the similarities and differences between the four languages? We make the following assumptions: at the relevant ages, V-to-I has been acquired. This is supported, for instance, by verb placement relative to adverbs and negation by the age of 2 in English and French (Pierce 1989, 1992; Déprez & Pierce 1993, 1994) and Friedemann (1993/1994, 2000), and by the mastery of VP-ellipsis in verbal answers in European Portuguese (Santos 2006). We further assume that, at this stage, subject movement outside the VP is still difficult, and children avoid it whenever possible. This assumption is supported by the observation that children leave subjects in the VP-internal base-generated position in several languages (Déprez & Pierce 1993 and Friedemann 1993/1994, 2000). These two assumptions of early acquisition of V-to-I and late acquisition of subject movement apply to all children acquiring language, and are independently motivated in the literature. We depart, however, from some literature in assuming that the fact that V-to-I has been acquired does not necessarily mean that the verb is moved in all contexts. In fact, it is well known that in languages with V-movement, the verb may remain in situ under certain circumstances. For instance, there are clear connections with inflectional morphology determining V-movement, and it is known that root infinitives, in the relevant languages, are not moved to I (see Guasti 2002 for a summary). As for unaccusative verbs, we assume that VP-internal movement applies, displacing the argument from object to subject position. This VP-internal movement operation is independently argued for in Friedmann (2007) and Silva (2004) for

 João Costa & Naama Friedmann

Hebrew and European Portuguese, respectively. As suggested by many authors for several languages, the argument of unaccusatives may remain in situ in adult grammar, which makes this movement internal to VP optional. We will assume its existence and optionality to hold both in adult and in child grammar. Bearing these assumptions in mind, let us look at how each of the word order patterns identified can be explained. We contend that the main difference between the two language groups has to do with verb movement to I. We propose that in Hebrew and European Portuguese, the verb does not move to I, whereas in Spanish and Palestinian Arabic, it does. Obviously, this difference calls for motivation. Before justifying this difference, let us consider how the word order patterns follow from this proposal and the set of assumptions spelled out above. Let us start with the SV/VS alternation in unaccusative contexts in Hebrew and European Portuguese. If the verb is not moved from within VP, then because the subject stays within the VP, either in object or in subject position, the two word orders follow. A similar analysis accounts for Palestinian Arabic and Spanish, but, since there is V-to-I movement and the subject cannot leave the VP, independently of the position of the argument within the VP, the moved verb will always precede it. Therefore the two positions for the subject will not yield different word orders, and only VS arises. With respect to unergative verbs, in Hebrew and European Portuguese, all constituents stay within VP, whereas Spanish and Palestinian Arabic display VS order, because the verb moves to I, and the subject stays in situ. The question is why V-to-I applies only in Spanish and Palestinian Arabic, but not in Hebrew and European Portuguese, considering the assumption made above that V-to-I is already available for all these languages. In order to account for why the verb does not move to I in Hebrew and European Portuguese at this stage, we follow the proposal by Fox & Pesetsky (2004a), according to which the word order at the clause level must follow the linearization established at the first Spell-Out domain. Consider a case in which the VP is created, and the constituents are linearized with the order SVO. In this context, subject movement out of VP will not destroy the linear order established within VP, because the subject still precedes the verb. However, if the object is moved out of the VP, the precedence relation is destroyed, since at the VP domain the object follows the V, but at the higher domain it precedes the V. For this reason, Fox & Pesetsky (2004a) propose that movement operates through the edge, enabling preservation of the linear order. If the object moves through the edge of VP, the order O>S is established within the VP phase, and it is preserved at the higher phase. This approach derives successive cyclicity from linearization. Importantly, this type of analysis predicts that VSO orders are not good, since moving an element that is not at the edge will disrupt the linear order created at the VP. Fox & Pesetsky (2004b) explain a similar problem, VS orders in



Hebrew and Arabic children going Romance 

Scandinavian V2 languages, by suggesting that the domain for Spell-Out can vary across languages. According to them, linearization in Scandinavian V2 is established only at the CP domain. We will explore the idea that Spell-Out domains can differ not only between different adult languages, but also between child and adult grammars, and between stages of acquisition. We suggest that Hebrew and European Portuguese children first assume that the Spell-Out domain is VP, since there is no reason to assume a higher domain, whereas Spanish and Palestinian Arabic children have evidence to assume that linearization is determined at least at the IP level. The crucial difference between the two sets of languages is the availability of clitic doubling only in the latter two languages, as shown in (5) and (6):

(5) Spanish:1

a. Lo vi a Juan ayer. Him saw-1-sg to Juan yesterday Palestinian Arabic (adapted from Aoun 1996): b. shuft-o la-Ayman. saw-1-sg-him to-Ayman

(6) European Portuguese:

a.

Vi (*-o) ao João ontem. Saw-1-sg him to João yesterday

Hebrew: b. ra’iti (*-v) et Yoni etmol. saw-1-sg him acc Yoni yesterday

According to proposals in the literature on clitic doubling (e.g. Torrego 1998), in clitic doubling languages, clitics and their doubles are generated as a single constituent, but the clitic must undergo movement to the IP domain, since it is licensed there. Contrarily, in languages without clitic doubling, DPs are the arguments bearing Case and theta-role and there is no part of the internal argument waiting for licensing at the IP-level. For this reason, in languages without clitic doubling, argument licensing is solved at the VP domain, whereas in clitic doubling languages, only at the IP-level, where clitics are licensed, theta-role assignment is completed. Accordingly, in Spanish and Palestinian Arabic, only IP counts as a phase, since it is the first level at which the predicate saturates its thematic 1.  Not all varieties of Spanish are equally permissive with clitic doubling. What is crucial for our point is that in all varieties of Spanish clitic doubling is allowed and present in the child’s input. Evidence for early sensitivity to clitic doubling even in Iberian Spanish – the variety with least clitic doubling – is given in Torrens and Wexler (1999).

 João Costa & Naama Friedmann

array. Therefore, children acquiring clitic-doubling languages have evidence to assume IP as the first domain for linearization, whereas children acquiring non clitic-doubling languages start out with the VP as the spell-out domain, since this is the domain at which thematic roles are fully assigned. Naturally, for this proposal to hold, it is necessary that children can interpret clitics and are sensitive to the existence of clitic doubling in the relevant languages. In fact, longitudinal data from Torrens and Wexler (1996) reveals that Spanish children master clitic doubling at the age of 1;7. Based on these considerations, we can now explain why at the first stage of acquisition VS with V in I and subject in situ is impossible for Hebrew and European Portuguese children (recall that VS will be possible with unaccusatives in these languages, since it does not involve movement out of VP), but available for Spanish and Palestinian Arabic children. Children who take VP as their Spell-Out domain cannot move the verb to I, because this does not preserve the linearization established within VP, the SpellOut domain. The only way for them to preserve linearization domain and move the verb to I would be to move the subject to Spec,IP, but at this stage they cannot do that, so they leave both the verb and the subject within VP. According to Fox & Pesetsky (2004a, 2004b), another possibility for achieving V-movement without movement of any of the arguments is VP deletion, since linearization, a PF matter, will be irrelevant for unpronounced material. Thus, an option that is open for the children, if they move the verb to I, is to delete the VP, because linearization applies only to phonologically overt material. This predicts that children who take VP as their Spell-Out domain can either leave both the subject and the verb within VP, or do VP ellipsis. We have already seen that in the majority of cases they leave the V in VP. In fact, there are also cases in which we can see they use the other option and delete the VP. According to Santos (2006), Portuguese-speaking children use VP ellipsis in answers to yes/no questions.2 Because VP ellipsis in Portuguese necessarily involves V-to-I, the fact that children use VP Ellipsis forms the strongest evidence for the early acquisition of

2.  An example of VP-ellipsis in verbal answers in European Portuguese: (i) A: Tu vais comprar o livro? You go-2ND buy-INF the book “Are you going to buy the book?” B: Vou. Go-1ST “Yes.”



Hebrew and Arabic children going Romance 

V-to-I movement. Adopting the linearization hypothesis, we can explain why in some languages children do not move the verb in all constructions, although they can perform this operation, as shown in other structures. This also allows us to predict the correlation between the use of VP ellipsis and V-to-I. In contrast, children who speak Spanish and Palestinian Arabic already assume a wider Spell-Out domain, and therefore moving the verb to I and leaving the subject within VP does not create a linearization problem for them, because the level at which precedence relations are established is the IP. As a result, they move the verb to I because they already acquired V-to-I and nothing in their syntax prevents it. Because subject movement to Spec,IP has not been acquired yet, they produce the order VS for both unaccusative and unergative/transitive verbs. This analysis describes the first stage of sentence construction. It must be explained now what causes the change enabling all children to use all possible word orders in their languages. We need to explain how Spanish and Palestinian Arabic children start using SV orders with unaccusative and unergative/transitive verbs, and how Hebrew and European Portuguese children start using VS orders with unergative and transitive verbs. The key for all the changes in the second stage is that children acquire movement of the subject to Spec,IP. The acquisition of subject movement immediately guarantees that Spanish and Palestinian Arabic children start producing SV orders for all types of verbs. As for Hebrew and European Portuguese children, acquiring subject movement makes it possible to move the verb to I in all constructions. This allows generating SV orders with both the subject and the verb in the IP domain, which competes with the alternative representation for SV, in which both are left in situ. Knowing that V-to-I is obligatory in the languages with a positive setting for this parameter, the acquisition of subject movement provides the necessary condition for children to posit obligatory V-to-I and abandon the in-situ representation for SV orders. The fact that both languages have the rich inflection generally assumed to drive obligatory V-to-I makes it legitimate to assume that the target languages include a positive setting for this parameter. Because V now moves obligatorily – even when the subject does not, creating VS orders in European Portuguese – the Spell-Out domain must be widened at least to IP, in a way similar to what is proposed by Fox and Pesetsky (2004b) for Scandinavian obligatory V2. A crucial difference between Hebrew and European Portuguese is that transitive and unergative VS in Hebrew involves I-to-C and the subject in Spec,IP, and a filled specifier in CP, whereas in European Portuguese, VS is possible with the verb in I and the subject in situ (and no trigger in spec,CP). This difference makes the prediction that VS orders with unergatives/transitives in European Portuguese are acquired by the time subject movement is acquired, but in Hebrew they will wait

 João Costa & Naama Friedmann

until I-to-C is acquired (at this stage VS is possible with unaccusatives, because it does not require the movement of the verbs to C). This is consistent with the proposal made above, according to which acquisition of specific movement types triggers the widening of the Spell-Out domain. Specifically, acquisition of subject movement triggers the widening of the Spell-Out domain from VP to IP for European Portuguese, and acquisition of Wh movement and I-to-C triggers the widening of the Spell-Out domain from VP to CP in Hebrew. In fact, when data from language acquisition in Hebrew and European Portuguese is examined, it is possible to observe that I-to-C is acquired later than VS in European Portuguese (Soares 2003), and that in Hebrew, VS orders only appear after there is independent evidence for Wh-movement to Spec,CP (Friedmann & Lavi 2006). The stages of acquisition can thus be summarized as follows (HEP: Hebrew and European Portuguese; SPA: Spanish and Palestinian Arabic):

Stage of acquisition Stage 1: V-to-I available Object-to-subject movement within VP available Subject movement out of the VP unavailable    HEP: VP as Spell-Out domain →        V does not move to I

Children’s performance

   SPA: IP as Spell-Out domain →       V moves to I Stage 2: Subject movement to spec-IP acquired →    HEP: V moves to I →    Obligatory V-to-I established →    VS sentences in Portuguese →    Spell-Out domain widened to IP in Portuguese

VS

   Obligatory subject movement in Hebrew →    no widening    SPA:

Hebrew: SV unergative

Stage 3: Verb movement to C acquired →    HEP: Spell-Out domain widened to CP

SV and VS unaccusative SV unergative

SV and VS unaccusative Portuguese: SV and VS unergative

SV and VS unaccusative and unergative Hebrew transitive and unergative and Portuguese Whquestions: V in C acquired



Hebrew and Arabic children going Romance 

5.  Conclusion We hope to have shown in this paper that there are robust crosslinguistic differences in the acquisition of word order of related languages. Methodologically, our results show that a comparison of the acquisition paths of languages sharing specific constructions using the same methods may prove fruitful. In particular, we argued that the existence (or absence) of clitic doubling in the target languages is the key factor to explain why the four languages under consideration cluster in two distinct groups. Finally, we showed how the acquisition of specific movement types, in this case the acquisition of subject movement, V-to-I, and I-to-C, may trigger differences in the acquisition of prerequisites for linearization.

References Adragão, Maria do Mar. 2001. Aquisição da inversão numa criança entre os dois e os três anos. Ms., Universidade Nova de Lisboa. Adragão, Maria do Mar & João Costa. 2004. “On the Status of Preverbal Subjects in Null Subject Languages: Evidence from Acquisition”. Proceedings of GALA 2003 ed. by Jacqueline van Kampen & Sergio Baauw, 69–80. The Hague: Utrecht University, Lot Publications. Aoun, Joseph. 1996. “Clitic-Doubled Arguments”. Ms., University of Southern California. Berman, Ruth A. 1985. “The Acquisition of Hebrew”. The Crosslinguistic Study of Language Acquisition ed. by Dan I. Slobin. Hillsdale, NJ.: Erlbaum. Vol. I. 255–371. Berman, Ruth A. & Esther Dromi. 1984. “On Marking Time Without Aspect in Child Language”. Papers and Reports on Child Language Development 23. 23–32. Bibi, Hana. 2003. Maternal input and its relation to the acquisition of early verbs in Hebrew. MA thesis, Tel Aviv University. Costa, João. 1998. Word Order Variation. A Constraint-Based Approach. Ph.D. dissertation, Leiden University. Davidson, Sagit. 2002. The language profile of Hebrew-speaking children with SLI. MA thesis, Tel Aviv University. Déprez, Vivian & Amy Pierce. 1993. “Negation and Functional Projections in Early Grammar”. Linguistic Inquiry 24. 25–68. Déprez, Vivian & Amy Pierce. 1994. “Crosslinguistic Evidence for Functional Projections in Early Child Grammar”. Language acquisition studies in generative grammar ed. by. Teun Hoekstra & Bonnie Schwartz. Philadelphia & Amsterdam: John Benjamins. 56–84. Fox, Daniel & David Pesetsky. 2004a. Cyclic Linearization and the Typology of Movement. Ms., MIT. Fox, Daniel & David Pesetsky. 2004b. “Cyclic Linearization of Syntactic Structure”. Theoretical linguistics 31. 1–46. Friedemann, Marc-Ariel. 1993/1994. “The Underlying Position of External Arguments in French: A Study in Adult and Child Grammar”. Language Acquisition 3. 209–255. Friedemann, Marc-Ariel. 2000. “Early French Postverbal Subjects”. The acquisition of syntax: Studies in comparative developmental linguistics ed. by Marc-Ariel Friedemann & Luigi Rizzi, 63–83. Geneva, Switzerland: Longman Linguistics Library Series.

 João Costa & Naama Friedmann Friedmann, Naama. 2007. “Young Children and A-chains: The acquisition of Hebrew Unaccusatives”. Language Acquisition 14. 377–422. Friedmann, Naama & João Costa. 2007. “Acquisition of SV and VS Order in Hebrew, European Portuguese, Palestinian Arabic, and Spanish”. Manuscript submitted for publication. Friedmann, Naama & Hedva Lavi. 2006. “On the Order of Acquisition of A-movement, Wh-movement and V-C movement”. Language Acquisition and Development ed. by Adriana Belletti, Elisa Bennati, Cristiano Chesi, Elisa Di Domenico & Ida Ferrari. Cambridge, UK.: Cambridge Scholars Press/CSP. Guasti, Maria Teresa. 2002. Language Acquisition: The Growth of Grammar. Cambridge: MIT Press. MacWhinney, Brian. 2000. The CHILDES Project: Tools for analyzing talk. 3rd Edition. Vol. 2: The Database. Mahwah, NJ: Lawrence Erlbaum Associates. Mohammad, Mohammad A. 2000. Word Order, Agreement, and Pronominalization in Standard and Palestinian Arabic. (= Current Issues in Linguistic Theory, 181). Philadelphia & Amsterdam: John Benjamins. Ouhalla, Jamal. 1994. “Verb Movement and Word Order in Arabic”. Verb Movement ed. by David Lightfoot & Norbert Hornstein, 41–72. Cambridge: Cambridge University Press. Pierce, Amy. 1989. On the Emergence of Syntax: A Crosslinguistic Study. Ph.D. dissertation, MIT, Cambridge. Pierce, Amy. 1992. Language Acquisition and Syntactic Theory: A Comparative Analysis of French and English Child Grammars. Dordrecht, the Netherlands: Kluwer. Santos, Ana Lúcia. 2006. Minimal answers. Ph.D. dissertation, Universidad de Lisboa. Shlonsky, Ur. 1997. Clause Structure and Word Order in Hebrew and Arabic. New York: Oxford University Press. Silva, Cláudia. 2004. A Natureza de Agr e as suas Implicações na Ordem VS. Ph.D. dissertation, UFAL. Soares, Carla. 2003. “Computational Complexity and the Acquisition of the CP Field in European Portuguese”. Paper presented at ConSole XII. Torrego, E. 1998. The dependencies of objects. Cambridge, MA: MIT Press. Torrens, Vicenç & Ken Wexler. 1996. “Clitic Doubling in Early Spanish”. BUCLD Proceedings, 20, 780–791. : Sommerville, MA: Cascadilla Press. Zuckerman, Shalom. 2001. The Acquisition of ”Optional” Movement. (= Groningen dissertations in linguistics, 34.) Groningen: Rijksuniversiteit Groningen.

Adjectives and deleted nominals in Spanish* Luis Eguren

Universidad Autónoma de Madrid In this paper I argue against previous strictly formal accounts of adjectival modification in nominal expressions with ellipsis in Spanish, which make use of structural conditions, such as different syntactic positions for adjectives, X-bar constraints on the target of ellipsis and/or minimality effects within the DP, and propose instead that the restricted distribution of adjectives in Spanish elliptical DPs can be reduced to a contrastive focus condition on nominal ellipsis, whereby only adjectives that create Roothian sets of alternatives, giving rise to a partitive set-subset relation, can combine with an elided noun.

1.  Introduction Just like in other Romance languages, not all kinds of adjectives can modify an elided noun in Spanish. It has often been noticed, for instance, that (postnominal) restrictive qualitative adjectives are compatible with nominal ellipsis, whereas (prenominal) non-restrictive evaluative adjectives block it: (1) a.

Los (escaladores) entrenados llegaron a la cima. the  climbers trained got to the top

b. Los entrenados *(escaladores) llegaron a la cima. the trained climbers got to the top “The trained climbers got to the top.”

Both this fact, and some other relevant patterns of adjectival modification in elliptical DPs, have generally been accounted for by resorting to structural conditions, such as minimality effects created by intervening adjectives, or constraints on the X-bar categorial status of the target of ellipsis combined with the assignment of particular structural positions to different types of adjectives.

*I thank two anonymous reviewers for their valuable comments on the content of this paper. Research for this work has been supported by a grant to the project EDU2008–01268/EDUC and the MICINN grant PR2008–0005.

 Luis Eguren

In this paper I will criticize these (strictly) formal approaches, and explain the (non-)occurrence of adjectives in nominal expressions with ellipsis in Spanish as the result of an information-structural condition stating that the non-elided material has to be interpreted as contrastive focus. The content of the paper is distributed as follows. Section 2 contains a discussion on the role contrastive focus plays in nominal ellipsis. Section 3 shows how the contrastive focus condition on the remnant of ellipsis explains the whole paradigm of adjectival modification in Spanish elliptical DPs. Finally, in Section 4, previous structural accounts are reviewed and discarded.

2.  Focus in nominal ellipsis Current mainstream research takes all forms of ellipsis in the sentential domain to be a focus-based PF-phenomenon (see, amongst others, Merchant 2001 for Sluicing, Winkler 2005 for VP-ellipsis and Gapping, and Gengel 2007 for Pseudogapping). This insight has been extended to the nominal domain, and it has been claimed that nominal ellipsis also obeys a semantic focus condition (see, e.g. Giannakidou & Stavrou 1999, Ntelitheos 2004, Corver & van Koppen 2005, 2006, 2007, Gengel 2006). This condition could be formulated as follows:

(2) Focus condition on nominal ellipsis:

Noun ellipsis can take place when focus is overtly expressed in the remnant constituent and the noun is given.1 (Corver & van Koppen 2005:21)

The notion of focus researchers on ellipsis usually make use of is taken from Rooth’s (1992, 1996) Alternative Semantics theory of focus. In Rooth’s theory, focus always identifies a relevant alternative or subset in a set of alternatives, which is the set of all the interpretations that can be obtained by replacing the focused item(s) by possible equivalent alternatives. To give an example of how this theory might apply to constructions with nominal ellipsis, in the sentence in (3), for example, the focused constituents, which are marked with a Focus feature (F) in the syntax, specify alternatives or subsets in a given set of dresses of different colours:

1.  Focus on the remnant of nominal ellipsis can be overtly expressed either by means of an affix, like the adjectival Dutch suffix -e(n) (cf. Corver & van Koppen 2005, 2006, 2007), or through focus stress more generally, as in Spanish for instance.



Adjectives and deleted nominals in Spanish 

(3) a.

A María le gusta el vestido [rojo]f, pero Ana prefiere el __ [azul]f. to Mary dat likes the dress red, but Ann prefers the __ blue “Mary likes the red dress, but Ann prefers the blue one.”

b. ‘x dresses’ (= the set of dresses of different colours)

In addition to the focus condition, some authors have pointed out that nominal ellipsis also complies with a contrastiveness requirement, whereby the remnant in an elliptical DP must not be semantically identical to its counterpart in the antecedent DP (see, e.g. Giannakidou & Stavrou 1999). The sentences in (4) illustrate this fact: (4) a. *María ha leído dos cuentos de Ribeyro y Ana ha leído dos __ Mary has read two stories by Ribeyro and Ann has read two __

de Ribeyro. by Ribeyro

b. María ha leído dos cuentos de Ribeyro y Ana ha leído tres Mary has read two stories by Ribeyro and Ann has read three

__ de Bolaño. __ by Bolaño

It can then be concluded that nominal ellipsis is subject to the contrastive focus condition in (5):

(5) Contrastive focus (in nominal ellipsis):

Contrastive focus identifies a relevant alternative or subset in a set of contextually or situationally given alternatives, and the focused constituent(s) in the remnant cannot be (semantically) identical to the corresponding part(s) in the antecedent phrase.

In (5), a weaker non-identity condition has replaced exhaustivity in the definition of contrastive (or identificational) focus by Kiss (1998:245), the reason being that exhaustivity is not a prerequisite for nominal ellipsis to take place, as shown by the fact that also-phrases and even-phrases are available in this case (cf. John has read these papers and I have {also/even} read those__.). This is not to say, however, that exhaustivity, or stronger forms of contrast, like the explicit identification of complementary subsets in closed sets, do not license nominal ellipsis. They certainly do in some cases, as will be seen below. To complete this succint overview of the role contrastive focus plays in nominal ellipsis, I would finally like to emphasize that it is precisely focus marking on the remnant which guarantees the recoverability of deletion, and so makes nominal ellipsis possible. The idea I have in mind goes as follows. A contrastive focused constituent in the remnant chooses an alternative in a set of given alternatives, and in so doing, a partitive (Sleeman 1996, Martinho 1998, Bouchard 2002) or

 Luis Eguren

D-Linking (López 2000) relation between a known set and a subset of the same descriptive class is created: this set-subset relation implies that the class at issue needs not be mentioned again in the elliptical DP, since its content can be unequivocally recovered in the context.2 In the next section, I will show how the contrastive focus condition on the remnant helps us explain why certain adjectival modifiers are either accepted or excluded in nominal expressions with ellipsis in Spanish.3 3.  Adjectival modification and nominal ellipsis in Spanish As illustrated in (6) and (7), serial adjectives and ordinals are fully compatible with nominal ellipsis in Spanish (and in other languages as well). Serial adjectives, like siguiente ‘following’ or anterior ‘previous’, denote a precedence or a succession relation in a sequence of members of the same descriptive class, whereas ordinals convert a given set into an ordered series. Both kinds of adjectives are therefore inherently contrastive relational items, and their use presupposes the existence of a set of reference from which they select an alternative. (6) a.

El Real Madrid ganó los primeros partidos, pero perdió the Real Madrid won the first games, but lost



los siguientes __. the following __



“Real Madrid won the first games, but lost the following.”

b. Fuimos a Grecia el año pasado y el __ anterior. went.1pl to Greece the year last and the __ previous “We went to Greece last year and the previous one.”

2.  It has recently been proposed that the semantic focus condition on nominal ellipsis is syntactically encoded by means of overt movement of the focused remnant to a DP-internal Focus Phrase (cf. Ntelitheos 2004, Corver & van Koppen 2005, 2006, 2007, Gengel 2006). In my view, this analysis cannot be applied to nominal ellipsis in Spanish: amongst other things, given that focus is regularly assigned in situ to members of DPs without ellipsis in this language, there seems to be no reason for DP-internal overt movement of focused constituents only in the case of elliptical DPs. 3.  Although contrastive focus is the crucial (information-structural) licenser for noun ellipsis in Spanish, morphonological factors are also at play in some cases, as in the alternation uno/un ‘a/one’, alguno/algún ‘some’, primero/primer ‘first’, etc. (e.g. He leído {el primer capítulo/ el primero__/el capítulo primero} ‘I have read the first (chapter)’), which could be captured by means of an apocope rule that reduces the full forms in certain contexts, or the limited distribution of the definite article (e.g. {el __ {de Pedro/*con gafas}lit. ‘the __{of Peter/with glasses}’), which has often been attributed to its clitic nature.



(7) a.

Adjectives and deleted nominals in Spanish 

Me gustó el primer capítulo. El segundo __ era muy aburrido. dat liked.3sg the first chapter. The second __ was very boring “I liked the first chapter. The second was very boring.”

b. Acertó las dos primeras preguntas, pero falló la tercera __. answered the two first questions, but failed the third __ “She answered the first two questions, but could not answer the third.”

Due to their semantic import, referential and quantitative adjectives also freely co-occur with an elided nominal in Spanish. Referential adjectives, like mismo ‘same’ and otro ‘other’ in the examples in (8), establish a relation of identity or contrast between the noun they introduce and another member of a contextually given class. The predicative cardinals and vague numerals in (9), as the corresponding quantifiers, convey a partitive meaning that licenses nominal ellipsis. The focal adjective único ‘only’ exhaustively identifies an alternative in a given set while excluding any other equivalent alternative (cf. (10a)). And, finally, the predicate of quantity demás ‘remaining’ singles out the complementary subset in a closed set of alternatives (cf. (10b)). (8) a.

Vieron varios coches y al final se compraron el mismo __. saw.3pl various cars  and to the final refl bought.3pl the same __ “They looked at various cars and finally bought the same (one).”

b. A Juan le gusta este traje, pero Pedro prefiere el otro __. to John dat likes this suit, but Peter prefers the other __ “John likes this suit, but Peter prefers the other (one).” (9) a.

He leído todos los libros de Cortázar y los dos __de Rulfo. I-have read all the books by Cortázar and the two __ by Rulfo “I have read all the books by Cortázar and the two by Rulfo.”

b. Las pocas preguntas a las que respondió no fueron las más the few questions to the that answered.3sg not were the most

interesantes de las muchas __que se le hicieron. interesting of the many __ that refl dat made.3pl

“The few questions he answered were not the most interesting of the many he was asked.” (10) a.

De todas esas estudiantes, la única __que sabe inglés es Ana. of all those students, the only __ who knows English is Ann “Of all those students, the only one who speaks English is Ann.”

b. Fueron a la huelga pocos maestros. Los demás __ no went.3pl to the strike few teachers. The rest  __ not

querían perder dinero. wanted lose money



“Few teachers went on strike. The rest did not want to lose money.”

 Luis Eguren

As is well-known, restrictive and non-restrictive qualitative adjectives do not behave alike, however, in nominal expressions with ellipsis in Spanish (and in Romance more generally). Postnominal restrictive adjectives modify an empty noun, whereas prenominal non-restrictive adjectives reject nominal ellipsis:4 (11) a.

Se viaja mejor en estos trenes rápidos que en esos __lentos. refl travels better in these trains fast than in those __ slow “It is better to travel in these fast trains than in those slow ones.”

b. Me gustan más estos cuentos cortos que esos __ largos. dat like.3pl more these stories short than those __ long “I prefer these short stories to those long ones.” c.

Este año he leído tres novelas buenas y dosˉ__ˉmalas. this year have.1sg read three novels good and twoˉ__ˉbad “This year I have read three good novels and two bad ones.”

(12) a. *Se viaja mejor en estos rápidos trenes que en esos lentos __. refl travels better in these fast trains than in those slow __ b. *Me gustan más estos cortos cuentos que esos largos __. dat like.3pl more these short stories than those long __ c. *Este año he leído tres buenas novelas y dos malas__. This year have.1sg read three good novels and two badˉ__

This contrast has a straightforward explanation in terms of the focus condition on the remnant of ellipsis. Restrictive adjectives can be readily marked as (contrastive) focus because they are partitive lexical items that select a subset in the class denoted by the noun they are predicated of. Non-restrictive evaluative adjectives highlight a defining feature of a descriptive class instead, and do not make a partition in that class. Being non-partitive, this kind of qualitative adjectives can only be interpreted DP-internally, i.e. they never belong to a DP-external set of alternatives built around a given shared descriptive class, and the content of the elided nominal cannot be recovered from the context.5

4.  An anonymous reviewer reminds me that the difference between the examples in (11) and (12) involves different intonation patterns: in his/her words, “prenominal adjectives cannot bear phonological stress and resist noun-ellipsis, contrary to their postnominal counterparts”. It is certainly true that adjectives in the remnant of ellipsis bear focus stress in Spanish (see Fn.  1). However, as indicated by another reviewer, it is not totally impossible to place pitch accent on non-restrictive prenominal adjectives, as in his/her example %He viajado en este LENTO tren ‘I have travelled in this SLOW train’, which is acceptable, but has either an emphatic interpretation or a sort of a metalinguistically corrective reading. 5.  As an anonymous reviewer points out, this analysis of prenominal adjectives raises the question of why appositive relative clauses can survive ellipsis, given that the semantics of both types of modifiers are quite similar:



Adjectives and deleted nominals in Spanish 

Independent evidence in support of a focus-based account for the contrast between restrictive and non-restrictive adjectives in the Spanish elliptical DPs in (11)–(12) comes from the fact that a similar contrast also obtains in other constructions in which focus has been claimed to play a relevant role, like Discontinuous DPs and Determiner Spreading in Greek or the Romanian cel construction (see Ntelitheos 2004, Marchis & Alexiadou, this volume, and the references therein). As shown in the English glosses, the qualitative focused adjectives in the examples in (13) can only have a contrastive restrictive interpretation: (13) a.

I mikres i gates itan pinasmenes. (Det. Spreading, Greek) the young the cats were hungry “The YOUNG cats were hungry. [The OLD ones were not].”

b. To prasino forese I Maria fustani. (Discont. DP, Greek) the green put on Maria dress “Maria put on the GREEN dress [and not the RED one].” c.

Legile cele importante n-au fost votate. (Cel constr.,Romanian) laws.the cel important not-have been voted “The IMPORTANT laws have not been voted. [But other laws have].”

The different behaviour of distinctive (or classifying) and non-distinctive (or non-classifying) restrictive qualitative adjectives in elliptical DPs in Romance (see, amongst others, Jones 1993, Sleeman 1993, 1996, Martinho 1998 and

(i) a. *Se viaja mejor en estos rápidos trenes que en esos lentosˉ__. refl travels better in these fast trains than in those slowˉ__ b.

Se viaja mejor en estos trenes, que son rápidos, que en esosˉ__,ˉ refl travels better in these trains, that are fast, than in those __,



que son lentos that are slow

The contrast in (i) could be attributed to the fact that appositive relatives, being adjuncts to the whole DP, do not interfere in the (contrastive) D-linking relation created by the demonstrative, whereas in the case of non-restrictive adjectives, that are DP-internal noun modifiers, there is a semantic clash between the partitive interpretation of the determiner and the non-D-linking import of the adjective. The paradigm in (ii) corroborates this idea: (ii) a.

[Ella]dp {sola/misma} lo ha conseguido. she {alone/herself} it has made

b.

[Esa (mujer)]dp {sola/misma} lo ha conseguido. that (woman) {alone/herself} it has made

Both the aspectual adjective solo and emphatic mismo are (non-focused) DP-adjuncts (cf. (iia)), which are also fully compatible with nominal ellipsis (cf. (iib)).

 Luis Eguren

Bouchard 2002) can also be derived from the contrastive focus condition on nominal ellipsis: (14) a.

Me compré el __verde. dat bought.1sg the __ green “I bought the green one.”

b. Escogí el __ pequeño. chose.1sg the __ small “I chose the small one.” c.

Leí el  __ corto. read.1sg the __ short “I read the short one.”

(15) a. #Me gustó la __ interesante. dat liked.3sg the __ interesting “I liked the interesting one.” b. #Prefiero los __difíciles. prefer 1sg the __ difficult “I prefer the difficult ones.” c. #Me cansan las __ aburridas. dat tire.3pl the __ boring “I am tired of the boring ones”. (16) a.

Escuché la charla que no tenía interés y la __ interesante. I. listened the talk that not had interest and the __ interesting “I listened to the uninteresting talk and to the interesting one.”

b. Resolví los problemas fáciles y los __ difíciles. I.solved the problems easy and the __ difficult “I solved the easy problems and the difficult ones.” c.

Vi las películas divertidas y las __ aburridas. I.watched the films amusing and the __ boring “I watched the amusing films and the boring ones.”

Distinctive restrictive adjectives denote discriminating easily recognizable properties, like colour or qualities expressed in paradigmatic binary contrasts (grande/pequeño ‘big/small’, largo/corto ‘long/short’, viejo/nuevo ‘old/new’), and their use is naturally associated to a set of references from which they choose a subset. That is why they can modify an empty noun even in an impoverished context (cf. (14)). Non-distinctive qualitative adjectives, like interesante ‘interesting’, difícil ‘difficult’ or aburrido ‘boring’ in (15), denote more abstract or subjective non-discriminating properties. These restrictive adjectives also create subsets, but these subsets are not sufficiently recognizable in cognitive terms to be unequivocally associated to a set of reference (to give a clear example, anything can be interesting



Adjectives and deleted nominals in Spanish 

from one point of view or another, but only some entities are green). As a result, the occurrence of non-distinctive adjectives in nominal expressions with ellipsis is pragmatically anomalous if the sentence containing the DP is uttered out of context (cf. (15)).6 However, the combination of a non-distinctive qualitative adjective with a deleted nominal turns out to be fully acceptable in an enriched contrastive setting in which an equivalent alternative is made explicit (cf. (16)): the non-distinctive adjective is now (contrastively) focused (i.e. it selects an alternative in a given set of alternatives), and the elided content can be recovered in the context. This account also applies, I would like to claim, to the case of relational adjectives, elatives and transitive adjectives. Grammaticality judgements on nominal expressions with ellipsis including a relational adjective in Romance are not uniform: for some linguists, this construction is well-formed, whereas other authors take it to be deviant, or even ungrammatical.7 In my view, this lack of agreement on the data is due to the fact that both thematic (cf. (17)) and classifying (cf. (18)) relational adjectives, just like the abstract qualitative adjectives in (15)–(16), are non-distinctive restrictive adjectives, that create subsets, but denote non-discriminating properties that are not clearly associated to a set of references. Accordingly, relational adjectives in elliptical DPs are fully felicitous only within a context in which equivalent alternatives belonging to a known set of alternatives are explicitly contrasted: (17) a.

La invasión alemana de Polonia causó tantos destrozos como the invasion German of Poland caused so many damages as



la __ rusa. the __ Russian



“The German invasion of Poland was as bad as the Russian one.”

b. La propuesta británica fue más sensata que la __ italiana. the proposal British was more sensible than the __ Italian “The British proposal was more sensible than the Italian one.” c.

El avance polaco en Europa es mucho mayor que el __ español. the rise Polish in Europe is much bigger than the __ Spanish “The Polish rise in Europe is much stronger than the Spanish one.”

6.  Sleeman (1993, 1996) considers that sentences like the ones in (15) are ungrammatical in French. This cannot be the case though, since, as in Spanish, they are also acceptable in French if they are inserted in a discriminating contrastive context (see Sleeman 1993:Fn. 16 and Bouchard 2002:231). The pattern in (15)–(16) also obtains in Portuguese (cf. Martinho 1998:121). 7.  See, for instance, the divergent judgements and the comments on this issue in Demonte (1999a:§3.3.1.2) for Spanish, in Sleeman (1993:Fns. 6 and 12) for French, and in Martinho (1998:118–119) for Portuguese.

 Luis Eguren

(18) a.

La energía atómica es tan cara como la __ solar. the energy atomic is as expensive as the __ solar “Atomic energy is as expensive as solar energy.”

b. Los reconocimientos terrestres son más arriesgados que the surveys terrestrial are more risky than

los__aéreos. the __ aerial



“Terrestrial surveys are more risky than aerial surveys.”

c.

La programación televisiva es peor que la __ radiofónica. the programming on TV is worse than the __ on the radio “TV programming is worse than radio programming.”

Lexical elatives (i.e. extreme degree adjectives without explicit morphology), like enorme ‘enormous’, excelente ‘excellent’ or espléndido ‘splendid’, have a subjective evaluative import, and might well be included in the semantic class of non-distinctive restrictive adjectives. As in the cases above, elatives cannot be used to fix the reference of a DP with an elided nominal if the sentence containing the DP stands in isolation, and examples such as the ones in (19) are odd. However, as noted by Brucart & Gràcia (1986), the acceptability of these sentences considerably improves, as before, when there is a previous agreement between speaker and hearer on the referent of the nominal expression with ellipsis, i.e. when a discriminating context has already been provided (or is presupposed) making explicit a closed set of contrasting alternatives to which the elative belongs (cf. (20)). (19) a. #A mí me gustó la __enorme. to me dat liked.3sg the __ huge “I liked the huge one.” b. #Me comí la __excelente. dat ate.1sg the __ excellent “I ate the excellent one.” c. #Fue a ver la __ espléndida más gente. went to see the __ splendid more people “More people watched the splendid one.” (20) a.

Nos enseñaron dos casas: una casa pequeña y una casa enorme. us showed.3pl two houses: a house small and a house huge



A mí me gustó la __ enorme. to me dat liked.3sg the __ huge

“We had a look at two houses: a small house and a huge house. I liked the huge one.”



Adjectives and deleted nominals in Spanish 

b. Había dos tartas de postre: una tarta buena y una tarta there-were two cakes of dessert: a cake good and a cake

excelente. Me comí la __ excelente. excellent. dat ate.1sg the __ excellent

“There were two cakes for dessert: a good cake and an excellent cake. I ate the excellent one.” c.

Ponían dos películas: una era muy mala, pero la otra era put.3pl two films: one was very bad, but the other was



espléndida. Fue a ver la __ espléndida, cómo no, más gente. splendid. went to see the __ splendid, how not, more people

“Two films were on: one was very bad, but the other was splendid. Not surprisingly, more people watched the splendid one.”

Transitive adjectives (i.e. adjectives that select a complement) are also non-distinctive, since they denote abstract properties that do not unambiguously identify a particular referent. The paradigm in (15)–(16) is therefore reproduced in this case as well: the isolated sentences in (21) are again pragmatically anomalous, but become fully acceptable in a contrastive setting (cf. (22)), where the transitive adjective selects an alternative in a set of alternatives.8 (21) a. #Ana detesta a los __ llenos de prejuicios. Ann hates to the __ full of prejudices “Ann hates [judges] full of prejudices.” b. #Solo llegarán a la cima los __capaces de arriesgarse. only will  get to the top the __ capable of taking risks “Only [climbers] capable of taking risks will get to the top.” c. #Yo prefiero a los __ orgullosos de sus hijos. I prefer to the __ proud of their children “I prefer the [parents] proud of their children.” (22) a.

Hay jueces que tienen prejuicios contra las mujeres y there are judges who have prejudices against the women and



otros que no los tienen. Ana detesta a los __llenos de prejuicios. others who not them have. Ann hates to the __ full of prejudices

“There are some judges that have a prejudice against women and others that do not. Ann hates those full of prejudices.”

8.  Ronat (1977) and Sleeman (1993) take nominal expressions with ellipsis containing a transitive adjective to be ungrammatical in French. However, this does not seem be the case: all the examples these authors offer are decontextualised sentences, and as in Spanish, these sentences are felicitous when uttered within a rich contrastive context. The same phenomenon is also observed in Portuguese (cf. Martinho 1998:129).

 Luis Eguren

b. A algunos de estos escaladores les gusta arriesgarse y a to some of these climbers dat like.3sg taking.risks and to

otros no. Solo llegarán a la cima los __capaces de arriesgarse. others not. Only will.get to the top the __ capable of taking risks

“Some of these climbers like taking risks and others do not. Only those capable of taking risks will get to the top.” c.

Hay dos tipos de padres: los que están orgullosos de sus there.are two types of parents: the that are proud of their



hijos y los que no. Yo prefiero a los __ orgullosos de sus hijos. children and the that not. I prefer to the __ proud of their children

“There are two types of parents: those who are proud of their children and those who are not. I prefer those proud of their children.”

The (non-)occurrence of so called ‘adverbial’ adjectives (i.e. intentionally oriented, modal epistemic and circumstantial adjectives)9 in elliptical DPs in Spanish can finally be attributed to the focus condition on nominal ellipsis as well:10 (23) a.

Es un simple *(niño). is a mere child

b. Acerté por pura *(casualidad). got it right by pure  chance c.

Has cometido un claro *(error). have.2sg made a plain  mistake

(24) a.

En la sala estaba el presunto #(asesino). in the room was the alleged  murderer

b. El supuesto #(culpable) salió bien parado. the supposed  guilty person came off well c.

Ni saludó al antiguo #(presidente). nor greeted the old  president

(25) a.

En la sala estaban el verdadero asesino y el presunto __. in the room were the true murderer and the alleged __ “Both the true murderer and the alleged one were in the room.”

b. El auténtico culpable salió mejor parado que el supuesto __. the real guilty person went better done than the alleged __ “The real guilty person came out of it better than the supposed one.”

9.  These are the sub-classes of prenominal adverbial adjectives in Demonte (1999b). 10.  The contrasts in (23)–(25) are also observed in Discontinuous DPs and Determiner Spreading in Greek and in the Romanian cel construction (Marchis & Alexiadou, p.c.).



c.

Adjectives and deleted nominals in Spanish 

El actual primer ministro ni saludó al antiguo __. the present prime minister nor greeted to.the old __ “The present prime minister did not even greet the old one.”

As shown in (23), some adverbial adjectives, like simple ‘mere’, puro ‘pure’ or claro ‘plain’ are completely out in nominal expressions with ellipsis, the reason being, I believe, that they can never be focused, i.e. these adjectives cannot be members of a contextually given set of contrasting equivalent alternatives. In contrast, some other adverbial adjectives, like presunto ‘alleged’, supuesto ‘supposed’ or antiguo ‘old’, do allow for nominal ellipsis: adverbial adjectives like these make up contrastive pairs (verdadero/presunto ‘true/alleged’, auténtico/supuesto ‘real/supposed’, actual/ antiguo ‘present/old’), and thus may belong to a set of equivalent alternatives, and be focused. Nevertheless, as they denote abstract non-discriminating notions, these adjectives behave like non-distinctive restrictive adjectives in this respect, and only felicitously modify an empty noun within an explicitly contrastive context (see the contrast in (24)–(25)).11 In this section, I have presented and examined all the relevant paradigms of adjectival modification in elliptical DPs in Spanish. The discussion above leads me to conclude that a semantic contrastive focus condition on the remnant of ellipsis explains why certain classes of adjectives modify an elided nominal, while others do not: only adjectives that select an alternative or subset in a given set of equivalent contrasting alternatives allow for the implicit content to be contextually recovered in Spanish (and in Romance more generally, I would dare say). In the next section, with all this background in mind, I will finally critically revise some structural accounts of the phenomenon under study.

11.  Demonte (1999b) takes examples like the ones in (25) to be “almost completely grammatical”. Sánchez (1996) and Ticio (2005) consider, however, that the Spanish sentence Ayer vi a la verdadera terrorista y a la supuesta ‘Yesterday I saw the true terrorist and the alleged one’ is ungrammatical. I agree with Demonte, and do not share Sánchez’ and Ticio’s judgement. For me, and for all the native speakers I have asked on the issue, this sentence is fairly OK as it is, and gets even better if the closed set of contrasting alternatives is made explicit beforehand, as expected: (i) Nunca había visto a ninguna de esas dos mujeres, ni a la que supuestamente había cometido el asesinato, ni a la verdadera asesina, pero por fin ayer vi a la verdadera asesina y a la supuesta __ en televisión. “I had never seen any of those two women, neither the one who had allegedly committed the crime nor the true murderer, but yesterday I finally saw both the true murderer and the alleged one on TV.”

 Luis Eguren

4.  Structural approaches: A critical review As mentioned in the introduction, the limited distribution of adjectival modifiers in nominal expressions with ellipsis in Romance has generally been analysed in strictly structural terms. These proposals might well be grouped as follows: some linguists mainly ascribe the fact that certain kinds of adjectives are excluded in elliptical DPs to minimality effects on proper government of [e]N created by intervening adjectives (cf. Bernstein 1993, Sleeman 1993 and Demonte 1999b),12 while other authors claim that this incompatibility results from the position different types of adjectives occupy within the DP plus the specification of the X-bar categorial status of the target of ellipsis (cf. Ronat 1977 and Ticio 2005).13 I will go through these structural accounts in turn. Bernstein (1993) holds that word markers (WM) are able to head-govern, and thereby license, an empty noun in Spanish, and generates prenominal adjectives like mero ‘mere’ or pobre ‘poor’ as heads of an Adjectival Phrase (AP) above the Word Marker Phrase (WMP). In this structural configuration, as depicted in (26), the intervening adjective blocks head movement of the affixal WM to its D host, and the Head Movement Constraint is violated: (26)

*DP D

AP A

un-oi mero/pobre

WMP WM

NP

ti

e

Sleeman (1993) adopts Cinque’s (1993) proposal, and assumes that all intransitive adjectives in French are generated as specifiers of different functional projections located between DP and NP. Within this framework, the ungrammaticality of French nominal expressions with ellipsis that include a non-distinctive adjective, like the one represented in (27), is the result of Minimality: “the binding relation

12.  See also Contreras (1992) and Lobeck (1995:137, Fn. 6). 13.  Sánchez (1996) and Giannakidou & Stavrou (1999) also resort to the structural positions they assign to different kinds of adjectives in order to (at least in part) account for their (non-) occurrence in elliptical DPs.



Adjectives and deleted nominals in Spanish 

between the [+Q] specifier and the base-generated variable is blocked by the presence of the intransitive adjective in an intervening Spec” (Sleeman 1993: 291). (27)

*DP D

QP Q′

Spec

Q

FP

Spec

F′ F

les

deux

intéressantes

NP e

Finally, as shown in the (simplified) structure in (28), Demonte (1999b) considers that prenominal non restrictive qualitative adjectives are heads adjoined to a Degree (Deg) head taking the np as a complement,14 and suggests that the interference of this closer governing head (A-Deg) creates a minimality barrier, so that the determiner cannot formally license the empty noun. (28)

*DP D

DEGP DEG A

la

alegre

NP DEG N e

In my view, all these proposals face three main problems: a. They crucially depend on how adjectives that block ellipsis are generated: these adjectives must be either heads or specifiers, but cannot be adjuncts

14.  To be more precise, in Demonte’s proposal prenominal qualitative adjectives coming from the lexicon with a [+Deg] formal feature are first projected in a small-clause within the NP, and are then adjoined to Deg0 through Adjective Raising.

 Luis Eguren

(if this were the case, relativized minimality could not be invoked).15 The question of how adjectival modifiers are generated is not, however, a settled issue at all.16 b. It is taken for granted that [e]N must be formally licensed, but this need not be the case. Nominal ellipsis is possible in Spanish, for example, in bare mass nouns and bare plurals, like the ones in (29), and it has convincingly been argued that bare nouns of this sort are not quantified expressions (see Laca 1999 and NcNally 2004, amongst many others): this construction would therefore lack a determiner, and the empty noun would not be properly governed. (29) a.

Nunca bebo cerveza caliente. Solo bebo __ fría. never drink.1sg beer warm. Only drink.1sg __ cool “I never drink warm beer. I only drink cool beer.”

b. Solo bebe vinos españoles. Nunca bebe __ franceses. only drink.3sg wines Spanish. Never drink.3sg __ French “She only drinks Spanish wines. She never drinks French wines.”

c. It is not easy to see how all the data in Section 3 can be accounted for in strictly structural terms (in particular, the fact that non-distinctive restrictive adjectives and some adverbial adjectives felicitously modify an elided noun in a contrastive setting), unless a condition is added, as in Sleeman (1996), stating that only intervening adjectives with a non-partitive interpretation give rise

15.  An anonymous reviewer notes that if adjectives are adjuncts, they ought to be sensitive to weak island boundaries. The contrast below suggests, in his/her view, that this is correct: (i) a.

m. ha comprado un traje negro, pero [rosa]i, ha comprado [DPuna camisa ti]. m. has bought a suit black, but pink, has bought  a shirt “M. has bought a black suit, but he has bought a pink shirt.”

b. ??m. ha comprado un traje negro, pero [rosa]i, no ha comprado m. has bought a suit black, but pink, not has bought

[DPuna camisa ti] a shirt



“M. has bought a black suit, but he has not bought a pink shirt.”

16.  In his analysis of adjectival modification in French, Bouchard (2002) adopts, for instance, a minimal version of Bare Phrase Structure in which the structures involving an adjective are not distinguished by different bar levels, or different types of structural combinations (such as Merge vs. node-splitting Adjoin), and all adjectives are uniformly merged with the noun or with a nominal projection. If it is really the case that there is no distinction between substitution and adjunction operations, minimality effects on proper government of [e]N created by adjectives occupying a specifier position, as opposed to adjectives generated as adjuncts, could not even be formulated, so it seems.



Adjectives and deleted nominals in Spanish 

to minimality effects. But, in that case, the syntactic analyses above would be at best redundant with the semantic (contrastive) focus condition on nominal ellipsis (see Bouchard 2002:267 Fn. 4 for a similar idea). More recently, Ticio (2005) has given a somewhat different structural explanation for the (in)compatibility of adjectival modifiers with an elided noun in Spanish. She first claims that nominal ellipsis (which she treats as a PF-phenomenon) targets the maximal NP projection (and can affect, in particular, the lower segment of an NP), and then assigns the positions in (30) to different types of nominal modifiers, so that only adjectives situated outside the ellipsis site (i.e. postnominal adjectives) can be left stranded after NP-ellipsis has applied:17 (30)

(…)

nP n’ n

Ellipsis site

R-As/AGENT NP

NP PreN As

ADJUNCTS/PostN As N’

N

OBJ

There are also some serious problems with the proposal in (30), amongst them the following: a. It only works if prenominal adjectives are generated within the lower segment of NP, but this view is not shared by many linguists (see the analyses above, for instance). b. Asserting that nominal ellipsis targets the lower segment of NP is a stipulation. c. The fact that non-classifying adjectives, relational adjectives, elatives and transitive adjectives are fully acceptable only in contrastive settings is not taken into consideration. d. And most significantly, it is just not the case that all prenominal adjectives in Spanish reject nominal ellipsis. For example, as shown in (25) and repeated in (31), some intentionally oriented and circumstantial adjectives do co-occur with an implicit nominal (cf. discussion in Fn. 11).

17.  Ticio accounts for the possibility of having objects as remnant elements by means of an stylistic PF-rule that moves the object out of the ellipsis site prior to NP-deletion.

 Luis Eguren

(31) a.

En la sala estaban el verdadero asesino y el presunto __. “Both the true murderer and the alleged one were in the room.”

b. El actual primer ministro ni saludó al antiguo __. “The present prime minister did not even greet the old one.”

Ticio’s ideas are reminiscent of the well-known proposal by Ronat (1977) on the limited distribution of adjectival modifiers in elliptical nominal expressions in French. Ronat claims that nominal ellipsis is subject to a constraint imposing that N’ has to dominate something at S-structure, and locates those adjectives that she takes to be incompatible with an elided noun (i.e. elatives and transitive adjectives) in Spec of NP. In my opinion, Ronat’s analysis runs into the same difficulties as Ticio’s. First, the N’-constraint is just an ad hoc condition. Second, her account crucially relies on assigning specific structural positions to different kinds of adjectives. And third, there is also an empirical problem: French elatives and transitive adjectives can co-occur with an elided noun if they are inserted in a contrastive setting (see the discussion in Fns. 6 and 8). As compared with these strictly structural approaches, the semantic analysis of the (in)compatibility of adjectives with an elided noun based on the contrastive focus condition on nominal ellipsis that I developed in §3 explains all the relevant paradigms of adjectival modification in Spanish elliptical DPs, and does not depend on adhering to any particular theory on the location of adjectives in the architecture of the nominal domain.

5.  Conclusions In this paper, I have studied the restricted distribution of adjectival modifiers in DPs with an elided nominal in Spanish. Two main conclusions can be drawn from the discussion above. First, the (non-)occurrence of adjectives in elliptical DPs is not related to strictly structural factors, such as minimality effects created by intervening adjectives, or the combination of constraints on the X-bar categorial status of the target of ellipsis with the assignment of particular structural positions to different classes of adjectives. And second, an information-structural contrastive focus condition on the remnant of ellipsis helps us explain why just certain types of adjectives modify an elided noun in Spanish (and in Romance more generally): only adjectives that can be focused (i.e. only those adjectives that identify an alternative or subset in a contextually given set), allow for nominal ellipsis, since they give rise to a partitive set-subset relation within a known descriptive class that makes it unnecessary to mention the class at issue in the elliptical DP again.



Adjectives and deleted nominals in Spanish 

References Bernstein, Judy. 1993. “The Syntactic Role of Word Markers in Null Nominal Constructions”. Probus 5. 5–38. Bouchard, Denis. 2002. Adjectives, Number and Interfaces. Why Languages Vary. Amsterdam: Elsevier. Brucart, José María & Lluïsa Gràcia. 1986. “I Sintagmi Nominali Senza Testa: Uno studio comparato”. Rivista di Grammatica Generativa 11. 3–32. Cinque, Guglielmo. 1993. “On the Evidence for Partial N-movement in the Romance DP”. University of Venice Working Papers in Linguistics 3:2. 21–40. Contreras, Heles. 1992. “Descripción y Explicación en la Teoría Lingüística. La Sintaxis de los Cuantificadores”. Revista de Lingüística Teórica y Aplicada (Concepción, Caracas) 30. 17–38. Corver, Norbert & Marjo van Koppen. 2005. “Microvariation and Ellipsis in the wat voorconstruction”. Paper presented at the workshop Sounds of Silence, Tilburg University. Corver, Norbert & Marjo van Koppen. 2006. “Let’s focus on Noun Ellipsis”. Glow Newsletter 57. Corver, Norbert & Marjo van Koppen. 2007. “Micro-variation and Macro-variation: Same or different? A comparative view on diversity in Noun Phrase Ellipsis”. Paper presented at the colloquium Formal models in linguistic diversity, University of Trieste. Demonte, Violeta. 1999a. “El Adjetivo: Clases y Usos. La Posición del Adjetivo en el Sintagma Nominal”. Gramática descriptiva de la Lengua Española, ed. by Ignacio Bosque & Violeta Demonte, vol. 1, 129–212. Madrid: Espasa. Demonte, Violeta. 1999b. “A Minimal Account of Spanish Adjective Position and Interpretation”. Grammatical Analyses in Basque and Romance Linguistics, ed. by Jon Franco, Alazne Landa & Juan Martín, 45–75. Amsterdam: John Benjamins. Gengel, Kirsten. 2006. “Phases and Ellipsis: vP, CP, and DP”. Paper presented at LUCL Wednesday Syntax Meeting, Leiden. Gengel, Kirsten, 2007. “Focus, Exhaustivity, and Deletion in English Pseudogapping”. Coreference, Modality, and Focus. Studies on the Syntax-Semantics Interface, ed. by Luis Eguren & Olga Fernández Soriano, 72–88. (= Linguistics Today, 111.) Amsterdam: John Benjamins. Giannakidou, Anastasia & Melita Stavrou. 1999. “Nominalization and Ellipsis in the Greek DP”. The Linguistic Review 16. 295–331. Jones, Michael. 1993. Sardinian Syntax. London: Routledge. Kiss, Katalin É. 1998. “Identificational Focus versus Information Focus”. Language 74:2. 245–273. Laca, Brenda. 1999. “Presencia y Ausencia de Determinante”. Gramática descriptiva de la lengua española, ed. by Ignacio Bosque & Violeta Demonte, vol. 1, 891–928. Madrid: Espasa. Lobeck, Anne. 1995. Ellipsis: Functional Heads, Licensing and Identification. New York: Oxford University Press. López, Luis. 2000. “Ellipsis and Discourse-Linking”. Lingua 110. 183–213. Marchis, Mihaela & Artemis Alexiadou. this volume. “On the Distribution of Adjectives in Romanian: the cel construction”. Martinho, Fernando Jorge dos Santos. 1998. A Elipse Nominal em Português e em Francês. MA thesis, Universidade do Oporto. McNally, Louise. 2004. “Bare Plurals in Spanish are Interpreted as Properties”. Catalan Journal of Linguistics 3. 115–133. Merchant, Jason. 2001. The Syntax of Silence: Sluicing, Islands and the Theory of Ellipsis. Oxford: Oxford University Press.

 Luis Eguren Ntelitheos, Dimitrios. 2004. Syntax of Elliptical and Discontinuous Nominals. MA thesis, UCLA. Ronat, Mitsou. 1977. “Une Contrainte sur l’Effacement du Nom”. Langue: Théorie générative étendue, ed. by M. Ronat, 153–169. Paris: Collection Savoir. Rooth, Mats. 1992. “A Theory of Focus Interpretation”. Natural Language Semantics 1:1. 75–116. Rooth, Mats. 1996. “Focus”. The Handbook of Contemporary Semantic Theory, ed. by Shalom Lappin, 271–297. Oxford: Blackwell. Sánchez, Liliana. 1996. Syntactic Structures in Nominals: A Comparative Study of Spanish and Southern Quechua. Ph.D. dissertation, University of Southern California. Sleeman, Petra. 1993. “Noun Ellipsis in French”. Probus 5. 271–295. Sleeman, Petra. 1996. Licensing Empty Nouns in French. Ph.D. dissertation, University of Amsterdam. Ticio, M. Emma. 2005. “NP-Ellipsis in Spanish”. Selected Proceedings of the 7th Hispanic Linguistics Symposium, ed. by David Eddington, 128–141. Somerville, Mass.: Cascadilla Proceedings Project. Winkler, Susanne. 2005. Ellipsis and Focus in Generative Grammar. Berlin/New York: Mouton de Gruyter.

On the nature of covert operations The case of focus in Spanish pseudoclefts* Olga Fernández-Soriano

Universidad Autónoma de Madrid The phenomenon analyzed in this paper is instantiated by a type of pseudocleft structure in (oral) Spanish, characterized by the fact that when the element which is clefted (also marked as focus) is an internal argument, it can ‘extend’ its focal status to the constituents dominating it (VP/IP). As a consequence, in these constructions syntactic structure does not match semantic interpretation. This is (obligatorily) accompanied by a phenomenon of ‘relative reduction’: the relative pronoun in the relative clause does not spell out the features of the clefted element but those of the constituent which is actually interpreted as focus (i.e. the VP or the whole IP). The analysis proposed for this phenomenon is that there is a covert process of focus extension preceded by (also covert) reconstruction. In particular, we adapt the monoclausal analysis of English cleft structures (Kiss 1998) to pseudoclefts and claim that (identificational) focus propagates in a direct path to VP or IP thus accounting for the interpretation obtained. This approach will be extended to other structures including a type of wh-structure, the so called ‘split interrogatives’.

1.  Introduction The point of departure of this paper is the behavior of pseudocleft structures in (oral varieties of) Spanish. The properties displayed by these structures will allow us to show that there is a special type of covert operation. In regular pseudocleft sentences in Spanish, the clefted element which follows the copulative verb ser is interpreted as focused (as is the usual case). The grammatical subject of the

*I would like to thank Violeta Demonte, Luis Eguren, Gemma Rigau and José María Brucart for their encouragement and for their comments and suggestions on earlier versions of this paper. I also thank the audience of Going Romance 2007. The two anonymous reviewers made very important observations which have crucially contributed to improve this paper. The research behind this work has been partially supported by a research project (EDU2008–01268) by the Spanish Ministerio de Ciencia e Innovación. Errors are all my own.

 Olga Fernández-Soriano

construction is a relative clause headed by a form which contains two parts: a referential component (a determiner) and a quantificational component (a wh-element). This is exemplified in (1): (1) Lo que quiero es un coche. the that I.want is a car “What I want is a car.”

As can be seen in (2), not only DP’s but also PP’s and adverbial phrases can be clefted. In this case, the relative form used in Spanish is complex: a PP or an adverbial relative, depending on the type of clefted phrase. This holds both for arguments and for adjuncts:1 (2) a.

A quien lo entregué fue a Maria. to whom it I.gave was to Maria

b. Donde lo puse fue debajo de la mesa where it I.put was under the table c.

Como habla es muy despacio. how s/he.speaks is very slowly

d. Por loˉque te lo digo es por eso. for what you it I.tell is for that

Another property of pseudoclefts is that the element that is clefted is interpreted as identificational focus. As Kiss (1998) argues, this element “represents the value of the variable bound by an abstract operator expressing exhaustive identification” and thus has to occupy a scope position.2 The rest of the structure is then interpreted as presupposed material. I will come back to the movement process and the position of the informational focus later. An important fact for our study is that in the structures under examination, together with verbal complements and adjuncts, the whole predicate and even the whole sentence can be clefted (and thus interpreted as focus). Crucially, in these cases in Spanish the relative clause is always headed by the neuter simple form lo que (lit. the that, “what”) and it contains a support verb, usually hacer “to do/ make” in the case of clefted VP (see (3)) and pasar or suceder “to happen” in the case of CP (see (4)). In the following examples both the support V and the clefted constituents are underlined:

1.  I only provide literal translations for the sentences in (2), since English does not easily allow complex relatives in these cases. 2.  Information focus, on the other hand, does not involve operator-variable relations.



(3) a.

On the nature of covert operations 

En esos casos, lo que hago es empezar de nuevo. in those cases, the that I.do is start of new “In those cases what I do is start from scratch.”

b. En vez de arreglarlo, loˉ que hice fue comprar otro coche. instead of repair.it, the that I.did was buy another car “Instead of repairing it, what I did was to buy a new car.” c.

Lo que tienes que hacer es echarle otro vistazo. “What you have to do is take another look at it.”

(4) a.

Lo que pasaba es que todos tenían ideas parecidas. “What happened is that they all had similar ideas.”

b. Quiero ir, lo que pasa es que no tengo tiempo. “I want to go, what happens is that I don’t have time.”

With this background in mind, the phenomenon I would like to analyze in this paper is the following. In spoken Spanish pseudocleft constructions display two parallel properties: one of a formal (syntactic) nature and the other of a semantic kind. To my knowledge, the second (semantic) property has so far gone unnoticed to grammarians. As to the formal property, which extends to all varieties of (spoken) Spanish, it has to do with the form of the relative pronoun. Even in cases of clefted PP’s or adverbial phrases, the relative pronoun is ‘reduced’ so that the simple form lo que (lit. the that, “what”) appears, instead of the required complex prepositional or adverbial relative (compare (2) with (5)). All the examples presented below are taken from real speech. (5) a.

Lo que estamos acostumbrados es a otra cosa. what we.are used.to is to other thing “What we are used to is something else.”

b. Lo que me habló fue de irnos de vacaciones. what me he.spoke was of going on vacation “What he spoke to me was about going on vacation.” c.

Ya somos mayorcitos y lo que estamos es en casa already we.are quite.old and what we.are is at home



a gusto viendo el televisor. comfortably watching the TV



“We are quite old and where we stay is at home comfortably watching TV.”

But the main property I would like to analyze has to do with the interpretation of these structures. To my knowledge, this property has so far not been noted by theoretical or descriptive studies. In a nutshell, not only is there a mismatch with respect to the relative form used (simple lo que) and the focalized constituent (a PP or an adverbial, which would require a complex form), the important thing is that what is interpreted as focus does not coincide with the clefted phrase either.

 Olga Fernández-Soriano

On the contrary, the focus of the structure may be a higher constituent. In other words, cases are frequently found in which (together with ‘relative reduction’ to lo que “what”) a verbal complement is clefted but the whole VP or even the whole sentence is interpreted as focus. Some real examples are provided in (6), where the English glosses correspond to the interpretation actually given to the sentences: (6) a.

(No me he disculpado) lo que le he traído es un regalo. (I didn’t apologize), what her I’ve.brought is a present “I did not apologize, what I did is bring her a present.”

b. Lo que no tienes ganas es de ir. what no you.have mood is to go “What happens is that you don’t feel like going.” c.

Lo que no puedes prorrumpir es en sollozos. what not you.can burst is into tears “What you cannot do is burst is into tears.”

Contrary to what is suggested by the structure, the focus of the sentence in (6a) is not on the object but on the entire VP: it is not presupposed that the subject brought something (exhaustively identified as a present), as would be the case if only a present is interpreted as focus, but that he did something (i.e bring a present to someone), This is shown by the translation. The case of (6c), is even clearer: the intended interpretation must be that in (6’), where focus is on the whole IP, since the presupposition is not that there is something the hearer has to burst into, with focus on en sollozos. (6′) Lo que no puedes (hacer) es prorrumpir en sollozos. “What you cannot (do) is burst into tears.”

In a similar fashion, sentence (6b) is interpreted with focus on the whole IP. What these facts seem to indicate is that the focus has the possibility to covertly extend up to the next branching node. For this extension to take place, we will claim, the clefted elements covertly reconstruct to their original position. Before getting into the concrete analysis I will propose, something needs to be said about focus interpretation.

2.  Focus ‘extension’ It is an extended assumption in the literature on focus that there is a direct relation between focus marking and nuclear stress assignment. In particular, it is claimed that a focal element must bear nuclear stress at PF (Selkirk 1982; Zubizarreta 1998;



On the nature of covert operations 

among others). It is also assumed that in the unmarked case nuclear stress falls on the most embedded element (Cinque 1993). An important fact for our purposes is that it has also been shown that in sentences with unmarked word order a process or ‘(informational) focus projection’ takes place. By means of this process, the element that bears the nuclear stress (the most embedded constituent) projects its focal status up to the nodes that dominate it (Reinhart 2006, for instance). Thus, many different focus structures can be obtained, proceeding in a direct path from the most embedded element (usually the direct object) up to the whole sentence. This is illustrated in (7): (7)

CP C′ C

TP Juan

T T

VP V′ V told



NP the truth



Note that the subject, crucially, is left aside in this ‘extension’ process, since it occupies an external position. I will come back to this issue later. On the other hand, according to Reinhart (2006) among others, in order to mark focus on a phrase that cannot receive nuclear stress by this unmarked process (that is, when the sentential focus is not the most embedded constituent), some special strategy must be employed. There are at least two possibilities: the application of a deaccentuation rule followed by a marked stress rule or different types of scrambling operations (see Vallduví 1988). Pseudoclefting is a syntactic means of focalizing by extraposing a constituent which is interpreted as focus.3 3.  See Irurtzun (2007) for a detailed description of the two types of foci. This author makes a distinction between ‘out of the blue’ and ‘contextualized’ utterances. The former contain broad focus and the later narrow focus. It is also shown in this work that there is a distinction between categorially different pitch accents for both types of foci. See the mentioned work for details.

 Olga Fernández-Soriano

In this work, I will adapt Kiss (1998)’s monoclausal structure for cleft structures to pseudoclefts. As can be seen in (8), a Focus Phrase headed by the copulative verb is generated under IP. The focused constituent moves to Spec,FP. In a pseudo cleft, the CP complement of FP further moves to Spec,IP, as shown in the structure below (adapted from Kiss 1998: (41)):

(8)

CP IP

I′

CPj Spec

C′ C

Lo

I IP

FP PP

F′ F

CP

que no puedes prorrumpir ti esk en sollozosi tk

tj

As we saw, different constituents can be clefted (and thus focalized): arguments, adjuncts, the whole VP and even the whole sentence. See (9): (9) a.

Lo que le dio Juan a María en Navidad fuel el regalo. what CL gave Juan to Mary for Christmas was the present “What John gave Mary for Christmas was the present.”

b. Cuando le dio Juan a María el regalo fue en Navidad. when CL gave Juan to Mari the present was for Christmas c.

Lo que hizo J. fue darle el regalo a María en Navidad. what did J. was give.CL the present to Mary for Christmas

d. Lo que pasó fue que J. le dio el regalo a M en Navidad. what happened was that J. gave the present to M. for Xmas

With this in mind, let us go back to our example (6) repeated below in (10a), with the interpretation in (10b): (10) a.

Lo que no puedes prorrumpir es en sollozos. what not you.can burst is into tears

b. Lo que no puedes (hacer) es prorrumpir en sollozos. what not you.can do is burst into tears “What you cannot do is burst into tears.”

In view of data such as (10), I would like to claim that the extension of the domain of focus to the projections dominating the constituent bearing the pitch accent is not restricted to informational focus (Kiss 1998) but, under certain circumstances,



On the nature of covert operations 

it can operate in the case of identificational focus. The analysis I would like to propose for this phenomenon is that in the constructions under study, there is a possibility of ‘reconstructing’ and ‘propagating’ the sentential focus in the interpretive (covert) component. This process takes place, as is always the case, for focus extension, from the most embedded phrase up to the whole IP. In a structure such as (8) the element in Spec,FP would covertly reconstruct to its original position inside CP (in bold in the tree in (8)) and from there focus would propagate up to VP thus accounting for the interpretation obtained. As was mentioned above, in the Spanish cases considered here, it is the simple neuter form lo que “what” which appears in Spec,CP. Crucially, this is the form that is used when a VP or an IP are clefted (see examples above). That amounts to saying that the referential component of the relative form refers to VP or IP (the one actually interpreted as focus), although it is an NP or PP inside it what is clefted. What the phenomenon under analysis seems to suggest is that, contrary to what has usually been accepted, the constituent interpreted as focus may not coincide with the one which is both phonologically and syntactically marked as focus. Moreover, in the structures under study it is the contrastive / identificational focus that extends up to higher nodes. I provide some more real examples in (11): (11) a.

Pensábamos que estábamos anticuados y lo que abrimos fue we.thought that we.were out.of.date and what we.opened was



un camino que después han seguido muchos. a path that afterwards have followed many

“We thought we were out of fashion and what we did was open a path that many people have pursued afterwards.” b. No hagas una presentación en power point si no quieres, not you.make a presentation in power point if not you.want

pero lo que no puedes repetir es el papel. but what not you.can repeat is the paper

“Do not make a power point presentation if you don’t want to, but what you cannot do is just repeat the printed text.” c.

Lo que tienen que abrir es los ojos y darse cuenta de what they.have to open is the eyes and realize of



lo que pasa. what happens

“What they have to do is open up their eyes and realize what is happening.” d. No le castigaría, loˉque le pagaría es un psicólogo. not him I.would.punish, what him I.would.pay is a psychologist “I would not punish him, what I would do is pay for a psychologist for him.”

 Olga Fernández-Soriano

In all these sentences, as a consequence of focus extension, the presupposed material becomes less restricted: it is not presupposed that there is something that has been opened (11a) or that should be repeated (11b), opened (11c) or paid (11d). Note that ‘relative reduction’ is mandatory for the extended focus interpretation to be obtained. In the sentences in (12), for instance, the variants with a complex relative are odd since the extended focus interpretation is highly preferred (otherwise it would be presupposed that there is some period of time during which the speaker has to be substituted or that the subject has arrived at a certain moment) and this requires the simple form lo que with no preposition: (12) a. #Despedirme no, por loˉque me releva es por tres meses fire.me not, for what me he.substitutes is for three months “He will not fire me, what he will do is substitute for me for three months.” (vs. lo que me releva es por tres meses) b. #No ha cometido ningún delito: cuando ha llegado es tarde. not he.has committed any crime: when he.has arrived is late “He did not commit any crime: all he did was arrive late.” (vs lo que ha llegado es tarde)

2.1  Island effects An interesting property displayed by the constructions under study is that if the clefted element is reconstructed (and the focus is extended), the sentence constitutes a strong island and further extraction is not allowed. In (13a, b) we have grammatical sentences because there is no covert extension process. In contrast, (13c, d) are ungrammatical. (13) a.

¿Qué es lo que comiste/quieres comer? what is.it what you.ate/want.to eat

b. ¿Adónde es adonde fuiste/quieres ir? where is.it where you.went/want.to go “Where is it that you want to go.” c. *¿Qué/adónde es lo que fuiste/quieres ir? where is.it what you.went/want.to go d. *¿En qué es lo que prorrumpe siempre? into what is what s/he.bursts always

This contrast is expected, since focus cannot extend if the clefted element moves to the (highest) CP position. The idea is that once the clefted element is moved to the higher Spec,CP it cannot reconstruct inside the relative clause and propagate.



On the nature of covert operations 

2.2  Fixed and periphrastic expressions, idiom chunks There is an important fact that I would like to note: given that there is a covert focus reconstruction process, it is possible to cleft any embedded constituent, independently of how tight its relation with the verb is. That is why in some cases ‘undetachable’ phrases that cannot be affected by any syntactic operations (such as passive or wh-movement) can nevertheless be clefted, if and only if the relevant phrase reconstructs and focus is extended. This is the case, for example of periphrastic and idiomatic expressions. The structures in (14) would, as expected, be impossible without focus covert extension: (14) a.

Lo que voy a empezar es de nuevo. what I.am.going to start is of new “What I am going to do is start from scratch.” (vs. *Como voy a empezar)

b. Ahora lo que tenemos que seguir es aplicando lo que nos han now what we.have to keep is applying what us they.have

enseñado. taught



“Now what we have to do is continue applying what we have been taught.”

It is also possible to cleft parts of idiom chunks. All the sentences in (15) can only be interpreted with focus on the VP (15a, c) or on the whole IP (15b): (15) a.

Loˉque no puedes meter esˉ la pata hasta elˉpuntoˉdeˉque te what not you.can put is the foot until you



echen. they.fire



“What you cannot do is put your foot in it until you get fired.”

b. No es que no me guste, lo que cuesta es un riñón. not it.is that not me it.like, what it.costs is a kidney “It is not that I don’t like it, what happens is that it costs a fortune.” c.

Lo que no se puede usted ir es por las ramas, sin what not CL can you go is by the branches without



centrarse en el tema a debatir centering in the topic to debate



“What you cannot do is beat about the bush avoiding the point at issue.”

d. Lo que tienes que seguir dando es la tabarra; verás what you.have to keep giving is the pain, you.willl see

como te hacen caso. how you they.pay attention.

“What you have to do is keep on being a nuisance, you’ll see how they pay you attention.”

 Olga Fernández-Soriano

In sum, in Spanish pseudocleft structures the constituent which is clefted may not coincide with what is actually interpreted as focus. In the covert component, the postcopular phrase may reconstruct inside the relative clause and extend up to the whole predicate or even the whole sentence. The presupposition of the sentence is consequently modified. This focus extension is restricted in such a way that it operates from the most embedded constituent in a direct path until it reaches the sentence boundary, as has been described for (informative) focus propagation. An interesting corollary of this hypothesis is that it seems that idioms can be affected by syntactic processes if and only if they are kept untouched in the semantic component. 3.  Arguments supporting the covert focus extension hypothesis 3.1  Subjects and VP adverbs One piece of evidence in support of the hypothesis just sketched comes from the behavior of the subject. From the analysis I am defending it follows that, since the subject occupies an external position, focus cannot propagate from it in a direct path up to VP or IP (although it can of course be focalized), so no extended interpretation should be found with subjects. This prediction is confirmed, and a sentence like (16b) cannot be interpreted, in any variety of Spanish, as (16a). For the same reason, sentence (16c) (with the corresponding neuter form lo que) is ungrammatical. The same holds for (17): (16) a.

Quien tiene que arreglarlo es Juan. who has to fix.it is John

b. #Lo que Juan tiene que hacer es arreglarlo. what John has to do is fix.it c. *Lo que tiene que arreglarlo es Juan. (17) a.

Quien tiene que hablar es tu padre. who has to speak is your father

b. #Lo que tiene que hacer tu padre es hablar. what has to do your father is speak up c. *Lo que tiene que hablar es tu padre.

The reason for this fact is, according to what I have been claiming, that the subject occupies a higher Spec position, and thus focus cannot propagate from it. An interesting fact that supports the analysis defended here is that, as expected, the preceding generalization does not hold for internal (unaccusative) subjects, since in this case the propagation line may be obtained. This seems in fact to be the case: sentences like (18) can be found, where focus extends from the subject



On the nature of covert operations 

of an unaccusative verb. In these cases the non agreeing neuter form lo que, is also obligatory: (18) a.

No es que tuviera suerte, loˉque seˉle apareció fue la virgen. not is that he.had luck what to.him appeared was the Virgin “It is not just that he was lucky: what happened to him was a miracle.”

b. Lo que se le vino fue el mundo encima. what him came was the world on top “What happened is that the whole world came down on him.” c.

No se ha estropeado: lo que han crecido son flores not it.has ruined what have grown are flowers



amarillas por todas partes. yellow everywhere

“It is not ruined: what happens is that yellow flowers have grown everywhere.”4

Due to similar reasons, another prediction of this analysis is that focus cannot extend from adverbs modifying the whole VP or, of course, from sentential adverbs. Therefore sentences like (19a’, b’) are impossible: (19) a.

El chico [toca el piano] maravillosamente. the boy plays the piano beautifully

a′. *El chico no es guapo, (pero) lo que toca el piano es the boy not is handsome but what he.plays the piano is

maravillosamente beautifully

b. El chico [se abstuvo] inteligentemente. the boy abstained intelligently (from voting) b′. *El chico no estuvo muy afortunado, (pero) lo que se abstuvo the boy not was very brilliant, but what he.abstained

fue inteligentemente. was intelligently.

4.  An anonymous reviewer points out correctly that sentences with extended focus with clefted (internal) subjects are not always well formed. As María Luisa Zubizarreta (p.c.) notes, this is specially clear with human DP’s. Compare, in this respect (18c) with: (i) ?*Sí los reconocí, pero lo que ha crecido mucho es el niño. I.did them recognize, but what has grown a.lot is the boy I cannot provide an appropriate account for this fact.

 Olga Fernández-Soriano

3.2  Focus ‘restriction’ An indirect piece of evidence in favor of our proposal about covert focus extension is that the inverse process of focus ‘restriction’ never takes place, that is, in all varieties of Spanish interpretations such as the ones in (20b) and (20d) for sentences (20a) and (20c) are impossible: (20) a.

Lo que tienes que hacer es hablar con tu padre. “What you have to do is talk to your father.”

b. #Con quien tienes que hablar es con tu padre. with whom you.have to talk is with your father c.

Lo que pasa es que has comprado un coche usado. “What happens is that you have bought a used car.”

d. ≠ Lo que has comprado es un coche usado. “What you have bought is a used car.”

3.3  Double objects An interesting clue about how the phenomenon of covert focus extension works is provided by ditransitive and double object constructions. In these configurations, the two objects seem to form a unit that cannot be divided. In particular extended focus is impossible if only the indirect object is clefted, unless the direct object is cliticized. Of course, focus can extend from the constituent formed by the two objects (as in (21a)). (21) a.

Lo que tienes que dar es [el regalo a tu hermana]. what you.have to give is the present to your sister “What you have to do is give the present to your sister.”

b. Lo que se lo tienes que dar es a tu hermana what CLdat CLacc you.have to give is to your sister c. *Lo que le tienes que dar el regalo es a what CLdat you.have to give the present is to

tu hermana your sister

The same contrast obtains with locative constructions, as seen in (22): (22) a.

Lo que llevó fue [el libro al sitio equivocado]. what he.took was the book to.the place wrong “What he did was he took the book to the wrong place.”

b. Lo que lo llevó fue al sitio equivocado. what CLacc he.took was to.the place wrong



On the nature of covert operations 

c. *Lo que llevó el libro fue al sitio equivocado. what he.took the book was to.the place wrong d. Loˉque no puedes ponerˉ es [el coche en otro garaje]. what not you.can put is the car in another garage “What you cannot do is put the car in another garage.” e.

Lo que no puedes ponerlo es en otro garaje what not you.can put.it is in another garage

f. *Lo que no puedes poner el coche es en otro garaje. what not you.can put the car is in another garage

In all these cases focus cannot extend from the direct object alone either, but the sentence is not ungrammatical since there is no relative mismatch. The sentences below are grammatical, but only with a narrow focus interpretation. (23) a.

Lo que tienes que darle a tu hermana es el libro. “What you have to give to your sister is the book.”

b. Lo que llevó al sitio equivocado fue el libro. “What he took to the wrong place was the book.” c.

Lo que no puedes poner en otro garaje es el coche. “What you cannot put in another garage is the car.”

Idiomatic expressions that contain double objects behave in the same way: (24) a. *Lo que no puedes poner una pica es en Flandes. what not you.can put a pike is in Flanders “What you cannot do is do a totally new heroic thing.” (vs: Lo que no puedes poner es una pica en Flandes) b.

*Lo que no puedes dar tres cuartos es al pregonero. what not you.can give three quarters is to.the crier “What you cannot do is give reasons for people to gossip.” (vs. Lo que no puedes dar es tres cuartos al pregonero)

c. *Lo que quiere llevarse el gato es al agua. what he.wants to.take the cat is into.the water “What he wants to do is to be the winner.” (vs. Lo que quiere llevarse es el gato al agua) d.

*Lo que tiene que arrimar el ascua es a su sardina. what he.has to put.near the ember is to his sardine “What he wants to do is to defend his own interests.” (vs. Lo que quiere arrimar es el ascua a su sardina)

In conclusion, focus cannot extend from one of the objects of a ditransitive construction. This might constitute indirect evidence for an analysis in Hale & Keyser’s (1993) terms, in the sense that it might indicate that both objects form a single constituent. Something similar happens with secondary predicates, which can

 Olga Fernández-Soriano

also be claimed to form a single constituent (namely a small clause) with the main object:5 (24′) a.

No es que no quiera venir, loˉque tiene es a su madre enferma. not is that not wants come, what has is her mother sick

b. ?*… lo que tiene enferma es a su madre. c.

??…lo que tiene a su madre es enferma.

d. No es que se ocupe demasiado de su madre, lo que la not is that cares too.much about her mother, what her

tiene es enferma. has is sick

Going back to the main issue, the most important argument in my opinion, in favor of the analysis presented here is that the described phenomenon does not restrict to pseudocleft sentences. Once identified, one can find other cases in which focus (and scope) markers seem to raise and affect higher constituents than the one they appear with. The data I will present are all taken from real oral (and even written) speech and have, to my knowledge, also gone unnoticed so far. 4.  Other cases In what follows I will analyze other structures which demonstrate the silent processes of ‘extension’. More specifically, I will show that there are some particles that can covertly widen their scope to higher elements than the ones they explicitly appear with in the sentence. All the examples presented in this section are also taken from real written and oral speech. 4.1  Either, neither A first type of example is provided by cases like (25): (25) a.

Speaker A



-Yo creo que va a llover. I think that is.going to rain



-Yo creo que también. I think that also6 “I also think (that it is going to rain) / I think so too.”

5.  I thank an anonymous reviewer for pointing this out to me. 6.  Note that the word that in the gloss corresponds to the complementizer, not to the demonstrative (in which case the sentence would be grammatical in English).



On the nature of covert operations 

b. Speaker B

-Yo creo que no ha venido. “I think that not he.has come.”



-Yo creo que tampoco. I think that neither “I also do not think that he has come / Neither do I.”

As can be seen in the examples above, adverbs like también “too/also” and tampoco “neither” are interpreted as affecting the main verb creer “believe” and not the embedded one with which they appear. Examples like (26) can also be found: (26) a.

Yo creo que tampoco va a hacerlo. I think that neither is.going to do.it “I also do not think s/he is going to do it.”

b. Yo creo que también lo negará. I think that also it will.deny “I also think s/he will deny it.”

4.2  NPI’s, (not) even Negative polarity items like ni (siquiera) “not (even)” also give rise to this covert scope extension process: (27) a.

No voy a probar ni el postre. not am.going to try even the dessert “I am not even going to try the dessert.”

b. No es que no lo vaya a leer, es que no va a abrir not is that not it is.going to read, is that not is.going to open

ni el sobre even the envelope



“It is not that s/he will not read it, s/he will not even open the envelope.”

In all these cases the negative element ni is interpreted as negating the predicate (the VP) as a whole, not the element it is adjacent to. I would like to note that we are again facing a phenomenon which takes into account only hierarchical structure: idioms and frozen expressions can be ‘broken’ in these cases. See (28): (28) a.

No me dirige ni la palabra. not me he.addresses even the word “He does not even talk to me.”

b. Se fue sin decir nada: no me dio ni las gracias. he. left without say nothing: not me he.gave even the thanks “He left without a word: he did not even thank me.”

 Olga Fernández-Soriano

c.

Si no le tomas ni las medidas, no sé cómo vas if not her you.take even the measures, not I.know how you.go



a decidir. to decide.

“If you don’t even take her measures, I don’t know how you can make a decision.” d. Engaña a la gente muy bien, no se le ve ni el plumero he.lies to the people very well, not SE him see even the duster

ni nada. or anything



“S/he lies to people very well; one cannot even see what s/he is up to.”

4.3  Focus markers A third type of structure I would like to present is instantiated by certain focus marking particles such as hasta “even” o solo “only, just”, whose scope can be covertly extended:7 (29) a.

Desde que se ha ido fumo hasta menos. since that has left I.smoke even less “Since s/he left I have even reduced smoking.”

b. No he venido a traerte solo las croquetas sino a hablar not I.came to bring.you just the croquettes but to talk

contigo. with.you



“I didn’t come just to bring you the croquettes but to talk to you.”

c.

No tengo frío, de hecho me voy a quitar hasta el abrigo not I.have cold in fact refl I.go to take.off even the coat “I am not cold; in fact I am even going to take off my coat.”

Particles of the type of casi “almost” give raise to an analogous phenomenon. The example in (30), taken from a newspaper, is quite conclusive: (30)

El diputado de turismo vuelve al trabajo tras morir the deputy of tourism goes.back to work after dying

casi ahogado. almost drown “The tourism deputy goes back to work after almost drowning.”

7.  Brody (1990) assumes that these particles are operators in situ which move in LF to adjoin to FP. Kiss (1998) argues against this hypothesis.



On the nature of covert operations 

Again it seems that these particles ‘reconstruct’ and extend their scope from the most embedded constituent up to VP in a covert way. In the next section I will try to show that the proposed hypothesis can be extended to other cases of wh-movement.

5.  Wh movement: Split interrogatives An important consequence which follows from the hypothesis defended in this paper is that it is predicted that one should also find constructions with wh-movement (like interrogatives) which display the described phenomenon of covert extension. More precisely, cases should be possible where an interrogative element is fronted and, again, the constituent which is interpreted as the focus of the question is hierarchically superior. In what follows I will show that this prediction is born out. The examples I am about to present are sentences which involve a fronted wh-word and contain a phrase that constitutes a possible answer. The overall interpretation of the sentence is that of a yes/no question with focus on the wh-element (as indicated by the gloss). These structures have been called ‘split interrogatives’. Phonologically, they involve two intonational curves, as if there were two separate questions: a wh-question and a yes/no question.8 I provide some examples in (31): (31) a. ¿Qué quieres, que te dé una bofetada? what you.want, that you I.give a slap “What do you want, for me to slap you?” b. ¿Adónde vas, a tu casa? where you.go, to your house “Is it home that you are going?” c. ¿Qué vas a hacer, pagarle lo que le debes? what you.go to do, pay.him what him you.owe “Is it to pay what you owe him that you are going to do?” d. ¿Cómo lo has resuelto, con una calculadora? how it you.have solved, with a calculator “Is it with a calculator that you solved it?”

8.  Camacho (2002) calls these constructions ‘wh-doubling’ and notes some differences between them and other interrogatives. He points out, for example, that these structures differ from regular yes/no questions in not allowing the continuation o no “or not”. (i) *¿Dońde vas, a casa o no? Where you.go to home or not

 Olga Fernández-Soriano

Interestingly enough, as was the case for pseudoclefts, it is common in spoken Spanish to ‘reduce’ the wh-word to the simple form qué “what” (see López Cortina 2003). Actually the real reported sentences are the following:9 (32) a. ¿Qué vas, a tu casa? what you.go, to your home b. ¿Qué lo has resuelto, con una calculadora? what it you.have solved, with a calculator c. ¿Qué vino, en avión?10 what he.came, by plane?

Based on the fact that in these cases there is a single construction with two interdependent parts, rather than two juxtaposed questions, López Cortina (2003) proposes a monoclausal analysis of these constructions and claims that the element after the comma moves to a focus position in the left periphery.11 The structure would be like (32c’) (his (18)): (32c′) FocP Foc′

PP en avión

F

VP V′





V

PP

vino

t

9.  In Camacho (2002)’s analysis, the element after the comma appears as an adjunct to the VP trace. I refer the reader to the mentioned work for more details. 10.  López Cortina (2003) points out that in certain varieties of English there is a similar construction, with the particularity that the wh-word remains in situ. The author provides the example in (i) (his (6)):

(i)

You are going what, by train?

11.  According to the author, a strong argument in favor of the single construction approach is that corresponding structures without the focus are not grammatical if taken separately, as in (i) (to be compared to (32)). (i)

a. b. c.

*¿Qué lo mandaste? *¿Qué lo resolviste? *¿Qué vino?



On the nature of covert operations 

The crucial fact that I would like to point out here is that, when that happens (i.e. if the wh-element is reduced to qué “what”), the interrogative can ‘propagate’ up to VP or to IP, giving interpretations like the following.12 (33) a. ¿Qué huyó, fuera del país? what he.ran.away, out of the country “What did s/he do, run away from the country?” b. ¿Qué han matado, al presidente? what they.have killed, to.the president “What has happened, that they killed the president?”

Accepting the analysis by López Cortina (2003), our claim is that, once moved to FocP in (32’), the focal element reconstructs to the position of its trace inside VP in the covert component and expands up to VP or IP, just as was the case for pseudoclefts. Idioms can also be affected by this type of wh-movement, as can be seen in (34): (34) a. ¿Qué lo mandaste, al infierno? what him you.sent, to.the hell “What did you do, send him to hell?” b. ¿Qué me chupo, el dedo? what refl I.suck, the finger “What happens, (do you think that) I am an idiot?”

Some other real examples are provided in (35). (35) a. ¿Y qué te tienes que quitar, toda la ropa? and what refl you.have to take.off, all the clothing “And what do you have to do, take off all your clothes?” b. Tienes dos hijos: ¿qué te vas a hacer cargo, de un you.have two children, what you go to make charge, of a

tercero? third.one

“You have two children, what are you going do, get in charge of a third one?” c.

Ya lo has intentado todo: ¿qué vas a ir ahora, already it you.have tried everything, what you.go to go now,



a ver al rey? to see the king?

“You have already tried everything, what are you going to do now, go to see the king?”

López Cortina also claims that qué is the spell out of the trace of the focal DP, and moves to a higher position in the left periphery. We refer the reader to the mentioned work for further details. 12.  This fact has also never been observed in the literature before.

 Olga Fernández-Soriano

It thus seems that there are other constructions instantiating the special type of covert movement process that was shown to apply to pseudoclefts. The result is that some structures have readings that do not correspond to what would be obtained by computing overt superficial constituent order.

6.  Conclusions The phenomenon analyzed in this paper involves a special type of covert operation. More specifically it has been shown that, in oral Spanish pseudocleft structures, when the element which is actually clefted is an internal argument, it can ‘extend’ its focal status to the constituents dominating it (VP/IP), thus instantiating a case of mismatch between syntactic structure and semantic interpretation. This is (obligatorily) accompanied by a phenomenon of ‘relative reduction’, by means of which the relative pronoun in the relative clause does not match the features of the clefted element but those of the constituent which is actually interpreted as focus (i.e. the VP or the whole IP). The proposal that has been put forward states that there is a covert process of focus extension preceded by (covert) reconstruction. In particular we have adopted Kiss’ (1998) monoclausal analysis of English cleft structures and adapted it to pseudoclefts. We claim that the clefted element moves covertly from its position in Spec,F(ocus)P to its most embedded position inside the relative CP. From there (identificational) focus propagates in a direct path up to VP or IP thus accounting for the interpretation obtained. It has also been shown that idiom chunks can be affected by clefting (contrary to what could be expected) just because this process of (covert) reconstruction takes place. Other structures have been analyzed which instantiate the same phenomenon of covert focus extension and finally the proposal has been extended to a special type of wh-structures (called ‘split interrogatives’). An important question that is left unresolved here is whether there is a (macro) parameter involved. From the analysis presented here it follows that the option described should be available for other languages but it is still to be determined under which circumstances.

References Brody, Michael. 1990. “Some Remarks on the Focus Field in Hungarian”. UCL Working Papers in Linguistics 2. 201–225. Camacho, José. 2002. “Wh-Doubling: Implications for the Syntax of Wh-Movement”. Linguistic Inquiry 33. 157–164.



On the nature of covert operations 

Cinque, Guglielmo. 1993. “A Null Theory of Phrase and Compound Stress”. Linguistic Inquiry 24. 239–297. Hale, Kenneth & Samuel J. Keyser. 1993. “On Argument Structure and the Lexical Expression of Syntactic Relations”. The view from Building 20: Essays in linguistics in honor to Sylvain Bromberger ed. by Kenneth Hale & Samuel J. Keyser, 53–110. Cambridge, Mass: MIT Press. Irurtzun, Aritz. 2007. The Grammar of Focus at the Interfaces. Ph.D. Dissertation, University of the Basque Country. Kiss, Katalin É. 1998. “Identificational Focus versus Informational Focus”. Language 74. 245–273. López Cortina, Jorge. 2003. “The Structure of Split Interrogatives”. Theory, Practice, and Acquisition ed. by P. Kempchinsky and C. Piñeros, 140–55. Somerville, Mass: Cascadilla Press. Reinhart, T. 2006. Interface Strategies: Optimal and Costly Computations. Cambridge, Mass: MIT Press. Selkirk, Elisabeth. 1982. Phonology and Syntax. Cambridge, Mass: MIT Press. Vallduví, Enric. 1988. “Functional Load, Prosody, and Syntax: Left-detachment in Catalan and Spanish”. Papers from the 24th Annual Regional Meeting of the Chicago Lingusitic Society ed. by L. McLeod et al., 391–403. Chicago: CLS. Zubizarreta, M. Luisa. 1998. Prosody, Focus and Word Order. Cambridge, Mass: MIT Press.

Ellipsis and Restructuring in European Portuguese* Anabela Gonçalves & Gabriela Matos Universidade de Lisboa, CLUL

In this paper we propose that, in European Portuguese, the complementary distribution between Null Complement Anaphora (NCA) and Restructuring (also noted by Brucart 1999 and Depiante 2001, for Spanish and Italian) results from the selectional properties of the verbs that accept these constructions: NCA corresponds to active T domains at the level of semantic interpretation, and is selected by verbs with intrinsic lexical content, while Restructuring involves a biclausal sentential structure, headed by a functional or a lexical verb occupying a T projection, and requires the embedded T to be syntactically inactive.

1.  Introduction Null Complement Anaphora (NCA), illustrated in (1) for English and Portuguese, is assumed to be a lexically-driven construction, occurring with verbs that do not seem to share any syntactic or semantic property underlying their ability to license the null complement (Brucart 1999; Depiante 2001; Cyrino & Matos 2006): 

(1) a.

I play cards and shoot dice, and my wife doesn’t approve __. (Hankamer & Sag 1976:412)

b. Eu queria que eles ficassem em casa e eles I want.pst.1sg that they stay at home and they

concordaram __. agree.pst.3sg



“I wanted that they stayed at home and they agreed.”

*We thank the participants of Going Romance 2007, as well as Enoch Aboh, Reineke Bok-Bennema, Jairo Nunes and Eduardo Raposo for comments on the topics tackled in this paper. We especially thank two anonymous reviewers of a previous version of this paper for their comments and suggestions. The research leading to this study has been developed within Onset-CEL (POCTI-SFA-17-745) and financially supported by Fundação para a Ciência e a Tecnologia.

 Anabela Gonçalves & Gabriela Matos

Verbs presenting similar content vary as far as the licensing of NCA is concerned. For instance, in European Portuguese (EP), not all modals expressing necessity allow this construction: dever “should” accepts it (2a), but ter (de) “have to” does not (2b). Likewise, continuar (a) “ continue/keep on” is an NCA licensor (3a), while estar (a) “be Ving” is not (3b), although both verbs are durative aspectuals: (2) a. Ele trabalha muito, mas não devia __. he work.prs.3sg much, but not should.3sg “He works hard, but he should not.” b. *Ele trabalha muito, mas não tinha __. he work.prs.3sg much, but not should.3sg (3) a.

Ela  já começou a estudar, e vai continuar __ amanhã. she  already begin.pst.3sg to study, and will continue tomorrow “She has already begun studying and she will keep on doing it tomorrow.”

b. *Ela já começou a estudar, e vai estar __ she already begin.pst.3sg to study, and will be

amanhã também. tomorrow too

From a syntactic point of view, identical contrasts show up: verbs selecting similar complements behave differently with respect to NCA – e.g., concordar (em) “agree (on)”, in (1b), contrasts with pensar (em) “think about”, in (4), despite the fact that both verbs take sentential prepositional complements: (4) *Ela trabalha aos fins de semana, mas não pensavaˉ __. she work.prs.3sg during.the weekends, but not think.pst.3g

This suggests that NCA constitutes an idiosyncratic property of some predicates and that each verb should be specified in the Lexicon as an NCA licensor or not. Such an approach is problematic since it relies on a loose conception of the Lexicon. In fact, following Chomsky (1995:6), the Lexicon should not be redundant and, in its optimal form, should only contain the information required for the computational system. Ascribing a specific construction to each lexical entry will overload the Lexicon in an unwelcome way, since constructions are not computational entities of the system. Attempting to devise some general properties underlying the distribution of NCA, Brucart (1999) claims that, in Spanish, this construction is licensed by verbs that do not accept the clitic lo, which denotes sentences or predicates (see the contrast in (5) between olvidar “forget” in (5a), which may take the clitic lo as its



Ellipsis and Restructuring in European Portuguese 

direct object, and olvidarse (de) “forget (about)” in (5b), which selects an oblique complement): (5) a.

Habían quedado en ir a comer pero *(lo) olvidó. have stay.pst.part on go for eat but   it forget.pst “They had agreed to go out for a meal but he forgot it.”

b. Habían quedado en ir a comer pero se have stay.pst.part on go for eat but himself (*lo) olvidó.   it forget.pst

“They had agreed to go out for a meal but he forgot (about it).”

Furthermore, Depiante (2001) claims that NCA predicates are in complementary distribution with Restructuring (cf. (6)):1 (6) *Juan las quiere ver, y Maria también las Juan [them.cl] want.prs.3sg see, and Maria also [them.cl] quiere __. want.prs.3sg

Correlating this property with Brucart’s generalisation on the ban of lo, she claims that NCA is a Deep Anaphora in Spanish and Italian, as proposed for English by Hankamer and Sag (1976): the omitted constituent would be a null pronominal, the null counterpart of lo. However, this correlation is challenged by EP, since in this language NCA occurs with verbs that admit the

1.  Brucart (1999:2841) presents examples that illustrate the complementarity between NCA and Restructuring, but does not correlate the two constructions. He provides evidence for the incompatibility between NCA and Long Object Movement (with the passive clitic se) but claims that the resulting marginality is due to the passive effects, (iB): (i) A: Se deben pagar estos trabajos? B: *Se deben.   pass.cl should pay these works? B:  pass.cl should The fact that in EP the passive clitic, in (iia)B, contrasts with the verbal passive, in (iib)B, leads us to impute the marginality of (iB) and (iia)B not to passive, but to Restructuring: (ii) a. A: Devem-se pagar estes trabalhos? B: *Devem-se   should-pass.cl pay these works? B:   should-pass.cl b. A: Estes trabalhos devem ser pagos.   these works should be paid “These works should be paid.”

B:  Mas podem? B:  but can? “But can they (be paid)?”

 Anabela Gonçalves & Gabriela Matos

invariable clitic o “it” (7b), but behaves like Spanish and Italian with respect to Restructuring (8): (7) a.

Ele viaja muito, embora não aprecie __. he travel.prs.3sg much, although not enjoy.prs.3sg “He travels a lot, although he does not enjoy travelling.”

b. Ele viaja muito, embora não o aprecie. he travel.prs.3sg much, although not [it.cl] enjoy.prs.3sg “He travels a lot, although he does not especially enjoy it.” (8) *Eu também as quero ver, mas não as posso__. I also [them.cl] want.prs.1sg see, but not [them.cl] can.1sg

Thus, the main goals of the current study are to isolate the selectional properties of NCA licensing verbs and to understand how the above mentioned selectional properties account for the complementary distribution between NCA and Restructuring. The paper is structured as follows: Section 2 re-examines the arguments for and against the status of NCA as a Deep Anaphora. Section 3 analyses the selectional properties of the verbs that admit NCA in EP. Section 4 focuses on the correlation between NCA and Restructuring and on the structures required by these constructions. Finally, Section 5 presents the concluding remarks.

2.  NCA as a Deep or Surface anaphora and the distribution of Restructuring The main questions about NCA concern the structure and recovery of the omitted constituent. Depiante (2001) assumes that in Spanish and Italian it is a pronominal, the null counterpart of the clitic lo, which denotes sentences or predicates. She argues that, in these languages, NCA is a Deep Anaphora (Hankamer & Sag 1976), i.e., a null constituent with no internal structure that picks up its denotation from a salient linguistic or situational context. In favour of her claim, Depiante shows that in Spanish (and Italian) the content of the omitted constituent can be pragmatically recovered (9); there is no requirement of strict parallelism between a linguistic antecedent and the structure of the recovered constituent (see (10), where the antecedent in the passive form is recovered by an active sentence); and the construction shows Missing Antecedent effects – cf. (11), which, according to Depiante, is marginal, because the null subject pronoun, pro, cannot find an antecedent in the NCA gap, since this gap corresponds to a pronoun and it has no internal structure: (9) [Javier jumps into the icy cold sea] Juan says: Yo también puedo __! I also can.1sg

(Depiante 2001:206)



(10)

Ellipsis and Restructuring in European Portuguese 

Los pacientes del tercero tienen que ser llevados a the patients of.the third have.prs.3pl to be taken to

terapia intensiva aunque la enfermera con más fuerza therapy intensive even.if the nurse with more strength no pueda __. not can.3sg “The patients of the third floor have to be taken to intensive therapy although the strongest nurse cannot (take them).” (id:207) (11) *Juan no pudo asesinar a Pablo con un cuchillo Juan not could.3sg kill to Pablo with a knife pero Pedro sí pudo __ y pro estaba oxidado. but Pedro yes could.3sg and pro was rusted

(id:208)

According to this approach, the complementary distribution between NCA and Restructuring derives from the pronominal nature of the null complement – since Restructuring operates on verbal or sentential complements – and its lack of internal structure, which would prevent the moved constituents from being related to their copies in Clitic Climbing (12a) and Long Object Movement (13a):2 (12) a. *Juan las quiere ver, y María también Juan [them.cl] want.prs.3sg see, and Maria also

las quiere __. [them.cl] want.prs.3sg

b. Juan quiere verlas, y María también quiereˉ__. Juan want.prs.3sg see.[them.cl], and Maria also want.prs.3sg “Juan wants to see them, and Maria also wants to (see them).” (13) a. *Le nuove case si cominceranno a costruire e the new houses [pass.cl] begin.fut.3pl to build and

anche i nuovi appartamenti si cominceranno. also the new flats [pass.cl] begin.fut.3pl

b. A Roma si comincerà a costruire le nuove case in Rome [nom.cl] begin.fut.3sg to build the new houses

ma a Milano non si comincerà. but in Milano not [nom.cl] begin.fut.3sg

“In Rome one will begin building the new houses but in Milano one will not (begin building the new houses).”

2.  Long Object Movement (cf. Rizzi 1982) captures the agreement relation between the embedded object and the matrix verb. Thus, in (13a), this verb occurs in the 3PL (cominceranno). In contrast, in (13b), it is the Nominative clitic that triggers agreement on the verb, 3SG (si comincerà).

 Anabela Gonçalves & Gabriela Matos

Yet, this approach is not adequate for Portuguese. Cyrino & Matos (2006) show that, in EP, NCA occurs with verbs that accept the clitic o “it”, as in (14), which exemplifies the free variation between the omitted constituent and the clitic pronoun. Besides, in Portuguese, the omitted constituent in NCA exhibits internal structure: it co-occurs with Antecedent Contained Deletion (15) and wh- extraction (16), and does not show Missing Antecedent sensitivity (17): (14) Só sairás, se {desejares __ /o desejares}. only go.out.fut.2sg, if {wish.fut.2sg /[it.cl] wish.fut.2sg} “You will only go out, if you {wish/wish it}.” (15)

Eu não me recuso a fazer qualquer coisa I not [me.cl] refuse.prs.1sg to do anything



Op que ele mande __.3 ( __= fazer (Op) ) Op that he order.prs.3sg ( __= to do (Op) )

“I do not refuse to do anything he orders (me to do).” (16)

O amigo [a quem tu querias telefonar __ mas não the friend  to whom you want.pst.2sg phone __ but not

conseguiste __ˉ] chegou. (ˉ__ˉ=ˉtelefonar (a quem)) succeed.pst.2sgˉ __ˉ  arrive.pst.3sg (ˉ__ˉ=ˉto phone (to whom)) “The friend to whom you wanted to phone but did not succeed in doing so arrived.” (17)

Ele ainda não escreveu nenhumˉ livro e a Ana só agora he yet not write.pst.3sg no book and the Ana only now

começou __, mas [pro] está a ficar óptimo. start.pst.3sg __, but pro is to become great “He has not yet written any book and Ana has just started to (write a book), but it is becoming great.”

Cyrino & Matos (2006) impute the internal structure of NCA in Portuguese to the properties of the null constituent and of the correlated pronominal. They propose that the gap in NCA may result from two different categorial sources: a pronominal complement, as in (14), or a sentential projection, as is clearly shown

3.  In (15), the causative mandar “order” should not be confused with its homonymous mandar “send”, a plain transitive verb that selects a DP as direct object as well as an indirect object. Thus, in (15), the relative constituent is not the direct object of mandar but the complement of the omitted verb, fazer “do”, its antecedent being qualquer coisa “anything”. If the relative constituent were the direct object of mandar, the right interpretation would not obtain, since the infinitival verb would not be recovered.



Ellipsis and Restructuring in European Portuguese 

in (18a) by ousar “dare”, a complementation verb that has lost its ability to select a pronominal complement (cf. (18b)):4 (18) a.

Ela tencionava ajudar-me, mas não ousou __. she intend.pst.3sg help-[me.cl], but not dare.pst.3sg “She intended to help me, but she did not dare (to help me).”

b. *Ela nunca o ousou. she never [it.cl] dare.pst.3sg

They provide evidence that in EP the invariable clitic o “it” is a Surface Anaphora, i.e., a proform that must be replaced by the relevant internal structure, through Reconstruction; thus, in the adversative clause in (19), the interpretation of the subject in the passive clause, the reports, as an internal argument of put depends on this Reconstruction: (19)

O livro foi posto na pasta, mas os relatórios the book be.pst.3sg put on.the briefcase, but the reports

não o foram. not [it.cl] be.pst.3pl “The book was put in the briefcase but the reports were not.” (o = put (the reports) in the briefcase)

(EP)

Hence, the internal structure of NCA is predictable: the gap is either a null sentential constituent or the null propositional clitic reconstructed as a sentential domain at the level of semantic interpretation. Yet, in spite of the internal structure of the omitted constituent, NCA in Portuguese, as in Italian and Spanish, is in complementary distribution with Restructuring (cf. (20a) vs. (20b)): (20) a.

Ela não as podia ver, e ele não queria __. she not [them.cl] could see and he not want.pst.3sg “She could not see them and he did not want to (see them).”

b. *Ela não as podia ver e ele não as queria __ she not [them.cl] could.3sg see and he not [them.cl] want.pst.sg

The fact that NCA in Spanish, Italian and Portuguese behaves alike with respect to Restructuring, but differs concerning the structure of the omitted constituent,

4.  Notice that this verb is a full predicate, since it does not present the properties that characterise auxiliary verbs: It keeps its thematic greed and its infinitival complement behaves as an autonomous sentence, since it accepts sentence negation (cf. Ele ousou não os ler “he dared not to read them”) and does not allow Clitic Climbing (cf. *Ele não os ousou ler “he not [them. CL] dared read”).

 Anabela Gonçalves & Gabriela Matos

indicates that its lack of internal structure is not the relevant factor for the complementary distribution between NCA and Restructuring. Assuming Cyrino & Matos’s (2006) proposal for NCA in EP, we will extend their analysis in two ways: We will show that at least as far as EP is concerned the distribution of NCA can be deduced from the selectional properties of the verbs and that the same kind of properties will be crucial to explain the complementary distribution between NCA and Restructuring, a fact not explored by these authors. Thus, in the remainder of this paper we will mainly concentrate on answering the following questions: Which are the relevant selectional properties of the NCA licensing verbs that predict the occurrence of this construction? How do they explain the complementary distribution between NCA and Restructuring?

3.  The selectional properties of NCA licensing verbs 3.1  The lexical content of NCA licensors Following Bosque (1984), Brucart (1999) claims that, from a semantic point of view, NCA in Spanish tends to occur with predicates that may be included in the following meaning classes: modal and aspectual verbs, predicates that express predisposition or purpose, causatives of permission or influence, and verbs that express psychological attitudes or states of the subject. In EP, NCA occurs with the same kind of predicates (see Table 1), as illustrated in (21)–(26): Table 1.  Predicates accepting NCA in EP Modals

Aspectuals

Verbs of predisposition or purpose

Causatives and influence verbs

Verbs denoting psychological states

Verbs denoting psychological attitudes

dever “shall/ should”, necessitar (de) /precisar(de) “need”, parecer “seem”, poder “can”, querer “want”

acabar (de) “finish”, começar (a) “begin/start”, conseguir “manage”, continuar (a) “continue/ keep on”, costumar “use to”, parar (de) “stop”, voltar (a) “do again”

disponibilizarse (a) / oferecer-se (para) / prontificar-se (a) “volunteer”

deixar “let”, mandar “order”, dissuadir (de) “dissuade”, incitar (a) “instigate”, insistir (em) “insist”

adorar “adore”, apreciar “enjoy”, desejar “wish”, detestar “hate”, duvidar (de) “doubt”, esquecer-se (de) “forget”, gostar (de) “like”, lembrar-se (de) “remember”

aceitar “accept”, aprovar “approve”, compreender “understand”, concordar (em) “agree on”, lamentar “regret”, ousar “dare”, persistir (em) “persist”, queixar-se (de) “complain”, recusar/ recusarse (a)/negar-se (a) “refuse”, tentar “try”



Ellipsis and Restructuring in European Portuguese 

(21) Ele trabalhava muito, mas não {precisava /devia} __. he work.pst.3sg much, but not need.pst.3sg /should.3sg “He worked hard, but he {did not need to/should not}.” (22) Ela está a estudar, e vai {continuar/parar} __ amanhã. she be.prs.3sg to study, and will  continue/stop tomorrow “She is studying and she will {continue/stop} (studying) tomorrow.” (23) Como ninguém tinha feito o jantar, eu ofereci-me __ since nobody had made the dinner, I offer.pst.1sg-myself “Since nobody had cooked dinner, I volunteered.” (24) Ele queria ver TV, mas eu {não deixei/dissuadi-o} __. he want.pst.3sg watch TV, but I  not allow/dissuade.pst.1sg-him “He wanted to watch TV, but I {did not allow him /dissuaded him} to do it.” (25) Ele adora viajar, mas ela {odeia/não gosta} __. he adore.prs.3sg travel, but she  hate/not like.prs.3sg “He loves to travel, but she {hates it/does not like it}.” (26) Ele come demasiado, e eu {lamentoˉ/ˉnãoˉaprovo} __. He eat.prs.3pl too.much, and I regret/not.approve.prs.1sg “He eats too much, and I {regret/do not approve}.”

However, the inclusion in one of these classes, by itself, does not constitute a decisive factor for NCA licensing. As we pointed out in Section 1, predicates with close content vary in their ability to accept this construction. This is what happens with the necessity modals dever “should” vs. ter (de Vinf) “have to” (see (21) vs. (27a)); with the durative aspectuals continuar (a) “continue/keep on” vs. estar (a Vinf) “be Ving”(cf. (22) vs. (27b)), or with the inceptive aspectual começar (a) “begin/start” vs. near future auxiliary verb, ir (Vinf) “go=will” (cf. (28)): (27) a. *Ele trabalhava muito, mas não tinha __. he work.pst.3sg much, but not should.3sg b. *Ela já começou a estudar, e vai estar__amanhã she already begin.pst.3sg to study, and will be   tomorrow

também. too

(28) a.

Ele está a aprender Sueco e ela começa  __ amanhã he be.prs.3sg to learn Swedish and she start.prs.3sgˉtomorrow “He is learning Swedish and she is starting (to learn Swedish) tomorrow.”

b. *Ele quer aprender Sueco e vai __! he want.prs.3sg learn Swedish and go.prs.3sg

We take these contrasts as evidence that NCA is licensed by verbs with intrinsic lexical content. This is the case with dever, continuar, or começar, whose

 Anabela Gonçalves & Gabriela Matos

modal or aspectual meaning is supplied by the verb stem, but not of ter (de Vinf ), estar (a Vinf ) or ir (Vinf).5 In fact, the modal and aspectual values of the latter kind of verbs are compositionally obtained. For instance, the aspectual estar (a Vinf ) exhibits a durative value only when it combines with the particle a plus the infinitival complement (29a); in other contexts it behaves like a copular verb (29b). In contrast, continuar keeps its durative value, independently of the nature of its complement (30): (29) a.

Ele está a ler o livro. he be.prs.3sg to read the book “He is reading the book.”

b. Ele está preocupado. he be.prs.3sg worried “He is worried.” (30) a.

Ele continua a ler o livro. he continue.prs.3sg to read the book “He continues reading the book.”

b. Ele continua preocupado. / Ele continuouˉ o trabalho. he continue.prs.3sg worried / he continue.pst.3sg the work “He is still worried.” “He continued the work.”

In the same way, the modal value of ter (de Vinf ) derives from the combination of the verb with the particle de plus the infinitival complement and does not occur in other selection contexts of ter “have”, while dever keeps its modal value. Likewise, the temporal value of ir arises when it selects an infinitival complement; in other contexts, ir “go” is a motion verb. In sum, the data analysed lead us to the following generalisation: (31) NCA licensors must exhibit inherent lexical content.6

This predicts that most auxiliary verbs exclude NCA. As we will show, not only meaning, but also syntactic properties determine this exclusion.

5.  The fact that this construction in English is licensed by full predicates and not by auxiliary verbs (in opposition to VP ellipsis) seems to corroborate this claim. 6.  Still, a residual problem remains: explaining how some verbs with lexical content do not license NCA (e.g., pensar “think”, see (4)). Tentatively, we raise the following hypothesis: verbs presenting an epistemic reading do not allow NCA. An argument in favour of this hypothesis is presented by the modal verbs poder “can” and dever “shall”: they only license NCA when they display deontic modality and exclude this construction when they present an epistemic reading. The same contrast shows up in Spanish (López 1994:350; Depiante 2001:201) and Italian (Depiante 2001:201). We will explore the Portuguese data in future work.



Ellipsis and Restructuring in European Portuguese 

3.2  The categorial nature of the gap in NCA constructions On the basis of examples like (32), Brucart (1999) suggests that in Spanish, the NCA licensing verbs cannot co-occur with the propositional clitic lo “it”: (32) a.

Vendrá si puede __. (__ = venir) come.fut.3sg if can.prs.3sg __ (__= to.come) “He/She will come if he/she can.” (Brucart 1999:2839)

b. *Su mujer aprendió el ruso, pero él no consiguió __. his wife learn.pst.3sg the Russian but he not manage.pst.3sg  (id:2838)

In fact, poder “can”, in (32a), admits NCA and its infinitival complement can not be replaced by lo (*Vendrá si lo puede.); in turn, conseguir “manage”, in (33b), bans NCA and its complement can be pronominalised (Su mujer aprendió ruso, pero él no lo consiguió.). He correlates this impossibility with the fact that NCA licensors generally select a sentential complement introduced by a preposition (see the contrast between olvidarse de “forget about” and olvidar “forget” in (33)). (33) a.

Habían quedado en ir a comer pero *(lo) olvidó. have stay.pst.part on go for eat but   it forget.pst “They had agreed to go out for a meal but he forgot it.”

b. Habían quedado en ir a comer pero se (*lo) olvidó. have stay.pst.part on go for eat but himself   it forget.pst “They had agreed to go out for a meal but he forgot (about it).”

Yet, these selectional properties do not account for the distribution of NCA in Portuguese. Cyrino & Matos (2006) show that, in this language, the NCA licensors select three different kinds of complements: direct object CPs (34a), (34b); oblique CP complements (35a), (35b); and TP complements (36). If a complement is or includes a CP, it occurs in free variation with o/isso (34c), (35c); but TP complements may not be replaced by a pronominal (36b): (34) a.

Essa lei aumentará o desemprego e os that law increase.fut.3sg the unemployment and the



sindicatos não aceitam __. unions not accept.prs.3pl



“That law will increase unemployment and the unions do not accept (it).”

b. Essa lei aumentará o desemprego e os sindicatos that law increase.fut.3sg the unemployment and the unions

não aceitam que ela aumente o desemprego. not accept.prs.3pl that it increase.prs.3sgˉ the unemployment.

“That law will increase unemployment and the unions do not accept that it increases unemployment.”

 Anabela Gonçalves & Gabriela Matos

c.

Essa lei aumentará o desemprego e os that law increase.fut.3sg the unemployment and the



sindicatos não {o aceitam / aceitam isso} unions not [it.cl] accept.prs.3pl / accept.prs.3pl it.



“That law will increase unemployment and the unions do not accept it.”

(35) a.

Os alunos não vão frequentemente à biblioteca, the students not go.prs.3pl often to.the library ,



mas precisavam __. but need.pst.3pl



“The students do not often go to the library, but they need to.”

b. Os alunos não vão frequentemente à biblioteca, the students not go.prs.3pl often to.the library,

mas precisavam de ir frequentemente à biblioteca but need.pst.3pl of go often to.the libray

“The students do not often go to the library, but they need to go to the library often.” c.

Os alunos não vão frequentemente à biblioteca, the students not go.prs.3pl often to.the library



mas precisavam disso. but need.pst.3pl of.that



“The students do not often go to the library, but they need to.”

(36) a.

Eles não gostam de ler, mas deviam __. they not like.prs.3pl of read, but should.3pl “They do not like reading, but they should.”

b. *Eles não gostam de ler, mas deviam {isso/(n)o} they not like.prs.3pl of read, but should.3pl {that/it}

Still, this description does not provide an answer to our major question: Which are the selectional properties of NCA licensors? Since the data show that these licensors in EP have distinct subcategorization properties (CP, Prep+CP and TP complements), we, apparently, have to maintain that NCA is a lexically-driven construction, not clearly determined by any specific lexical property. However, all the above mentioned verbs have a common subcategorization property: in all their complements at least TP projects. In other words, in NCA the omitted constituent must correspond to a proposition, not to a single predicate. Therefore, we propose the following generalisation: (37) NCA licensors select a complement where T projects.



Ellipsis and Restructuring in European Portuguese 

This would account for the fact that the NCA gap is always interpreted as a T domain. If we adopt Cyrino & Matos’s (2006) proposal that the null pronominal in NCA is reconstructed as a CP, all the instances of the omitted constituent in this construction count as complete T domains, at the level of semantic interpretation. The fact that verbs that select vP or AspP, such as Perfect Tense and Passive auxiliaries, do not license NCA, as in (38), confirms the adequacy of the generalisation in (37): (38) a. *Ele não costumava ir ao teatro, mas agora tem __. he not use.pst.3sg go to.the theatre, but now has b. *Ela nunca repreende os filhos, mas hoje eles she never blame.prs.3sg the children, but today they

foram be.prs.3sg

__.

In sum, the distribution of NCA is not entirely unpredictable: it occurs with verbs with inherent lexical content which select sentential complements with T.

4.  NCA and Restructuring Although the generalisations in (31) and (37) accommodate the data analysed in Section 3, a problem remains unsolved: Why do Restructuring verbs license NCA only when Restructuring does not operate (cf. the contrast in (39))? (39)

Ela cumprimentou-os, porque {queria / *os she greet.pst.3sg-[them.cl], because want.pst.3sg / [them.cl]

queria } __. want.pst.3sg “She greeted them, because she wanted (to greet them).”

In the next subsections, we will briefly examine Restructuring structures in EP (Section 4.1.), and the interaction of this construction with NCA (Section 4.2.). 4.1  Restructuring in EP Since Rizzi (1978, 1982), Restructuring is a term that applies to structures where an infinitival complement does not count as a boundary to such phenomena as Clitic Climbing (40a) and Long Object Movement (40b), which suggests that the verbs involved form a complex predicate and that there is a single sentential

 Anabela Gonçalves & Gabriela Matos

domain. Still, restructured sentences have full sentential correlates clearly exhibiting a bi-sentential structure (cf. (41)):7 (40) a.

Todos os jornalistas o quiseram entrevistar. all the journalists [him.cl] want.pst.3pl interview “All journalists wanted to interview him.”

b. Começaram-se a construir as novas casas. begin.pst.3pl-[se.pass.cl] to build the new houses “The new houses began to be built.” (41) a.

Todos os jornalistas quiseram entrevistá-lo. all the journalists want.pst.3pl interview-[him.cl] “All journalists wanted to interview him.”

b. Começou-se a construir as novas casas. begin.pst.3sg-[se.nom.cl] to build the new houses “Someone began building the new houses.”

The transparency effects of these infinitival complements were captured in the late seventies by Rizzi’s Restructuring rule, which merges the infinitival and the embedding sentences into a single sentential domain. Within the Principles and Parameters Theory, alternative explanations have been devised, relying on the categorial nature or on the defective status of the complement of the verb (cf. Baker 1988; Bok-Bennema & Kampers-Manhe 1994; Gonçalves 1999; Wurmbrand 2001, 2004, 2006; Cinque 2004, 2006; Bok-Bennema 2006). According to Wurmbrand (2001, 2004, 2006) and Cinque (2004, 2006), T does not project in Restructuring infinitival complements.8 This hypothesis apparently captures the fact that the embedded infinitive temporally depends on the tense specifications of the higher domain (Stowell 1982; Raposo 1987; Ambar 1992; Gonçalves 1999), which would account for the marginality of (42), where two antonymous time adverbials co-occur within the Restructuring domain, one affecting the finite domain and the other the infinitival complement. Yet, this hypothesis fails to explain (43), where Clitic Climbing (hence, Restructuring) has applied, despite the fact that the infinitival domain exhibits its own temporal features, as indicated by the scope of the first instance of amanhã “tomorrow”: 7.  Rizzi (1982) also presents Auxiliary Change in Italian as evidence for Restructuring. Yet, in EP this phenomenon does not occur: Composed tenses always use the auxiliary verb ter “have”. 8.  The main difference between Wurmbrand’s and Cinque’s proposals is that the former argues that Restructuring verbs may be lexical or functional, thus heading lexical or functional projections, whereas the latter claims that these verbs are always functional and head functional projections of a single sentential domain. In EP, Restructuring verbs may be lexical verbs. This is why we do not classify Restructuring as a case of the Serial Verb Construction, as proposed by Aboh (2009).



Ellipsis and Restructuring in European Portuguese 

(42) *Eles, ontem, quiseram entrevistar esse candidato hoje. they yesterday want.pst.3pl interview that candidate today (43) Hoje ele não te quer [ver amanhã]. today he not [you.cl] want.prs.3sg see tomorrow “Today he does not want to see you tomorrow.”

Example (43) shows that, from a semantic point of view, two distinct situations are involved in the Restructuring construction: one is denoted by the embedding clause and the other by the infinitival complement.9 So, in a sentence like A Maria não os queria ver “Mary did not want to see them”, two time intervals can be recognized: the one where the situation denoted by want and its arguments is located and the one where see and its arguments takes place. Taking into account that tense heads are not only syntactically but also semantically relevant,10 in the sense that they are the locus of the tense/event structure (Chomsky 2001:9), we propose that, in EP, T projects even when Restructuring effects are visible.11 However, based on Gonçalves (1999), we consider that the following properties must be met: (44) T in Restructuring infinitival complements (i) T has non-valued uninterpretable T-features (Pesetsky & Torrego 2001, 2004; Duarte et al. 2005; Ambar 2007), whose value depends on the value of the matrix T; (ii) T is syntactically inactive, so it cannot delete the uninterpretable features of the goal, since for Agree to occur both the goal and the probe must be active (Chomsky 2001:6).12

This proposal correctly captures the fact that Restructuring may apply when the matrix verb sets the temporal location of the embedded event as posterior to (45a) or simultaneous with (45b) the matrix one (Stowell 2004), but it does not apply when the embedded infinitival sentence has a free temporal location w.r.t. the matrix: overlapping (46), anteriority (47) or posteriority (48).

9.  This is one crucial difference between Restructuring and auxiliary verbs, which are partly identical concerning the clitic position. 10.  For the semantic relevance of T, see Zagona (1990), Stowell (1996), Pesetsky & Torrego (2004). 11.  Bok-Bennema (2006) also shows that Restructuring involves two sentential domains in the initial step of the derivation, presenting arguments against the VP nature of the infinitival complement. 12.  The way the subordinate external argument gets Case seems to be unproblematic, if we assume that obligatory Control is raising to the matrix [Spec,T], as in Hornstein (1999).

 Anabela Gonçalves & Gabriela Matos

(45) a.

O João não os quis fazer. the João not [them.cl] want.pst.3sg do “John did not want to do them.”

b. O professor não os estava a corrigir. the teacher not [them.cl] be.pst.3sg to correct “The teacher was not correcting them.” (46) Ele não afirmou conhecê-la. vs. *Ele não a afirmou conhecer. he not claimed know-[her.cl]  He not [her.cl] claimed know “He did not claim that he knew her.” (47)

Ele não afirmou tê-la visto vs. *Ele não a afirmou he not claimed have-[her.cl] seen   he not [her.cl] claimed



ter visto. have seen

“He did not claim that he had seen her.” (48) Ele não afirmou ir vê-la vs. *Ele não a afirmou ir ver. he not claimed go see-[her.cl]   he not [her.cl] claimed go see “He did not claim that he would see her.”

In our proposal, temporal dependence means that the situation expressed in the embedded clause is temporally interpreted w.r.t. the matrix event time, not to the utterance time. This explains the contrasts in (42)–(43) and in (49). Although in both cases of (49) the embedded domain is interpreted as a future w.r.t. the matrix, only in (49b) is this future related to the event time. In (49a), the adverbial amanhã “tomorrow” denotes a future w.r.t. the utterance time, thus the sentence is ungrammatical. (49) a. *Ele quis entrevistar o candidato amanhã. he want.pst.3sg interview the candidate tomorrow b. Ele quis entrevistar o candidato no dia seguinte. he want.pst.3sg interview the candidate in.the day after “He wanted to interview the candidate the day after.”

For the same reason, the example in (43), above, is grammatical, although the two time adverbials (hoje “today” and amanhã “tomorrow”) are antonymous: In (43) the matrix event time and the utterance time overlap, so the adverbial amanhã “tomorrow” expresses a future w.r.t. the matrix event time. Still, there is a subset of verbs (such as prometer “promise”) that also have dependent T in the embedded domain, but do not allow Restructuring (50). The contrasts in (50) find an explanation in (44ii): prometer “promise” only takes an embedded CP/TP with active T. In opposition, Restructuring verbs (e.g., querer



Ellipsis and Restructuring in European Portuguese 

“want”) take infinitival complements headed by active T or inactive T (see the examples without and with Restructuring, respectively, in (51) and (52)): (50)

Ele não prometeu vê-la vs. *Ele não a prometeu ver. he not promised see-[her.cl]   he not [her.cl] promised see

(51) a.

O João não quis comprá-lo. the João not want.pst.3sg buy-[it.cl] “John did not want to buy it.”

b. Quis-se votar várias propostas. want.pst.3sg-[se.nom.cl] vote several proposals “One wanted to vote on several proposals.” c.

O Pedro só quis não o magoar. the Pedro only want.past.3sg not [him.cl] hurt “Peter just wanted not to hurt him.”

(52) a.

O João ainda o quis comprar. the João still [it.cl] want.past.3sg buy “John still wanted to buy it.”

b.ˉ Quiseram-se votar várias propostas. want.pst.3pl-[pass.cl] vote several proposals “They wanted to vote on several proposals.” c. *O Pedro só o quis não magoar. the Pedro only [him.cl] want.pst.3sg not hurt “Peter just wanted not to hurt him.”

In sum, the properties of the embedded T have consequences for the licensing of the elements within the embedded domain: When the embedded T is active, it qualifies as a probe for checking the φ-features of clitics, forbidding Clitic Climbing (51a). Since it is a complete domain, accusative Case is checked by TO (Pesetsky & Torrego 2004), excluding Long Object Movement (51b), and the embedded negation is licensed as in Zanuttini (1996) and Matos (2001) (51c). In contrast, when the embedded T is defective, the expected Restructuring effects show up – see the examples in (52): Given the defectiveness of the embedded T, Agree holds between the matrix complete T and the embedded T, under c-command, and the temporal features of the latter are set by those of the former. Then, a single tensed phase obtains and the complex predicate is formed. Assuming that Restructuring is actually the formation of a T-chain, Clitic Climbing and exclusion of the embedded negation (often in correlation with Neg Raising effects) are correctly predicted: They have to be licensed by the head of the chain – the matrix T – which is active. As far as Long Object Movement is concerned, we hypothesise that the well-known effects of the passive clitic (inhibition of the external theta role and of the accusative Case) also affect the embedded

 Anabela Gonçalves & Gabriela Matos

domain, so TO is absent from this domain, thus preventing the DP object of the embedded verb to check accusative case. Notice that neither Clitic Climbing nor Long Object Movement violate the Minimal Link Condition, because Move is internal to the same phase. The intervening T is not active, so it is not an appropriate probe, that is, it cannot value the features of the moved elements. 4.2  The complementary distribution of NCA and Restructuring Given that the NCA gap is always interpreted as a T domain (i.e., NCA licensors select at least a TP complement (Section 3.2)) and Restructuring verbs select a TP even when Restructuring effects show up, how can we account for the marginality of (53)? (53) *Ela cumprimentou-os porque os queria __. she greet.pst.3sg-[them.cl] because [them.cl] want.pst.3sg

We claim that the explanation relies on the properties of T. Since, for Restructuring to apply, the embedded infinitival complement must have inactive T, the complementary distribution of these constructions leads us to propose the generalisation in (54), which reformulates (37): (54) NCA licensors must select a sentential complement with active T.

If a verb is marked in the Lexicon as selecting an active TP, Restructuring is forbidden (55) and NCA is possible (56). In turn, when the verb selects a TP with inactive T, it bans NCA (53), but licenses this construction when Restructuring effects are not visible, i.e., when it selects a TP with active T (57). (55) Elas {odeiam visitá-lo /*odeiam-no visitar [-]}. they hate.prs.3pl visit-[him.cl] /hate.prs.3pl-[him.cl] visit “They hate to visit him.” (56)

Não gosto muito de sair à noite, mas não odeio __. not like.prs.1sg much of go.out at.the night, but not hate.prs.1sg “I do not like much to go out at night, but I do not hate (to go out at night).”

(57) Ela cumprimentou-os, porque queria __. she greet.pst.3sg-[them.cl] because want.pst.3sg “She greeted them, because she wanted to (greet them).”

In sum, the analysis developed in this paper shows that NCA is not randomly associated with verbs and explains the complementary distribution between NCA and Restructuring: The more defective the sentence domain is, the weaker the chance of NCA to appear, and the stronger the chance of Restructuring to occur.



Ellipsis and Restructuring in European Portuguese 

Table 2 below illustrates the complementary distribution between Restructuring and NCA licensing verbs in EP. Table 2.  Restructuring (REST)/Null Complement Anaphora (NCA) licensing verbs

Verbs selecting active T

REST

NCA

with lexical CP/TP content

No

Yes

Prep CP/TP

No

Yes

No

No

Yes

No

without TP lexical content Verbs selecting defective T

VERBS aprovar “approve”, costumar “use to”, decidir “decide”, deixar “let”, desejar “wish”, dever “should”, lamentar “regret”, mandar “order”, odiar “hate”, parecer “seem”, poder “can”, prometer “promise”, querer “want”, tentar “try” começar “begin”, continuar “continue/ keep on”, dissuadir “dissuade”, duvidar “doubt”, esquecer-se “forget”, gostar “like”, incitar “instigate”, insistir “insist”, lembrar-se “remember”, necessitar “need”, persistir “persist”, recusar-se / negar-se “refuse” estar a “be Ving”, haver de/ter de “have to”, ir “going to” começar a “begin”, conseguir “manage”, continuar a “continue/ keep on”, costumar “use to”, disejar “wish”, dever “should”, estar a “be Ving”, haver de/ter de “have to”, ir “going to”, poder “can”, querer “want”, tentar “try”

5.  Conclusions In this paper we have argued that NCA is not a property randomly associated with some verbs. Rather, it occurs under semantic and formal conditions: The licensing verb must have intrinsic lexical content and select an active T. We have also claimed that Restructuring effects occur when the embedded T is defective and, hence, the matrix T and the embedded T form a single tensed phase. This analysis explains why NCA and Restructuring are in complementary distribution: This is not because NCA is a null proform with no internal structure (Depiante 2001), but rather due to the properties of the infinitival constituent, that is, the more syntactically defective this domain is, the less NCA occurs and the more Restructuring effects appear. The fact that in Portuguese NCA exhibits internal structure (Cyrino & Matos 2006) is compatible with its complementary distribution with Restructuring: Restructuring effects only occur with a syntactically inactive TP, while NCA

 Anabela Gonçalves & Gabriela Matos

corresponds either to active TP complements or to propositional DP proforms that reconstruct as CPs.

References Aboh, Enoch. 2009. “Clause Structure and Verb Series”. Linguistic Inquiry 40.1. 1–33. Ambar, Manuela. 1992. “Temps et Structure de la Phrase en Portugais”. Structure de la Phrase et Théorie du Liage ed. by Hans-Georg Obenhauer & Anne Zribi-Hertz, 29–49. Saint-Denis: PUV. Ambar, Manuela. 2007. “Verb Movement and Tense – EPP and T-Completeness”. Proceedings of the XXXII Incontro di Grammatica Generativa ed. by M. Cecilia Picchi & Alan Pona, 1–20. Alessandria: Edizioni dell’Orso. Baker, Mark. 1988. Incorporation. Chicago: UCP. Bok-Bennema, Reineke. 2006. “Clitic Climbing”. The Blackwell Companion to Syntax ed. by Martin Everaert & Henk van Riemsdijk, vol. V, 229–343. Oxford: Blackwell. Bok-Bennema, Reineke & Brigitte Kampers-Manhe. 1994. “Transparency Effects in the Romance Languages”. Issues and Theory in Romance Linguistics: Selected Papers from the LSRL XXIII ed. by Michael Mazzola, 199–217. Washington: Georgetown University Press. Bosque, Ignacio. 1984. “Negación y elipsis”. Estudios de Lingüística 2. 171–199. Brucart, José M. 1999. “La Elipsis”. Gramática Descriptiva de la Lengua Española, ed. by Ignacio Bosque & Violeta Demonte, vol. 2, 2787–2863. Madrid: Espasa. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale. A Life in Language ed. by Michael Kenstowicz, 1–52. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 2004. “Restructuring and functional structure”. Structures and Beyond. The Cartography of Syntactic Structures ed. by Adriana Belletti, vol. 3, 132–191. Oxford: OUP. Cinque, Guglielmo. 2006. Restructuring and Functional Heads. The Cartography of Syntactic Structures, vol. 4. Oxford: OUP. Cyrino, Sonia & Gabriela Matos. 2006. “Null Complement Anaphora in Romance: Deep Or Surface Anaphora?”. Romance Languages and Linguistic Theory 2004 ed. by Jenny Doetjes & Paz González, 95–120. Amsterdam: John Benjamins. Depiante, Marcela. 2001. “On Null Complement Anaphora in Spanish and Italian”. Probus 13. 193–221. Duarte, Inês, Anabela Gonçalves & Matilde Miguel. 2005. “Propriedades de C em frases completivas”. Actas do XX Encontro Nacional da Associação Portuguesa de Linguística ed. by Inês Duarte & Isabel Leiria, 549–562. Lisboa: Colibri. Gonçalves, Anabela. 1999. Predicados Complexos Verbais em Contextos de Infinitivo não Preposicionado do Português Europeu. Doctoral dissertation, Universidade de Lisboa. Hankamer, Jorge & Ivan Sag. 1976. “Deep and Surface Anaphora”. Liguistic Inquiry 7. 391–426. Hornstein, Norbert. 1999. “Movement and Control”. Linguistic Inquiry 30. 69–96. López, Luis. 1994. “The Syntactic Licensing of VP-Ellipsis: A Comparative Study of Spanish and English”. Issues and Theory in Romance Linguistics. Selected Papers form the LSRL XXIII ed. by Michael Mazzola, 333–354. Washington: Georgetown University Press. Matos, Gabriela. 2001. “Negative Concord and the Minimalist Approach”. Romance Languages and Linguistic Theory 1999 ed. by Yves D’Hulst, Johan Rooryck & Jan Schroten, 245–280. Amsterdam: John Benjamins.



Ellipsis and Restructuring in European Portuguese 

Pesetsky, David & Esther Torrego. 2001. “T-to-C Movement: Causes and Consequences”. Ken Hale: A Life in Language ed. by Michael Kenstowicz, 355–426. Cambridge, Mass: MIT Press. Pesetsky, David & Esther Torrego. 2004. “Tense, Case and Syntactic Categories”. The Syntax of Time ed. by J. Guéron & J. Lecarme, 495–537. Cambridge, Mass: MIT Press. Raposo, Eduardo. 1987. “Case Theory and Infl-to-Comp: the Inflected Infinitive in European Portuguese”. Linguistic Inquiry 18. 85–109. Rizzi, Luigi. 1978. “A Restructuring Rule in Italian Syntax”. Recent Transformational Studies in European Languages ed. by Samuel Jay Keyser, 113–158. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. Stowell, Tim. 1982. “The Tense of Infinitives”. Linguistic Inquiry 13. 561–570. Stowell, Tim. 1996. “The Phrase Structure of Tense”. Phrase Structure and the Lexicon ed. by Johan Rooryck & Laurie Zaring, 277–291. Dordrecht: Kluwer. Stowell, Tim. 2004. “Tense and Modals”. The Syntax of Time ed. by Jacqueline Guéron & Jacqueline Lecarme, 621–635. Cambridge, Mass.: MIT Press. Wurmbrand, Susanne. 2001. Infinitives. Restructuring and Clause Structure. Berlin: Mouton de Gruyter. Wurmbrand, Susanne. 2004. “Two Types of Restructuring – Lexical vs. Functional”. Lingua 114. 991–1014. Wurmbrand, Susanne. 2006. “Verb Clusters, Verb Raising and Restructuring”. The Blackwell Companion to Syntax ed. by Martin Everaert & Henk van Riemsdijk, vol. V, 229–343. Oxford: Blackwell. Zagona, Karen. 1990. “Times as temporal argument structure”. Ms., University of Washington. Zanuttini, R. 1996. “On the Relevance of Tense for Sentential Negation”. Parameters and Functional Heads ed. by Adriana Belletti & Luigi Rizzi, 181–207. Oxford: Oxford University Press.

The early steps of modal and negation interactions Evidence from child Italian* Andrea Gualmini1 & Vincenzo Moscati2 1Utrecht

University/2University of Siena

Children’s interpretation of scopally ambiguous sentences has been the subject of much recent research. In this paper, we address this issue by looking at children’s interpretation of sentences containing an ambiguity between different scope-bearing elements, namely negation and a modal verb in Italian. We present experimental results suggesting that Italian-speaking children differ from Italian-speaking adults in that they choose strong inverse scope interpretations rather than weak surface scope interpretations.

1.  Introduction One difference between natural and formal languages is that natural languages are ambiguous in many respects, one of them being the ambiguity of sentences with more than one logical operator. In this paper we focus on sentences containing negation and modal operators and their interpretation by 4- and 5-year olds. In many cases, a sentence with multiple operators is ambiguous and grammars of human languages need formal means to express a one-to-many mapping from a sentence to its different possible meanings. A widely shared assumption is that such a mapping can be generated by the application of non-visible movement operations, which make available interpretations different from the ones that can be read off the surface syntax (see Fox 2000; May 1985; Reinhart 1997, 2006). To illustrate, a sentence like (1):

(1) Every horse didn’t jump over the fence

might be used to describe two different states of the world. According to one interpretation, (1) describes a situation in which none of the horses of a given set were *Andrea Gualmini is currently supported by a VIDI fellowship from the Netherlands Organization for Scientific Research (NWO) and Utrecht University.

 Andrea Gualmini & Vincenzo Moscati

able to cross an obstacle, but under a different interpretation the sentence in (1) is true also if some horses jumped but others did not. The availability of these two interpretations might be explained by assuming two logical representations correspond to a single surface structure. Whatever theoretical solution we adopt to derive the multiple interpretations of scopally ambiguous sentences, what is important is that the hearer of a sentence like (1) must be able to map this sequence of words onto two distinct logical forms. One research question which has been addressed in the last ten years concerns the learnability of such mapping between a sentence and its different possible meanings. The first results came from Musolino (1998). In a series of experiments, he investigated English-speaking children’s interpretation of sentences like (1) and other sentences which contain more than one logical operator, such as (2) and (3):

(2) The detective didn’t find some guys



(3) The detective didn’t find two guys

Each of the sentences in (1) through (3) contains two operators: negation and a quantified noun phrase. This yields two logically possible scope assignments for each sentence.1 To illustrate, as we saw above informally, (1) is ambiguous between the two interpretations listed below. (4) Every horse is such that it did not jump over the fence ∀x [horse (x) → ¬ jump over the fence (x)] ‘for every x, if x is a horse then x did not jump over the fence’ (5) Not every horse jumped over the fence ¬∀x [horse (x) → jump over the fence (x)] ‘it is not the case that for every x, if x is a horse then x jumped over the fence’

The two interpretations of (1) listed in (4) and (5) result from the relative scope assignment to negation and every, as suggested by the order of the operators ¬ and ∀ in the logical formulae. We will follow the semantic literature and refer to the interpretation in (4) as the ‘surface scope’ interpretation of (1), and to the interpretation in (5) as its ‘inverse scope’. This is because the scope bearing elements every and not in (4) are interpreted in the same order with which they appear in the overt syntax, whereas in (5) they are interpreted in the opposite order. The research question that Musolino (1998) and others have addressed is whether young children are capable of accessing both the surface scope and the inverse scope interpretation of sentences containing negation and another scope-bearing element. The experimental evidence collected by Musolino (1998) suggests that this is not the case and, for all the constructions in (1), (2) and (3) 1.  In the case of (2), the surface scope reading is arguably unavailable due to the polarity properties of some (see Ladusaw 1979).



The early steps of modal and negation interactions 

4- and 5-year-old children seemed to consistently resort to their surface scope interpretations. This finding led Musolino to propose the Observation of Isomorphism, the claim that children’s semantic scope coincides with overt syntactic scope (see also Musolino, Crain & Thornton 2000). More recent studies, however, have led to a revision of the picture suggested by Musolino’s early findings. In particular, recent works have shown that children’s interpretation of sentences like (1), (2) and (3) is affected by the context. For instance, Gualmini (2004a) has shown that, in a context in which the Troll is expected to deliver all the relevant pizzas, children easily access the inverse scope interpretation of (6).

(6) The Troll didn’t deliver some pizzas

In the same vein, Gualmini (2004b) demonstrated that children are perfectly capable of assigning wide scope to the indefinite some with sentences such as (7), and Gualmini (2007a) has shown that the same is true for Italian-speaking children.

(7) Every farmer didn’t clean some animal

Furthermore, Gualmini, Hulsey, Hacquard & Fox (2008) have recently shown that the same contextual maneuver discovered by Gualmini (2004a, b) leads children to select the inverse scope interpretation of sentences equivalent to (1) and (3) to a higher extent than observed in previous literature. Finally, Musolino & Lidz (2006) report that children access the inverse scope interpretation of (8) to a larger extent than they do for (1).

(8) Every horse jumped over the log, but every horse didn’t jump over the fence

The finding that children may select different logic representations for the same sentence depending on the context poses a challenge for the original view of Isomorphism and suggests that children have access to covert operations. This has led to a debate on whether surface scope interpretations have a special status in child or adult grammar (see Musolino 2006; Musolino & Lidz 2004; Hulsey, Hacquard, Fox & Gualmini 2004; Gualmini 2007b, 2007c, 2008 and Gualmini, Hulsey, Hacquard & Fox 2008). To sum up, the current debate on children’s interpretation of sentences like (1) through (3) focuses on the resolution of the relevant ambiguity. The reason is that for all of the sentences discussed above, children can generate both readings from the earliest stages of language development. Although this finding has enriched our understanding of children’s parsing abilities, it does not bring us any closer to children’s first hypothesis. In other words, the question remains whether children acquire scope ambiguities step-by-step. In this paper, we address this issue by looking at children’s interpretation of sentences containing an ambiguity between different scope-bearing elements, namely negation and a modal verb.

 Andrea Gualmini & Vincenzo Moscati

2.  Previous studies on children’s interpretation of modals and negation Modal verbs occurring in sentences containing negation give rise to ambiguities similar to the ones found with nominal quantifiers. This parallel has not gone unnoticed in the semantic literature, and modal operators have been reduced to quantification over sets of worlds rather than individuals (Hintikka 1962; Stalnaker 1968; Lewis 1995).2 The existence of scope ambiguities involving modals led Moscati & Gualmini (2008a) to investigate children’s interpretation of sentences like the following:

(9) The lion cannot be in the same cage as the tiger

a. It is not the case that the lion can be in the same cage as the tiger b. It is possible for the lion not to be in the same cage as the tiger

As the morphology suggests, the verbal form cannot can be decomposed into the modal operator can (◊) and negation (¬). Interestingly, these two scope-bearing elements are interpreted in a way that does not match surface syntax. In other words, adults’ preferred – if not their only – interpretation of (9) is the inverse scope interpretation paraphrased in (9a), whereas the interpretation paraphrased in (9b) is largely dis-preferred, if at all available to adults. Let us now consider the Italian sentence in (10): (10)

Il leone non deve stare nella stessa gabbia con la tigre the lion not must be into.the same cage with the tiger

a. It is not the case that the lion must be in the same cage as the tiger b. It is necessary for the lion not to be in the same cage as the tiger

Although both interpretations (10a)and (10b)are possible for sentence (10), if the sentence is uttered with a plain intonation, the inverse scope reading in (10b) is more prominent than the surface scope reading in (10a). Due to previous claims about the privileged status of surface scope interpretations, the English sentence in (9) and the Italian sentence in (10) were used in two experiments conducted by Moscati & Gualmini (2008a). The results of the

2.  It has been long recognized in logic that modality is not a homogeneous category and that different kinds of attitudes toward the truth of a given proposition may be distinguished. Albeit possible to distinguish between different kinds of modality (see Palmer 1986; Cinque 1999), a general opposition between epistemic and deontic modality will be sufficient here. When it comes to child language, children start using modals very early and they have been argued to initially produce modals in their deontic meaning (see Wells 1979). These results are not so surprising, given that children are likely to be exposed in their first years to utterances expressing the permission or the prohibition to do something. For these reasons, the sentences discussed in this paper are all compatible with a deontic interpretation.



The early steps of modal and negation interactions 

experiments suggest that 4-year-old children robustly interpret such sentences as expressing a prohibition. In both cases, this amounts to the inverse scope interpretation of the target sentence. This is an important finding by itself because it adds to the growing body of empirical evidence showing that 4-year-olds do not experience any sort of problems with inverse scope interpretations. Nevertheless, since children seem to behave like adults from very early on, we are still looking for a protracted stage of non-adult behavior that could help us understand how children approach the acquisition of scope ambiguities. A possible reason behind children’s early mastery of the ambiguities investigated by Moscati & Gualmini (2008a) is their frequency in the primary linguistic data. For this reason Moscati & Gualmini (2008b) decided to test a construction which is unambiguously interpreted as a permission not to do something, rather than a prohibition. The study by Moscati & Gualmini (2008b) used the standard Truth Value Judgment Task (see Crain & Thornton 1998; Crain & McKee 1985) to investigate children’s interpretation of sentences such as (11). (11) To be a good farmer, you need not feed the zebra a. To be a good farmer, it is necessary not to feed the zebra b. To be a good farmer, it is not necessary to feed the zebra

Children were asked to evaluate whether the target sentence accurately described the outcome of a story. The story was constructed in such a way that the target sentence (11) was true under its weak inverse scope reading (11b), but false under its strong surface scope reading(11a).3 Fifteen English-speaking children participated in the experiment. They ranged in age from 3:9;5 to 6:2;26 (mean age: 4:10;5). Each child was presented with four target trials interspersed with an equal number of fillers, to balance the number of ‘yes’ and ‘no’ responses. Let us illustrate one of the trials. (12) This is a story about a boy, a pirate and a little girl. The boy is a farmer and wants to know if his friends are good at farming. He asks them to feed some animals to find out who is a good farmer. He has two presents to give to his friends if they are good at farming. The pirate goes first and feeds the pony, but not the zebra (because the zebra doesn't look hungry). Then the little girl goes, and she feeds both the pony and the zebra. The boy awards both the pirate and the little girl a prize for being a good farmer.

At this point, the puppet uttered the target sentence in (11). This sentence is true on its weak inverse scope reading, since it is not necessary to feed the zebra in

3.  When the two interpretations of an ambiguous sentence stand in an entailment relation, it is standard practice to label as ‘weak’ the reading that is entailed by the other, and as ‘strong’ the reading that entails the other.

 Andrea Gualmini & Vincenzo Moscati

order to be a good farmer. If that was the case, then only the little girl would have received a prize. By contrast, the sentence is false on its strong surface scope interpretation. Here are the results: children rejected the statement in (11) 41 times out of 60 trials (68.3%). When children rejected the target sentence, they were asked to explain ‘what really happened.’ Children’s responses suggested that they were accessing the strong surface scope interpretation of the target sentence. For instance, children would point out that the puppet was wrong because the pirate had also received a prize. By contrast, a group of fourteen adult controls accepted the target sentences 87% of the time. To sum up, the experiment conducted by Moscati & Gualmini (2008b) provides us with one instance of non-adult behavior in children’s interpretation of sentences containing negation and a modal operator. The next step is to determine what accounts for children’s non-adult behavior. Assuming that the context used by the experiment was an appropriate context for children to select the adult interpretation of the target sentence, the response uncovered by Moscati & Gualmini (2008b) might indeed be indicative of children’s initial hypothesis. In particular, it is possible that children approach the acquisition of privative ambiguities by initially positing only the strong interpretation (see Crain, Conway & Ni 1994) To corroborate this hypothesis, we conducted an experiment with Italianspeaking children on their interpretation of sentences that express the same proposition targeted by Moscati & Gualmini (2008b). 3. An experimental study on children’s interpretation of sentences containing puo’ and non in Italian In this experiment, we tested Italian-speaking children’s interpretation of sentences that adult speakers of Italian can only interpret on their surface scope interpretation. Consider the following example. (13)

Il contadino può non dare le carote all’ elefante the farmer can not give the carrots to.the.elephant

a. *it is not possible that the farmer gives carrots to the elephant b. it is possible that the farmer doesn’t give carrots to the elephant

In this sentence, only the surface scope reading (13b) is allowed in adult Italian.4 Notice that this is the weak reading, since it is entailed by the other reading in (13a).

4.  It is unclear why the reading (13a) is blocked. One possible explanation is that potere belongs to the class of restructuring verbs, which take a sentential complement and through a restructuring rule (Rizzi 1982) a biclausal structure is transformed into a monoclausal one. The presence of negation might block such a rule (Rizzi 1982; Kayne 1989) making the deletion



The early steps of modal and negation interactions 

The reason is that if it is not possible to give carrots to the elephant, then it is also possible not to give carrots to the elephant. The entailment does not go through in the opposite direction, however. In particular, if we know that it is possible not to give carrots to the elephant, then we cannot conlude that doing so is prohibited, but simply that it is not necessary. 3.1  Materials We conducted a standard Truth Value Judgement task (Crain & Thornton 1998) and we presented children with four sentences like (13), reported in the Appendix, in a context like the following. (14) There is a farmer who has to feed his animals. He has turnips and carrots and there are three animals: one tiger and two elephants. But the tiger says that she doesn’t like carrots. Thus the farmer decides to give a turnip to the tiger. The elephants do not have any preference, and the farmer decides to give a turnip to the first elephant and a carrot to the second one.

We then asked a puppet to describe what had happened in the story and the puppet uttered the target sentence (13). Since there is one elephant who did not receive any carrots, the weak surface scope reading makes the sentence true. By contrast, the strong inverse scope reading (13a) is false since one elephant did indeed receive some carrots and no prohibition exists in the context. Adult speakers of Italian were expected to select the latter interpretation, arguably the only interpretation licensed by their grammar, while we expected children’s responses to be dictated by the strong inverse scope interpretation, in accordance with the results documented by Moscati & Gualmini (2008b). In addition to the target sentences, children were also presented with four control items. The control items were of two kinds. In the first type of controls (Type A) we gave children affirmative sentences containing the modal potere after a brief story: (15) Semola knows that there is a king who has objects with magic properties and he wants to visit the king to have the magic objects. When he arrives, the king says that he has a harp and two swords: the harp makes one able to run very fast, but only if you can play it, while the swords give the power to fly if someone can lift them and hold them. First, Semola tries with the harp, he plays it and he is now able to run very fast. Then he gives it a try with the swords. He tries the first sword, but it is too heavy and he fails. He tries with the second one, and this time he manages to lift it and he flies away.

of the clause boundary impossible. The negative operator is than unable to cross a CP border (Moscati 2006, 2007) and its scope is bounded to the lower clause. Inverse scope reading of negation over the modal will then be ruled out.

 Andrea Gualmini & Vincenzo Moscati

At this point, a puppet is asked whether Semola can lift the sword and the puppet utters the following sentence: (16)

Semola può sollevare la spada Semola can lift the sword

If children are cooperative, they should assume that the relevant sword is the sword that was actually lifted and they should accept the target sentence. Let us illustrate the reason behind this type of control sentences. As the reader may have noticed, the target sentence in (13) contains a singular definite determiner in order to avoid the insertion of another scope-bearing element such as an indefinite. This choice may be a source of confusion and children’s rejection of the target sentence might be due to the inappropriateness of the definite determiner. If this is indeed the case, then children’s confusion should also surface in response to sentences like (16). In the second control (Type B), we gave children sentences in which a negative operator was also present, preceding the modal potere. Here is an illustrative trial. (17) There is a Pilot who wants to go out and he has a motorbike and an F1 car. He decides to take the motorbike, but when he tries to climb onto it, he falls down since it is too high. He tries again, but he falls again. Thus, he decides to take the car and goes away.

At the end of the story, the puppet was asked whether the pilot could drive the motorbike. The puppet then answered as follows: (18)

Il pilota non può guidare la moto the pilot not can drive the motorbike

a. it is not possible that the pilot drives the motorbike b. *it is possible that the pilot does not drive the motorbike

This type of control sentences were inserted to ensure that children are able to process the negative operator with modal verbs and to test if they have problems in accessing strong surface scope readings. 3.2  Participants Twenty monolingual Italian-speaking children (age: 3;9 – 5;7. mean age: 4;5) from two kindergartens in the Siena area participated in the experiment. 3.3  Results Let us first look at the results of the control sentences. In the Type A controls, children did not show any problems with declarative modal sentences, nor with



The early steps of modal and negation interactions 

the use of the singular definite article in the experimental context. The results are illustrated in the table below: Table 1.  Children’s acceptance of Type A control sentences Condition

Type A control

Acceptance

Trial 1

Trial 2

Total

17/20

16/20

33/40 (82,5%)

The results from the Type B control condition were somewhat surprising. In particular, the rate of acceptance we recorded was lower than we expected: children only accepted Type B control sentences 65% of the times. Results are given in Table 2: Table 2.  Children’s acceptance of Type B control sentences Condition

Type B control

Acceptance

Trial 1

Trial 2

Total

13/20

13/20

26/40 (65%)

Upon inspection of the individual scores, however, we noticed that two children failed all control sentences and three children always reject true negative sentences. These children possibly show a ‘no’ bias. Since, the experimental hypothesis is indeed associated with a ‘no’ response, we decided to exclude these children from the analysis of the results. This leaves us with 15 children out of the initial twenty (age: 3;9 – 5;7. mean 4;5) which pass at least one control for each Type. These are children who do not have problems with (i) the modal potere, (ii) the use of the definite determiner or (iii) negative modal sentences. These fifteen children heard four target stories, and at the end of each story they were asked to judge the puppet’s statement. Recall that the target sentences were always true under the weak interpretation, the only reading permitted in adult Italian. If children can access this reading, which doesn’t involve covert operations, they are expected to accept the puppet’s statement. Table 3 summarizes children’s response: Table 3.  Children’s acceptance of the target sentences Acceptance

Trial 1

Trial 2

Trial 3

Trial 4

Total

6/14

5/14

4/15

4/15

19/58 (32,7%)

 Andrea Gualmini & Vincenzo Moscati

Looking at Table 3, we can see that children tend to reject the target sentence.5 In the experiment, every time the children rejected the target sentence, we asked for an explanation. In the vast majority of cases, the explanation was consistent with the strong reading, confirming that a non-adult interpretation was the source of children’s rejection. Even if we exclude the cases in which children gave an inconsistent explanation (for example: ‘because elephants are strange’), we are left with only 19 acceptances out of 55 trials i.e., 34.5%). Children’s responses stand in contrast with the responses of a group of ten adult controls, who always accepted the target sentence. To sum up, the experimental results suggest that Italian speaking children initially interpret sentences like (13) on a non-adult inverse scope interpretation. Although recent developments in the study of ambiguities remind us that our interpretation of the findings would have to be revised as soon as a successful contextual manipulation eliciting adult behaviour is put forward, for the time being we must register this finding together with English-speaking children’s behaviour documented by Moscati & Gualmini (2008b). A first attempt to single out the role of the pragmatic context is reported in Moscati (2008) where the target sentence is presented in response to different kinds of requests for information (see Gualmini et al. 2008). However, the results show that in the present case this manipulation has no effect on children’s preference, which still conforms to the results just presented.

4.  Concluding remarks In this paper we have considered experimental evidence suggesting that indeed children do not assume that sentences containing negation and a modal verb are ambiguous right off the bat. In particular, it looks as if children initially start off with the strong reading. We would like to conclude by acknowledging that our description of children’s non-adult behavior is still in need of an explanation. The same is true for other cases in which it looks as if children initially start off from the hypothesis that a sentence only has one reading, which turns out to be informationally stronger than the adult reading (see for instance Crain et al. 1994; Goro & Akiba 2004 and possibly the youngest children in Unsworth, Gualmini & Helder in press). One

5.  Notice that if we include also the answers of the five children excluded on the bases of their responses to Type A and Type B control sentences, we would have the same acceptance rate: 34.7% (25/72).



The early steps of modal and negation interactions 

explanation that is often invoked to explain this sort of findings is the learnability argument advanced by Crain et al. (1994). According to these authors, children must start off from the strong reading, because this is the only way for them to add a weaker reading on the basis of truth-conditional evidence. This is the Semantic Subset Principle. However, Gualmini & Schwarz (2007a, 2007b) recently showed that this claim ignores a wide range of sources of alternative evidence as well as truth-conditional evidence from sentences containing downward entailing operators (whose frequency in the primary linguistic data still needs to be evaluated). We must leave to further research the task of finding an explanation for the fact that strong interpretations seem to have a special place in the early stages of language acquisition. In fact, an interesting question for future research is why strong interpretations seem to play a crucial role for adults when it comes to many linguistic phenomena (e.g., among many others, reciprocals Dalrymple, Kanazawa, Kim, Mchombo & Peters 1998; plurals Winter 2001; implicatures Chierchia 2004 and accommodation Singh 2008).

Appendix – Target sentences I.

Il contadino può non dare le carote all’ elefante the farmer can not give the carrots to.the.elephant



a. *it is not possible that the farmer gives carrots to the elephant b. it is possible that the farmer doesn’t give carrots to the elephant

II.

L’ indiano può non dare la palla piccola al soldato romano the Indian can not give the ball little to.the soldier Roman



a. *it is not possible that the Indian gives the little ball to the Roman soldier b. it is possible that the Indian doesn’t give the little ball to the Roman soldier

III. Il pittore può non dare il verde all’ ippopotamo the painter can not give the green to.the.hippopotamus

a. *it is not possible that the painter gives the green brush to the hippopotamus b. it is possible that the painter doesn’t give the green brush to the hippopotamus

IV.

Il pilota può non dare la macchina grande al nano the pilot can not give the car big to.the dwarf



a. *it is not possible that the pilot gives the big car to the dwarf b. it is possible that the pilot doesn’t give the big car to the dwarf

 Andrea Gualmini & Vincenzo Moscati

References Chierchia, Gennaro. 2004. “Scalar Implicatures, Polarity Phenomena and the Syntax-Pragmatics Interface”. Structures and Beyond ed. by Adriana Belletti. Oxford: Oxford University Press. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford: Oxford University Press. Crain, Stephen, Laura Conway & Weijia Ni. 1994. “Learning, Parsing, and Modularity”. Perspectives on Sentence Processing ed. by Charles Clifton, Lyn Frazier & Keith Rayner, 443–467. Hillsdale, NJ: Lawrence Erlbaum Associates. Crain, Stephen & Cecile McKee. 1985. “The Acquisition of Structural Restrictions on Anaphora”. Proceedings of NELS 15, 94–110. Amherst, MA: GSLA Crain, Stephen & Rosalind Thornton. 1998. Investigations in Universal Grammar: A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge, Mass: MIT Press. Dalrymple, Mary, Makoto Kanazawa, Yookyung Kim, Sam Mchombo & Stanley Peters. 1998. “Reciprocal Expressions and the Concept of Reciprocity”. Linguistics and Philosophy 21. Fox, Danny. 2000. Economy and Semantic Interpretation. Cambridge, Mass: MIT Press. Goro, Takuya & Sachie Akiba. 2004. “Japanese Disjunction and the Acquisition of Positive Polarity”. Proceedings of the Tokyo Conference on Psycholinguistics. Tokyo: Hituzi Publishing Company. Gualmini, Andrea. 2004a. “Some Knowledge Children Don’t Lack”. Linguistics 42. 957–982. Gualmini, Andrea. 2004b. The Ups and Downs of Child Language: Experimental Studies in Children’s Knowledge of Entailment Relationships and Polarity Phenomena. New York: Routledge. Gualmini, Andrea. 2007a. “Mechanisms of Scope Resolution in Child Italian”. Romance Languages and Linguistic Theory 2005 ed. by Sergio Baauw, Frank Drijkoningen & Manuela Pinto, 149–163. Amsterdam: Benjamins. Gualmini, Andrea. 2007b. “And Now for Some Facts About Adults’ Interpretation of Ambiguous Sentences”. Proceedings of the Fifth Semantics in the Netherlands day 1–11 Groningen: University of Groningen. Gualmini, Andrea. 2007c. “Negation and Polarity: The View from Child Language”. Proceedings of the Workshop on negation and polarity 43–49. Tübingen: University of Tübingen. Gualmini, Andrea. 2008. “The Rise and Fall of Isomorphism”. Lingua 118. 1158–1176. Gualmini, Andrea, Sarah Hulsey, Valentine Hacquard & Danny Fox. 2008. “The Question–Answer Requirement for Scope Assignment”. Natural Language Semantics 16. 205–237. Gualmini, Andrea & Bernhard Schwarz. 2007a. “Negation and Downward Entailingness: Consequences for Learnability Theory”. Proceedings of the Workshop on negation and polarity, 50–56. Tübingen: University of Tübingen. Gualmini, Andrea & Bernhard Schwarz. 2007b. “Solving Learnability Problems in the Acquisition of Semantics”. Ms., Utrecht University and McGill University. Hintikka, Jaakko. 1962. Knowledge and Belief. Ithaca, NY: Cornell University Press. Hulsey, Sarah, Valentine Hacquard, Danny Fox & Andrea Gualmini. 2004. “The Question-Answer Requirement and Scope Assignment”. Plato’s Problem: Problems in Language Acquisition ed. by Aniko Csirmaz, Andrea Gualmini & Andrew Nevins, 71–90, Cambridge, Mass.: MIT Papers in Linguistics 48. 71–90. Ladusaw, William. 1979. Polarity Sensitivity as Inherent Scope Relations. Ph.D. Dissertation, University of Texas.



The early steps of modal and negation interactions 

Lewis, David. 1995. Counterfactuals. Oxford: Basil Blackwell. May, Robert. 1985. Logical Form: Its Structure and Derivation. Cambridge, Mass: MIT Press. Moscati, Vincenzo. 2008. Strength and Weakness of Children’s Interpretation of Modals. The Proceedings of the Ninth Tokyo Conference on Psycholinguistics ed. by Yukio Otsu. Tokyo: Hituzi Publishing Company. Moscati, Vincenzo & Andrea Gualmini. 2008a. “Negation and Modality in Child Language”. Paper presented at the Incontro di Grammatica Generativa, Padova. Moscati, Vincenzo & Andrea Gualmini. 2008b. “More Facts that Isomorphism Cannot Explain”. Proceedings of SALT XVII ed. by Tova Friedman & Masayuki Gibson. CLC Publications, Cornell University, Ithaca. Musolino, Julien. 1998. Universal Grammar and the Acquisition of Semantic Knowledge: An Experimental Investigation into the Acquisition of Quantifier-Negation Interaction in English. Ph.D. Dissertation, University of Maryland. Musolino, Julien. 2006. “On the Semantics of the Subset Principle”. Language Learning and Development 2. 195–218. Musolino, Julien, Stephen Crain & Rosalind Thornton. 2000. “Navigating Negative Quantificational Space”. Linguistics 38. 1–32. Musolino, Julien. & Jeffrey Lidz. 2004. “The Syntactic Basis of Isomorphism”. Ms., Indiana University and Northwestern University. Musolino, Julien & Jeffrey Lidz. 2006. “Why Children Aren’t Universally Successful with Quantification”. Linguistics 44. 817–852. Palmer, Frank R. 1986. Mood and Modality. Cambridge: Cambridge University Press. Reinhart, Tanya. 1997. “Quantifier Scope: How Labor Is Divided between QR and Choice Functions”. Linguistics and Philosophy 20. 335–397. Reinhart, Tanya. 2006. Interface Strategies: Reference-Set Computation. Cambridge, Mass: MIT Press. Singh, Raj. 2008. “Economy and Intermediate Accomodation”. Paper presented at the DIP Colloquium, Amsterdam. Stalnaker, Robert. 1968. “A Theory of Conditionals”. Studies in Logical Theory ed. by Nicholas Rescher, 98–122. Oxford: Blackwell. Unsworth, Sharon, Andrea Gualmini & Christina Helder. in press. “Children’s Interpretation of Indefinites in Sentences Containing Negation: A Re-Assessment of the Cross-Linguistic Picture”. Language Acquisition. Wells, Gordon. 1979. “Learning and Using the Auxiliary Verb in English”. Cognitive Development: Language and Thinking from Birth to Adolescence ed. by Victor Lee, 250–270. London: Croom Helm. Winter, Yoad. 2001. “Plural Predication and the Strongest Meaning Hypothesis”. Journal of Semantics 18. 333–365.

Structural patterns blocking plural in Romance nominalizations* Gianina Iordăchioaia1 & Elena Soare2 1University

of Stuttgart/Tübingen/2University of Paris 8

Some Complex Event Nominals (CENs) like the Romanian infinitive contradict the generalization in Grimshaw (1990) according to which CENs do not pluralize. But many CENs like the Romanian supine and the Spanish nominal infinitive obey this generalization. In this paper we explain the availability and the blocking of pluralization in relation to the internal structure of the CEN. We argue that pluralization in CENs is blocked by the syntactic encoding of the unbounded outer or inner aspect. The Romanian supine projects a verbal structure with a [–bounded] (outer) Aspect projection. In the Spanish nominal infinitive, the unbounded/atelic inner aspect is accommodated by a [–count] Classifier projection in a nominal structure. Both patterns prevent the projection of Number, and thus, pluralization. But this does not exclude the availability of plural for CENs with a bounded/telic inner aspect which project a [+count] ClassifierP and implicitly, NumberP, as in the case of Romanian infinitival CENs.

1.  Introduction Grimshaw (1990) argued that Complex Event Nominals (CENs) disallow plural marking (1b), as opposed to the corresponding R(esult) Nominals (1a). Among deverbal nouns, CENs can be distinguished by the fact that they express an event (indicated in (1b) by the expression took a long time) which is ‘complex’, since they inherit (at least parts of) the argument structure of the base verb (the theme of the problems in (1b)): (1) a. The assignments were long. b. The assignment(*s) of the problems took a long time. *We are thankful to the audience of Going Romance 2007, two anonymous reviewers, and the editors of this volume for insightful remarks and useful suggestions. We also thank Antonio Fàbregas for his help with the Spanish data. The research of the first author in alphabetical order was supported by a DFG grant to the project B1, The formation and interpretation of derived nominals, as part of the Collaborative Research Center 732, Incremental Specification in Context, at the University of Stuttgart.

 Gianina Iordăchioaia & Elena Soare

Empirical evidence from Romance languages contradicts this generalization (see Roodenburg 2006, for French and Italian, Iordăchioaia & Soare 2008, for Romanian).1 In particular, the Romanian infinitive CEN in (2), as opposed to the supine CEN, does not block pluralization:2 (2) demolă-r-i-le /*demola-t-uri-le frecvente ale demolish.inf.pl-the3 /demolish.sp.pl-the frequent.pl of cartierelor vechi de către comunişti pentru a ridica turnuri quarters.g old by communists for to raise towers “the frequent demolitions of old quarters by the communists in order to raise towers (instead)”

In order to account for this contrast, we closely consider the fact that the semantics of plural may take several syntactic forms. In a nominal structure, plural is realized in the morphology or as a mass noun. In a verbal structure, plural appears as an unbounded aspect: either an imperfective outer aspect or an atelic inner aspect. CENs have a mixed categorial status which is due to their verbal internal make-up combined with a nominal external behavior. From this perspective, we expect plural in various CEN patterns to be realized in different ways, depending on their mainly nominal/verbal structure and their aspectual properties. Grimshaw’s generalization led to the idea that CENs are mass nouns. Since this is obviously not always the case, we propose a close investigation of the internal syntactic structure of the CENs in order to describe their behavior with respect to plural marking. In general, plural is blocked by an unbounded aspect, but it is allowed with a telic/perfective (i.e. bounded) aspect (cf. Mourelatos 1978). Moreover, we assume that morphological plural is hosted by Num(ber)P which is in complementary distribution with (outer) Asp(ect)P. Thus, we follow Iordăchioaia & Soare (2008) and explain the contrast in (2) by the fact that the infinitive CEN realizes morphological plural in a nominal structure with a NumP and the supine

1.  Pluralization of event nominals with argument structure seems to be also possible in other languages like Dutch, English, German, and Portuguese (Sleeman & Brito 2008 and references therein, Alexiadou et al. 2009). 2.  Note that in (2), the presence of the theme ale cartierelor vechi, of the by-phrase de către comunişti, and the availability of control in the purpose clause clearly indicate the CEN status of the infinitive and the supine nominals. To save space, in what follows we will often express only the theme argument, but adding a by-phrase and a purpose clause is always possible, so their absence should not question the CEN status of the nominals. For a detailed discussion of this issue, see also Cornilescu (2001). 3.  Abbreviations: inf ‘infinitive’, sp ‘supine’, sg ‘singular’, pl ‘plural’, g ‘genitive’, a ‘accusative’, n ‘nominative’, f ‘feminine’, m ‘masculine’, n ‘neuter’.



Structural patterns blocking plural in Romance nominalizations 

rejects morphological plural because it has a verbal structure with an AspP. The NumP in the infinitive hosts the plural marker and the AspP in the supine hosts a pluractional operator introducing an (unbounded) plurality of events. This syntactic solution is independently motivated by the nominal properties of the infinitive and their absence with the supine. In consequence, unbounded outer AspP constitutes one syntactic pattern that explains Grimshaw’s generalization. It appears with the Romanian supine and it was argued to also characterize the English verbal gerund (Alexiadou et al. 2009). In this paper, we provide evidence for a second syntactic pattern which blocks plural in CENs. It comes from Spanish Nominal Infinitives (SNIs, see Miguel 1996) like in (3b) below: (3) a.

el construir casas de los albañiles the.sg build.inf houses.a of the workers “the constant building of houses by the workers”

b. *los (constantes) construi-r-es casas de los albañiles the.pl constant.pl build.inf.pl houses.a of the workers “the constant buildings of houses by the workers”

Apparently, SNIs are unbounded like the Romanian supine, as they exclusively select atelic VPs.4 But an AspP is not motivated in the structure of SNIs, since unlike the supine, they have no aspectual contribution of their own, reject adverbs, and carry (nominal) gender features. We will argue that SNIs realize unboundedness in a nominal structure under Class(ifier)P (a category proposed in Picallo 2006), and not in a verbal one under AspP. In Section 2, we start by presenting the properties of the two Romanian CEN patterns in (2) in order to trace the origin of their contrasting behavior with respect to plural marking. In Section 3, we address the aspectual properties of the supine CEN and we argue that they indicate the presence of a syntactic projection AspP. We complete the picture with the Spanish nominal infinitive which shares properties with both the Romanian infinitive and the supine (Section 4). We present our conclusions in Section 5. 2.  Plural marking in Romanian CENs In this section, we provide evidence for the nominal properties of the infinitive CEN in (2), generally absent in the supine.

4.  Note that the bare plural casas makes the whole VP atelic, although the verb construir is telic.

 Gianina Iordăchioaia & Elena Soare

The infinitive and the supine are the most productive deverbal nouns in Romanian and they derive from the infinitive and the past participle stem, respectively. Unlike -t/s, which derives both the past participle and the supine form, as it is originally a verbal suffix, the suffix -re attaches to the short infinitive demola and derives infinitival nouns, as it is exclusively a nominalizer. We employ the hash symbol to stand for ‘unavailable in the relevant use/meaning’:

(4) a.



The Infinitive: a demola: demola–r –e /demolă–r–i to demolish: demolish.inf.f.sg /demolish.inf.pl

b. The Supine:

a demola: demola–t /#demola–t–uri to demolish: demolish.sp /  demolish.sp.pl



a conduce: condu-s/#condu-s-uri to drive: drive.sp/ drive.sp.pl

As can be observed in (4), an important difference between the two nominal forms in general concerns plural marking, possible with the infinitive (4a), but not with the supine (4b).5 For CENs, in particular, the selection of discrete (vs. mass) quantifiers further suggests that the infinitive (unlike the supine) is a count noun projecting NumP: (5) a.

Prea multe spălări /o spălareˉ a(le) rufelor too many wash.inf.pl /one wash.inf of laundry.g



distrug(e) ţesătura. destroy(s) fabric.the

b. Prea mult /*un spălat al rufelor … too much /  one wash.sp of laundry.g “Too much washing of the laundry destroys the fabric.”

According to Picallo (2006), gender features and noun class information, hosted by a ClassP projection, are obligatory for the projection of NumP. This idea helps us to bring further support for the assumption that the infinitive CEN projects NumP, while the supine does not, to the extent that the former exhibits

5.  The default plural ending -uri is associated with the supine by analogy with common nouns originating from the supine/past participle (e.g. tuns – tunsuri ‘haircut(s)’, venit – venituri ‘income(s)’). These nouns either express simple events, or acquire a special lexical meaning (Iordăchioaia & Soare 2008), but the homonymous CEN does not accept plural: (i) tunsul /*tunsurile părului cut.sp.the / cut.sp.pl.the hair.g “cutting the hair”



Structural patterns blocking plural in Romance nominalizations 

nominal properties which are absent in the latter: infinitive CENs display feminine gender and a full case declension, while supine CENs have default neuter gender and a defective case paradigm. With respect to gender, the data in (6) indicate that infinitive CENs successfully establish anaphoric relations with the feminine demonstrative aceasta (6b), while the supine rejects the masculine-neuter form acesta and can only be referred to by the genderless form asta (6c), the common anaphor for CPs in Romanian (6a):6 (6) a.

Că Ion a venit, asta/*aceasta/*acesta ştiu. “That John came, I know it/this.f/this.m.”

b. Am vorbit despre interpretarea rolului Hamlet în general. Se pare ca aceasta/??asta îi consacră indubitabil pe actorii tineri. “We spoke about the interpretation.inf of Hamlet in general. Apparently, this.f/??it undoubtedly validates the young actors.” c. Am vorbit despre interpretatul rolului Hamlet în general. Se pare ca *acesta/asta îi atrage pe toţi actorii tineri. “We spoke about the interpretation.sp of Hamlet in general. Apparently, *this.m.n/it attracts all the young actors.”

Defectiveness characterizes the supine CEN also with respect to case declension. The data in (7) show that unlike the infinitive, the supine is uncommon in the genitive-dative form.7 It usually exhibits the non-oblique form for nominativeaccusative (8): (7)

Alunecările de teren au apărut dinˉcauza flows.the of earth have occurred becauseˉof

tăierii/ *tăiatului pădurilor. cut.inf.g/ cut.sp.g woods.g “The earth flows occurred because of cutting the woods.”

6.  Although the anaphor asta is not excluded in (6b), this is an instance of coercion, since in this case the infinitive interpretarea is understood as a fact, and not as an event. 7.  An apparent counterexample is the genitive supine in (i). However, note that tăiatul porcului here is not a CEN, but almost an idiom, since a by-phrase makes it infelicitous. The nominative supine in (ii) allows a by-phrase: (i) ritualul tăiatului porcului (*de către bunicul familiei) ritual.the cut.sp.the.g pig.g ( by grandfather family.g) “the ritual of killing the pig (*by the grandfather of the family)” (ii) Tăiatul porcului de către bunicul familiei e obiceiul local. cut.sp.the.n pig.g by grandfather family.g is custom local “The grandfather of the family killing the pig is the local custom.”

 Gianina Iordăchioaia & Elena Soare

(8) Tăierea /tăiatul pădurilor este interzis(ă). cut.inf.n.the /cut.sp.n.the woods.g is forbidden “Cutting the woods is forbidden.”

As a consequence of the observations above, we follow Iordăchioaia & Soare (2008) and take the structure of the infinitive CEN to be fully nominal, thus including both ClassP and NumP (see the tree in (9)). The suffix -r under N nominalizes the VP and, according to Picallo (2006), N moves to Class to check its gender features carried by -e. 2.1  Telicity and plural marking The realization of number in infinitive CENs may be related to their aspectual properties. Cornilescu (2001) argues that they express telic events, since they reject unergative verbs which are invariably atelic. These verbs can only appear in the supine (see the examples in (10)).8

(9) citirea “the reading” DP D

NumP Num

ClassP Class

-a

(10)

NP

N [±sg] -e[fem] -r-

a locui *locui-re to live   live.inf a munci *munci-re to work   work.inf

VP citi

locui-t live.sp munci-t work.sp

As a further confirmation of its telicity, the infinitive obligatorily projects the theme of the base verb – it cannot project only the external argument. On the assumption that the theme indicates the culmination of the event, the contrast in (11a–c) shows that the event in the infinitive CEN must always culminate (see Cornilescu 2001): (11) a. *citirea lui Ion read.inf.the John.g

8.  More examples are given in Cornilescu (2001: 489). The infinitives locuire, muncire are available only if derived from colloquial transitive forms of the verbs ‘live’, ‘work’.



Structural patterns blocking plural in Romance nominalizations 

b. citirea cărţii (de către Ion) read.inf.the book.g  by John “the reading of the book by John” c.

cititul lui Ion cu glas tare pentru a ne enerva read.sp.the John.g with voice loud for to us annoy “John’s reading loudly in order to annoy us”

The supine can project its agent without obligatorily projecting the theme (11c) and since it has no restrictions on the base verb, Cornilescu concludes that it denotes an atelic event. We will return to the supine in Section 3. For now, the availability of morphological plural in infinitive CENs correlated with their telic aspect confirms previous observations in Mourelatos (1978) and Borer (2005), according to which telic CENs can pluralize and only atelic ones cannot. 3.  The aspect projection in the Romanian supine In this section we present the unboundedness induced by a pluractional operator via aspect shift as the main characteristic of the supine CEN. By comparison with the infinitive CEN, this motivates the presence of an AspP in the supine CEN. We reformulate the aspectual differences between infinitive and supine CENs in terms of [±b](oundedness) as a wider notion allowing for generalizations across the nominal/verbal domains (Jackendoff 1991). The nominal plural, mass nouns, and the atelic aspect count as [–b], while the nominal singular, count nouns, and the telic aspect are [+b]. Given its telicity, the infinitive CEN expresses a [+b] event that cannot be bounded once again by a function like until which turns a [–b] event into [+b].9 This is possible with the supine CEN, which expresses a [–b] event: (12) arestatul /#arestarea lui Miron Cozma (iar şi iar) arrest.sp.the /  arrest.inf.the Miron Cozma.g again and again pînă la schimbarea puterii until at change.the regime.g “arresting M. C. again and again until the regime changed”

3.1  The pluractional operator (PO) The [–b] character of the supine is the effect of an implicit PO present in its structure which changes the aspectual value of the original event into [–b]. POs have

9.  The only possible reading for the infinitive in (12) is ‘the arresting of M. C. before the government changed’ where until means ‘before’ and does not act as a bounding function.

 Gianina Iordăchioaia & Elena Soare

usually been studied in polysynthetic languages (Lasersohn 1995; van Geenhoven 2004); however, implicit instances of POs have been identified also in English (van Geenhoven 2004), and Laca (2006) describes two phrasal constructions with the same value in Spanish. Two important properties of POs are the lack of multiplicity effects with indefinites and the distribution effects with plurals, illustrated by the West Greenlandic PO qattaar in (13a) and (13b), respectively: (13) a. ?Qaartartoq sivisuumik qaaqattaarpoq. “A bomb exploded again and again for a long time.” b. Qaartartut sivisuumik qaaqattaarput. “Bombs exploded again and again for a long time.”

Van Geenhoven shows that the telic verb equivalent to explode must combine with the PO qattaar in order to become [–b] and accept the atelic modifier for a long time. But the singular noun cannot be multiplied by the presence of the PO for us to understand that a different bomb explodes every time, so (13a) can only be accepted in the scenario of a magic bomb which explodes more than once. The grammaticality of (13b) with a plural argument indicates that the PO induces distributivity over the plural, so that we understand the sentence as referring to several explosions of individual bombs and not to one explosion of a group of bombs. The first property distinguishes POs from frequency adverbs like occasionally which are known to pluralize the event, too, but besides this, they also create multiplicity effects with singular indefinites. This explains the availability of the reading in (14b): (14) A bomb occasionally exploded. a. The same bomb exploded. (magic bomb) b. A different bomb exploded every time.

The two properties of POs discussed above can also be observed in the behavior of the Romanian supine CEN, as shown in (15). A one-time event like kill in the supine is ungrammatical with a singular theme, but grammatical with a plural one. This is because the PO in the supine suggests more killing events which cannot have the same theme; a plurality is needed which can be distributed over several events: (15) a. *ucisul unui jurnalist de către mafia politică kill.sp.the a.g journalist by mafia political b. ucisul jurnaliştilor de către mafia politică kill.sp.the journalists.the.g by mafia political “the killing of journalists by the political mafia”

The same properties have been observed with the pluractional construction ‘andar+gerund’ in Spanish (Laca 2006), in addition to other characteristics like



Structural patterns blocking plural in Romance nominalizations 

the incompatibility with ‘once only’ frequency adjuncts and habituality. These can also be observed in the behavior of the supine in (16) and (17): (16) cititul romanului (*dintr-o răsuflare) read.sp.the novel.g   in-one-breath “the reading of the novel in one sitting” (17) a.

Fumatul trabucurilor /acest obicei l-a îmbolnăvit. smoke.sp.the cigars.g /this habit him-has sickened “Smoking/this habit made him sick.”

b. fumatul lui Ion ori de cîte ori iese în pauză smoke.sp.the John.g whenever takes a break “John’s smoking whenever he takes a break”

The possibility to replace a supine CEN with the phrase this habit, as well as the compatibility with whenever-clausal modifiers show that the supine expresses habituality (Soare 2006), a typical effect of pluractionality. We conclude that it is the PO that gives the [–b] aspect of the supine CEN. But unlike in the case of the [+b] infinitive CEN, where the nominalizer -r(e) rejects atelic verbs, the supine can combine with both telic and atelic verbs, and the result is always [–b], so it triggers aspect shift. This can be observed in its interaction with semantic verb classes below.10 3.2  The interaction between the supine and the verbal base Achievements and accomplishments, which are telic, i.e. [+b], successfully appear with the supine, which expresses a [–b] plurality of events, and thus, a habitual interpretation occurs (18): (18) a.

Sositul lui Ion cu întîrziere mă enervează. arrive.sp.the John.g with delay me annoys “John’s (habit of) arriving late annoys me.”

b. mîncatul micului dejun pe terasă eat.sp.the breakfast.g on terrace “(the habit of) having breakfast on the terrace”

10.  A reviewer notes that the supine may also refer to a single event like in (i). In this case, we should mention that the supine suggests an event involving several smaller events, so it induces distributivity. As Lasersohn (1995) notes, the same PO may at times yield habituality or distributivity or other pluractionality effects. In Romanian, the use of supine in (i) usually has a colloquial flavor: (i) Ştersul prafului în sufragerie mi-a luat o jumătate de oră. wipe.sp.the dust.g in living.room me-has taken a half of hour “Dusting in the living-room took me half an hour.”

 Gianina Iordăchioaia & Elena Soare

Activities (19a) and states (19b), known as atelic/[–b], may combine with the supine only if the event is understood as bounded. Thus, the presence of a bounding function like until is sometimes obligatory :11 (19) a.

învăţatul lui Ion *(pînă la miezul nopţii) study.sp.the John.g (until midnight) “John’s (habit of) studying until midnight”

b. dormitul lui Ion *(pînă după-amiaza tîrziu) sleep.sp.the John.g (until afternoon late) “John’s (habit of) sleeping until late afternoon”

In conclusion, the [–b] character of the supine CENs comes from the presence of the PO which pluralizes the bounded event, independently of the inner aspect of the base verb. We propose with Iordăchioaia & Soare (2008) that the supine CEN expresses verbal plurality, hosted by an AspP in the syntax, so it has the structure in (20). An N head is not motivated in (20), given the lack of a bona fide nominalizer (-t/s marks verbal stems, see Section 2) and the absence of nominal properties. We take the determiner -(u)l to be alone responsible for the external nominal behavior of the supine CEN which thus qualifies as a ‘syntactic’ nominalization in the terms of Chomsky (1970): (20)

cititul “the reading” DP D

-(u)l

AspP Asp

VP

PO [–b]

citit

(21) reprodusul (constant al) tablourilor (constant) reproduce.sp.the constantly of paintings.g constantly “constantly imitating paintings”

AspP is independently motivated by the compatibility with the adverb constantly argued in Cinque (1999) to indicate an Aspect projection. Note that despite

11.  In contexts like (i), an originally [–b] state verb can be understood as [+b] even without being bounded by ‘until’. Note, however, that the supine is still interpreted as a plurality of events, since dormitul does not refer to an unbounded state of sleeping, but to a habit. We thank an anonymous reviewer for bringing this example to our attention: (i) Dormitul părintelui cu bebeluşul în pat prezintă riscuri. sleep.sp.the parent.g with baby.the in bed presents risks “The parent sleeping with the baby in the same bed is risky.”



Structural patterns blocking plural in Romance nominalizations 

the adjective – adverb homonymy in Romanian, constant in (21) is not an adjective, since if it were so, it should be able to appear in prenominal position, as it does with the infinitive CEN in (22b): (22) a. *constantul reprodus al tablourilor constant.the reproduce.sp.the of paintings.g b. constanta reproducere a tablourilor constant.the reproduce.inf of paintings.g “the constant imitation of paintings”

4.  Inner aspect in CENs: Spanish Nominal Infinitives (SNIs) The projection of [–b] Aspect as in the Romanian supine represents one source of plural blocking in CENs. In this section we discuss an additional case of unboundedness as blocking plural which concerns Spanish Nominal Infinitives as in (3b). We will argue that in SNIs we are dealing with a nominal structure and as a consequence it is not an AspP, but a ClassP which hosts the [–b] aspectual information in the syntax. (3) a.

el construir casas de los albañiles the.sg build.inf houses.a of the workers “the constant building of houses by the workers”

b. *los (constantes) construi-r-es casas de los albañiles the.pl constant.pl build.inf.pl houses.a of the workers “the constant buildings of houses by the workers”

Miguel (1996) distinguishes between verbal and nominal infinitive CENs in Spanish which primarily differ in the way they realize their external argument: with nominative case, or with de ‘of ’, respectively:12 (23) a.

el murmurar la gente the murmur.inf the people.n “the murmuring of the people”

b. el murmurar de las fuentes the murmur.inf of the fountains “the murmuring of the fountains”

Here, we are concerned with the nominal pattern in (23b).

12.  The internal argument receives accusative case (e.g. the theme casas in (3)) both in the verbal and in the nominal infinitive.

 Gianina Iordăchioaia & Elena Soare

Miguel (1996), Demonte & Varela (1997), Fábregas & Varela (2006) have shown that SNIs have atelic/unbounded aspect, since they systematically reject telic VPs. SNIs are ungrammatical with achievements and accomplishments, for instance: (24) a. *Elˉ llegar tardío de Juan nos preocupó a todos. the arrive.inf late of Juan us worried to all b. *El comprar una casa de Juan nos alegró. the buy.inf a house of Juan us made happy

In view of the correlation between atelicity, unboundedness, and mass nouns and our discussion with respect to the Romanian supine, the unavailability of plural morphology in (3b) is expected in SNIs. The next concern is to find out what exactly blocks the projection of Number in the syntax of SNIs and to see if we can argue for an AspP like in the case of the supine. We address this issue below. 4.1  Nominal properties in SNIs Miguel (1996) enumerates a number of properties which motivate the need to distinguish between the two infinitives in (23). These properties clearly indicate that SNIs (23b) present a rather nominal structure. Here, we address only a few, which are relevant in establishing a relation between SNIs and the two CEN patterns in Romanian. First, SNIs carry gender features which − although not visible in the suffix -r − become obvious in anaphoric contexts, where an SNI can be referred to only by the masculine pronoun él and not by the default neuter pronoun ello usually employed with non-nominal CP structures. In this respect, SNIs are similar to the Romanian infinitive CEN (cf. (6b) above): (25)

Accostumbrado al dulce mirar de su amada, ya used to.the sweet gaze.inf of his beloved, now

no podía vivir sin él/*ello. not could live without him/it “Used to the sweet gaze of his loved one, he could no longer live without it.”

Second, SNIs never combine with adverbs, only with adjectives. Modifiers with aspectual value also appear as adjectives, like in the case of the Romanian infinitive in (22b): (26) a.

El andar errabundo /*errabundamente de Juan the go.about.inf aimless /*aimlessly of Juan “John’s aimless going about”

b. El constante temer (*constantamente) de Juan the constant fear.inf ( constantly) of Juan “John’s constant fear (of something)”



Structural patterns blocking plural in Romance nominalizations 

As a consequence of these properties and others, Miguel analyzes the suffix -r in SNIs as a nominalizer to distinguish it from the inflectional affix -r which appears with verbal infinitives like in (23a). 4.2  Inner vs. outer aspect A comparison between the two Romanian CENs and the SNI indicates that the latter resembles the Romanian infinitive in what concerns the nominal character: they both carry gender features, reject adverbs, and their affix has been analyzed as a nominalizer. But unlike Romanian infinitives, SNIs do not accept plural marking since they are atelic, and in this respect, they pattern with the supine, although an Aspect projection is not motivated, given their nominal character. The similarities and the differences between the three CENs are summarized in Table 1 below: Table 1.  The properties of Romanian and Spanish CENs CEN

Plural marking

Unbounded aspect

Verbal structure

Infinitive

yes

no

no

Supine SNI

no no

yes yes

yes no

In order to account for the mixed properties of SNIs, we propose that their aspectual information is encoded within the Classifier projection together with its nominal features. We first make a distinction between inner and outer aspect (after Verkuyl 1993). The former − also called Aktionsart − is closely related to the lexical properties of an event and its value is composed within the VP. Outer aspect − also called grammatical aspect − concerns the aspectual properties introduced at a higher level than the VP and is oblivious to the value of the inner aspect. The progressive -ing in English is an instance of outer aspect: it turns both telic and atelic events into imperfective: (27) a. John is arriving/dying. b. John is running/sleeping.

Telicity is most often used in relation to inner aspect, while perfectivity refers to outer aspect. We have taken boundedness to cover both notions, including the mass/count distinction in the nominal domain. In the study of CENs, we assume that the presence of outer aspect is directly correlated with aspect shift: independently of the inner aspect, if outer aspect is contributed, its value is consistent and can be delimited from that of the inner aspect (see Alexiadou et al. 2009 for a similar analysis of the English verbal gerund). The PO in the Romanian supine has been shown to have this property: it always refers

 Gianina Iordăchioaia & Elena Soare

to a plurality of events. The aspectual properties of infinitive CENs and SNIs are of a different nature. The two nominalizers -r(e) and -r do not introduce any aspectual information, they are only sensitive to the inner aspect of the VP: it should be telic, or atelic, respectively. We conclude that Romanian infinitives and SNIs carry only inner aspect information. 4.3  The analysis On the basis of the parallelism between the nominal Classifier projection and the verbal inner aspect (see Fassi Fehri 2005; Alexiadou et al. 2008), we propose that CENs with clear nominal structure like the Romanian infinitive and the SNI encode their aspectual properties under ClassP by means of a semantic feature [±count]. This feature is independently present under the ClassP of common nouns and it distinguishes between countable and uncountable nouns. The value [+count] motivates a further projection of Number in countable nouns in the spirit of Picallo (2006). The [–count] value occurs with mass nouns and it prevents the projection of NumP. Since Picallo’s main concern is to argue that the Classifier hosting gender features is the input to a Number projection, our analysis correctly predicts that both mass and count nouns have gender, but only the latter project NumP. Within this frame, we can analyze SNIs as projecting a [–count] ClassP and no NumP. The difference between them and mass nouns stands only in their eventive nature triggered by the VP that they nominalize. The atelicity of the event is of the same nature as the uncountability of mass nouns. In (28) below we give the functional structure of SNIs. To complete the picture, the Romanian infinitive CEN projects a [+count] ClassP and NumP. The former encodes the telicity of the base verb and the latter hosts the plural marker, as indicated in (9) above. (28) el construir “the building” DP D

ClassP Class

el

N [–count] -r

NP VP construi-

Note that in comparison to Miguel (1996), our analysis of SNIs can better explain the distinction between verbal infinitives and SNIs. In order to account for the aspectual properties of the SNI, Miguel places an AspP above the NP hosting the nominalizer -r. The same projection is then assumed in the structure of the verbal infinitive, this time on top of the VP level. But she does not explain why a verbal projection should be allowed in a nominal structure with NP (contra Borsley & Kornfilt 2000)



Structural patterns blocking plural in Romance nominalizations 

and why the Asp value should be unspecified in one case and obligatorily [–perfective] in the other. We believe that a distinction between inner and outer aspect, on the one hand, and a delimitation of the nominal projections from the verbal ones, on the other hand, is the right solution for this problem. 5.  Conclusions In this paper we argued for two structural patterns of CENs which explain the plural blocking effect identified in Grimshaw (1990). We showed that from a semantic point of view unbounded aspect – a property reminiscent of the verbal origin of CENs – prevents the realization of morphological plural. In the syntax, we argued that unboundedness may be encoded either in a verbal or in a nominal structure, under AspP or ClassP, respectively. These distinctions allowed us to characterize three different instances of CENs in Romance: the Romanian infinitive CEN, the Romanian supine CEN, and Spanish nominal infinitives. The first one is mainly nominal, expresses bounded/telic (inner) aspect and thus, projects [+count] ClassP and NumP. Plural morphology is available. The second one is mainly verbal, carries unbounded aspect, so it projects an (outer) AspP which blocks Number and implicitly, pluralization. The last CEN structure combines nominal properties with unbounded/atelic (inner) aspect, so it projects a [–count] ClassP which prevents the projection of NumP and thus, plural morphology is unavailable.

References Alexiadou, Artemis, Gianina Iordăchioaia & Elena Soare. 2009. “Plural Marking in Argument Supporting Nominalizations”. To appear in Layers of Aspect ed. by Patricia Cabredo Hofherr & Brenda Laca. Stanford: CSLI Publications. Alexiadou, Artemis, Gianina Iordăchioaia & Elena Soare. 2008. “Nominal and Verbal Parallelisms”. Paper presented at the workshop DP Types and Feature Syntax, 31st GLOW Colloquim, Newcastle, March 2008. Borer, Hagit. 2005. Structuring Sense. Oxford: Oxford University Press. Borsley, Robert & Jaklin Kornfilt. 2000. “Mixed Extended Projections”. The Nature and Function of Syntactic Categories (= Syntax and Semantics, vol. 32) ed. by Robert Borsley. 101–131. San Diego: Academic Press. Chomsky, Noam. 1970. “Remarks on nominalization”. Readings in Transformational Grammar ed. by Roderick A. Jacobs & Peter S. Rosenbaum, 184–221. Waltham, Mass.: Ginn and co. Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A crosslinguistic perspective. New York: Oxford University Press. Cornilescu, Alexandra. 2001. “Romanian Nominalizations: Case and aspectual structure”. Journal of Linguistics 37:3. 467–501.

 Gianina Iordăchioaia & Elena Soare Demonte, Violeta & Soledad Varela. 1997. “Spanish Event Infinitives. From Lexical-Semantics to Syntax-Morphology”. Theoretical Issues at the Morphology-Syntax Interface, supplement of the International Journal of Basque Linguistics and Philology XI ed. by Amaya Mendikoetxea & Myriam Uribe-Etxebarria, 253–277. Fábregas, Antonio & Soledad Varela. 2006. “Verb Classes with Eventive Infinitives in Spanish”. Selected Proceedings of the 9th Hispanic Linguistics Symposium ed. by Nuria Sagarra & Almeida Jacqueline Toribio, 24–33. Somerville, Mass.: Cascadilla Proceedings Project. Fassi-Fehri, Abdelkader. 2005. “Verbal and Nominal Parallelisms”. Documents & Reports 8. Rabat: Publications IERA. Geenhoven, Veerle van. 2004. “For-adverbials, Frequentative Aspect, and Pluractionality”. Natural Language Semantics 12. 135–190. Grimshaw, Jane. 1990. Argument structure. (= Linguistic inquiry: Monographs, 18.) Cambridge, Mass.: MIT Press. Iordăchioaia, Gianina & Elena Soare. 2008. “Two Kinds of Event Plurals: Evidence from Romanian Nominalizations”. Empirical Issues in Syntax and Semantics 7 ed. by Olivier Bonami & Patricia Cabredo Hofherr, ISSN1769–7158. Jackendoff, Ray. 1991. “Parts and Boundaries”. Cognition 41. 9–45. Laca, Brenda. 2006. “Indefinites, Quantifiers and Pluractionals: What scope effects tell us about event pluralities”. Non-definiteness and Plurality ed. by Liliane Tasmowski & Svetlana Vogeleer, 191–217. Amsterdam: John Benjamins. Lasersohn, Peter. 1995. Plurality, Conjunction and Events. Dordrecht & Boston: Kluwer. Miguel, Elena de. 1996. “Nominal Infinitives in Spanish: An aspectual constraint”. Canadian Journal of Linguistics 41:1. 29–53. Mourelatos, Alexander P.D. 1978. “Events, Processes and States”. Linguistics and Philosophy 2:3. 415–434. Picallo, M. Carme. 2006. “Some Notes on Grammatical Gender and l-Pronouns”. Proceedings of the workshop “Specificity and the evolution/emergence of nominal determination systems in Romance” ed. by Klaus von Heusinger, Georg A. Kaiser & Elisabeth Stark, Universität Konstanz, (= Arbeitspapier, 119.) 107–121. Roodenburg, Jasper. 2006. “The Role of Number within Nominal Arguments”. Paper presented at the 36th Linguistic Symposium on Romance Languages. Rutgers, March-April 2006. Sleeman, Petra & Ana Maria Brito. 2008. “Aspect and Argument Structure of Deverbal Nominalizations: A split vP analysis”. To appear in Nominalizations across languages and frameworks ed. by Artemis Alexiadou & Monika Rathert. Berlin: Mouton de Gruyter. Soare, Elena. 2006. “Why Smoking is a (Bad) Habit”. Bucharest Working Papers in Linguistics. VIII. Bucharest: Bucharest University Press. Verkuyl, Henk. 1993. A Theory of Aspectuality. The Interaction between Temporal and Atemporal Structure. (= Cambridge Studies in Linguistics, 64) Cambridge: Cambridge University Press.

On the distribution of adjectives in Romanian The cel construction* Mihaela Marchis & Artemis Alexiadou Universität Stuttgart

This paper deals with the variable position of adjectives in the Romanian DP. As all other Romance languages, Romanian allows adjectives to appear in both prenominal and postnominal position. In addition, however, Romanian has a third pattern: the so-called cel construction, in which the adjective in the postnominal position is preceded by a determiner-like element, cel. This pattern is superficially similar to Determiner Spreading in Greek. In this paper we contrast the cel construction to Greek DS and discuss the similarities and differences between the two. We then present an analysis of cel as involving an appositive specification clause, building on De Vries (2002). We argue that the same structure is also involved in the context of nominal ellipsis, the second environment in which cel is found.

1.  Introduction: Adjectival modification and cel As is well known, a general characteristic of Romance languages is that in cases of ambiguous adjectives, prenominal placement of adjectives correlates with a strictly unambiguous interpretation, while postnominal placement correlates with ambiguity (Cinque 2005). This is illustrated below with a Romanian example with respect to the ambiguity between the restrictive vs. non-restrictive interpretation of adjectives. In (1a) the postnominal adjective is ambiguous between a restrictive and a non-restrictive interpretation, while in (1b) the prenominal adjective is unambiguously interpreted as non-restrictive.

*We would like to thank the participants of the Going Romance Conference in December 2007 in Amsterdam for their questions. Special thanks to two anonymous reviewers and to the editors of this volume for their comments and suggestions. Our research was supported by a DFG grant to the project C1: The syntax of nominal modification and its interaction with nominal structure, as part of the Collaborative Research Center 732 Incremental Specification in Context at the Universität Stuttgart.

 Mihaela Marchis & Artemis Alexiadou

(1) a.

Legile importante n-au fost votate (Romanian) laws.the important not-have been passed “The laws which were important were not passed.”

b. Importantele legi n-au fost votate Important.the laws not-have been passed “Important laws were not passed.”

1. all of them were important 2. #The laws which were important were not passed

As Cinque argues in detail, in this respect English and Romance are mirror images of one another, since in English it is the prenominal position which is ambiguous between the two interpretations, while the postnominal one is interpreted as strictly restrictive. (2) a. All of his unsuitable acts were condemned b. All of his acts were condemned; they were unsuitable c. All of his acts that were unsuitable were condemned (3) a. Every word unsuitable was deleted b. #Every word was deleted; they were unsuitable c. Every word that was unsuitable was deleted

In addition to (1), Romanian has a third pattern, illustrated in (4), where the demonstrative article cel follows the definite noun and precedes the adjective in postnominal position (see also Coene 1994; Cinque 2004; Cornilescu 2005). The word order in (4) is strict, i.e. no re-arrangement of the elements (i.e. the DP and the cel+Adjective sequence) is permitted: (4) Legile cele importante n-au fost votate. Laws.the cel important not-have been voted.

As (4) contains multiple determiners, it seems superficially similar to Greek Definiteness Spreading (DS), see (5b) (Androutsopoulou 1995; Alexiadou & Wilder 1998 and many others). The behaviour of adjectives in Greek is similar to that of the English ones in (2)–(3) above. In (5a) the adjective is ambiguous between a restrictive and a non-restrictive interpretation. In (5b), where the postnominal adjective is obligatorily preceded by a determiner, it is unambiguously interpreted as restrictive. (5) a. i simandiki nomi the important laws b. i nomi (afti) i simandiki the laws that the important

Unlike (1a), however, the adjective in (4) is not ambiguous: it is interpreted as restrictive, although it is found in postnominal position. In this respect (4) is again



On the distribution of adjectives in Romanian 

similar to DS in Greek. As just mentioned, DS also only involves a restrictive interpretation of the adjective (see Alexiadou 2001; Kolliakou 2004 and others for further discussion). We believe that the cel construction in Romanian represents the key element for understanding the syntax-semantics interface of postnominal adjectives in Romance languages. Therefore, in this paper we focus on the nature and distribution of cel by comparing this pattern with both Greek DS and postnominal adjectives in Romance languages. After introducing the distribution of cel, we compare the cel construction to Greek DS. We show that in spite of certain similarities between the two constructions, several differences exist, the main one being that cel has the status of a maximal projection. We then analyse cel strings as an appositive relative clause with a specification function, building on De Vries (2002, cf. Cinque 2004). The proposed analysis, departing from Kayne’s analysis of adjectives as involving reduced relative clauses, explains the differences between the cel construction and DS and argues that the same structure is also involved in the context of nominal ellipsis, the second environment in which cel is found.

2.  Descriptive facts about cel in Romanian 2.1  Historical development Historically, cel is derived from the Latin distal demonstrative ille preceded by the demonstrative adverb ecce. A number of phonological changes led to the presentday form:

(6) a Cl.Lat. ecce-ĭllum > Late Lat. eccíllum > eccéllum >16th cent. čelu >Mod. Rom. Cel (masc., sg.)

In Modern Romanian all forms contain an invariant ce and an -l morpheme which shows inflexion for gender, number and case and which is similar to the definite suffixal article -l (Coene 1994). This is shown in the two tables below:

Table 1.  the cel paradigm number

case

masculine

feminine

neuter

singular

N/A G/D N/A G/D

cel celui cei celor

cea celei cele celor

cel celui cele celor

plural

 Mihaela Marchis & Artemis Alexiadou

Table 2.  the definite article number

case

masculine

feminine

neuter

singular

N/A G/D N/A G/D

Băiat-ul Băiatu-lui Băieţi-i Băieţi-lor

Fat-a Fete-i Fete-le Fete-lor

Tablo-ul Tablou-lui Tablouri-le Tablouri-lor

plural

2.2  The distribution of cel in Romanian There are two main environments in which cel is found in Romanian: following a lexically expressed noun and in the absence of a lexically expressed noun. In both cases, it can precede adjectives, cardinals,1 PPs (but see Fn. 2), and superlatives. The former environment is restricted by two conditions. First, cel is an optional element. Second, the noun obligatorily takes a suffixal definite article when it precedes cel. This means that cel cannot occur in indefinite NPs (7b), in NPs with preposed demonstratives (7c), in front of the noun without an article (7d), but

1.  Both of our anonymous reviewers point out that the fact that cel can precede cardinals suggests that it has a head status, see (ib). But note that (ib) differs in interpretation from (ia): (i) a.

Am văzut trei băieti. (I).have seen three boys. “I randomly saw three boy, I don’t know who they are.”

b.

I-am văzut pe cei trei băieti. CL-have seen PE cel three boys “I saw those three boys, you know who I am talking about.”

First, (ib) has a restrictive interpretation and triggers the presupposition that the hearer knows about whom the speaker talks. Second, the presence of cel in (ib) is associated with the presence of transitive marker pe and clitic-doubling which reinforce the specific interpretation of the construction. The problem, however, is why cel can precede the cardinal in the presence of the noun instead of an adjective. Vulchanova & Giusti (1998:338) note that this is also the pattern we find with the demonstrative in Romanian: (ii)

aceste două femei frumoase these two women beautiful

Marchis (2007) argued that cel is inserted in Spec,DP in such cases to check definiteness, as the numeral itself cannot check this feature. We leave a full description and explanation of the distribution of numerals and demonstratives in Romanian for further research.



On the distribution of adjectives in Romanian 

it is fine in NPs with postposed possessives (preceded by N + suffixal definite article, (7e)): (7) a.

băiatul (cel) frumos boy.the cel beautiful “the boy cel nice”

b. *un om cel rău a man cel bad c. *acest om cel rău this man cel bad d. *cel cel e.

rău om bad man

fiul meu cel mic son.the my cel younger “my son, the young one”

The latter environment, where all elements preceded by cel agree in phi-features (gender, number, case) with it, has been described as a case of nominal ellipsis:2 (8) a.

cel frumos (adjectives) cel beautiful “the beautiful one”

b. cei mai frumoşi (superlatives) cei most beautiful

2.  As an anonymous reviewer correctly points out, the distribution of cel in the context of ellipsis is considerably more extended. The arguments of a head are excluded in the postnominal construction, but are ok in ellipsis: (i) a.

dependenţa de părinţi si apoi, cea de droguri. dependence.the of parents and then cel of drugs

b. *dependenţa cea de părinţi dependence.the cel of parents If we are right in analysing cel as part of an appositive relative clause, we can account for the ungrammaticality of (ib) as cel cannot intervene between the head and its complement. The grammaticality of (ia) is also covered by our analysis, as here cel behaves similarly to the demonstrative pronoun: c.

producţia de maşini şi aceea/ cea de avioane production.the of cars and that/ cel of aircrafts

 Mihaela Marchis & Artemis Alexiadou

The data in (7a) raise a legitimate question: does the cel construction involve DS? If so, cel should be analysed as a kind of determiner. In principle, such an analysis does not contradict the historical development of cel, described in Section 2.1. The development of definite articles out of demonstratives is a common path of grammaticalization (see Coene 1994; Giusti 2002). In the next sections we turn to a comparison of the two constructions.

3.  Similarities between the cel construction and Greek DS To begin with, the cel pattern, like Greek DS, does not give rise to adjectival ambiguity. We saw in (4) that no ambiguity is present in the cel construction. In (9) we provide further examples for Greek DS. As Kolliakou (2004) discusses in detail, (9a) is ambiguous between two readings: on reading 1, only the efficient researchers will be fired; on reading 2, the efficient researchers happen to be part of the larger group that will be fired. Crucially, (9b) is not ambiguous, it only has reading 1:3 (9) a.

o diefthindis ipe oti i kali erevnites tha apolithun the director said that the efficient researchers will be.fired

b. o diefthindis ipe oti i kali i erevnites tha apolithun the director said that the efficient the researchers will be.fired

Consider next the stage level vs. individual level ambiguity. As discussed in Cinque (2005), in Romance the prenominal adjective has an exclusive individual interpretation, while the postnominal adjective is ambiguous between an individual and stage level reading (10a–b). In English, on the contrary, it is the postnominal 3.  Our claim was that in Romanian (4) is not ambiguous. As Coene (1994) discusses in detail, adjectives that are exclusively non-restrictive can never be combined with cel. However, as correctly pointed out by an anonymous reviewer, there are occurrences of cel with proper names and adjectives expressing characterizing properties which are not restrictive: (i) Radu cel Frumos Radu cel handsome (ii) Admiram iarba cea verde I admired grass cel green Following Radatz (2001), we take (i) and (ii) to involve lexicalizations of a synthetic construction, hence they constitute a case which does not affect our main argument. Note that patterns like the one in (i) can also be found in a language like English e.g. Alexander the Great. We leave a full investigation of these constructions for further research.



On the distribution of adjectives in Romanian 

position in which the adjective is unambiguously interpreted as stage level. Importantly, in the cel construction the adjective has only the stage level reading, much like the postnominal adjective in Greek DS (10b)–(11b). In the prenominal position the adjective is ambiguous in Greek, similar to its English counterpart (see Campos & Stavrou 2004; Alexiadou 2006):4 (10) a.

Luminoasele stele sunt foarte îndepărtate “The stars, which are generally bright, are very far away.”

b. Stelele cele vizibile sunt foarte îndepărtate “The stars which are now visible are far away.” (11) a.

ta orata asteria ine poli makria the visible stars are far away



1. stars which are generally visible, are very far 2. stars, which happen to be visible now, are very far

b. ta asteria ta orata ine poli makria the visible the stars are very far

1. #stars, which are generally visible, are very far 2. stars, which happen to be visible now, are very far

Second, both constructions are similar in that they imply a contrast and are organized around the presupposition/focus distinction; the article-noun sequence constitutes the context or presupposition and the articled adjective constitutes the focus. This is shown in (12a–b) for Greek, see Kolliakou (2004), Campos & Stavrou (2004) and in (12c–d) for Romanian. In both languages the Det+N sequence can be omitted (a case of noun ellipsis); we note here that in Romanian when the noun is overt, pe is impossible, but we do not proceed to a discussion on the status of pe:

4.  As an anonymous reviewer points out, a different pattern is found in the case of participles. Here it is the participle with the habitual interpretation that is favoured, and not the one with the episodic interpretation: (i) a.

??cartea cea citită aseară book.the cel read last.night

b.

cartea cea mult citită book.the cel much read

The data point to the fact that more needs to be said about the type of participle involved (±eventive) and the reasons why cel can never combine with an eventive participle. The discussion will lead us too far afield, so we leave it for further research.

 Mihaela Marchis & Artemis Alexiadou

(12) a.

pia pena-ti hrisi i tin asimenia? (Greek) which pen-the golden or the silver

b. nomizo tin asimenia (tin pena) I.think the silver the pen c.

Pe care stilou l-ai cumpărat, (Romanian) PE which pen CL-have.you bought,



pe cel argintiu sau pe cel auriu? PE cel silver or PE cel golden?

d. Cred că pe cel argintiu. I think that PE cel silver.

Third, both constructions are illicit with non-intersective adjectives such as former (as known, these occur strictly prenominally in Romance): (13) a. *preşedinte-le cel fost (Romanian) president-the cel last b. *O monos tu o erotas ine i dulja tu. (Greek) the only his the love is the work his “His only love is his work.”

All of the above suggests that the two constructions should be analysed on a par with one another. Though several analyses of the DS pattern exist in the literature, we briefly focus here on the one put forth in Alexiadou & Wilder (1998), building on Kayne (1994). On this analysis, adjectives are generated as predicates within a reduced relative clause, and various movement operations guarantee the surface orders that are available in Greek (14): (14) a. [DP D [CP [IP DP AP]]]5 the the book red b. [DP D [CP AP [IP DP the red the book c. [DP DP D [CP AP the book the red

t ]]] [IP t t ]]]

According to Alexiadou & Wilder (1998), DS is a sub-case of indirect modification (Sproat & Shih 1988) and hence involves a relative clause analysis, see also recent

5.  As the extra determiner in DS does not introduce new reference, Alexiadou (2006) argues that extra determiners in DS are familiarity markers, much like the cases of object clitic doubling, which involves familiar definite NPs only, see Anagnostopoulou (1994):

(i)

DP1(familiarity/presupposition)[CP….[IPDP2/DefP(definite)]]]



On the distribution of adjectives in Romanian 

work by Cinque (2005).6 But can this analysis be used for Romanian cel? (see Cinque 2004 for a proposal along these lines, where he claims that cel is a realization of indirect modification). While most researchers (Campos 2005; Cornilescu 2005) agree that the cel pattern marks predicative adjectives in Romanian, it is not clear that cel has the same status as the definite determiner in Greek. For instance, Cornilescu (2005:9) analyses cel as a Pred head which selects a non-verbal phrase XP and marks it as a predicate. In this view, cel is a functional element which spells out a (+def) feature just like the definite article in Romanian.7 Obviously, if cel is a head in the spirit of Cornilescu (2005), Campos (2005) and Campos & Stavrou (2004), then the constructions with optional postnominal cel represent instances of DS. However, a number of properties suggest that the cel construction is not a case of DS, since it involves an XP.

4.  Differences between DS and the cel construction First, unlike Greek DS, the presence of cel is not obligatory, when the adjective occurs in postnominal position. When it is present, however, a semantic effect arises, described in Section 1. (15) a.

fluture-le (cel) frumos butterfly-the cel beautiful

b. to vivlio *(to) megalo the book the big

This is unexpected, as the definite article is not an optional element. Furthermore, the fact that the cel construction has different semantics from the postnominal position of adjectives in Romanian casts doubts on a possible analysis of the two 6.  Actually, Alexiadou & Wilder (1998) note that DS cuts across indirect and direct modification. Like the former it involves predicative adjectives, but like the latter it is subject to ordering restrictions. In syntactic analyses such as Alexiadou & Wilder’s and Cinque’s, a relative clause is involved only in indirect modification. For cases of direct modification, e.g. the former president, a different structure is needed, one in which the adjective is closer to the head noun, see the discussion in the aforementioned papers. On double source analyses of adjectives, the ambiguity found in (1a) is because in the post-nominal position, the adjective can have both structural base generated positions, i.e. both a reduced relative clause source and a direct modifier source. Movements of certain parts of the structure ensure the surface word order. 7.  A similar proposal is given in Campos (2005), who regards cel as a pseudo-article spelled out as a last resort operation in order to check phi-features and EPP.

 Mihaela Marchis & Artemis Alexiadou

patterns as involving similar structures. In this respect, cel resembles afti, the demonstrative pronoun that can optionally appear in DS in Greek: (16) a.

i pena (afti i asimenia (Greek) the pen this the silver

b. afto to vivlio this the book c.

fluturele cel frumos (Romanian) butterfly.the cel nice

d. fluturele acela butterfly.the that

Second, only one cel can appear per DP; the definite determiner in DS can have multiple occurrences, if a further adjective is present: (17) a.

fluturele cel frumos, *(cel) colorat si *(cel) zglobiu butterfly.the cel beautiful, colourful and lively “the beautiful, colored and lively, butterfly”

b. i petaluda i orea i polihromi the butterfly the beautiful the colourful

Third, cel does not only precede adjectives but also PPs or other modifiers; this is not possible for Modern Greek DS, though such orders were possible in Classical Greek (18b–c):8 (18) a.

casa cea de piatră house.the cel of stone

b. *to spiti to apo petra the house the from stone c.

oi anthropoi oi para sou (Classical Greek) the people the near you

Fourth, as already mentioned, the word order is strict, cel always follows the definite noun. Greek DS, however, permits word order permutations: (19) a.

to vivlio to kokino the book the red

b. to kokin to vivlio the red the book

8.  As an anonymous reviewer notes, the fact that in Classical Greek articles could introduce PPs brings these constructions closer to the cel patterns, and suggests that the definite article in Greek has undergone a full change, while cel is in a case of transition.



On the distribution of adjectives in Romanian 

Finally, cel behaves like the demonstrative acest, in that it is able to have a deictic/ anaphoric interpretation. According to Giusti (2002), the enclitic definite article in Romanian is a mere functional category with no semantic content. Giusti (2002) argues that D, the locus of the enclitic definite article, is instantiated for syntactic reasons whereas the specifier of the DP is occupied by the demonstrative which has a semantic value. This property is triggered only by those categories in Spec,DP and is crucial for the interpretation of the referential index of the noun phrase: (20) a.

Directorul de departament şi preşedintele de facultate Director.the of department and president.the of faculty



a venit/ au venit. has come/have come.

b.

Acest This au have

director de departament şi acel preşedinte de facultate director of department and that president of faculty venit. come.

The examples in (20) show that the double definiteness within the nominal phrase does not necessary lead to two referents, the construction is ambiguous between one and the same referent and two different individuals. Unlike the definite article, the demonstrative in Romanian triggers not only reference but also deictic interpretation. The complementary distribution between the definite article and the demonstrative can be explained by Giusti’s (2002) doubly-filled DP Filter. Importantly, cel shows the same syntactic and interpretive effects as the demonstrative from which it originates: (21)

fluturele cel albastru si cel roz au zburat Butterfly.the cel blue and cel pink have flown.

Unlike the definite article, cel introduces reference to the nominal phrase and apart from that, it triggers a restrictive meaning as well. As a result, the behaviour of cel is very unlike Greek DS where the presence of multiple determiners does not create independent reference. The general consensus in the literature is that demonstrative elements are XPs located in Spec,DP; they are relevant for the interpretation of the referential index of the noun phrase.9 This would mean that cel is also an XP element.10 We thus conclude that the cel construction is very unlike Greek DS.

9.  See Dobrovie-Sorin (2000:4) who argues that cel is a maximal projection in Spec,DP. 10.  Dobrovie-Sorin (1987) proposes that cel is a phrasal element, pretty much like French celui. The context of cel is larger than that of celui. See the appendix for discussion.

 Mihaela Marchis & Artemis Alexiadou

5.  The analysis of the cel construction11 As already mentioned, both in the presence of a noun (7) and in the context of ellipsis (8), cel can occur only with predicative/intersective adjectives. Our analysis must capture this behaviour. If the source of predicative adjectives is within a relative clause structure, then such a structure should also be present in the cel construction and be involved in both environments where cel appears. We have shown, however, that the cel-construction differs from DS in important respects. Moreover, the cel-construction cannot be analysed on the basis of the structure proposed in e.g. Cinque (2005) for the restrictive reading of postnominal adjectives, which is again a reduced relative clause with the adjective in predicative position, though generated in prenominal position. In this case the noun would have to move over the relative clause in order to combine with the (suffixed) article. Importantly, recall that cel accepts only one adjective, while in the other two constructions, DS and postnominal adjectives in Romance, we can have more than one adjective. Hence we cannot make use of base structures of this kind. Recall further that cel is available with all kinds of modifiers, adjectives as well as PPs (Cornilescu 2005:18). This behaviour in connection to its XP status constitutes important clues in understanding the role of cel. We believe that cel introduces a reduced relative clause which has a specifying function, rendered in English via the use of e.g. the adverb namely in (22): (22) the horse, namely the beautiful one

In other words, the cel construction is an appositive specification clause, in the form of a reduced relative clause.12 The analysis of cel as introducing a relative clause is supported by the fact that it can generally be found in relative clauses (23): (23)

băiatul cel care a plecat boy.the cel who left

In (23) cel introduces the relative clause. In the typological literature it is also widely known that demonstratives, which cel was originally, have often grammaticalized as subordinate, including relative, complementizers in a number of language groupings around the world, including Germanic. 11.  The analysis proposed here differs from the one presented in Marchis & Alexiadou (2007), where we attempted to make use of Kayne’s analysis of celui for cel. In the appendix, we discuss some differences and similarities between cel and celui. 12.  Lekakou & Szendroi (2007) analyse DS as a form of apposition, which crucially involves adjunction. We argue here that DS involves restrictive clause modification, while the cel construction involves a specification clause.



On the distribution of adjectives in Romanian 

Building on the above, we claim that cel clauses are appositive specification relative clauses. Discussion of the syntax of such clauses is found in De Vries (2002), upon which we build here. As De Vries (2002:211ff.) discusses in detail, restrictive appositions do not only involve adjectives, but can also include PPs. In his analysis, appositive relatives involve specifying co-ordination. Semantically, in such a clause the second element specifies the first, and necessarily the second is a logical subset of the first element. The semantics crucially involves set intersection, giving thus the restrictions on the adjectives presented here. In addition this structure explains the fact that the adjective is never ambiguous, it is always generated in the predicative position of the relative clause. It also explains why it cannot be iterated, since the NP can receive only one specification. The relevant structure is as in (24). (24) [[&:P [XP ] &: [DP [CP [C’ [IP ]]]]

As in the languages De Vries discusses, and in the case of Romanian, a relative clause must be introduced by a relative element, we propose that cel is this element. On the basis of structure (24), there are two options for cel: either it is located in Spec,CP in a state of reanalysis from a demonstrative pronoun to a relative marker, see also the references in De Vries (2002), or it is placed in Spec,DP (cf. Cornilescu’s claim that cel is a predicative head). This analysis is faithful to the fact that cel is a complex phrasal determiner (pronoun-determiner) which contains an invariant form ce and a morpheme similar to the suffixal definite article (Section 2.1). We come back to that in the following paragraphs. If we are right in analyzing cel clauses as appositive clauses with a specification function, this explains the strict word order associated with the pattern as well as the single occurrence of cel, and also the differences between the cel construction and DS. Since there can only be one specification, no iteration is allowed. The structure of (25a) is given in (25b). (25) a.

băiatul cel frumos

b. [&:P [DP baiatul]i &: [DP celi [CP [C’ [IP Øi AP]]]]

The adjective is generated in the predicative position within the relative clause. The second conjunct is co-indexed with the first one, hence they both agree in all features and have unique reference. As mentioned above, several options are available for cel: It can either be inserted in DP or be moved there from inside the clause. To the extent that it moves via Spec,CP to Spec,DP again either the subject of the small clause is a pronominal zero element co-indexed with cel or pronominal cel is generated as the subject of the small clause. An anonymous reviewer suggested that there is a way

 Mihaela Marchis & Artemis Alexiadou

to choose between the different analyses, and provided evidence against a raising analysis of cel. This comes from the observation that cel agrees in Case with the antecedent. In addition cel can co-occur with wh-pronouns in Spec,CP, see (23) and (26) below, suggesting that it is actually external to the clause: (26) a.

fiul eroului celui/?cel proaspăt decorat son.the hero.gen cel.gen/cel.nom recently decorated “son of the recently decorated hero”

b. omul cel pe care nimeni nu-l mai iubeşte man.the cel PE whom nobody any more loves “the man who nobody loves anymore”

We now turn to the presence of cel in the context of nominal ellipsis, where it precedes APs. The relevant example is repeated below: (8) cel frumos cel beautiful “the beautiful one”

As already mentioned, the same restrictions on the adjectives apply to the ellipsis environment, i.e. the adjective has to be an intersective, predicative one. This is illustrated in (27) below:13

13.  An anonymous reviewer points out that ellipsis is more encompassing than the cel construction. Note that both classifying and thematic adjectives are out with cel, but occur in the ellipsis case. Such adjectives are sub-types of so called denominal relational adjectives. Classifying adjectives (sub)classify the denotation of the noun they modify and modify common nouns, while thematic ones fill a thematic slot in the theta-grid of the noun they modify and modify derived nominals (see Bosque & Picallo 1996): (i) a.

tragedia ˉclasică/*tragedia cea clasică tragedy classic/ˉ  tragedy celˉ classic

b.

tragedia tragedy

clasică si (nu) cea romantică classic and not cel romantic

(ii) a. b.

Atacul attack.the atacul attack.the

american/ * atacul cel american American/ *attack.the cel American american si nu cel sovietic American and not cel Soviet

We expect the ungrammaticality of (ia) and (iia) in the cel pattern, as such adjectives are not predicative. Concerning the grammaticality of (ib) and (iib) we note that the adjective has a contrastive reading. We further feel that it receives a slightly different interpretation from the patterns in (a), thus making a predicative analysis of the adjectives viable.



On the distribution of adjectives in Romanian 

(27) a. *L-am văzut pe preşedintele cel fost. PRO.CL-have seen PE president.the cel former. b. *L-am văzut pe cel fost. PRO.CL-have seen PE cel.masc former.

In order to capture these similarities, we propose that (25b) is involved also in the context of ellipsis. This is supported by the fact that the English translation contains the pronoun one in both (25b) and (8), the form English generally uses in ellipsis. When the first conjunct has previously been mentioned in the discourse it can be elided; ellipsis is licensed as the remaining part is informative/contrastive (see Section 3). (28) [&:P [DP băiatul]i &: [DP celi [CP [C’ [IP Øi frumos]]]]

6.  Summary and conclusions Although the distribution of cel in the presence of an N in Romanian is superficially similar to the phenomenon of DS in Greek, the syntactic differences between DS in Greek and cel in Romanian lead us to the conclusion that cel is not a definite article and, consequently, the optional cel construction is not a case of DS in Romanian. In our analysis, cel is a complex phrasal determiner (pronoun-determiner) which contains an invariant form ce and a morpheme similar to the suffixal definite article. Syntactically, cel is an element located in Spec,DP in a state of reanalysis from a demonstrative pronoun to a relative marker, which introduces an appositive specification clause (see also the references in De Vries 2002). The proposed analysis, departing from Kayne’s analysis, explains the differences between the cel construction and DS and argues that the same structure is also involved in the context of nominal ellipsis, the second environment in which cel is found.

Appendix: cel and celui Dobrovie-Sorin (1987) and Cornilescu (2005) discuss certain distributional differences between cel and celui. Importantly, celui cannot be followed by bare adjectives, it can precede present and past participles as well as PPs: (29) a.

celui + full relative



Celui que tu as lu the.one that you have read

 Mihaela Marchis & Artemis Alexiadou

b. Celui + past participle

Celui envoyé à Jean the.one sent to Jean

c.

Celui + present participle



Ceux parlant quatre langues those speaking four languages

d. Celui + PP

Celui de Jean that of John

e.

Celui + complement taking AP



ceux contents de leur sort those happy with their kind

f. *celui + Bare AP

*celui rouge this red

Kayne (1994:101) points out that celui is not an N0, rather it is an XP composed of ce+lui. This is similar to what we said about Romanian cel. But one cannot analyse celui similarly to cel, as (29f) is out. While we do not have a detailed analysis of the above pattern, we would like to mention here that the bare adjectives in French seem to have a special status. For instance, Ronat (1974, 1977) in her study of noun ellipsis in French makes a distinction between intransitive, i.e. bare, adjectives and transitive ones, i.e. adjectives followed by complements. While the transitive ones are reduced relative clauses, the intransitive ones have a very different analysis (see also Sleeman 1996 for further discussion). In particular, Sleeman (1996:55), citing joint work by Sleeman & Verheugd, states that bare adjectives in French can only be attributes within the extended projection of the NP. The ungrammaticality of (29f) is related to the fact that intransitive adjectives cannot appear in predicative position in French reduced relatives. If correct, this raises questions as to the general analysis of bare adjectives in French, as bare adjectives can appear in postnominal position: (30)

le livre jaune the book yellow

On Kayne’s analysis, (30) involves a reduced relative clause. On the basis of the remarks in Sleeman, however, (30) should be rather analysed as involving an adjective within the extended projection of NP, which is the analysis Sleeman herself gives. See also the discussion in Alexiadou & Wilder (1998), who pointed out that adjectives are not always amenable to a reduced relative clause basic construction.



On the distribution of adjectives in Romanian 

References Alexiadou, Artemis. 2001. “Adjective Syntax and Noun Raising. Word Order Asymmetries in the DPs as the Result of Adjective Distribution”. Studia Linguistica 55. 217–248. Alexiadou, Artemis. 2006. “On the Cross-linguistic Distribution of (In)definiteness Spreading” Ms., University of Stuttgart. Alexiadou, Artemis & Chris Wilder. 1998. “Adjectival modification and multiple determiners”. Possessors, Predicates and Movement in the DP ed. by Artemis Alexiadou & Chris Wilder, 303–332. Amsterdam: John Benjamins. Anagnostopoulou, Elena. 1994. Clitic Dependencies in Modern Greek. Ph.D. dissertation, University of Salzburg. Androutsopolou, Antonia 1995. “The licensing of Adjectival Modification”. Proceedings of WCCFL 13. 17–31. Bosque, Ignacio & Carmen Picallo 1996. “Postnominal Adjectives in Spanish DPs”. Journal of Linguistics 32. 349–385. Campos, Hector, 2005. “Noun Modification, Pseudo-articles, and Last Resort Operations in Arvantovlaxika and in Romanian”. Lingua 115. 311–347. Campos, H. & M. Stavrou. 2004. “Polydefinite Constructions in Modern Greek and in Aromanian”. Topics in Balkan Sprachbund Syntax and Semantics ed. by Olga Tomic, 147–189. Amsterdam: John Benjamins. Cinque, Guglielmo. 2004. “A Phrasal Movement Analysis of the Romanian DP”. Studia Linguistica et Philologica in Honorem D. Irimia ed. by Ana-Maria Minut & Eugen Munteanu, 129–142. Iasi: Editura Universitatii ‘Al.I. Cuza’. Cinque, Guglielmo. 2005. “The Dual Sources of Adjectives and XP vs. N-raising in the Romance DP”. Ms., University of Venice. Coene, Martine. 1994. “Adjectival Articles” Ms., Antwerp University. Cornilescu, Alexandra. 2005. “Modes of Semantic Combinations: NP/DP-adjectives”. Ms., University of Bucharest. Dobrovie-Sorin, Carmen. 1987. Syntaxe du Roumain. Ph.D. dissertation, Université de Paris. Dobrovie-Sorin, Carmen 2000. “Indefiniteness Spread: from Romanian to Hebrew Construct State”. Comparative Studies in Romanian Syntax ed. by Virginia Motapanayane, 177–226. Oxford: Elsevier. Giusti, Giuliana. 2002. “The Functional Structure of Noun Phrases: A bare phrase structure approach”. Functional Structure in DP and IP ed. by Guglielmo Cinque, 54–99. New York: Oxford University Press. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press. Kolliakou, Dimitra. 2004. “Monadic Definites and Polydefinites: Their form meaning and use”. Journal of Linguistics 40. 263–333. Lekakou, Marika & Krista Szendroi. 2007. “Eliding the Noun in close Apposition, or Greek Polydefinites Revisited”. Paper presented at the MIT Workshop on Syntax and Semantics, May 2007. Marchis, Mihaela. 2007. The Semantics and Syntax of Adjectives in Spanish and Romanian. M.A. Thesis, Universität Stuttgart. Marchis, Mihaela & Artemis Alexiadou. 2007. “On the Properties of the cel Adjectival Construction in Romanian”. Paper presented at the Going Romance 2007 conference, Amsterdam. Radatz, Hans-Ingo. 2001. Die Semantik der Adjektivstellung: eine Kognitive Studie zur Konstruktion ‘Adjektiv + Substantiv’ im Spanischen, Französischen und Italienischen. Tübingen: Niemeyer.

 Mihaela Marchis & Artemis Alexiadou Ronat, Mitsou. 1974. Echelles de Base et Mutation en Syntaxe Française. M.A. thesis, Vincennes University, Paris. Ronat, Mitsou. 1977. “Une Contrainte sur l’Effacement du Nom”. Langue ed. by Mitsou Ronat, 153–169. Paris: Hermann. Sleeman, Petra. 1996. Licensing Empty Nouns in French. Ph.D. dissertation, University of Amsterdam. Sproat, Richard & Chilin Shih. 1988. “Prenominal Adjectival Ordering in English and Mandarin”. Proceeding of NELS 12. 465–489. Vries, Mark de. 2002. The Syntax of Relativization. Ph.D. dissertation, University of Utrecht. Vulchanova, Mila & Giuliana Giusti. 1998. “Fragments of Balkan Nominal Structure”. Possessors, Predicates and Movement in the DP ed. by Artemis Alexiadou & Chris Wilder, 333–360. Amsterdam: John Benjamins.

Subject doubling in European Portuguese dialects The role of impersonal se* Ana Maria Martins University of Lisbon

This paper discusses a subject doubling construction found in European Portuguese dialects where the impersonal clitic se is doubled by a strong pronoun/DP. The availability of subject doubling is explained under the hypothesis that dialectal se escapes the Case filter because it is ϕ-incomplete (plural, but person-less). Se and its doubler begin as a single constituent (a “big DP”); later, the clitic se head-moves to T while the doubling pronoun/DP takes one of the positions available to EP subjects (preverbal or post-verbal). This analysis rightly predicts that: (i) there is no fixed word order between se and its doubler, (ii) the doubling pronoun/DP controls subject-verb agreement, (iii) the interpretation of the double subject is compositional (the doubler establishes the inclusive or exclusive reading of impersonal se). The analysis also enlightens why the clitic se can behave as the universal plural anaphora or the expletive subject of impersonal predicates in some dialects.

1.  Introduction European Portuguese (EP) dialects display a double subject construction where the impersonal clitic se shares with a (plural) strong pronoun, or less commonly a full DP, the role of sentential subject. In the illustrative sentences in (1) the subject clitic se is doubled by a gente, a former DP (“the people”) grammaticalised as a first person plural pronoun (see Menuzzi 2000; Lopes 2003). In the English translations of the examples in (1), we is to be understood as non-specific, roughly, “people here, including myself ”.

*My thanks to Jairo Nunes, an anonymous reviewer and the audience of Going Romance 2007 (Amsterdam) for comments and helpful discussion of issues addressed herein.

 Ana Maria Martins

European Portuguese dialects: (1) a.

A gente chama-se rãs a isto.  pron.1pl calls-SE rãs to that “We call these ones rãs (frogs).”

b. Chama-se-lhe a gente espigas. calls-SE-it.dat pron.1pl espigas “We call it espigas (spikes of cereal).”

(CORDIAL-SIN.1 FLF)

(CORDIAL-SIN. AAL)

Standard European Portuguese does not allow the double subject construction exemplified in (1). Thus while a gente and se can separately act as an indefinite subject, they are not allowed to co-occur in that role. See the paradigm in (2). Standard European Portuguese: (2) a. *Aˉgente chama-se rãs a isto. pron.1pl calls-SE rãs to that “We call these ones rãs (frogs).”

(ˉ≠ˉEPˉdialectsˉunderˉdiscussion)

b. Aˉgente chama rãs a isto. pron.1pl callsˉ rãs to that “We call these ones rãs (frogs).”

(ˉ=ˉEPˉdialectsˉunderˉdiscussion)

c.

Chama-se rãs a isto. calls-SE rãs to that “We call these ones rãs (frogs).”

( = EP dialects under discussion)

The aim of this paper is to offer an analysis of the double subject construction exemplified in (1),2 and so clarify what makes it available in the EP dialects under observation, while standard EP (like standard Spanish or Italian) rejects it. The facts are interesting from a theoretical perspective since within the generative

1.  CORDIAL-SIN: Syntax-oriented Corpus of Portuguese Dialects. See http://www.clul.ul.pt. 2.  The double subject SE construction is found in the archipelagos of Azores and Madeira as well as in continental Portuguese. It is much more common in the Centre and South of Portugal than in the North (where nonetheless it is also attested). It is fully ungrammatical in standard EP and has gone totally unobserved by philologists and linguists who dealt with dialect variation in European Portuguese. All the dialects where the double subject SE construction is found also display the impersonal nominative SE construction of standard European Portuguese, which is similar to the “non-agreeing” SE/SI construction of Spanish and Italian (see Cinque 1988; Dobrovie-Sorin 1998; Rivero 2000; and D’Alessandro 2004, among others). The data offered in this paper come from CORDIAL-SIN (Syntax-oriented corpus of Portuguese dialects).



Subject doubling in European Portuguese dialects 

framework, on which this work hinges, sentences with two subjects (neither of which is left dislocated) should, in principle, be disallowed.3 The paper is organized in seven sections. Section 2 deals with word order. The fact that there is no fixed ordering between the two elements that express the subject makes previous analyses of subject doubling unfit to characterize the dialectal EP construction. Section 3 describes agreement properties of the construction showing that subject-verb agreement does not depend on impersonal se but is set by the doubling nominative pronoun or DP. In Section 4, the interpretative effects of subject doubling in the impersonal se construction are considered. It is proposed that the double subject has a compositional interpretation with the strong pronoun or DP acting as a restrictor on the denotative set of impersonal se and establishing its inclusive or exclusive reading. Section 5 draws an integrative analysis of the syntactic and semantic facts introduced in

3.  The type of se under discussion in this paper differs from the aspectual reflexive found in Spanish (Strozer 1976; Zagona 1996; De Miguel 1999; Otero 1999) in that it is compatible with intransitive (unergative) verbs and is not restricted to sentences with perfective predicates. See (i) below, with aspectual se, in contrast to the examples above. (i) a.

Juan (*se) come normalmente en JuanˉSE eats usually in



este bar. this bar



“Juan usually eats in this bar.”

b.

Juan (*se) come tortilla siempre que puede. (idem) Juan SE eats omelette always that can.3sg “Juan eats omelette whenever he can.”

c.

Juan (se) comió una tortilla él solo. (idem) Juan SE ate.3sg an omelette he alone “Juan ate an omelette all by himself.”

(Spanish.ˉFromˉDeˉMiguelˉ1999:2995)

It also differs from both the aspectual reflexive and the benefactive/malefactive dative (Strozer 1978, among others), exemplified in (ii), as it surfaces exclusively with the form se while the aspectual and the benefactive/malefactive clitics display person distinctions (e.g. me.1SG, te.2SG, etc.).

(ii) Pepe me

le comió la manzana al niño. Pepe me.dat him.dat ate the apple to the.boy “Pepe ate the boy’s apple.”

Moreover, the dative clitic is never phonologically realized as se in European Portuguese. In this respect standard EP and dialectal EP are alike.

 Ana Maria Martins

the previous sections. The compositional interpretation is derived under the hypothesis that the double subject results from the splitting of an earlier single constituent, a complex DP, where se is the nuclear element. The particular property of se that makes it irrelevant for subject-verb agreement and Case is taken to be its lack of person. Hence nominative Case rests available for the doubling pronoun or DP. In Section 6 independent evidence is brought into consideration in order to support the approach developed in the paper. It is shown that the very same dialects where the double subject se construction is attested display other peculiar features pointing to the defectiveness of se with respect to ϕ-features, in sharp contrast with what is observed in standard EP.

2.  Word order as evidence for different types of subject doubling In the double subject se construction, the two subject items do not appear in a fixed order. The subject clitic se may follow or precede the doubling strong pronoun/ DP, as illustrated by the root sentences in (1) – here repeated as (3) – and by the embedded clauses in (4).4 (3) a.

A gente chama-se rãs a isto. pron.1pl calls-SE rãs to that “We call these ones rãs (frogs).”

b. Chama-se-lhe a gente espigas. calls-SE-it.dat pron.1pl espigas “We call it espigas (spikes of cereal).”

(CORDIAL-SIN. FLF)

(CORDIAL-SIN. AAL)

(4) a.

Este pode ser a coisa que a gente se this can be the thing that pron.1pl SE



diz peixe-cavalo. says peixe-cavalo



“This one could be what we call peixe-cavalo (a kind of fish).”

(CORDIAL-SIN. CLC)

b. O muge é o.que se chama aˉgente quando anda dentro the muge is what SE calls pron.1pl when is inside

do rio. of.the river



“Muge is the name we give to that kind of fish when it is found in the river.”

(CORDIAL-SIN. ALV)

4.  I take the preverbal non-clitic subject in sentences like (1) to sit in Spec,ΣP/SubjP (Martins 1994; Cardinaletti 1997, 2004 respectively), a position of the high IP field. The post-verbal non-clitic subject in sentences like (2) stays in Spec,vP.



Subject doubling in European Portuguese dialects 

This variable word order sets the EP double subject se construction apart from other cases of subject doubling reported in the literature and is left unaccounted for by previous analyses. The construction falls outside the scope of the analyses that handle subject doubling as a case of multiple realisation of copies of a movement chain (Craenenbroeck & Van Koppen 2002; Van Koppen 2007, for Dutch; Holmberg & Nikanne 2006, for Swedish). It is also not covered by the analyses that derive subject doubling from multiple external merges of distinct lexical items in different subject positions made available by clause structure (Holmberg & Nikanne 2006, for Finnish). The sentences in (5), taken from Van Koppen (2007), illustrate subject doubling in Wambeeck Dutch. Here the strong pronoun or DP, functioning as subject, necessarily precedes the co-referring weak pronoun. The Wambeeck Dutch construction is analysed by Craenenbroeck & Van Koppen (2002), and Van Koppen (2007) as an instance of A’-movement coupled with multiple spell-out of chain links (see Nunes 2004; Corver & Nunes 2007). The subject moves from Spec,IP to a topic position in the CP-domain. Concomitant spell-out of the higher and lower copies (in Spec,TopP and Spec,IP respectively) is an available option because the head of chain differs from the other copies in that the latter do not contain internal structure.5 (5) a.

Zaai gui zaai nuir ojsh. she goes she to home “She is going home.”

b. Dei vrou gui zaai nuir ojsh. that woman goes she to home “That woman is going home.”

(Wambeek Dutch)

(Wambeek Dutch)

This type of subject doubling clearly differs from what is found in European Portuguese dialects.6 That the doubling pronoun/DP in the EP impersonal se construction is not a topic is clearly shown by the fact that it may surface in postverbal position (see (3b), (4b)), being in this case preceded by the impersonal subject clitic. Another type of subject doubling was reported in the literature where the weak subject pronoun precedes the subject DP or strong pronoun. This is the Finnish pattern (see examples (6a–b)), which Holmberg & Nikanne (2006) analyse as represented in (7).

5.  According to Van Koppen (2007), lower copies are reduced in the sense that they only contain the features of the maximal projection of the moved item (the reduced copy theory). 6.  The Wambeek Dutch type of subject doubling is also found in Swedish (see Holmberg & Nikanne 2006) and does not seem to significantly differ from Brazilian Portuguese or French left dislocation of the subject with resumption.

 Ana Maria Martins

(6) a.

Se on Jari lopettanut tupakoinnin. he has Jari quit smoking “Jari has quit smoking.”

b. Ne sai kaikki lapset samat oireet. they got all children same symptoms “All the children got the same symptoms.” (7) [FinP se [on+Fin [PredP SE has “Jari has quit smoking.”

Jari lopettanut tupakoinnin]]] Jari quit smoking

According to Holmberg & Nikanne (2006), Finnish clause structure makes two positions available for subjects: a position inside the information Focus domain, that is PredP, and a position inside the Presupposition domain, that is FinP. In the Finnish double subject construction, the subject is interpreted as information focus because it remains inside PredP. The doubling weak pronoun directly merges in Spec,FinP checking the EPP.7 According to Holmberg & Nikanne (2006), the fact that Finnish has two pronouns with number but no person (i.e. se-SG and nePL) is crucial to account for the availability of subject doubling: Why is subject doubling of the Finnish kind not found in all languages? Presumably either because the other languages lack the requisite categories, i.e. deficient pronouns, or because they lack the requisite ‘structural positions’ (Holmberg & Nikanne 2006)

As for the EP double subject impersonal se construction, given the non rigid order of the two subject items, it is not reducible to the Finnish type of subject doubling. Nevertheless, it will be proposed in Section 5 that the idea that person-less pronouns lie behind subject doubling is to be maintained. The EP double subject impersonal se construction thus lends further empirical support to the proposals in Holmberg & Nikanne (2006).

7.  The weak pronoun can only follow the subject DP if the latter is A’-moved to the CP domain because it is a contrastive focus or a wh- phrase, as illustrated below. (i) [CP Sinäkini [FinP se [olet+Fin [PredPˉSinäkinˉi vihdoinkin you-too SE have finally

lopettanut tupakoinnin]]]] quit smoking

“Soˉyou,ˉtoo,ˉhaveˉfinallyˉquitˉsmoking.”ˉ(contrastiveˉfocusˉonˉSinäkin)

(ii) [CP kukai [FinP se [on+Fin [PredP kukai vihdoinkin lopettanut tupakoinnin]]]] who SE  has finally quit smoking?



Subject doubling in European Portuguese dialects 

3.  Agreement in the double subject impersonal se construction In the double subject impersonal se construction, the verb displays agreement with the strong doubling pronoun or DP, as shown in (8) below, where the verb agrees with the nominative first person plural pronoun nós “ we”. Sentences (9a) and (9b) have the first person plural pronoun a gente (literally, “the people”) in subject position on a par with se. The pronoun a gente induces variable patterns of subject-verb agreement (see fn. 15), which allows the verb to surface in 3rd person singular (see (1) above), in 1st person plural (see (9a) below), and in 3rd person plural (see (9b) below). The two latter options make subject-verb agreement visible. (8) Há várias qualidades que até ainda nós não there.is several species that even already we not se conhecemos. SE know.1pl

(CORDIAL-SIN.ˉALV)

“There are so many species among fish that even we (fishermen) do not know all of them yet.” (9) a.

Não sabem o.que a gente not know.3pl what pron.1pl

se SE



passámos aí. passed.1pl there



“You don’t know all we’ve been through.”

b. A tesoura que a gente se tosquiavam the scissors that pron.1pl SE shear.3pl

A gente pegavam-lhe pron.1pl would.hold.3pl-them.dat



pela cabeça…  by.the head

(CORDIAL-SIN. CLC)

as ovelhas! the sheep!

(CORDIAL-SIN. MIG)

“This is the scissors that we would shear the sheep with! We would hold their head…”

Moreover, a null pronoun may be the doubler of se and control subject-verb agreement.8 In the sentences in (10), a 1st person plural null subject pronoun is responsible for the 1st person plural morphology on the verb. (10) a.

Nunca se vimos este peixe. never SE see.1pl this fish “We never saw that kind of fish.”

(CORDIAL-SIN. CLC)

8.  European Portuguese is uniformly a null subject language; no dialectal variation is found in this respect.

 Ana Maria Martins

b. Com a licença de caça e tudo, não se with the licence of hunting and all not SE

pode mos caçar…  can.1pl hunt.INFIN



“Even having a hunting licence, we are not allowed to hunt…”

(CORDIAL-SIN. EXB)

These same dialects display non-doubling impersonal se sentences similar to those found in standard EP, Spanish or Italian. In this case the verb invariably surfaces in third person (singular or plural), like in (11) below.9 The fact that in the double subject impersonal se construction the verbal inflection always has its ϕ-features valued by the doubling strong pronoun or DP (see the ungrammaticality of (12)) shows that se is unable to enter the kind of Agree relation on which nominative Case assignment is dependent (see Chomsky 2000, 2001). It will be proposed in Section 5 that this is because the impersonal clitic se is person-less in the relevant EP dialects. (11) Há várias qualidades que ainda não se conhece(m). there.is several species that already not SE know.3sg(/3pl) “There are many species among fish that are still unknown.” (12) *Há várias qualidades que até ainda nós não se conhece(m). there.is several species that even already we not SE know.3sg(/3pl) “There are so many species among fish that even we (fishermen) do not know all of them yet.”

4.  Interpretative effects in the double subject impersonal se construction The bipartite subject of the dialectal impersonal se construction usually has a generic reading.10 The impersonal subject denotes an unspecified group of

9.  On this matter, see Burzio (1986), Raposo and Uriagereka (1996), Rivero (2000) and D’Alessandro (2004). 10.  A definite reading is also attested in the corpus but appears to be infrequent. We find but the two examples given below: (i) Então, aˉgente deu-se o jantar e then pron.1pl gave.3sg-SE the dinner and

ficou melhor. stayed better



“Then we offered that dinner to our neighbours and things started to go better.” [‘we’ = ‘our family’]

(CORDIAL-SIN.ˉFLF)



Subject doubling in European Portuguese dialects 

humans, which may include or exclude the speaker. In this respect, it does not differ from the standard impersonal se construction of EP, Spanish and Italian (see the references in fn. 2 and Chierchia 1995). But the doubling pronoun/DP acts as a restrictor on the denotative range of se, and establishes its inclusive or exclusive reading. 4.1  Inclusive reading (the doubling pronoun/DP is 1st person plural) The inclusive reading, which sentences (13) and (14) illustrate, arises when the doubling pronoun/DP is 1st person plural, be it a null or an overt pronoun, as shown in (13) and (14) respectively. These and the other examples throughout the paper come from the CORDIAL-SIN dialectal corpus (see fn. 1) and arose in the context of spontaneous conversation. In (13) we can see how the speaker reformulates the sentence to make clear that he is mentioning what local people, including himself, used to do, not what anyone would do. This is unambiguously conveyed by the double subject impersonal se sentence, not by the standard impersonal se sentence that the speaker uttered in first place. Note that the 1st person plural agreement morphology on the verb indicates the presence in the relevant sentence of a null pronoun with those same ϕ-features. (13) Sabe às.vezes o.que se faz, o.que se know sometimes what SE does, what SE fazíamos, antigamente? would.do.1pl, in.older.times

(CORDIAL-SIN. OUT)

“Do you know what one would do, what we would do, in older times?”

In (14) the contrast between the fishermen of Câmara de Lobos (Madeira island) and the inhabitants of Lisboa with respect to their eating habits can be set with no need for a locative argument in the first member of the coordinate structure because the presence of the 1st person plural pronoun a gente is enough to restrict

(ii)

Leis! Como a gente se falámos laws as pron.1pl SE talked.1pl



ainda agora. just now



“That’s what laws are made for, as we have just talked about.” [‘we’ = ‘the four of us who are here talking about this and that’]

(CORDIAL-SIN. CLC)

 Ana Maria Martins

the range of the group of humans denoted by se, and thus establish its inclusive reading.11 (14) A gente pron.1pl

não se come, mas os de Lisboa diz que comem not SE eats but the.ones of Lisbon says that eat.3pl

daquele peixe. of.that fish

(CORDIAL-SIN. CLC)

“Here we don’t eat that fish but we heard that in Lisbon people eat it.”

4.2  Exclusive reading (the doubling pronoun/DP is 3rd person plural) When an exclusive reading of se is intended, this is obtained by placing a 3rd person plural pronoun (overt or null) as the doubler/restrictor of se,12 as illustrated by (15). Although the speaker did not himself work with the farm tool whose name he is being asked, he knows exactly what people used it for in former times. The sentence conveys a generic reading that excludes the speaker. Sentences like (15) confirm that both members of the bipartite subject contribute to interpretation since the generic reading would not be available if the pronoun eles “they” expressed alone the sentential subject. (15) Não sei lá a certeza também do nome, também não digo. “I’m not sure about the name, so I won’t spell it out.” Sei é de real certeza que isto era com know.1sg is of real certainty that this was with o.que se eles batiam o centeio. what SE they would.beat.3pl the rye

(CORDIAL-SIN. FLF)

“But I am totally sure that this was the thing that people used when husking the rye.”

11.  The pronoun a gente is semantically 1st person plural even when it does not induce 1st person plural agreement morphology on the verb. 12.  The doubler may also be a full DP, like in (i) below, with the same effect of blocking the inclusive reading. Here the speaker is describing the traditional cheese making process, but he is not himself a cheese maker. (i) A minha mãe e os outros todos tiravam-se aquele punhadozinho the my mother and the others all would.take.3pl-SE that handful

de coalhada e depois iam espremendo aquilo… (CORDIAL-SIN.ˉMTM) of curd and then went.3pl pressing that

“My mother and everybody else would take a portion of curd in their hands and would then go on pressing the curd (until it was ready to pack into the mould).”



Subject doubling in European Portuguese dialects 

In (16), it is the null 3rd person plural pronoun that induces an unambiguous exclusive reading (the 3rd person plural features are visible on the verbal inflection). The sentential subject refers to the people in town who used to catch birds (but would not use birdlime for that purpose); the speaker previously made clear that he does not belong to the group of local bird-catchers. (16) Cá nunca se usaram isso. here never SE used.3pl that “People would never use that here (the birdlime).”

(CORDIAL-SIN. LVR)

5.  Accounting for the availability of subject doubling in EP dialects In order to account for the facts described in the previous sections, I will rely on the hypothesis that the doubling pronoun/DP and the subject clitic se belong to the same constituent, a complex DP where the doubler is an adjunct modifier of the clitic se (the latter, a minimal/maximal category in the terms of Chomsky 1995). (17) [DP DP [ se ] ]

(cf. Uriagereka 1995; Kayne 2000, among others)

In this way the interpretative effects described in Section 4 can be smoothly accommodated and the syntactic features of the construction can be thoroughly derived as well, as will be clarified in a moment. Although I am here following the ‘complex DP’ analysis devised by Uriagereka (1995) and Kayne (2000) to account for clitic doubling, it should be noted that the double subject impersonal se construction does not reduce to regular clitic doubling. The crucial difference is that in the latter there must be a secondary source for Case assignment (see how the lack of the Case-assigning preposition a “to” makes (18b) ungrammatical), whereas in the former there is no such requirement. (18) a.

Não foi ele que a beijou a ela, foi ela que o not was he that her kissed to she, was she that him



beijou a kissed to

ele he

b. *Não foi ele que a beijou ela, foi ela que not was he that her kissed she, was she that

o beijou ele him kissed he



“It wasn’t him who kissed her, it was she who kissed him.”

Going back to the derivation of subject doubling in dialectal EP, after the complex DP is assembled (see (17)), it is merged in the structure as the external argument

 Ana Maria Martins

of transitive and unergative verbs, or (less often) as the internal argument of unaccusative verbs, and is assigned a theta-role in its merging position.13 The two elements of the complex DP later follow separate routes. The clitic head-moves to T and undergoes cliticization according to the general pattern of clitic placement in EP. The adjunct modifier, as part of the remnant DP, either stays in its base-generated position or takes the regular path of EP subjects. In this way, we account for the word order facts. As the complex DP splits, each one of its constitutive parts just behaves as a regular clitic and a regular DP-subject with regard to movement operations. The clitic se has number (plural) but does not have a person feature or Case. Though it has semantic content, it is syntactically expletive-like and that is what makes the double subject construction available. Because se is inert to enter the type of agreement relation that underlies Case assignment (cf. Chomsky 2000, 2001), it is the doubling pronoun/DP that induces subject-verb agreement and displays nominative Case. The doubling pronoun/DP has a complete set of ϕ-features. It therefore values the ϕ-features of T and has its Case feature valued. In the Agree based system of Chomsky (2000, 2001), the feature person endows a given probe with Case-valuation properties; hence, finite T can value a given Case feature (as nominative), but a participial head, for example, which does not have a person feature, cannot. I tentatively propose that, conversely, a nominal without a person feature may be devoid of Case-licensing requirements. This hypothesis could be implemented by taking a nominal without person to be an incomplete DP or, to be more precise, an NP articulated with just some functional structure (e.g. NumP). Crucially, lack of person would signal lack of

13.  Double subject sentences with SE are found with transitive, unergative and unaccusative verbs as exemplified respectively in (i), (ii) and (iii). (i)

Aˉgente sega-se esses olhos todos e deixa-se pron.1pl cuts.off-SE those eyes all and leaves-SE



este só. this.one only



“(When it is time for the pruning), we remove all these sprouts leaving only this one.”

(ii)

Mas, enfim, cá se vai a gente but after.all here SE goes pron.1pl “But nevertheless we keep going.”

(iii)

A gente chega-se à noite, eu faço o pron.1pl arrives.SE at.the night I do the



meu trabalho… my work



“When we arrive home at night-time, I do my work…”

(CORDIAL-SIN.PST)

andando. walking

(CORDIAL-SIN. AAL)

(CORDIAL-SIN. PIC)



Subject doubling in European Portuguese dialects 

the higher functional categories of the DP domain, in particular lack of D°, the locus of referentiality and, hypothetically, the locus of Case (cf. Longobardi 1994; Chomsky 2001, 2004). As for the contrast between the EP dialects that allow subject doubling and standard EP, it can be simply derived under the assumption that in standard EP the impersonal pronoun se is not a deficient pronoun, but a full referential expression with a complete set of ϕ-features and Case (cf. D’Alessandro 2004). The current proposal thus converges with Holmberg & Nikanne (2006) in the observation that the grammars that possess the right kind of deficient pronoun are expected to allow subject doubling.14 In the next section independent additional evidence will be offered to support the contention that the pronoun se has the feature plural and no person feature in the EP dialects where the double subject impersonal se construction is attested.

6.  Further evidence supporting the analysis A cluster of contrasts between EP dialects and standard EP with respect to the behaviour of se can be easily handled under the approach developed in this paper. This section will briefly consider: adjectival agreement with se in predicative contexts, anaphoric binding, and the alternation between null and overt subjects in expletive constructions. 6.1  The pronoun se is plural in dialectal EP: adjectival agreement Adjectival agreement in predicative contexts shows that in standard EP the impersonal pronoun se has the number feature ‘singular’. Thus the adjective novo “young” 14.  The subject doubling impersonal se construction is a grammatical option in dialectal EP because it satisfies (a), (b)-2. and (c)-2. of (i):

(i)

On the availability of doubling (Holmberg & Nikanne 2006): (a) (b)

A category a doubles a DP b iff they share a theta-role. Doubling is possible iff

1. 2. 3.

(c)



a and b are copies (the case of A and A’-movement); a is deficient. [there is an extra source for case assignment/valuation (cf. Schoorlemmer 2006). Addition mine]

Deficient pronouns include 1. 2.

pronominal clitics (the case of clitic doubling); personless pronouns.

 Ana Maria Martins

in (19) is not allowed to display plural inflection, as the grammaticality contrast between (19a) and (19b) proves. (19) Standard European Portuguese: a.

Quando se é novo… when SE is young.masc.sg “When one is young…”

b. *Quando se é novos… when SE is young.masc.pl “When one is young…”

In sharp contrast with the standard variety, European Portuguese dialects allow plural agreement between se and an adjectival predicate as illustrated by (20a–b). (20) a.

Na idade é que é; uma pessoa quando in.the age is that is; a person when



se é novos, poder SE is young.masc.pl can.INFIN

(CORDIAL-SIN. ALV)

“There is a right time for everything; when one is young, one must have the strength to work hard.” b. Não se andava calçados (CORDIAL-SIN. CDR) not SE would.go.3sg with.shoes.on.the.feet.adj.masc.pl “We wouldn’t go around with shoes on our feet (we would go barefooted).”

These data confirm that dialectal se may differ from standard se as for the value of its number feature. This is certainly so in the dialects where the double subject impersonal se construction is attested. Hence, in this construction the doubling pronoun/DP is always plural, like se itself.15 The plural number of se has further grammatical consequences as will be discussed in the next section.

15.  The conclusion that dialectal se is plural and therefore can only be doubled by a plural pronoun/DP seems to be at odds with the fact that in sentences like (i) below the subject pronoun a gente doubles impersonal se but at the same time shows third person singular agreement with the verb: (i) a.

A gente chama-se rãs a isto. pron.1pl call.3sg-SE rãs to that “We call these ones rãs (frogs).”

(CORDIAL-SIN. FLF)

In fact the pronoun a gente displays (obligatorily in standard EP and optionally in dialectal EP) a mismatch between its semantic and syntactic features, being semantically (1st person) plural but syntactically (3rd person) singular (cf. Costa & Pereira 2005). Crucially, it is the semantic features that matter with respect to agreement between se and its doubler. This is



Subject doubling in European Portuguese dialects 

6.2  The pronoun se is person-less in dialectal EP: anaphoric binding In a sub-group of the dialects where the double subject impersonal se construction is found, anaphoric se (reflexive or inherent) can be bound by a 1st person plural pronoun (null or overt), as exemplified in (21). Such sentences are excluded in standard EP (as in many EP dialects). (21) a.

No mar ainda às.vezes se orientamos in.the sea still sometimes SE orientate.1pl



pela vaga. by.the wave



“In the sea we are able to orientate ourselves by watching the waves.”

b. A gente casámos-se novos. pron.1pl married.1pl-SE young.masc.pl “We were young when we married each other”

(CORDIAL-SIN. MLD)

(CORDIAL-SIN. GRC)

The availability of sentences like (21) is expected under the analysis put forth in this paper. In the relevant dialects, defective se can be anaphorically bound by any plural pronoun/DP because it has the feature plural and no person feature.

confirmed by clitic doubling structures in standard EP (where the same type of big DP is at stake – cf. Uriagereka 1995; Kayne 2000). Although in standard EP a gente always sets subject-verb agreement in 3rd person singular (see (ii) below), it can double 1st person plural pronoun nos “us” (see (iii)). The accusative pronoun nos “us” is the clitic correlate of the strong nominative pronoun nós “we” which only permits 1st person plural agreement with the verb (see (iv)). Thus nos “us” and a gente can be part of the same big DP because they match in semantic features, although they do not match in syntactic features. (ii) a.

A gente chega tarde. pron.1pl arrive.3sg late

b. *A gente chegamos tarde. pron.1pl arrive.1pl late “We arrive late.” (iii)

Fazer-nos isto à gente ?! É incrível. do.INFIN-us this to pron.1pl is unbelievable “It’s unbelievable that he/she would do this to us.”

(iv) a.

Nós chegamos tarde. we arrive.1pl late

b. *Nós chega tarde. we arrive.3sg late “We arrive late.”

(Standard EP)

 Ana Maria Martins

The fact that the number feature of dialectal se is plural also explains why it cannot take a 1st or 2nd person singular pronoun as antecedent, as the ungrammaticality of the sentences in (22) illustrates.16 (By the same token, se is never doubled by a 1st or 2nd person singular nominative pronoun in the subject doubling impersonal se construction). (22) *Eu casei-se cedo I married-SE early “I was young when I got myself married.”

Se cannot have a 2nd person plural pronoun as antecedent either, because the relevant dialects, like standard EP, lack a 2nd person plural pronoun as well as 2nd person plural verbal inflection. There appears to be a problem for this approach though. The question that comes to mind is why the clitic nos (“us”, “ourselves”), specified for number (plural) and person (1st), does not block the insertion of underspecified se. The answer lies in the fact that the relevant dialects virtually replaced nós (“we”) with the new 1st person plural pronoun a gente (a former DP: “the people”) and presumably also replaced nos (“us”, “ourselves”) – see (23).17 (23) Consequences of the evanescence of nós/nos-1pl in some EP dialects:   STAGE 1 a. Nós vimos-te we saw-you

STAGE 2 → A gente vimos-te pron.1pl saw-you

b. Eles viram-nos they saw-us

→ Eles viram a gente they saw pron.1pl

c.

(Nós) vimo-nos → (A gente) vimos-se we saw-ourselves pron.1pli saw-SEi



na televisão on TV

na televisão on TV

The innovation a gente replaced the former 1st person plural pronoun nós/nos “we/us”. But the pronoun a gente obeys Principle B of Binding Theory (see (24)) and is therefore unable to act as a substitute for the 1st person plural anaphora

16.  Note that all dialects also have the non deficient 3rd person se of standard EP. 17.  The evidence given by the CORDIAL-SIN dialectal corpus is not as clear-cut as we would want because the speakers of EP dialects also know standard EP to a certain degree and may produce forms of both varieties in spontaneous speech (especially when involved in conversation with standard EP speakers).



Subject doubling in European Portuguese dialects 

(i.e. nos “ourselves”). The person-less se thus became the universal plural anaphora in the relevant dialects (see (23c)).18 (24) a. *[Eu e o Pedro]i vimos [a gente]i na televisão. [I and the Pedro]i saw [pron.1pl]i in.the television “I and Pedro saw ourselves on TV.” b. [Eu e o Pedro]i dissemos que [aˉgente]i adorámosˉ a festa. [I and the Pedro]i said that [pron.1pl]i loved the party “I and Pedro said that we loved the party.” c.

[A gente]i sabemos que a Maria criticou [a gente]i. [pron.1pl]i know that the Maria criticized [pron.1pl]i “We know that Maria criticized us.”

For anaphoric se to be bound by a 1st person plural pronoun (overt or null) two conditions must be fulfilled: (i) se must be plural and person-less; (ii) the 1st person plural anaphora nos “ourselves” must not be available. Otherwise, the more specified form (i.e. nos-1PL) would have preference over the underspecified one (i.e. se-PL). We thus expect that some dialects allow the double subject impersonal se construction but do not allow binding of se by a 1st person plural pronoun. Such dialects fulfil condition (i), they have a plural person-less se, but not condition (ii), they have not lost nos “ourselves”. The data offered by the CORDIAL-SIN dialectal corpus confirm this prediction.19

18.  We could hypothesize that sentences (21a–b) would instantiate the impersonal subject doubling construction but either the impersonal or the reflexive clitic se would have been deleted because the sequence *se se is not allowed in European Portuguese (similarly to most Romance varieties), as illustrated by (i) below – cf. Wanner (1977), Bonet (1995), among others. (i) *Orienta-se-se bem. orientates-SE(impersonal)-SE(reflexive) well “One knows how to take care of one’s life.” This hypothesis faces two problems. First, in the dialects that do not have the impersonal subject doubling construction the option of deleting one of the clitics se is not available. Second, sentences like (21a–b) are not found in all the dialects that have the impersonal subject doubling construction but only in a sub-group of those dialects. This is expected under the view that the dialects that have a person-less plural se but did not lose the 1st person plural pronoun nós-NOM/nos-ACC will not display sentences like (21a–b). 19.  The double subject se construction can be found in some regions of the North of Portugal but binding of anaphoric se by a 1st person plural pronoun is not found anywhere in the North.

 Ana Maria Martins

6.3  Pure expletive se with impersonal predicates The syntactic deficiency of dialectal se usually does not extend to its semantic content. Thus ϕ-incomplete impersonal se still denotes a group of humans. Some dialects, however, reveal that the weakening of se that changed it into a person-less pronoun can go a step further depriving se from semantic content. At that point the semantic [+ hum] feature of se is lost and the clitic starts to appear as the expletive subject of impersonal predicates, as the examples in (25) illustrate. Sentences (25a) and (25c) are particularly revealing as they show that the overt expletive se can freely alternate with a null expletive. (25) a.

Às.vezes acontece; já se aconteceu. sometimes happens already SE happened “Sometimes it happens; it has already happened.”

b. Nunca se me aconteceu isso. never SE me.dat happened.3sg that “It never happened to me a thing like that.”

(CORDIAL-SIN. PIC)

(CORDIAL-SIN. CRV)

c.

Chega a pontos custa-se caro. Custa caro uma arrives to a.point costs-SE expensive costs expensive a



parelha a arrancar uma charrua. pair.of.animals to pull a plough

(CORDIAL-SIN. CPT)

“At a certain point it is really hard for the animals to go on pulling the plough.” d. Vou também ao mar apanhar uns peixes (...), go.1sg also to.the sea catch some fish

quando se calha. when SE happens



“I also go fish occasionally.”

(CORDIAL-SIN. PIC)

The appearance of the pure expletive se is again restricted to the dialects where the double subject impersonal se construction is found. It thus constitutes one more indicator that in the relevant EP dialects the pronoun se gradually lost its referential content. Finally, it is worth noting that expletive se does not occur with weatherpredicates. This is exactly what is expected if the subject of weather-predicates is a quasi-argument with a complete set of ϕ-features, Case, and a theta-role (see Chomsky 1981:325, 1995:288; Vikner 1995).

Moreover, only a sub-region of the South of Portugal has se as the universal plural anaphora while the double subject se construction is pervasive in southern dialects. Cf. Footnote 3.



Subject doubling in European Portuguese dialects 

7.  Consequences of the proposed analysis This paper focuses on the subject doubling impersonal se construction which is found in certain EP dialects but disallowed in standard EP. The main proposal in the paper is that a nominal that lacks a person feature may also be devoid of Case. This is how the clitic se of the relevant EP dialects is like; it has number (plural), but no person or Case. Its deficient nature (in the spirit of Holmberg & Nikanne 2006) is what makes subject doubling a grammatical option in dialectal EP. In a few dialects, the expletive-like se drifted in the direction of a pure expletive. If the analysis put forth in the paper is on the right track, the subject doubling se construction of European Portuguese dialects brings new empirical evidence that supports the contentions (i) to (iii) below. i. “If all other conditions are right, a given argument can be linked to, and coherently composed with, more than one expression”. (Chung & Ladusaw 2004:118, on the extra object in Chamorro).20 ii. The impersonal se/si of standard EP, Spanish and Italian is not deficient with respect to ϕ-features and Case (cf. Cinque 1988), but a full referential expression (D’Alessandro 2004). Therefore, subject doubling with impersonal se/si is not an option in standard EP, Spanish and Italian.21

20.  “Multiple linking, as we have just characterized it, is obviously not the norm. Typically, the syntax and the semantics conspire to make available exactly one expression to be composed with a targeted argument, so that the mapping between arguments and expressions appears to be one to one.” (Chung & Ladusaw 2004:76). 21.  A reviewer notes that a type of subject doubling is found in Italian (especially Tuscan Italian) in sentences like (i), where noi “we” doubles impersonal si, but agreement between 1st person plural noi and the verb is not allowed: (i) a.

Noi si mangia sempre alle 12. we.nom SI eat.3sg always at 12

b. *Noi si mangiamo sempre alle 12. we.nom SI eat.1pl always at 12 “ We always eat at twelve o’clock.” Burzio (1986:81, Fn. 47) briefly comments on the construction illustrated by (ia): “SI appears in non impersonal uses only dialectally (Tuscan dialects, mostly). In such cases SI has the force of a first person plural pronominal, which sometimes appears overtly in addition to SI, as in (Noi) si vorrebe vedere i nostri amici “(We) si would like to see our friends” (notice that whereas nostri agrees with noi, the verb still fails to agree: a fact for which we have no account)”. If si in Tuscan Italian is a first person plural pronominal as suggested by Burzio, it might be similar to EP a gente in bearing different values for its semantic and syntactic features, respectively 1st person plural and 3rd person singular (cf. Fn. 15). It is very revealing in this

 Ana Maria Martins

iii. Overt pure expletives are not necessarily blocked when in competition with null pure expletives in a particular grammar (cf. Chomsky’s 1981 Avoid Pronoun Principle).

References Bonet, Eulàlia. 1995. “Feature Structure of Romance Clitics”. Natural Language and Linguistic Theory 13. 607–647. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. Cardinaletti, Anna. 1997. “Subjects and Clause Structure”. The New Comparative Syntax ed. by Liliane Haegeman, 33–63. London & New York: Longman. Cardinaletti, Anna. 2004. “Toward a Cartography of Subject Positions”. The Structure of CP and IP. (= The Cartography of Syntactic Structures vol. 2) ed. by Luigi Rizzi, vol. 2, 115–165. Oxford & New York: Oxford University Press. Chierchia, Gennaro. 1995. “The Variability of Impersonal Subjects”. Quantification in Natural Language ed. by Emmon Bach, Eloise Jelinek, Angelika Kratzer & Barbara Partee, 107–143. Dordrecht: Kluwer. Chomsky, Noam. 1981. Lectures on Government and Binding. The Pisa Lectures. Dordrecht: Foris. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. “Minimalist Inquiries: The framework”. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik ed. by Roger Martin, David Michaels & Juan Uriagereka, 89–155. Cambridge & London: MIT Press. Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale: A Life in Language ed. by Michael Kenstowicz, 1–52. Cambridge & London: MIT Press. Chomsky, Noam. 2004. “Beyond Explanatory Adequacy”. Structures and Beyond (= The Cartography of Syntactic Structures, vol. 3) ed. by Adriana Belletti, 104–131. Oxford & New York: Oxford University Press. Chung, Sandra & William A. Ladusaw. 2004. Restriction and Saturation. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1988. “On si Constructions and the Theory of arb”. Linguistic Inquiry 19. 521–582.

respect that: “In certain Tuscan dialects most typically Fiorentino, si has virtually replaced the ordinary 1st pers. PL. ending of the verb (*andiamo → si va)” (cf. Cinque 1988:55, Fn. 34 and references therein). Again the parallelism with “a gente+3SG verbal form” competing with “nós+1PL verbal form” in (European and Brazilian) Portuguese is striking. The Tuscan subject doubling construction exemplified in (ia) appears to differ from the subject doubling construction discussed in this paper in two respects: (i) si in Tuscan Italian does not behave like a deficient (person-less & Case-less) pronoun; (ii) as far as I know, in Tuscan Italian the doubling pronoun noi necessarily precedes the doubled si. On these grounds, I tentatively take (ia) and similar sentences to be instantiations of the Hanging Topic Left Dislocation construction, with the hanging topic noi escaping the Case Filter (because it is base-generated in an A-bar position) and eventually receiving Default Case, which in Italian is Nominative (cf. Schütze 2001; Legate 2008 on Default Case).



Subject doubling in European Portuguese dialects 

Corver, Norbert & Jairo Nunes, eds. 2007. The Copy Theory of Movement. Amsterdam & Philadelphia: John Benjamins. Costa, João & Sandra Pereira. 2005. “Phases and Autonomous Features: A case of mixed agreement”. Perspectives on Phases (= MIT Working Papers in Linguistics, 49.) ed. by Martha McGinnis & Norvin Richards, 115–124. Cambridge, Mass.: MIT Press. Craenenbroeck, Jeroen & Marjo van Koppen. 2002. “Subject Doubling in Dutch Dialects”. Proceedings of Console 9 ed. by Marjo van Koppen, Erica Thrift, Erik Jan van der Torre & Malte Zimmermann, 54–67. (Also published at http://www.sole.leidenuniv.nl/). D’Alessandro, Roberta. 2004. Impersonal si Constructions. Agreement and interpretation. Ph.D. dissertation, University of Stuttgart (= Electronic Dissertation Series, Stuttgart University Library), http://elib.uni-stuttgart.de/opus/volltexte/2004/1630. Dobrovie-Sorin, Carmen. 1998. “Impersonal se Constructions in Romance and the Passivization of Unergatives”. Linguistic Inquiry 29. 399–437. Holmberg, Anders & Urpo Nikanne. 2006. “Subject Doubling in Finnish: The role of deficient pronouns”. Edisyn – Papers of Workshop on Doubling 2006. (= Electronic Publication), http://www.meertens.knaw.nl/projecten/edisyn/. Kayne, Richard. 2000. “A Note on Clitic Doubling”. Parameters and Universals, 163–184. Oxford & New York: Oxford University Press. Koppen, Marjo van. 2007. “Agreement with (the Internal Structure of) Copies of Movement”. The Copy Theory of Movement ed. by Norbert Corver & Jairo Nunes, 327–350. Amsterdam & Philadelphia: John Benjamins. Legate, Julie Anne. 2008. “Morphological and Abstract Case”. Linguistic Inquiry 39. 55–101. Longobardi, Giuseppe. 1994. “Reference and Proper Names: A theory of N-movement in syntax and Logical Form”. Linguistic Inquiry 25. 609–665. Lopes, Célia. 2003. A Inserção de a gente no Quadro Pronominal do Português. Frankfurt: Iberoamerica – Vervuert. Martins, Ana Maria. 1994. Clíticos na História do Português. Ph.D. dissertation, University of Lisbon. Menuzzi, Sérgio. 2000. “First Person Plural Anaphora in Brazilian Portuguese: Chains and constraint interaction in binding”. Portuguese Syntax: Comparative Studies ed. by João Costa, 191–240. Oxford & New York: Oxford University Press. Miguel, Elena de. 1999. “El aspecto léxico”. Gramática Descriptiva de la Lengua Española ed. by Ignacio Bosque & Violeta Demonte, 2977–3060. Madrid: Espasa. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge & London: MIT Press. Otero, Carlos Peregrín. 1999. “Pronombres Reflexivos y Recíprocos”. Gramática Descriptiva de la Lengua Española ed. by Ignacio Bosque & Violeta Demonte, 1427–1517. Madrid: Espasa. Raposo, Eduardo & Juan Uriagereka. 1996. “Indefinite Se”. Linguistic Inquiry 14. 749–810. Rivero, María-Luisa. 2000. “On Impersonal Reflexives in Romance and Slavic and Semantic Variation”. Paper presented at 30th Linguistic Symposium on Romance Languages, University of Florida, Gainesville, February 2000. Schoorlemmer, Erik. 2006. “Double Agreement in Complex Inversion”. Romance Languages and Linguistic Theory 2004 ed. by Jenny Doetjes & Paz González, 275–295. Amsterdam & Philadelphia: John Benjamins. Schütze, Carson T. 2001. “On the Nature of Default Case”. Syntax 4. 205–238. Strozer, Judith R. 1976. Clitics in Spanish. Ph.D. dissertation, University of California at Los Angeles. Strozer, Judith R. 1978. “On the So-called “Dative of Interest”. Hispania 61. 117–123. Uriagereka, Juan. 1995. “Aspects of the Syntax of Clitic Placement in Western Romance”. Linguistic Inquiry 26. 79–123.

 Ana Maria Martins Vikner, Sten. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford & New York: Oxford University Press. Wanner, Dieter. 1977. “On the Order of Clitics in Italian”, Lingua 43. 101–128. Zagona, Karen. 1996. “Compositionality of Aspect: Evidence from Spanish Aspectual Se”. Aspects of Romance Linguistics. Washington D.C.: Georgetown University Press. 475–488.

On the Quebec French interrogative particle tu* Annick Morin

University of Toronto

The Quebec French interrogative particle tu is phonologically identical to the second person singular subject clitic (disregarding stress) and shows a distribution similar to that of clitic doubling, which is optional in this dialect. More specifically, it is subject to a definiteness effect, previously unnoticed in the literature, in that only definite subjects are acceptable in questions with tu. In this paper I argue that the distribution of the question particle tu and that of clitic doubling are both due to the presence of one feature, namely a Definiteness feature.

1.  Introduction In Quebec French (henceforth QF), yes/no questions are marked by the particle tu (pronounced [tsy]). If we disregard stress, this particle is phonologically identical to the second person singular pronoun, tu. Example (1) shows both in the same sentence, the pronoun being preverbal and the particle being postverbal. (1) Tu viens -tu? you come q “Are you coming?”

Although the particle tu has been previously studied by several authors (Vinet 2001, Noonan 1992, among others), an interesting peculiarity about it has gone unnoticed until now, namely a definiteness effect. In fact, tu’s distribution is reminiscent *I wish to thank Marie-Thérèse Vinet for the Pied Noir data, Diane Massam, Yves Roberge, Michela Ippolito, Kenji Oda, the audiences of Going Romance 2007, the Concordia Symposium on Generative Grammar and the Toronto Syntax Project, as well as the editors of this volume and two anonymous reviewers for their useful comments. I am grateful to the Social Sciences and Humanities Research Council of Canada for its support through a CGS Doctoral Scholarship (767-2007-2220), a Standard Research Grant (410-2005-1112, Massam) and an MCRI grant (412-2003-1003, DiSciullo), and to the estate of the late Margaret and Nicolas Fodor for the Fodor Fellowship.

 Annick Morin

of that of clitic doubling in QF, the observation being that for tu-questions to be grammatical, clitic doubling has to be possible. In this paper, I propose an analysis of the particle tu which accounts for its definiteness effect and then I show how it can apply to other Romance dialects as well. The paper is organized as follows: in Section 2, I present a brief historical background of the particle tu and give its distribution; in Section 3, I give the distribution of clitic doubling in QF, followed by an overview of the previous analyses of French subject clitics. In Section 4, I propose an analysis of the definiteness effect of the particle tu, which must also take into account the fact that clitic doubling shares this restriction. This analysis is compatible with existing accounts of other distributional facts about the particle tu, which I also comment on. In Section 5, I explore data from two other Romance dialects and show how my proposal can apply to them. Section 6 presents the conclusions. 2.  Tu: Interrogative particle or pronominal clitic? The particle tu not only looks like a pronominal clitic, it also comes from one and behaves like one, as we will see in this section. I first present a brief history of its development, and then I give its distribution. 2.1  A brief history of tu According to Picard (1992), tu originates in French inversion, of which an example is given in (2a). Final obstruent deletion, which occurred in the 12th century, gave rise to the resyllabification of (2a) as (2b). In the 14th century, complex inversion arose (Bourciez 1967), which gave utterances like (2c). Picard also points out that, by the 16th century, the linking [t] was extended to all verbs in the third person before a pronoun (Brunot 1927). One century later, the /l/ of il was also dropped, so that (2b) was actually pronounced as (2d). By analogy with complex inversion, forms like (2e) arose. Then, since it always co-occurred with a preverbal subject (full NP or pronoun), ti became redundant and lost its φ-features. It acquired an interrogative feature, which was extended to the whole paradigm.1 (2) a.

Vient -il? (inversion) comes he “Is he coming?”

1.  For a discussion on why it is assumed that ti has developed from complex inversion, see Roberts (1993).

On the Quebec French interrogative particle tu 



b. Vien t-il? (resyllabification) c.

Jean vien t- il? Jean comes he “Is Jean coming?”

(complex inversion)

d. Vien t-i? (/l/ deletion) e.

I vien t-i? (by analogy to complex inversion) he comes-he “Is he coming?”

f.

I viennent-ti? (loss of φ-features) they come-q “Are they coming?”

Now, to explain how ti became tu, probably during the 20th century, Picard suggests that ti first extended from third person masculine singular and plural to third person feminine singular and plural, and then to first person singular and plural. Then, according to Morin (1985), forms with ti in the first and third persons would be found alongside subject clitic inversion in the second person singular and plural, so that the paradigm was as in (3). (3)

1s je 2s 3ms il 3fs elle 1p on 2p 3mp ils 3fp elles

viens viens vient vient vient venez viennent viennent

-ti -tu -ti -ti -ti -vous -ti -ti

(Picard 1992)

Then tu replaced ti, and eventually vous as well, and at the same time tu and vous were retained preverbally (Morin 1985). The fact that tu is a pronoun might have caused its substitution of ti, as suggested by Picard (1991). In other words, tu, which only occurred in the second person singular, and still as a pronominal clitic – which is clear as it does not yet appear preverbally – replaced ti as a question particle, which was already established as such in all the persons of the paradigm but the second. 2.2  Distribution As already noted by, among others, Vinet (2001), the particle tu must always immediately follow the finite verb, is restricted to main clauses and is incompatible with wh-elements and with negation. This can be seen in (4). In (4a) the particle follows the auxiliary, est, and in (4b) tu follows the verb va. Example (4c) shows

 Annick Morin

that tu is not allowed in embedded clauses. Sentences (4d) and (4e) show that tu is incompatible with wh-elements. In (4f), we see that negative questions with tu are impossible. Notice that sentences (4c-f) are grammatical without the particle tu. (4) a.

L’ autobus est -tu passé? the bus is q passed “Did the bus come?”

b. On va -tu à Amsterdam? we go q to Amsterdam “Shall we go to Amsterdam?” c.

Je me demande [si elle va (*-tu) appeler]. I myself ask if she will   q call “I wonder if she will call.”

d. Qui a (*-tu) mangé mon chocolat? who has   q eaten my chocolate “Who ate my chocolate?” e.

Quelle sorte de moineau elle a (*-tu) rencontré? what kind of sparrow she has   q met “What kind of weirdo has she met?”

f.

Les politiciens tiennent (*-tu) pas leurs promesses? the politicians hold   q not their promises “Don’t politicians stick to their promises?”

However, tu is also subject to another distributional restriction, of which no mention is made in the literature: it is incompatible with preverbal indefinite and quantified subjects. Examples are provided in (5). (5) a.

?*Un autobus est -tu passé? a bus is  q passed “Did a bus come?”

b. ?*Des enfants jouent -tu dans la rue?   some children play  q in the street “Are some children playing in the street?” c.

*Plusieurs personnes ont -tu répondu? several people have  q responded “Did several people respond?”

d.

*Au moins deux étudiants ont -tu passé l’ examen? at least two students have  q passed the exam “Did at least two students pass the exam?”

Example (5a), as it forms a minimal pair with (4a) above, shows that tu is incompatible with indefinite subjects, as does (5b). Examples (5c) and (5d) show that quantified subjects with tu are also unacceptable to speakers of QF.

On the Quebec French interrogative particle tu 



Note that questions with only intonational raising, such as ?*Un autobus est passé?, are not common in QF, except in the case of echo questions (and such a question with a definite subject would also receive a ‘?*’ judgement); as for questions introduced by est-ce que, they do allow indefinite subjects.2 To ask a yes/no question about an indefinite or quantified NP using the particle tu, speakers use the form ‘Il y a-tu…’.3 Such a grammatical counterpart of Example (5a) is given in (6). (6) Il y a -tu un autobus qui est passé? it there has  q a bus that is passed “Did a bus come?”

These constructions are also used in declarative contexts, as shown in (7). (7) Il y a un autobus qui est passé. it there has a bus that is passed “A bus came.”

But while the sentences in (5) are degraded, turning them into declarative sentences as in (8) makes them perfectly grammatical. (8) a.

Un autobus est passé. a bus is passed “A bus came.”

b. Plusieurs personnes ont répondu. several people have responded “Several people responded.”

Although (8a) is not the preferred form, it is not ungrammatical as (5a) is. The particle tu is, in fact, compatible with one kind of indefinite subject, namely the generic indefinite.4 An example is given in (9). (9) Un linguiste peut -tu avoir une vie sociale? a linguist can  q have a life social “Can a linguist (generic) have a social life?”

2.  An anonymous reviewer suggested an interesting parallel between QF tu and the interrogative particle ESK (est-ce que) in Standard French. This goes beyond the scope of this paper, but will be addressed in further research. 3.  Il being a pronoun, it is treated as definite (but not specific as noted by an anonymous reviewer, hence my claim that the distinction is along definiteness and not specificity). 4.  I am aware that this seems to argue against my claim, but see Section 4.1 for a discussion on why I consider generic indefinites as belonging to the class of ‘definites’. Note that I am using this term in the absence of a better one. Thanks to two anonymous reviewers for pointing this out.

 Annick Morin

At first sight, the distributional restriction affecting tu seems parallel to that of clitic doubling, which we will see in the next section. 3.  Clitic doubling in QF In this section I give the distribution of clitic doubling in QF. Then I present an overview of the previous analyses of that phenomenon as well as of French clitics in general. 3.1  Distribution In QF, subject NPs can be doubled by a pronominal clitic, as in (10). (10) L’ autobusi ili est passé. the bus it is passed “The bus came.”

As can be seen in (11), the distribution of clitic doubling in QF, like that of tu, shows a definiteness effect. (11) a. *Un autobusi ili est passé. a bus it is passed “A bus came.” b. *Quelqu’uni ili a appelé. someone he has called “Someone called.” c. *Plusieurs personnesi ellesi ont répondu. several people they have responded “Several people responded.” d. *Au moins deux étudiantsi ilsi ont passé l’ examen. at least two students they have passed the exam “At least two students passed the exam.”

Hence, while we see in (10) that clitic doubling is possible with a definite subject like l’autobus, (11a) and (11b) show that an indefinite subject cannot be clitic-doubled. Examples (11c) and (11d) show that quantified subjects can also not be cliticdoubled. We saw in (5) that these subjects are not allowed in tu constructions. However, just like tu, clitic doubling is allowed with generic indefinite subjects. We see that un linguiste in (12) is doubled with ça. (12) Un linguistei çai peut avoir une vie sociale. a linguist that can have a life social “A linguist (generic) can have a social life.”

On the Quebec French interrogative particle tu 



And again like tu, clitic doubling is not allowed with wh-subjects, as we can see in (13); we saw in (4d) that tu cannot occur in this context. (13) *Quii ili a mangé mon chocolat? who he has eaten my chocolate “Who ate my chocolate?”

The distribution of tu is in fact not perfectly parallel to that of clitic doubling, however. As we would expect, clitic doubling can occur in embedded clauses and in negative contexts, as shown in (14). (14) a.

Il pense [que sa sœuri ellei a mangé son chocolat]. he thinks that his sister she has eaten his chocolate “He thinks that his sister ate his chocolate.”

b. Les politiciensi ilsi tiennent pas leurs promesses. the politicians they hold not their promises “Politicians don’t stick to their promises.”

Example (14a) shows that clitic doubling is possible in an embedded clause, whereas, as we have seen, tu is not allowed. In (14b) we see an instance of clitic doubling in a negative clause, while we saw in (4f) that tu cannot co-occur with negation. Despite these differences, the historical development of the particle tu and its quasi phonological identity with the second person singular pronoun lead to the hypothesis that their distributional similarities are not a coincidence and that pronominal clitics and the particle tu do have some properties in common. Since the analysis I propose here depends on the analysis of clitic doubling in QF, I will now give an overview of the major proposals that have been made with respect to clitic doubling in Romance languages and French clitics in general. 3.2  Previous analyses There is an ongoing debate as to the status of subject clitics (SCLs) in various dialects of French. Some authors have proposed that these clitics are independent syntactic entities, while others analyze them as agreement morphemes on the verb. 3.2.1  Rizzi (1986) Of the former opinion, Rizzi (1986) makes a structural distinction between Trentino SCLs, which can (and in fact must) double the subject NP, and Standard French (SF) SCLs, which cannot co-occur with the subject NP. Examples are given in (15) and (16).

 Annick Morin

(15) a.

E Gianni el magna. The John he eats “(The) John (he) eats.”

(Trentino, Rizzi 1986)

b. *El Gianni magna. The John he-eats “(The) John eats.” (16) a. *Jean il mange.  John he eats “John (he) eats.”

(SF,ˉRizziˉ1986)

b. Jean mange. John eats “John eats.”

According to Rizzi, Trentino SCLs are adjoined to I°, whereas French SCLs are in Spec,IP, and therefore the notion of SCL is spurious. He argues that what appears to be clitic doubling in dialects of French is in fact left-dislocation, as doubling is not allowed with quantified subjects. This is indeed predicted since quantified subjects cannot be topicalized. 3.2.2  Roberge (1990) Roberge (1990), on the other hand, is of the opinion that in all languages with SCLs, these clitics are base-adjoined to I°, and that the cross-linguistic differences with respect to clitic doubling are due to the setting of a parameter: either the clitic absorbs case, preventing a doubling subject from getting it, or it need not absorb case, making subject doubling possible. He presents data from Colloquial French, where clitic doubling is optional and restricted to definite subjects (as mentioned by Rizzi) and Pied Noir French, where clitic doubling is also optional, but unrestricted, to show how Rizzi’s distinction fails to account for these two “intermediate” languages with respect to clitic doubling. His classification is given in (17). (17)

Standard French: Colloquial French: Pied Noir French: Fiorentino:

no subject doubling (Roberge 1990) optional subject doubling (restricted) optional subject doubling (non-restricted) obligatory subject doubling (non-restricted)

Indeed, Rizzi’s distinction, as Roberge argues, is problematic in light of this classification. One would still have to explain why clitic doubling is possible in some languages and not in others, and why it is optional in some languages and obligatory in others. Moreover, Rizzi’s system gives rise to potential learnability problems, “since it is not obvious how the child could distinguish between the two groups once the presence of subject clitics has been detected” (Roberge 1990: 64).

On the Quebec French interrogative particle tu 



Given the cross-linguistic variation and the optionality of clitic doubling in some languages, Roberge’s solution, where in all languages with SCLs they are in the same position, is certainly more economical, and preferable from a learnability point of view. 3.2.3  Sportiche (1998) Sportiche (1998) proposes that the non co-occurrence of a subject and a clitic is due to a doubly filled phrase prohibition of the same kind as the Doubly filled Comp filter. He thus proposes the Doubly filled voice filter, given in (18). (18) Doubly filled voice filter

(Sportiche 1998)

*[HP XP [H…]] where H is a functional head and both XP and H overtly encode the same property P.

In his analysis of SF SCLs, preverbal SCLs are base-generated in the head of a distinct projection, NomP, the corresponding XP occupying the specifier. As for postverbal SCLs, they are morphologically incorporated into the verb, with the verbal complex raising covertly to Nom°. Since the Doubly filled voice filter is active, no clitic doubling is possible in SF. This, according to Sportiche, accounts for the asymmetry between French preverbal and postverbal SCLs shown in (19). (19) a.

Il ou elle connaît bien le problème. he or she knows well the problem “He or she knows the problem well.”

(Sportiche 1998)

b. *Mange -t-il ou (t-)elle? eats he or she “Does he or she eat?”

The fact that one can say (19a) but not (19b), which is referred to as the ‘proximity problem’, is accounted for, according to Sportiche, by the fact that, in constructions with preverbal SCLs, the verb does not raise to Nom° (where the SCL is basegenerated). Postverbal SCLs, on the other hand, are assumed to be morphologically incorporated to the verb, and then the verbal complex must raise to Nom° (which it does covertly) in order to license the SCL. Furthermore, Sportiche proposes that no overt I to C movement takes place in French complex inversion constructions, nor in SCL inversion constructions. In a brief “excursus on other varieties of French”, he extends his proposal for QF. He assumes that tu is a question morpheme that is incorporated into the verb in the same fashion as French postverbal SCLs and that it must be licensed by (covert) movement of the verb to C. While I accept that incorporating a SCL into the verb is coherent with the idea proposed by Chomsky (1995) that verbs (and nouns) are drawn from the

 Annick Morin

lexicon with all their inflectional features and that they must then check these features in the appropriate position(s), I do not see how this can be justified for an interrogative particle. Arguments and tense are requirements of the verb, clause type is not. Hence it makes sense to assume that the verb comes with some information about the former, but not the latter. Clause type is a requirement of I° (which bears an uninterpretable [ClauseType] feature to be checked by the interpretable [ClauseType] feature in C°, according to Adger 2003 and assumed here), but since the verb is not base-generated in I°, tu cannot be inserted there as an inflectional affix either. Therefore, it has to be base-generated in C°, where clause type is determined. That said, QF SCLs should not be analyzed in the same way as SF preverbal SCLs, because they differ in their behaviour. For instance, Example (19a), which is grammatical in SF, is unacceptable in QF. In fact, it looks like QF preverbal clitics have a status similar to that of SF postverbal clitics. Given what was observed by De Cat (2005) for object clitics, which also applies to QF, I am not following Sportiche’s analysis of postverbal SCLs to account for QF preverbal SCLs.5 Sportiche analyzes postverbal SCLs as morphologically incorporated into the verb (i.e. in the lexicon). I am instead assuming that QF SCLs are base-generated in T°. This is in line with Roberge (1990), LeBel (1998), and also Sportiche (1998), if we accept that projections of the inflectional domain can be conflated in cases where the language provides no evidence for them being distinct (see for example Liakin 2005). 3.2.4  De Cat (2005) De Cat (2005) follows Rizzi’s analysis for French, claiming that what appears to be clitic doubling in the varieties of colloquial French spoken in Belgium, France and Canada is in fact left dislocation. However, her arguments are contradicted by QF facts. For instance, the distribution of the preverbal negative marker ne cannot be explained, according to her, under an analysis where the SCLs are agreement markers. But the negative marker ne is absent in QF speech. De Cat also observes that inversion is productive in QF, but in her study she ignores questions with tu, and questions with est-ce que are relatively infrequent in her corpus. As for intonation, it is virtually never used as a strategy for asking questions in QF. She also gives examples of focussed subjects which, if clitic doubling is not allowed, give

5.  “If subject clitics are attached to the verb stem pre-syntactically, any clitic intervening between the subject clitic and the verb should also undergo affixation pre-syntactically. Given the absence of look-ahead […], the choice of which verb stem an object clitic is affixed to should not be allowed to depend on the future structural configuration.” (De Cat 2005: 1203)



On the Quebec French interrogative particle tu 

rise to ungrammaticality if they occur with a SCL, but such sentences are actually grammatical in QF. For a detailed discussion, see Morin (2009). Moreover, Culbertson et al. (2007) present experimental evidence of change toward agreement status of the clitics in spoken French.6 This is the view that I am adopting.

4.  Analysis I now propose an analysis of the QF interrogative particle tu following the hypothesis that it shares some features with SCLs. This analysis will involve a slight adaptation of Roberge’s analysis of SCLs in light of the facts about tu described above. This analysis is compatible with previous accounts of the other distributional facts about yes/no interrogative particles, of which I will then present an overview. 4.1  The definiteness effect To account for the distributional restriction of clitic doubling given in 3.1, Roberge (1990) argues that indefinites are in fact quantified expressions and thus undergo movement. He states that doubled arguments cannot be locally A’-bound, and therefore their movement is ruled out. In other words, the structure in (20) is ungrammatical. (20) *NPi [IP ti cli + V … ]

(Roberge 1990)

In (20), the NP has moved out of the IP, leaving a trace in Spec,IP. This analysis rules out both quantifier raising and wh-extraction. Roberge’s analysis, however, cannot account for the very similar behaviour of tu, since tu is not a SCL. As shown above, the interrogative particle tu originates from a pronominal clitic, and is phonologically very similar to the second person singular nominative one. It therefore seems intuitive to assume that their distributional similarity is due to their featural similarity. What I propose is that QF pronominal clitics bear a [Definite] feature in addition to their φ-features. This contradicts De Cat’s (2005) argument that the features marked by SCLs are identical to those marked by agreement suffixes. I propose that the particle tu, as it became an interrogative marker, lost its φ-features, but retained its [Definite] feature. That feature, however, has become uninterpretable. Following Rizzi (1997), Poletto (2000) and Vinet (2001), among others, I assume that clause type particles are hosted in the left periphery of the clause. More specifically, I assume, as in Morin (2006), that the particle tu is base-generated 6.  Culbertson et al.’s corpus of ‘spoken French’ comes from Picardy and Paris.

 Annick Morin

in the head of Fin. Assuming tu to be in Fin is justified, as it is sensitive to finiteness (it only appears in finite clauses). Since tu originates in the head of Fin and since it is postverbal, it follows that the verb, together with the SCL, must raise to its left. I follow Roberge (1990) and LeBel (1998) in assuming that the clitic is base-adjoined to I°.7 Then the subject raises to end up in preverbal position. I propose that tu bears a strong [Q] feature, which forces movement of I° to Fin° for the purpose of checking the uninterpretable [ClauseType] feature on I°.8 As for the subject, I propose that it raises to the Spec of Fin°. The structure is given in (21). (21) [FinP NPi [Fin° [cli + V] tu] [IP tNP tcl+V … ]]

Now, this analysis is problematic for Roberge’s account of clitic doubling, since in (21) the doubled argument (tNP) is A’-bound but the sentence is grammatical. I propose that what is crucial for clitic doubling is not that the doubled NP may not be locally A’-bound, but actually that the clitic and the doubled NP must remain in a Spec-head relationship at LF, as stipulated in Sportiche’s Clitic Criterion, given in (22). (22) Clitic Criterion

(Sportiche 1998)

1. A clitic must be in a Spec/head relationship with a [+F] XP at LF. 2. A [+F] XP must be in a Spec/head relationship with a clitic at LF.

Raising of the subject to Fin is thus necessary to restore its Spec-head relationship with the clitic in yes/no questions. This explains the ungrammaticality of (13), repeated in (23) below: in order to check the uninterpretable [wh] feature, the wh-word must move to a higher position, Focus according to Rizzi, therefore breaking the Spec-head relationship with the clitic. I propose that the definiteness effect of clitic doubling is due to the fact that pronominal clitics bear a [Definite] feature that has to agree with a [Definite] feature on the subject NP. Note that, since clitic doubling is optional, the SCL can be phonologically null, but its features are nevertheless present and must agree with the subject.9 In yes/no questions with the particle tu, then, the uninterpretable [Definite] feature on Fin ensures that the

7.  There are in fact other proposals in the literature as to the projection where the subject clitics are generated, which are also compatible with the analysis presented here. Whether the clitic is generated in T° or in a dedicated projection à la Sportiche (1998) is not crucial; what is crucial is that the clitic shares an X° position with the verb (and an eventual object clitic) and requires an XP in its Spec. 8.  I am using the concept of feature strength in the sense of Chomsky (1995), where ‘strong’ features trigger overt movement while ‘weak’ features do not. 9.  In other words, I claim that there is no optionality in the syntax: optionality rather manifests itself at PF, a clitic being always present in the syntactic structure.

On the Quebec French interrogative particle tu 



derivation will converge only if the I° that adjoins to it contains an interpretable [Definite] feature. (23) *Quii ili a mangé mon chocolat? who he has eaten my chocolate “Who ate my chocolate?”

The derivation is given in (24). We see that I° raises to Fin and gets its clause type feature valued as interrogative. The uninterpretable [Definite] feature on Fin° is satisfied by the clitic. Then the subject NP is attracted to the Spec of Fin by an EPP feature created by the Clitic Criterion, which requires the subject and the clitic to remain in a Spec-head relationship. (24)

FinP Fin′

NPi [Def] [φ]

Finº Iº

i [Def] [φ] V

IP Finº tu [uDef] [Q]

Iº

t

I′ tCL+V

…

Iº [uClause Type: Q]

I am assuming that generic indefinites like un linguiste seen in (9), repeated below in (25), also have a [Definite] feature. I take ‘definite’ to correspond to the class of what Milsark (1977) identified as ‘strong NPs’. Milsark’s classification is given in (26). (25) Un linguiste peut -tu avoir une vie sociale? a linguist can  q have a life social “Can a linguist (generic) have a social life?” (26) Strong determiners:

(Milsark 1977)

– “definites” (the, demonstratives, pronouns, possessive DET’s) – universals (all, every, each, any when not polarity item of some, null DET in universal reading) Weak determiners: – a, sm, number determiners, bare plural and mass determiner in nonuniversal reading

Indeed, un in un linguiste reflects the English ‘null determiner in a universal reading’. As for the other ‘universals’, though, they are not possible in questions with tu, but

 Annick Morin

then they are not possible with clitic doubling either. Examples are given in (27) to (29). (27) a.

*Tous les instruments sont -tu accordés? all the instruments are  q tuned “Are all the instruments tuned?”

b. *Tous les instruments ils sont accordés. all the instruments they are tuned “All the instruments are tuned.” (28) a.

*Chaque étudiant va -tu parler quinze minutes? each student will  q speak fifteen minutes “Will each student speak for fifteen minutes?”

b. *Chaque étudiant il va parler quinze minutes. each student he will speak fifteen minutes “Each student will speak for fifteen minutes.” (29) a.

*N’importe qui pourrait -tu faire ça? anyone could  q do that “Could anyone do that?”

b. *N’importe qui il pourrait faire ça. anyone he could do that “Anyone could do that.”

However, Culbertson et al. (2007) have proposed that doubled NPs in spoken French must be old information NPs. Since only old information NPs can be topicalized, this requirement would actually point toward the left-dislocation analysis of clitic doubling argued for by Rizzi (1986) and De Cat (2005). But, as argued in 3.2, such an analysis is inadequate for QF. Now if we assume QF SCLs to be basegenerated in I° (which also corresponds to the structure adopted by Culbertson et al.), the fact that only old information NPs are possible in such a construction should be related to something other than a Top feature, since the canonical subject position, Spec,TP, is not a Top position. Now turning to the tu-construction, if we look at the derivation in (24), one could argue that Spec,Fin is actually a Top position, and therefore all of Milsark’s weak determiners, as well as the strong determiners in (27) to (29), would be excluded because they cannot be topicalized. But recall that clitic doubling also disallows these determiners, and yet in clitic doubling constructions, we have seen that the subject is not in a Top position. Therefore, although it looks like subjects of QF yes/no questions are topics, the fact that clitic doubling disallows these same subjects indicates that Spec,FinP is also not a Top position. To summarize, assuming that both pronominal clitics and the particle tu are endowed with a [Definite] feature follows the hypothesis that the similarity in the distribution of clitic doubling and of the particle tu in QF is not a coincidence.



On the Quebec French interrogative particle tu 

A [Definite] feature on the clitic rules out indefinite and quantified clitic-doubled subjects, and does not prevent the subject from moving to the left periphery in questions. An uninterpretable [Definite] feature on tu ensures that only definite subjects will be allowed in interrogative clauses with tu in QF . 4.2  Incompatibility of tu with wh-elements According to Cheng (1997), languages which type yes/no questions by means of a particle also use that strategy to type wh-questions. Some languages have an overt yes/no particle as well as an overt wh-particle; others have an overt yes/no particle but a covert wh-particle (whereas no language, she claims, has an overt wh-particle but a covert yes/no particle). Following Cheng, then, I propose that QF has a covert wh-particle, tu being a yes/no particle. This is compatible with Aboh & Pfau’s (to appear) claim that interrogative clause typing is achieved via a question operator, never via wh-movement. 4.3  Restriction to main clauses To account for the fact that tu does not appear in embedded clauses, I assume that tu is always associated with a strong Clause Type feature, which triggers verb movement, and only a weak Clause Type feature can be selected. This is also the case in English, as there is no verb movement in embedded questions. 4.4  Incompatibility of tu with negation As for the incompatibility of tu with negation, it was proposed in Morin (2006) that both the interrogative and the negative markers check the Polarity feature on Fin/Σ, the interrogative particle tu merging in that position, as illustrated in (30a), and the negative marker pas raising to Spec,Fin/ΣP.10 I would like to propose here a slightly modified version of that analysis, namely that, instead of pas raising to Spec,Fin/ΣP, the phonologically empty Neg head moves to Fin/Σ°, as shown in (30b). This makes sense, since, under the influence of SF, it would be plausible, if we assume Pollock’s (1989) analysis, that some negative marker be represented in Neg° even if it is not pronounced. This is also consistent with the account of the same restriction in Gascon, given in Morin (2006).11

10.  I am assuming that in Quebec French the Fin projection is conflated with a Polarity (Σ) projection (in the sense of Laka 1994). 11.  I am presenting these facts about negation because they can shed light on the syntax of tu, as there are clear interactions between the two; however, space limitation does not permit a thorough analysis of negation.

 Annick Morin

(30) a.

b. …

Fin/ΣP

Fin/Σº [uPolarity] tu [+neg]

… IP

Fin/ΣP Fin/Σº

Negº ∅ [+neg]

IP

Fin/Σº [uPolarity]

NegP

pas

Neg′ Negº ∅ [+neg]

…

The analysis I present here is coherent with the hypothesis that the distributional similarities of tu and clitic doubling in QF are not a coincidence. I propose that both tu and pronominal clitics bear a [Definite] feature. Since clitic doubling has to be possible, for tu-questions to be grammatical, I assume that there is always a clitic position, even if a clitic is not phonologically realized. Moreover, I propose that the subject of yes/no questions raises to Spec,Fin. To explain why tu does not occur in negative contexts, I also suggest that the empty Neg° raises to Fin/Σ°, preventing tu from merging in Fin°. 5.  Other dialects In this section I explore some implications that my proposal may have for other dialects: Pied Noir French and Paduan. 5.1  Pied Noir French In Pied Noir French (a dialect spoken in North Africa), clitic doubling is optional, like in QF (Roberge 1990). However, unlike in QF, indefinites can also be doubled, as shown in (31). (Example 31c cited from Brua 1938) (31) a.

Le soleil il brille pour tout le monde. the sun it shines for all the people “The sun shines for everybody.”

b. Un homme il vient. a man he comes “A man comes.” c.

Personne il sait qui c’ est leur mère. nobody he knows who it is their mother “Nobody knows who their mother is.”

(Roberge 1990)

On the Quebec French interrogative particle tu 



In Pied Noir, the strategy for forming yes/no questions consists in a rising intonation, with no change in the word order (Vinet, p.c.), as in (32). (Note that, as in other French varieties, est-ce que forms are also possible.) (32) a.

Jean il est venu? John he is come “Has John come?”

(Vinet, p.c.)

b. Le train il arrive? the train it arrives “Is the train coming?”

Interestingly, Pied Noir French, like QF, seems to dislike indefinite or quantified subjects in questions. Examples of ungrammatical Pied Noir French questions are given in (33). (33) a. *Quelqu’un il est arrivé? somebody he is arrived “Has somebody arrived?”

(Vinet, p.c.)

b. *Un train il arrive? a train it arrives “Is a train arriving?” c. *Tous les instruments ils sont accordés? all the instruments they are tuned “Are all the instruments tuned?” d. *Au moins deux étudiants ils ont passé l’ examen? at least two students they have passed the exam “Did at least two students pass the exam?” e. *Personne il est venu? nobody he is come “Has nobody come?”

However the constraint does not seem to be the same as that affecting QF tu. Hence, examples in (34) are grammatical in Pied Noir French, whereas yes/no questions with these subjects are ungrammatical in QF. (34) a.

Plusieurs personnes elles sont venues?  several people they are come “Did several people come?”

b. La plupart des gens ils aiment voyager? most people they like to.travel “Do most people like to travel?” c.

Chaque étudiant il va parler quinze minutes? each student he will talk fifteen minutes “Will each student talk for fifteen minutes?”

(Vinet, p.c.)

 Annick Morin

d. N’importe qui il pourrait faire ça? anyone he could do that “Could anyone do that?”

The data in (34) show that the constraint at work in Pied Noir French is not definiteness, as plusieurs (“several”) in (34a) and la plupart (“most”) in (34b) are acceptable subjects in yes/no questions. Like QF, Pied Noir French also disallows negation in yes/no questions, as can be seen in (35). (35) *Les politiciens ils tiennent pas leurs promesses? the politicians they hold not their promises “Don’t politicians stick to their promises?”

Pied Noir French differs from QF in two relevant respects: it lacks a question particle equivalent to tu, and it allows doubling of indefinite subjects. The constraint on the subjects of yes/no questions in Pied Noir is not clear, but it seems clear that it is different from the one at work in QF. It is also clear that both languages show an asymmetry between declarative clauses and yes/no questions as to which kinds of subjects are allowed. Given this asymmetry and assuming clause type features to be in the left periphery, I suggest that in Pied Noir French the phonologically empty Fin head has a strong [Q] feature, like in QF, which attracts the verb and the clitic that is adjoined to it. What QF and Pied Noir seem to be telling us is that when clitic doubling arises in a dialect of French, something must happen to complex inversion: either it becomes a question particle, as in QF, or it disappears, as in Pied Noir French. Moreover, both languages also show that there are other constraints on yes/no questions. For example in some languages, negative yes/no questions are not possible, whereas they are possible in other languages like SF or English, which also allow indefinite subjects in yes/no questions.12 Examples are presented in (36) and (37). (36) a.

Plusieurs personnes sont -elles venues? several people are they come “Did several people come?”

b. Les politiciens ne tiennent -ils pas leurs promesses? the politicians neg hold they not their promises “Don’t politicians stick to their promises?”

12.  I am not giving an analysis of inversion in SF, but it would be interesting to see how the differences with QF can be accounted for.

On the Quebec French interrogative particle tu 



(37) a. Did several people come? b. Don’t politicians stick to their promises?

It would be interesting to see whether there is any correlation between the (un)grammaticality of negative yes/no questions and the (im)possibility for yes/no questions to have indefinite subjects, and if so, what it means. 5.2  Paduan Paduan, a Northern Italian dialect, has a set of interrogative SCLs which are postverbal in questions, and morphologically different, preverbal, non-interrogative SCLs only in the second person singular, third person singular and third person plural. Examples are given in (38). In (38a), we have a declarative clause, with the preverbal second person singular non-interrogative SCL te; in (38b) we have an interrogative clause with the postverbal second person singular interrogative SCL to. Interestingly, the interrogative SCLs are incompatible with negation, as is QF tu. An example is given in (38c). Negative questions are formed instead using the non-interrogative SCLs and no verb movement, as can be seen in (38d). (38) a.

Te magni. scl eat “You(sg) eat.”

(Zanuttini 1997)

b. Cosa màgni -to? what eat scl “What do you eat?” c. *No ve -to via? neg go scl away “Aren’t you going away?” d. No te ve via? neg scl go away “Aren’t you going away?”

The hypothesis that Zanuttini (1997) puts forth is that these two sets of SCLs are not the same type of elements. Of course more work is needed before drawing any conclusions, but the fact that the postverbal interrogative clitics are incompatible with negation leads me to suggest that they might have acquired a Polarity feature, and thus might be evolving from pronominal clitics to clause type particle(s). 6.  Conclusions In this paper I have shown the distributional similarities between the interrogative particle tu and clitic doubling in QF and argued that the particle tu, just like clitic

 Annick Morin

doubling, is subject to a definiteness effect. The historical development of the particle tu, which, apart from stress, is phonologically identical to the second person singular nominative pronoun, leads to the hypothesis that the particle shares some feature with pronominal clitics. I have therefore proposed an analysis that takes this observation into account and indeed assumes that both the interrogative particle and the pronominal clitics are endowed with a [Definite] feature, the one on the clitics being interpretable, and the one on tu being uninterpretable. To account for the definiteness effect of clitic doubling, I have proposed that the [Definite] feature on the pronominal clitic must agree with a [Definite] feature on the subject NP, and that the clitic and the doubled NP must remain in a Spec-head relationship at LF, according to Sportiche’s Clitic Criterion. I have also suggested that the empty Neg° raises to Fin/Σ°, in order to explain the fact that tu does not occur in negative contexts. Finally, I have given data on two other Romance dialects, Pied Noir French and Paduan, and speculated on how my analysis can apply to them.

References Aboh, Enoch O. & Roland Pfau. To appear. “What’s a Wh-word Got to do with it?”. The Cartography of CP ed. by Paola Benincà & Nicola Munaro. Oxford: Oxford University Press. Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford: Oxford University Press. Bourciez, Édouard. 1967. Éléments de Linguistique Romane. Paris: Klincksieck. Brua, Edmond. 1938. Fables bônoises. Alger: Typo-Litho. Brunot, Ferdinand. 1927. Histoire de la langue française des origines à 1900, Vol. II, Paris: Armand Colin. Cat, Cécile de. 2005. “French Subject Clitics are Not Agreement Markers”. Lingua 115. 1195–1219. Cheng, Lisa Lai-Shen. 1997. On the Typology of Wh-questions. New York: Garland. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Culbertson, Jenny, Géraldine Legendre, Nathalie Filippin & Thierry Nazzi. 2007. Experimental Evidence on Subject Doubling in Spoken French. Poster presented at the 4th Hopkins Workshop on Language: Grammar in Cognition, Baltimore, Md., October 2007. http://www.cog.jhu.edu/grad-students/culbertson/papers/SDoubling_HOWL07.pdf Laka, Itziar. 1994. On the Syntax of Negation. New York: Garland. LeBel, Marie-Élaine. 1998. Modularité et cliticisation : Les pronoms sujets du français du Québec. M.A. thesis, Université du Québec à Montréal. Liakin, Denis. 2005. “OVS pour une langue SVO. Pourquoi?”. Proceedings of the 2005 Canadian Linguistics Association Annual Conference, ed. by Claire Gurski, 11 pages. http://ling.uwo.ca/publications/CLA-ACL/CLA-ACL2005.htm Milsark, Gary L. 1977. “Toward an Explanation of Certain Peculiarities of the Existential Construction in English”. Linguistic Analysis 3:1. 1–29. Morin, Annick. 2006. On the Syntax of Clause Type Particles : Evidence from Gascon, Innu and Quebec French. M.A. thesis, Concordia University. Morin, Annick. 2009. “Clitic Doubling in Quebec French”. Ms., University of Toronto.



On the Quebec French interrogative particle tu 

Morin, Yves-Charles. 1985. “On the Two French Subjectless Verbs voici and voilà”. Language 61:4. 777–820. Noonan, Máire B. 1992. Case and Syntactic Geometry. Ph.D. dissertation, McGill University. Picard, Marc. 1991. “Clitics, Affixes, and the Evolution of the Question Marker ‘tu’ in Canadian French”. French Language Studies 1. 179–187. Picard, Marc. 1992. “Aspects synchroniques et diachroniques du tu interrogatif en québécois.” Revue québécoise de linguistique 21:2. 65–75. Poletto, Cecilia. 2000. The Higher Functional Field. Evidence from Northern Italian Dialects. Oxford: Oxford University Press. Pollock, Jean-Yves. 1989. “Verb Movement, Universal Grammar and the Structure of IP”. Linguistic Inquiry 20. 365–424. Rizzi, Luigi. 1986. “On the Status of Subject Clitics in Romance”. Studies in Romance Linguistics ed. by Osvaldo Jaeggli & Carmen Silva Corvalán, 391–421. Dordrecht: Foris. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. Elements of Grammar: Handbook of Generative Syntax ed. by Liliane Haegeman, 281–337. Dordrecht: Kluwer. Roberge, Yves. 1990. The Syntactic Recoverability of Null Arguments. Kingston & Montreal: McGill-Queen’s University Press. Roberts, Ian G. 1993. Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Sportiche, Dominique. 1998. Partitions and Atoms of Clause Structure. London: Routledge. Vinet, Marie-Thérèse. 2001. D’un français à l’autre : La syntaxe de la microvariation. Montreal: Fides. Zanuttini, Raffaella. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: Oxford University Press.

Autonomous typological prosodic evolution versus the Germanic superstrate in diachronic French phonology* Roland Noske

Université Lille 3 & CNRS UMR 8163 ‘STL’ The traditional idea that in Early French, the Frankish superstrate had a major influence on the phonology is mistaken. Facts from Old Frankish itself show that this cannot have been the case. In addition, the phonetic concept on which the idea is based, i.e. the distinction between ‘expiratory’ and ‘melodic’ languages, is outdated and invalid. The facts for which the Frankish influence was invoked find a much more satisfactory explanation if one considers the evolution of the entire prosodic system of the language. The distinction between syllable languages and word languages, as proposed by Auer & Uhmann (1988), provides a fruitful framework for this.

1.  Introduction In Gallo-Romance, processes of vowel reduction, vowel deletion (syncope/apocope) and diphthongization became active in the 5th and 6th centuries. Their advent is traditionally explained by the development of a strong ‘expiratory’ accent: the ‘melodic’ accent of Latin would have been replaced by an accent which then would have grown progressively stronger, to the detriment of the effort of articulation for unstressed vowels. This assumed decrease in articulatory effort is presented as the explanation for the reduction or even deletion of these vowels. Examples are given in (1)–(3) (from Pope 1952:103–104, 112, transcribed into IPA): (1) vowel reduction a. ter:a > terə

“earth”

(2) vowel deletion a. perdere > perdrə “lose”

b. portas > portəs “doors” b. arborem > arbrə “tree”

*I wish to thank Wolfgang Kehrein, Robert Kyes, Stephen Laker, Pieter Muysken, Arend Quak, Norval Smith, Piet van Reenen, Martha Young-Scholten and two anonymous reviewers for their remarks. Remaining errors are mine.

 Roland Noske

In descriptions of the history of French, this development is usually attributed to the influence of the Germanic, particularly Frankish, superstrate. In these descriptions one finds passages like: The Frankish system of accentuation was a strong expiratory one and it was in the intensifying of the weak Latin tonic stress that the Germanic speech-habits, and in particular the Frankish, exercised their strongest influence in pronunciation. Directly resultant were: <…> the reduction, or effacement of the unstressed vowels <…>. (Pope 1952:15)

Further on in this text we find: Under the influence of the intensified tonic (=expiratory) stress of the Gallo-Roman period, atonic vowels in every type were ordinarily either effaced or reduced to e̥ (= ə, RN), <…>. (Pope 1952:112)

Similar passages can be found in various other texts, like Von Wartburg (1965:65), Zink (1986:37) and La Chaussée (1989:193). Apart from vowel reduction and vowel deletion, the development of final stress that characterizes Modern French is also – indirectly – attributed to the Germanic influence: Mais la prononciation des Germains, qui frappait à coups de marteau la syllabe accentuée et qui, par là, a provoqué la chute de la voyelle finale (non accentuée) au VIIe siècle, doit avoir créé des circonstances favorables à l’oxytonisme français. (Kukenheim 1971:319) (But the Germanic pronunciation, which hit the stressed syllable by hammer strokes and which as a consequence caused, in the 7th century, the deletion of the (unstressed) final vowel must have created a situation favorable for the advent of French final stress.) (translation mine, RN)

In these texts, a big part of the major phonological changes that French has undergone from its genesis until the Middle French period is indeed attributed to the influence of Germanic.1

1.  Although widely found in texts, the idea of a Frankish superstrate influence on Gallo-romance stress has not been accepted by all historical linguists of French. For instance, Cohen (1967:109) expresses some doubts: “Les grands effets de l’accentuation (disparition des voyelles après l’accent) ont quelquefois été attribués à l’influence germanique, plus particulièrement francique, bas-allemande. De même certaines diphtongues. Mais l’ensemble des altérations n’est pas de type germanique.” (The major effects of acccentuation (vowel loss after stress) has sometimes been attributed to the Germanic, more particularly Frankish or Low German,



Autonomous typological prosodic evolution 

In this article, I challenge this idea. I show that the arguments advanced in favor of a Germanic influence are invalid, that the very conception of the nature of stress on which these arguments are based has become outdated by modern phonetic research and that the evolution of stress in French forms part of an autonomous development regarding the prosodic organization of the language. This paper is organized as follows. In Section 2, I show that the assumption of an influence of a Germanic expiratory stress on Gallo-Romance is mistaken. Then, in Section 3, I briefly show that experimental phonetic research has shown that the very distinction between melodic vs. expiratory stress, as used in the aforementioned texts, is outdated. Then, in Section 4, I treat some additional changes in the histories of French and West Germanic. This then paves the way for Section 5, where I treat a general prosodic typology in which the changes in French are embedded and which offers a far more comprehensive framework for the explanation of the phonological changes in question than the putative influence of a Germanic superstrate.

2.  Arguments against a Frankish influence on the stress system For the sake of the argument, I assume in this section that expiratory stress is linguistically relevant (but see Section 3). The claim that Frankish had a ‘strong expiratory [stress]’ (Pope 1952:15), a stress of ‘excessive strength’ (Zink 1986:37) or even one that can be compared with ‘hammer strokes’ (Kukenheim 1971:319) is not supported by any reference to studies that show that Frankish, in what was the pre-literary period of the language, did indeed have the characteristics of a language with a strong expiratory stress. I show here that the supposed borrowing of a heavy expiratory stress from Frankish is extremely improbable, for the following reasons: i. borrowings of stress systems are rare or non-existent; ii. 5th century Frankish probably still had the initial Common Germanic root-initial stress, which, had there been transfer, would also have been borrowed; iii. given the absence of reduced vowels in Frankish itself, Frankish did not have a heavy expiratory stress. Let us look at these points in detail. influence. The same applies to certain diphthongs. But the changes as a whole are not of a Germanic type).

 Roland Noske

2.1  Entire stress systems are not borrowed from one language to another In the literature on language contact (e.g. Weinrich 1963; Appel & Muysken 1987; Thomasson 2001; Winford 2003) there is no mention of transfer of a stress system from one language to another. Linguists such as Muysken (p.c.) confirm that this sort of transfer is unattested in historical and current language contact situations. The only real type of stress transfer that has been attested is linked to borrowed words or morphemes.2 Another type of stress transfer cited in historical linguistics (e.g. by Halle & Keyser 1971:99–100; Árnason 1996:1) is that of English and other Modern Germanic languages (except Icelandic and Faroese), which would have borrowed their present stress systems (in morphologically simple words, trochaic stress is assigned from the right, with extrametricality of the final syllable) from Medieval Latin. However, this hypothesis has been recently refuted for English by Fournier (2007). It is also highly questionable whether Medieval Latin, a non-natural language because it had no native speakers, still had the stress system of Classical Latin, as tacitly assumed by the proponents of this hypothesis. Rather, one would expect that it was pronounced with the stress systems of the various native languages of the speakers. It seems that the Germanic shift from initial to final trochaic stress is simply an independent development (a claim implicitly made by Lahiri, Riad & Jacobs 1999), just as the shift from the Proto-Latin initial stress to the stress system of Classical Latin (for an Optimality-theoretic account of this, see Jacobs 2003). 2.2  The place of Frankish stress in the 5th century3 Apart from some words in the Salic Law of the early 6th century, we do not have any written source for Frankish for the period of the supposed borrowing of the

2.  German and Dutch provide examples of borrowed morphemes that have kept their stress, e.g. the verbal suffixes -ier/-eer, the nominal suffixes -ität/-iteit and the adjectival suffixes -el/-eel. Also, Samaraccan (a Surinamese creole) has a split prosodic system: words of European origin (English and Portuguese) have accentual prosody, while words of African origin have tonal prosody (Good 2009). But in both cases, that of German/Dutch and that of Saramaccan, the stress or tonal accents remain linked to the original words or morphemes and are not propagated to the rest of the lexicon. 3.  In this subsection and the following the issue arises of Old Frankish from the 5th to the 11th century. During this period, the second Germanic consonant shift took place, which split up Frankish into Low and High Frankish (the latter being subdivided into Central and Upper Frankish). However, this division is of no importance for the arguments which are advanced here, because the relevant facts are the same for all varieties of Frankish.



Autonomous typological prosodic evolution 

strong expiratory accent from the language by Gallo-Romance.4 We do know, however, that around the 5th century the Germanic dialects stressed the first root syllable (cf. Prokosch 1939:118–119).5 So, if Frankish stress had indeed had a strong influence on Gallo-Romance, one would expect that this feature would also have been transferred. Hence, Gallo-Romance would have had initial stress. But this, as we know, was not the case. 2.3  The supposed heavy expiratory stress in Frankish Pope (1952) and Zink (1986) explicitly attribute the heavy expiratory accent of Gallo-Romance that would have caused vowel reduction and deletion to the Frankish superstrate. We would then expect that the putative strong expiratory stress (‘with hammer strokes’) would have caused unstressed vowel reduction and deletion in early Germanic itself. However, Old Frankish did not have processes of reduction and deletion of unstressed vowels. These processes took place only much later in the history of High German and Dutch (both partial or total heirs of Old Frankish), i.e. from the end of the 11th century onwards. The advent of vowel reduction is generally taken as marking the transition between OHG and Middle High German (MHG) and Old Dutch and Middle Dutch (König 1978:73; Quak 1997:37). To illustrate this, I present here three examples of texts preceding the end of the 11th century, one in High Frankish and two in Low Frankish. Here are the first two lines of Our Father in Southern Rhenisch Frankish, a High Frankish (more precisely, Central Frankish) dialect, taken from the Weissenburg Catechism at the end of the 8th century (Braune 1994): (3) fater unser thu in himilom bist giuuihit si namo thin. father our you in heavens are hallowed be name your “Our Father in heaven, may your name be kept holy.”

In this example one notes that in the word himilom, where stress falls on the first syllable, the next two syllables contain full vowels. The second vowel in namo, where stress is also initial, has not been reduced either. Let us compare

4.  There are, however, records of Germanic loan words in Vulgar Latin (cf. Brüch 1913). 5.  Verner’s law had functioned in the 1st and 2nd centuries. Its effects became phonemic by the shift of the free Indo-European accent to the first root syllable. Hence, this shift took place after the working of Verner’s law, but had been completed before the 5th century (Prokosch 1939:62–64).

 Roland Noske

this example with the first two lines of the same prayer in MHG (from around 1300 A.D.): (4)

vater unser der da bist in den himeln geheiliget wert din name. father our who there are in the heavens hallowed is your name

Here, we see that (among other changes, like the introduction of articles) the second i of himilon has been reduced to e ([ə]), just like the o in namo. In addition, and in contrast to the 8th century text, the third vowel in himilom has been syncopated (the final n in the MHG text, compared to m in the Old Frankish one, is the result of a change of the form of the dative plural marker). Let us now look at a text from Eastern Low Frankish (Limburgish), i.e. the Wachtendonck Psalms, dating from the 10th century (Van Helten 1902; Cowan 1957; Kyes 1969). (5) Forchta in biuonga quamon ouer mi in bethecoda mi thuisternussi Fears and tremblings came over me and covered me darkness “I became afraid, started to shake, and was covered by darkness.”

In Middle and Modern Dutch (which are primarily Low Frankish), the o in quamon has been reduced to [ə] (kwamen [ˈkυaːmən] in Modern Dutch). Bethecoda corresponds to bedekte [bəˈdεktə] in Modern Dutch, showing two instances of vowel reductions as well as syncopation. My third example stems from Western Old Frankish (Schönfeld 1933): (6) Hebban olla vogala nestas hagunnan hinase hic enda thu uuat have all birds nests begun except I and you what

unbidan uue nu wait we now

“All birds have started building nests, except you and me. So what are we waiting for?”

If one compares the words in this sentence with their equivalents in Middle and Modern Dutch, one notes that many vowels have been replaced by e [ə]. Hebban thus corresponds to hebben [hεbən] in Middle Dutch, vogala to vogele, nestas to nesten, and the infinitival suffix as in unbiddan, -an, to -en. Hence, we see that in Frankish, reduction and deletion of unstressed vowels did not take place before the end of the 11th century, in contrast to Gallo-Romance, where, as mentioned in the introduction, vowel reduction as well as syncope and apocope took place in the 5th and 6th centuries (Richter 1934:202). Given these chronologies, it would be strange that 5th century Frankish would have had stress of ‘excessive strength’, and that this stress would have caused vowel



Autonomous typological prosodic evolution 

reduction and syncope/apocope in Gallo-Romance, but that it would not have had the same effects in Old Frankish itself.6 3.  ‘Melodic’ versus ‘expiratory’ accent 3.1  The role of stress distinction in historical linguistics As we have seen, the handbooks on Old French refer to an expiratory stress, which would have been part of the ‘Germanic speech-habits’ and which would have been taken over into Gallo-Romance. The concept of expiratory stress stems from early phonetic science, a stage when a general distinction was made between languages with a ‘melodic’ accent and languages with a ‘expiratory accent’. The distinction dates back to the 1870s and became known through Sievers’ textbook on phonetics (1876, 19015). Sievers (19015:217) describes melodic accent as relating to ‘wechselnden Tonhöhen’ (‘changing pitch heights’), whereas dynamic accent or expiratory accent involves ‘Stärkeabstufungen’ (‘alternations in strength’). The distinction is also adopted by other early phoneticians such as Sweet (1906). In traditional descriptions of the evolution of Indo-European ‘languages with melodic accent’ sometimes refer to what we now call pitch accent languages (like (supposedly) Proto-Indo-European, Ancient Greek and Modern Swedish), but can also refer to other accentual languages in which accent is simpler. For instance, it was assumed that Classical Latin was a language with a melodic accent. In traditional handbooks on the history of Germanic languages (e.g. Hirt 1929:88ff), one of the changes characterizing the genesis of Proto-Germanic would have been the change from melodic to expiratory stress. 3.2  The distinction confronted with modern experimental phonetics Today the major distinction regarding the accentual nature of languages is no longer one between melodic and expiratory stress, but rather between accentual and tone languages. All Indo-European languages are accentual languages. However, there are indeed differences in the realization of accent. As one can read in any modern introduction to phonetics (e.g. Ladefoged 2001; Rietveld & Van Heuven 2001), accent or stress in accentual languages is a mixture of modifications of tone height (F0), duration, intensity, spectral expansion of vowels and of spectral tilt (change in 6.  In an extensive study of accent in Germanic, D’Alquen (1988:17, 232) concludes that at the time of the working of Verner’s law (1st and 2nd century A.D.), Germanic accent was melodic. There is no indication at all that by the 5th century it had all at once become strongly ‘expiratory’.

 Roland Noske

the relative intensity of the signal in the audible frequency spectrum). Every language has its own specific mix of these correlates and this mix is usually different for primary, secondary and focal stress. But, as pointed out by Ladefoged (2001), In nearly every language (…) what we hear as stress is more a matter of increasing pitch and length of the syllables than of increasing their loudness (2001:23, emphasis mine, RN).

Moreover the perceptibility of intensity (hence strength of expiration) was already questioned early in the 20th century by Saran (1907). Subsequently it was fully dismissed by Mol & Uhlenbeck (1956) in an article written at a time when phonetic measurements had become much more refined. Therefore, from the 1950s onwards, reference to expiratory stress disappeared from the literature, except for some articles by historical linguists uninformed of developments in experimental phonetics.

4.  Other changes in French and Frankish The arguments advanced in Sections 2 and 3 suffice by themselves to refute the claims concerning the influence of Frankish stress on Gallo-Romance. They lead to the conclusion that the evolution towards vowel reduction and deletion as well as to diphthongization in Gallo-Romance was an autonomous development, which was not the result of influence by a putative expiratory stress. We therefore have to reconsider the developments of the prosodic system in the history of French and continental West Germanic. Then we will see that the phenomena in Gallo-Romance and Germanic discussed above fit into a much more general picture. I will show in Section 5 that they are part of more general, partially opposite, typological movements in Gallo-Romance and Germanic. But first, in order to understand this, we take a brief look in this section at other relevant aspects of the evolution of French and West-Germanic. 4.1  Other aspects of the evolution of French Apart from the processes of vowel reduction and vowel deletion mentioned in Sections 1 and 2, we should take a look at other changes in French. We will look at the general changes of syllable structure as well as at segmental changes. 4.1.1  A pendular movement in syllable structure As often observed (cf. Jacobs 1992), there was a pendular movement in syllabic structure in the evolution from Classical Latin to Modern French. The different stages of the development can be characterized as follows:



Autonomous typological prosodic evolution 

i. from Classical Latin to Late Latin: by a development towards a less complicated, more open, syllabic structure (by the working of several epenthesis processes); ii. from Late Latin to Gallo-Romance and further on to early Old French: by a more complex syllable structure; iii. from early Old French to late Old French, Middle French and Modern French: by a progressive movement towards a much more open syllable structure. This evolution can be schematized as: closed > open > closed > open. 4.1.2  Diphthongization A process of diphthongization was operative in Gallo-Romance and Old French in stressed syllables (Pope 1952:60–62, 103–104; La Chaussée 1989:182, 185, 187, 194). Examples: mel [mel] > [miɛ̯l] “honey”, cor [kɔr] > [kuɔ̯r] “heart”, mare [ˈmaːre] > [ˈmaɛ̯rə] “sea”. 4.1.3  Degemination The geminate consonants of Late Latin and those resulting from assimilation in the Gallo-Romance period were reduced to single consonants in Old French. This happened from the 9th century onwards (Bourciez 1930:305; Pope 1952:147). 4.1.4  The genesis of final devoicing As we have seen above in Subsection 4.1.1, stage (ii) is characterized by the reduction of unstressed vowels and by different deletion processes of unstressed vowels. In addition, there was, from the 7th century onwards, a process of final obstruent devoicing, (Pope 1952:98), cf. the examples given in (7), transcribed into IPA: (7)

a. b. c. d. e.

[luŋɡum] [ɡrɑndem] [perdo] [riːsum] [serwum]

> [luŋk] > [ɡrãnt] > [pεrt] > *[rizʊ] > [ris] > *[sɛrvə] > [sɛrf]

“long” “big” “lose” “laugh” “slave, servant”

Pope remarks that this process has taken place in a period characterized by a strong intensity accent. In language typology, so-called word languages tend to have more boundary signals (like devoiced obstruents at word endings) than so-called syllable languages (see the discussion on word versus syllable languages in Section 5). 4.1.5  Loss of productivity of vowel reduction During stage (iii), the evolution towards Middle French and further on to Modern French, vowel reduction was no longer productive: unstressed full vowels introduced into the language at the time, e.g. in loan words, were no longer reduced to schwa.

 Roland Noske

4.1.6  Loss of productivity of final consonant devoicing During the same period, final devoicing disappeared from the language (apart from a few remnants like the lexicalized alternations in neuf ~ neuve “new” and grand ami [ɡrɑ̃tami] ~ grande amie [ɡrɑ̃dami] “big friend” (masc. ~ fem.). Below, in Section 5, I show that the loss of consonant devoicing is not accidental, but, together with other changes, is the result of a typological development. 4.1.7  A change in the conditioning of final vowel deletion Vowel deletion (of schwa) has remained in Middle and Modern French, but has taken another role: instead of being conditioned by stress, it is conditioned by syllable structure optimization, like, among others, the tendency to arrive at a CV syllable, cf. the deletion of schwa in prevocalic position in (8):

(8) le + homme > l’homme (/lə+ɔm/ > [lɔm]) “the man, mankind”

I will show that this change in conditioning of schwa deletion is also a consequence of the typological evolution of French.7 4.2  Some aspects of the evolution of continental West Germanic Like French, Frankish and the other dialects of West Germanic have undergone substantial changes from the period of the first written sources. The relevant changes are: i. the genesis, already mentioned in Section 2.3, of the process of vowel reduction in the transition of Old High German and Old Dutch towards Middle High German and Middle Dutch; ii. the development of syncope/apocope in MHG and Middle Dutch;8 iii. the genesis in MHG, and partially already in Old Dutch, of final devoicing;9 iv. the evolution from an accentual system stressing the first root syllable in Common Germanic towards a system where stress was assigned from the right edge of the word (modulo the stress on the initial parts of compounds). v. the reduction of geminate consonants from the period of Middle High German and Middle Dutch onwards. 7.  Schwa deletion in modern French has been the object of a vast literature in modern phonology. Apart from syllabic structure, it mentions among other things, rhythm and morphology as conditioning factors, cf. Verluyten (1988:4–10). 8.  For High German, this was a continuous evolution during several centuries. From the MHG period onwards, more and more vowels were deleted, which gave rise to consonant clusters of increasing complexity (Werner 1978). This evolution is called Konsonantenhäufung ‘crowding of consonants’ by linguistic historians of German. 9.  According to Quak (1997) and other descriptions of the history of Dutch, the genesis of final devoicing in Dutch preceded that in High German.



Autonomous typological prosodic evolution 

5.  Syllable counting languages and stress counting languages It will become clear in this section that the processes in early French and West-Germanic can be understood not by the presence of a putative expiratory accent, but by the evolution of the prosodic systems of the languages in general. I make use of the theory of syllable languages vs. word languages as proposed by Auer (1993, 1994, 2001) and by Auer & Uhmann (1988). This theory is treated in Subsection 5.1. Then, in Subsection 5.2, I show that the evolution of French, as well as that of West-Germanic can be understood much more fully in this framework. 5.1  Syllable vs. word languages Auer (1993, 1994, 2001) and Auer & Uhmann (1988) propose a theory of syllable counting languages (or simply ‘syllable languages’) versus stress counting languages (or simply ‘word languages’). In order to understand this theory however, we should briefly dwell upon three previous theories regarding prosodic organization. We consider the theories on syllable vs. stress timed languages by Pike (1945; see also Abercrombie 1967) in Subsection 5.1.1, the perceptual theory by Dauer (1983, 1987) in Subsection 5.1.2 and the model of prosodic phonology by Nespor & Vogel (1986) in Subsection 5.1.3. 5.1.1  Syllable timed languages versus stress timed languages According to the theory of syllable timed languages vs. stress timed languages proposed by Pike (1945) and Abercrombie (1967), there are two types of languages: those which maintain an equal temporal distance between syllables and those which keep an equal distance between stressed syllables. The first group, syllable timed languages, would include languages like Spanish and French, whereas the second group, stress time languages, would contain among others English and German. With the further development of the study of acoustic phonetics, however, this idea was refuted by phonetic measurements. For instance, Wenk & Wioland (1982) show that the distances between French syllables are all but equal, and thus French does not neatly fit under the syllable-timed category. 5.1.2  The perceptive theory by Dauer Despite its multiple refutations, the idea of syllable vs. stress timed languages continues to persist, as it intuitively appeals to linguistically skilled listeners. When confronted with sounds of an unknown language, linguists seem to agree whether the language in question is more of the syllable timed sort or more of the stress timed sort (see Dauer 1983:52–54).

 Roland Noske

Therefore, it seems that perception is involved. Dauer (1983) noted this and showed that languages that are perceived as syllable timed share the following characteristics: (9) Characteristics of languages perceived as syllable timed: i. a relatively simple syllable structure ii. the possibilities of contrast between stressed and unstressed syllables are the same (no vowel reduction) iii. word accent is weak or non-existent

In contrast, characteristics of languages perceived as stress timed are: (10) Characteristics of languages perceived as stress timed: i. complex syllable structure ii. existence of vowel reduction in unstressed syllables iii. a clearly perceptible word accent, and, on top of that, grammatical rules referring to the place of stress

In a later paper, Dauer (1987) added that languages perceived as stress timed have more allophonic variation (e.g. final devoicing, intervocalic voicing) and have more length differences. Dauer concluded from this that languages perceived as syllable timed and as stress timed in fact differ in the above mentioned characteristics and that it is through these characteristics that the perceptual distinction between the two types of languages is in fact made. 5.1.3  The prosodic phonology of Nespor & Vogel Nespor & Vogel (1986) propose a system of phonological constituents which are organized in a hierarchical fashion. Under this system, one or more constituents are licensed by a constituent belonging to an immediately higher one, implying that a constituent cannot belong to two higher constituents at the same time. The categories of the constituents are: i. the phonological syllable (σ) ii. the foot (F) iii. the phonological word (ω) iv. the clitic group (C) v. the phonological phrase (φ) vi. the intonational phrase (I) vii. the utterance (U) 5.1.4  The theory by Auer and Uhmann Auer (1993, 1994, 2001) and Auer & Uhmann (1988) combine the ideas of Dauer and Nespor & Vogel. This enables them to propose a scalar, multifactorial typology



Autonomous typological prosodic evolution 

the extremes of which are syllable counting languages (or simply syllable languages), and languages that count stressed syllables, or word languages. For the syllable languages, the syllable is the main prosodic constituent and these languages have characteristics like the ones in (9). In the word languages, on the other hand, the prosodic word is the main prosodic constituent. These languages share the characteristics in (10). A more precise list of properties is given in Table 1, adapted from Nübling & Schrambke (2004): Table 1.  Prototypical properties of syllable (syllable counting) versus word (stress counting) languages nr.

indicator

syllable languages → syllable word / accent languages → stress counting counting phonological word as basic prosodic unit syllable as basic prosodic unit (syllable length variable) (foot length variable)

1

syllable structure

CV syllables (rarely closed syllables); all syllables equally long

2

syllable well defined, constant boundaries syllable boundaries geminates geminates possible

3

4

stress effects

5

stress mostly syllable based; assignment absence of fixed word stress possible phonotactics regular, stable phonotactics, no positionally determined allophones vocalism little discrepancy between strongly and weakly stressed vowels, relatively equal tenseness vowel possible harmony, umlaut vowel for reasons of syllable deletion structure optimization liaison yes (across morpheme boundaries)

6 7

8 9 10

no/few differences in structure of stressed vs. unstressed syllables

variable syllable types of different complexity, dependent on the stress position; often differences between medial and peripheral syllables ill-defined, variable, speech-rate dependent syllable boundaries geminate reduction, except in places where they are morphologically relevant, e.g. in internal compound boundaries e.g. German Schifffahrt [fː] stressed syllables are heavy, unstressed syllables are light; diphthongization of stressed vowels, aspiration of initial plosives of stressed syllables rules of stress assignment (complex) are morphologically/lexically/semantically determined word boundary (delimitative) signals, positionally determined allophones (initial, medial, final), phonotactic restrictions much discrepancy between strongly and weakly stressed vowels (German, Danish, English). Heavy stress: often difference in length, centralizations (reductions) rare conditioned by stress no (border signals/junctures, e.g. glottal stop)

 Roland Noske

Of course, word languages and syllable languages are prototypes. In reality languages are situated somewhere between the two extremes. Similarly, not every indicator in the table is relevant for every language. The ideas put forth by Auer and Uhmann have up till now received scant attention, perhaps because the majority of their publications is in German. 5.2  Modern French as a syllable language and Old French as a word language All but one of the indicators given above in Table 1 show that Modern French is a syllable language: Table 2.  Indicators of Table 1 showing that Modern French is a syllable language nr.

indicator

reason

1

syllable structure syllable boundaries

many open syllables (see Section 4.1.1)

2 4 5 6 7 9 10

syllable boundaries are not blurred like in Modern German and Modern English (cf. the co-called ambisyllabic consonants in these languages), but clear cut stress effects stressed syllables are not phonologically longer than unstressed syllables stress stress is based on syllabic structure, not on morphology; there are no assignment minimal pairs of words that contrast only in the place of stress, like in Germanic languages phonotactics no positionally determined allophones, no intervocalic voicing, no final devoicing (see Section 4.1.6) vocalism little discrepancy between strongly and weakly stressed vowels, no (synchronic) vowel reduction (see Section 4.1.5) vowel vowel deletion because of reasons of syllable structure optimization (see deletion Section 4.1.7) liaison for reasons of syllable structure optimization

If one applies the criteria of Table 1 to Old French, one sees that there are at least six criteria that militate in favor of its categorization as a word language: Table 3.  Indicators of Table 1 showing that Old French is a word language nr.

indicator

reason

1

syllable structure

3 4

geminates stress effects

6 7 9

phonotactics vocalism vowel deletion

many complex syllables, much variability between syllable types (see Section 4.1.1) geminate reduction (see Section 4.1.3) diphthongizations from the 3rd to the 6th centuries (see Section 4.1.2) final devoicing as boundary signal (see Section 4.1.4) reduction of unstressed vowels to schwa vowel deletion (syncope, apocope) conditioned by stress (see Sections 1 and 4.1.7)



Autonomous typological prosodic evolution 

As one can see by these criteria, between the periods of Old French and Modern French, the language changed from a word language to a syllable language. 5.3 Modern Dutch and German as word languages and Old High German and Old Dutch as syllable languages Let us now examine High German and Dutch by the same criteria. First, we look at Modern High German and Modern Dutch, cf. Table 4: Table 4.  Indicators of Table 1 showing that Modern High German and Modern Dutch are word languages nr.

indicator

reason

1

syllable structure

2 3

syllable boundaries geminates

4

stress effects

5

stress assignment

6 7 8

phonotactics vocalism vowel harmony, umlaut

9

vowel deletion

10

liaison

complex syllables, variable syllable types, determined by stress and morphology ill-recognizable syllable boundaries, ambisyllabicity geminate reduction from MHG and Middle Dutch onwards aspiration of initial plosives of stressed syllables in Modern High German, phonetic diphthongizations of stressed vowels in Modern Dutch complex stress rules, dependent on morphology, preference to stress heavy syllables boundary signals: glottal stop insertion, final devoicing productive vowel reduction in Modern Dutch phonological umlaut no longer productive in German, umlaut has become morphologized syncope and apocope conditioned by stress, like in German ich hab’ (for ich habe), Dutch vreeslijk (for vreselijk). in general non-existent, instead: glottal stop insertion

By contrast, OHG and Old Dutch occupied a place on the scale close to the syllable language prototype: Table 5.  Indicators of Table 1 showing that OHG and Old Dutch behaved like syllable languages nr.

indicator

reason

1

syllable structure

2 3 6 7 8

syllable boundaries geminates phonotactics vocalism vowel harmony, umlaut

much less complex than in Modern High German and Modern Dutch, many open syllables clear syllable boundaries exist in heterosyllabic position no final devoicing in OHG no vowel reduction in OHG and Old Dutch productive in OHG

 Roland Noske

It can be concluded that, while Modern High German and Modern Dutch are quite clearly word languages, Old High German and Old Dutch were still relative syllable languages.10 Hence, their evolution is the opposite of that of French after the Old French period.

6.  Conclusion In this article I have shown that the (putative) strong intensity accent of GalloRomance cannot have had a Frankish source, because around the 5th century, the period in which this putative borrowing would have taken place, Frankish, by the same criteria, cannot have had a strong intensity accent. Another, independent, reason that was advanced is that if the strong accent had indeed been borrowed from Frankish, the root-initial stress would also have been borrowed. And this did not happen. In addition to this, I showed that the entire distinction between ‘melodic’ and ‘expiratory’ accent is outdated, because it has been shown to be linguistically irrelevant due to the imperceptibility of intensity gradation. Using the typology introduced by Auer and Uhmann (1988), I then showed that evolution in French and in West-Germanic are changes in the entire prosodic systems of these languages, along the axis of syllable and word languages. It was shown that from the Early Middle Ages onwards, French and Frankish developed autonomously in opposite directions; while French has become a syllable language, High German and Dutch, the descendants of Frankish, became word languages. In retrospect, it is clear that the authors of works that attribute vowel reduction and deletion and a putative strong intensity stress in Old French to the influence of Frankish were the victims of the outdated ideas prevalent in phonetics and historical linguistics in the last quarter of the 19th century and the first half of the 20th century. However, given the advances in experimental phonetics and language typology since, there is no excuse not to abandon these ideas.

10.  Nübling and Schrampke (2004:290) refer to Frey (1988) who provides arguments that OHG is a syllable language. At the completion of this article, I became aware of a recently published book, Szczepaniak (2007), which is totally dedicated to the evolution of High German from a syllable to a word language.



Autonomous typological prosodic evolution 

References Abercrombie, David. 1967. Elements of General Phonetics. Edinburgh: Edinburgh University Press. Appel, René & Pieter Muysken. 1987. Language Contact and Bilingualism. London: Edward Arnold. Árnason, Kristján. 1996. “How to Meet the European Standard: Word stress in Faroese and Icelandic”. Nordlyd 24. 1–22. Auer, Peter. 1993. “Is a Rhythm-based Typology Possible? A Study of the Role of Prosody in Phonological Typology”. KontRI Working Paper 21. Ms., University of Konstanz. Downloadable from : http://www.germanistik.uni-freiburg.de/auer/?Personal:Prof._Dr._Peter_Auer:Publikationen. Auer, Peter. 1994. “Einige Argumente gegen die Silbe als universale prosodische Hauptkategorie”. Universale phonologische Strukturen und Prozesse ed. by Karl-Heinz Ramers et al. 55–78. Tübingen: Niemeyer. Auer, Peter. 2001. “Silben- und akzentzählende Sprachen”. Language Typology and Language Universals. An International Handbook ed. by Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible, 1391–1399. Berlin: de Gruyter. Auer, Peter & Suzanne Uhmann. 1988. “Silben- und akzentzählende Sprachen”. Zeitschrift für Sprachwissenschaft 7:2. 214-59. Bourciez, Edouard. 1930. Elements de linguistique romane. Paris: Klinksieck. Braune, Wilhelm. 1994. Althochdeutsches Lesebuch. 17th edition. Tübingen: Niemeyer. Brüch, Joseph. 1913. Der Einfluss der germanischen Sprachen auf das Vulgärlatein. Heidelberg: Carl Winter. Cohen, Marcel. 1967. Histoire d’une langue: le français. 3rd edition. Paris: Editions sociales. Cowan, Hendrik K.J. 1957. De Oudnederlandse (Oudnederfrankische) Psalmenfragmenten. Leiden: Brill. D'Alquen, Richard. 1988. Germanic Accent, Grammatical Change and the Laws of Unaccented Syllables. New York: Lang. Dauer, Rebecca. 1983. “Stress and Syllable-timing Reanalysed”. Journal of Phonetics 11. 51–62. Dauer, Rebecca. 1987. “Phonetic and Phonological Components of Language Rhythm”. Proceedings of the XIth International Congress of Phonetic Sciences, Vol. 5, 447–450. Tallinn: Academy of Sciences of the Estonian S.S.R. Fournier, Jean-Michel. 2007. “From a Latin Syllable-driven Stress System to a Romance versus Germanic Morphology-driven Dynamics: In honour of Lionel Guierre”. Language Sciences 29. 218–236. Frei, Evelyn. 1988. Wortteilung und Silbenstruktur im Althochdeutschen. Munich: CD Copy und Druck. Good, Jeff. 2009. “A Twice-mixed Creole? Tracing the History of a Prosodic Split in the Saramaccan Lexicon”. Studies in Language 33:2, 459–498. Halle, Morris & Samuel J. Keyser. 1971. English Stress: its form, its growth and its role in verse. New York: Harper & Row. Helten, Willem van. 1902. Die altostniederfränkischen Psalmenfragmente, die Lipsiusschen Glossen und die altsüdmittelfränkischen Psalmenfragmente. Groningen: Wolters. Hirt, Hermann A. 1929. Indogermanische Grammatik. Vol. 5: Der Akzent. Heidelberg: Carl Winter. Jacobs, Haike. 1992. “The Interaction between the Evolution of Syllable structure and Foot Structure in the Historical Phonology of French”. Theoretical Analyses in Romance Linguistics ed. by Morgan Terrel & Christiane Laueffer, 55–79. Amsterdam: Benjamins.

 Roland Noske Jacobs, Haike. 2003. “On the Change from Left to Right Word-Edge Main Stress in Icelandic, Polish and Latin”. Nordic Journal of Linguistics 26:2. 259–282. König, Werner. 1978. DTV-Atlas zur Deutschen Sprache. Munich: Deutscher Taschenbuch Verlag. Kukenheim, Louis. 1971. “Rôle de la prosodie dans l’histoire de la langue française”. Mélanges de Philologie romane dédiés à Jean Boutière ed. by Irénée Cluzel & François Pirot, vol 1, 317–331. Liège: Soledi. Kyes, R.L. 1969. The Old Low Franconian Psalms and Glosses. Ann Harbor: University of Michigan Press. La Chaussée, François de. 1989. Initiation à la phonétique de l’Ancien Français. Nouvelle édition revue et augmentée. Paris: Klincksieck. Ladefoged, Peter. 2001. Vowels and Consonants. Malden, Mass.: Blackwell. Lahiri, Aditi, Tomas Riad & Haike Jacobs. 1999. “Diachronic Prosody”. Word Prosodic Systems in the Languages of Europe ed. by Harry van der Hulst, 335–422. Berlin: de Gruyter. Mol, Hendrik & Eugenius M. Uhlenbeck. 1956. “The Linguistic Relevance of Intensity in Stress”. Lingua 5. 205–213. Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris. Nübling, Damaris & Renate Schrambke. 2004. “Silben- versus Akzentsprachliche Züge in Germanischen Sprachen und im Alemannischen”. Alemannisch im Sprachvergleich. Beiträge zur 14. Arbeitstagung für alemannische Dialektologie in Männedorf (Zürich) vom 16.–18.9.2002 ed. by Elvira Glaser, Peter Ott, & Rudolf Schwarzenbach, 280–320. Stuttgart: Franz Steiner. Pike, Kenneth. 1945. The Intonation of American English. Ann Arbor: University of Michigan Press. Pope, Mildred. 1952. From Latin to Modern French. 2nd edition. Manchester: Manchester University Press. Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia: University of Philadelphia Press & Linguistic Society of America. Quak, Arend. 1997. “Oudnederlands”. Geschiedenis van de Nederlandse taal ed. by Maarten C. van den Toorn, Wilhelmus J.J. Pijnenburg, Jan A. van Leuvensteijn & Joop M. van der Horst, 39–68. Amsterdam: Amsterdam University Press. Richter, Elise. 1934. Beiträge zur Geschichte der Romanischen. I. Chronologische Phonetik des Französischen bis zum Ende des 8. Jahrhunderts. Halle (Saale): Niemeyer. Rietveld, Anton & Vincent van Heuven. 2001. Algemene Fonetiek. Tweede, aangevulde druk. Bussum: Coutinho. Saran, Franz. 1907. Deutsche Verslehre. München: Beck. Schönfeld, Moritz. 1933. “Een Oudnederlandsche Zin uit de Elfde Eeuw (met reproduktie)”. Tijdschrift voor Nederlandsche Taal en Letterkunde 52. 1–8. Sievers, Eduard. 1876 (19015). Grundzüge der Lautphysiologie zur Einführung in das Studium der Lautlehre der indogermarmanischen Sprachen. Leipzig: Breitkopf & Härtel. (Later editions bear the title: Grundzüge der Phonetik.) Sweet, Henry. 1906. A Primer of Phonetics. 3rd, revised edition. Oxford: Claredon Press. Szczepaniak, Renata. 2007. Der phonologisch-typologische Wandel des Deutschen von einer Silbenzu einer Wortsprache. Berlin: de Gruyter. Thomason, Sarah G. 2001. Language Contact. Edinburgh: Edinburgh University Press. Verluyten, S. Paul. 1988. “Introduction”. La phonologie du schwa français ed. by S. Paul Verluyten, 1–13. Amsterdam: Benjamins.



Autonomous typological prosodic evolution 

Wartburg, Walther von. 1965. Evolution et Structure de la Langue Française. 7e édition révisée. Bern: Francke. Weinrich, Uriel. 1963. Languages in Contact. The Hague: Mouton. Wenk, Brian & François Wioland. 1982. “Is French Really Syllable-timed?” Journal of Phonetics 10. 193–216. Werner, Otmar. 1978. “Schwa-Schwund und Phonotaktik im Deutschen”. Studia Linguistica Alexandro Vasilii filio Issatschenko a Collegis Amicisque oblata ed. by Henrik Birnbaum et al. 471–486. Lisse: The Peter de Ridder Press. Winford, Donald. 2003. An Introduction to Contact Linguistics. Malden, Mass.: Blackwell. Zink, Gaston. 1986. Phonétique Historique du Français. Paris: Presses Universitaires de France.

Dummy prepositions and the licensing of null subjects in Brazilian Portuguese* Jairo Nunes

Universidade de São Paulo This paper discusses constructions in Brazilian Portuguese in which a null subject is adequately licensed only if the clause containing it is preceded by the dummy preposition de. Assuming that referential null subjects in Brazilian Portuguese are traces of A-movement (see Ferreira 2000, 2009 and Rodrigues 2002, 2004), I argue that the presence of de signals whether or not the embedded clause has received inherent Case and this ultimately determines whether the embedded subject is allowed to move from within its clause and whether it should undergo upward or sideward movement (see Nunes 2001, 2004).

1.  Introduction In this paper I examine the role that dummy prepositions appear to play in the licensing of null subjects in (Colloquial) Brazilian Portuguese (henceforth BP). I will be specifically concerned with the three types of constructions exemplified below. (1) a.

Ninguém concordou com a sugestão do João (de) que nobody agreed with the suggestion of.the João  of that



o Pedro devia escrever o relatório. the Pedro should write the report



“Nobody agreed with João’s suggestion that Pedro should write the report.”

b. Ninguém concordou com a sugestão d[o João]i *(de) que nobody agreed with the suggestion of.the João   of that eci devia escrever o relatório. should write the report

“Nobody agreed with Joãoi’s suggestion that hei should write the report.”

*Parts of this paper have been presented at Going Romance 2007, the XVIII Colloquium on Generative Grammar, and the Universities of São Paulo, Connecticut, and Leiden. I would like to thank these audiences for their comments and suggestions. Special thanks to Željko Bošković, Rerisson Cavalcante, João Costa, Mary Kato, Renato Lacerda, Marcello Modesto, Carol Petersen, Josep Quer, Leonor Simioni, and two anonymous reviewers. The results reported here are part of research projects supported by CNPQ (302262/2008-3) and FAPESP (2006/00965-2).

 Jairo Nunes

(2) a.

[o João]i convenceu [a Maria]k que eci/*k tinha que sair.  the João convinced  the Maria that had that leave “João convinced Maria that he/*she had to leave.”

b. [o João]i convenceu [a Maria]k deˉ queˉeci/k tinha que sair.  the João convinced  the Maria of that had that leave “João convinced Maria that he/she had to leave.” (3) a.

É difícil (d)esses professores elogiarem alguém. is difficult of.these teachers praise.inf.3pl someone “These teachers rarely praise someone.”

b. Esses professores são difíceis *(de) elogiarem alguém. these teachers are difficult   of praise.inf.3pl someone “These teachers rarely praise someone.”

(1) shows that the dummy preposition de may in general optionally precede a noun complement clause, but is obligatorily required if the subject of the noun complement clause is null. In turn, (2) shows that the presence of de may change the interpretation of a null subject, also allowing the matrix object as its antecedent. Finally, (3) shows that whereas some impersonal predicates optionally allow de to precede their infinitival complements, the hyper-raising counterpart of these constructions obligatorily requires this dummy preposition. The paper is organized as follows. In Section 2, I briefly present the proposal that referential null subjects in BP are traces of A-movement, rather than null pronominals. In Sections 3, 4, and 5 I analyze each of the constructions in (1)–(3), discussing why the presence of a dummy preposition preceding a CP in BP may interfere with the licensing of a null subject/A-trace within it. With respect to (1) and (2), I argue that the dummy preposition may signal whether or not the relevant CP is a complement and this may ultimately determine if the DP in its subject position should undergo upward or sideward movement (in the sense of Nunes 2001, 2004). It will be shown that sideward movement provides a licit derivation for constructions such as (2), but not for constructions such as (1b). As for the contrast in (3), I argue that as a marker of inherent Case, the preposition freezes the infinitival for purposes of A-movement, thereby allowing A-movement of its subject without yielding a violation of Chomsky’s (1964) A-over-A Condition. Section 6 presents some concluding remarks.

2.  Referential null subjects in BP as traces of movement Exploring Hornstein’s (2001) movement analysis of obligatory control, Ferreira (2000, 2009) and Rodrigues (2002, 2004) convincingly argue that with the weakening



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

of the verbal agreement paradigm in BP (see e.g. Duarte 1995), its finite Ts ceased to license “referential”1 pro and referential null subjects came to be analyzed as traces of A-movement. Consider, for instance, the BP data in (4). (4) a. *Comprou um carro novo. bought.3sg a car new “She/he bought a new car.” b. *Parece que comprou um carro novo. seems that bought.3sg a car new “It seems that she/he bought a new car.” c.

[O João]i disse [que eci/*j comprou um carro novo].  the Joãoi said  that bought.3sg a car new “Joãoi said that hei/*j bought a new car.”

d. *O pai da Maria acha [que está grávida]. the father of.the Maria thinks [that is pregnant.fem] “Maria’s father thinks she is pregnant.” e. *A Maria disse [que o médico acha [que está grávida]]. the Maria said  that the doctor.masc thinks   that is pregnant.fem “Mariai said that the doctor thinks shei is pregnant.” f. *Eu encontrei o livro que perdi. I found the book that lost.1sg “I found the book that I had lost.”

1.  The qualification is meant to exclude null expletives as well as null “arbitrary” third person subjects − both plural and singular (see e.g. Galves 1987; Nunes 1990 and Rodrigues 2004) − which are still available in BP, as illustrated in (i) and (ii), respectively. (i) a.

Tinha vários livros na mesa. had several books on.the table “There were several books on the table.”

b.

Choveu ontem. rained yesterday “It rained yesterday.”

(ii) a.

Telefonaram para você. called.3pl to you “Someone called you.”

b.

No Brasil não usa mais saia. in.the Brazil not wear.3sg more skirt “In Brazil people don’t wear skirts anymore.”

 Jairo Nunes

(4a)–(4c) show that a referential null subject in BP requires an appropriate antecedent in the sentence;2 (4d), that the antecedent must be in a c-commanding position; (4e), that the antecedent must be local; and (4f), that the antecedent cannot be outside a strong island. (4f) is especially interesting in that there is an island effect even though the embedded verb perdi is inflected for first person singular and could in principle identify the null subject without the help of an antecedent. If referential null subjects in BP are A-traces rather than null pronominals, as argued by Ferreira and Rodrigues (see also Martins & Nunes 2005 and Nunes 2008), the fact that they require a local c-commanding antecedent in order to be licensed follows from familiar independent requirements on movement. In this paper I will assume the gist of Ferreira’s and Rodrigues’s proposal, relying on the specific technical implementation advanced by Ferreira and further refined by Nunes (2008). Assuming Chomsky’s (2000, 2001) Agree-based framework, Ferreira (2000, 2009) proposes that finite Ts in BP are ambiguous in being associated with either a complete or an incomplete set of ϕ-features. When the Case-assigning version of a finite T is selected (i.e. a ϕ-complete T), it assigns nominative to the subject, freezing it for further A-movement. If the non-Caseassigning version of a finite T is selected instead (i.e. a ϕ-incomplete T), the subject of its clause remains Caseless and can undergo further A-movement. From this perspective, a sentence such as (4c) is to be derived along the lines of (5), where the embedded T is ϕ-incomplete and the matrix T is ϕ-complete. (5) [TP [o João]i Tϕ-complete [vP ti disse [CP que [TP ti Tϕ-incomplete [vP ti the João said.3sg that comprou um carro novo]]]]] bought.3sg a car new

2.  Referential null subjects in matrix clauses are only allowed in BP as instances of topicdeletion in the sense of Ross (1982) (see Modesto 2000; Ferreira 2000 and Rodrigues 2004). Thus, the null subject in (iB) below is to be analyzed as a variable bound by a null topic and the presence of a wh-element in (iB’) yields a minimality violation. (i) A: Cadê o João? where the João “Where’s João?” B: Acabou de sair. finished.3sg of leave “He’s just left.” B’: *O que fez dessa vez? what did.3sg of.this time “What did he do this time?”



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

Nunes (2008) has reinterpreted the ambiguity of T proposed by Ferreira in terms of how its person and number features are combined in the course of the computation. More specifically, Nunes proposes that finite Ts in BP may enter the numeration specified for number and person or for number only. When T is only specified for number, well-formedness conditions in the morphological component trigger the addition of the feature person in accordance to the redundancy rule informally sketched in (6) below. To put it differently, person features in BP may be dissociated features in the sense of Embick (1997). Crucially, the paradigm of verbal agreement morphology in BP given in Table 1 is such that the only form that distinctively encodes person and number is the syncretic inflection for first person singular; the other two inflections involve a default value (third) for the person feature.

(6) When T is only specified for number (N):



(i) Add [P:1], if N is valued as sg; (ii) otherwise, add [P:default].

Table 1.  Verbal agreement paradigm in (Colloquial) Brazilian Portuguese cantar “to sing”: indicative present eu (I)

canto

P:1.N:SG

você (you.SG) ele (he) ela (she) a gente (we)

canta

P:default; N:default (= 3SG)

vocês (you.PL) eles (they.MASC) elas (they.FEM)

cantam

P:default; N:PL (= 3PL)

Thus, the three different verbal inflections available in Table 1 can be obtained either if T is specified for both person and number throughout the derivation, as in Table 2, or if T is only specified for number and the feature person is associated with T in the morphological component in accordance with (6), as shown in Table 3. Table 2.  Person and number specified cantar “to sing”: indicative present Valuation of T in the syntactic component

Surface form of the verb

P:1.N:SG P:default; N:default P:default; N:PL

canto canta cantam

 Jairo Nunes

Table 3.  Only number specified Valuation of T in the syntactic component

Addition of [person] in the morphological component

Surface form of the verb

N:SG N:default N:PL

P:1.N:SG P:default; N:default P:default; N:PL

canto canta cantam

In this view, a more appropriate representation of the derivation of (4c) is given in (7) below, where the embedded T is specified only for number, whereas the matrix T is specified for number and person. In other words, the subject is assigned (nominative) Case by the matrix T rather than the embedded T. Common in both representations in (5) and (7) is the movement of the embedded subject to the matrix Spec,vP before reaching the matrix Spec,TP, which accounts for the fact that o João in (4c) is interpreted as associated with the external θ-roles of both the matrix and the embedded verb (see Hornstein 2001). (7) [TP [o João]i T[P:default; N:default] [vP ti disse [CP que [TP ti T[N:default] the João said that [vP ti comprou um carro novo]]]]] bought a car new

To sum up, BP exercises an option that is generally restricted to non-finite clauses in other languages. Namely, it allows raising out of a finite embedded clause when its T is not a Case assigner (i.e. when it only has a number feature as it enters the numeration).3 In the sections that follow, all sentences with referential null subjects will be examined under the derivation in which the T head of the clause containing the null subject is only specified for number in the syntactic component.4

3.  For the purposes of the current presentation, I assume Ferreira’s (2000:55) suggestion that if C selects a ϕ-incomplete T, it does not define a strong phase (see Nunes 2008 and Martins & Nunes forthcoming for relevant discussion). Under the assumption that successive cyclic movement to Spec,CP is driven by the Phase Impenetrability Condition (Chomsky 2000, 2001), A-movement from within a TP with ϕ-incomplete T does not need to pass through the Spec,CP, as the CP dominating such TP is not a strong phase. 4.  Recall that if the relevant T is fully specified, it will Case-mark its subject, which will then become frozen for purposes of A-movement. In turn, if the subject becomes frozen, no null subject construction of the relevant sort (see Fn. 1) will be derived. In other words, if the embedded T in (7) is specified for both person and number, the movement of the embedded subject to the matrix Spec,vP is to be ruled out for the same reason the movement of the Case-marked object to Spec,vP in (i) below is blocked (see Hornstein 2001 for discussion): In



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

3.  Finite control into noun complement clauses In BP, dummy prepositions such as de, which are required before nominal complements of verbs like gostar “like”, are optional when preceding CP complements, as illustrated in (8) below. The optionality in (8b) should be qualified, though. The version without the preposition is by far the most commonly used construction; the version with the preposition is generally associated with formal style and written language. Similar considerations apply to the noun complement clauses illustrated in (9). (8) a.

Eu gosto *(de) você. I like   of you “I like you.”

b. Eu gostaria (de) que você viesse. I would.like  of that you come.past.subj “I would like you to come.” (9) a.

A hipótese (de) que a Terra é chata não foi esquecida. the hypothesis   of that the Earth is flat not was forgotten “The hypothesis that the Earth is flat was not forgotten.”

b. A idéia do João (de) que você seja o candidato the idea of.the João  of that you be.subj the candidate

é bem boa. is rather good



“João’s idea that you become the candidate is very good.”

c.

O boato (de) que o Pedro seria despedido era falso. the rumor of that the Pedro would.be fired was false “The rumor that Pedro would be fired was false.”

d. Ele comentou a alegação do João (de) que a he commented the allegation of.the João  of that the

Ana era inocente. Ana was innocent



“He commented on João’s allegation that Ana was innocent.”

both derivations, the relevant DP becomes inactive for purposes of A-movement after it has its Case-feature valued (see Chomsky 2000, 2001). (i) a. *John saw. “John saw himself.”

b. *[tp Johni [vP ti [vp saw ti]]]

 Jairo Nunes

What has not been noticed in the literature is that the presence of the dummy preposition is obligatorily required in constructions involving a referential null subject inside the noun complement clause, as shown in (10) below, for all registers and styles. Interestingly, if the null subject is not referential (see Fn. 1), the preposition is again optional, as illustrated in (11). (10) a.

A hipótese do João *(de) que vai ser eleito é de rir. the hypothesis of.the João   of that goes be elected is of laugh “John’s hypothesis that he’s going to be elected is laughable.”

b. A afirmação do João *(de) que fez o trabalho é falsa. the affirmation of.the João   of that did the job is false “João’s statement that he did the job is false.” (11) a.

A afirmação (de) que nunca chove aqui é exagerada. the affirmation  of that never rains here is exaggerated “The claim that it never rains here is an exaggeration.”

b. A hipótese do João (de) que não existe movimento-wh the hypothesis of.the João of that not exists wh-movement

nessa língua parece estar errada. in.this language seems be wrong

“John’s hypothesis that wh-movement doesn’t exist in this language seems to be wrong.”

The paradigm in (7)–(11) raises the following questions: (i) why is it that referential and non-referential null subjects behave differently as far as obligatory presence of de is concerned?; and (ii) why is it that null referential subjects require the presence of the dummy preposition? Let us consider each of these questions in turn. The difference between referential and non-referential subjects should in principle be related to the nature of referential null subjects in BP, i.e. to their being traces rather than null pronominals (see discussion and references in Section 2). In fact, referential null subjects inside noun complement clauses also behave as A-traces, as can be seen in (12). (12) a. *O boato de que ec renunciou era falso. the rumor of that resigned was false “The rumor that she/he resigned was false.” b. [o João]i criticou a proposta d[o amigo d[o Pedro]j]k de  the João criticized the proposal of.the friend of.the Pedro of

queˉeck/*i/*j/*w devia reclamar com o diretor. that should complain with the director

“Joãoi criticized [Pedroj’s friend]k’s proposal that hek/*i/*j/*w should complain to the director.”



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

c. *[o João]i ficou chateado com os boatos que foram gerados  the João got upset with the rumors that were caused

pela notícia de que eci renunciou. by.the news of that resigned

“João got upset because of the rumors that were triggered by the news that he resigned.”

If the referential null subjects inside the noun complement clauses in (12) are A-traces, the ungrammaticality of (12a) should be due to the lack of an antecedent for the trace. In turn, (12c) should be out because movement from within the noun complement clause to the position occupied by o João crosses the relative clause island. Finally, if the moving element must land in the closest c-commanding A-position, (12b) should only be licit if the null subject is interpreted as o amigo do Pedro “Pedro’s friend”, which is indeed the case. That movement of the subject of the noun complement clause targets a θ-position is shown by the data in (13) and (14). (13) a.

A afirmação do João de [que a Maria fez o trabalho] the affirmation of.the João of  that the Maria did the job



é falsa. is false



“João’s claim that Maria did the job was false.”

b. *A probabilidade do João de [que a Maria tenha feito the probability of.the João of  that the Maria has.subj done

o trabalho] é alta. the job is high



“*João’s probability that Maria did the job is high.”

(14) a.

A afirmação do João de [que fez o trabalho] é falsa. the affirmation of.the João of  that did the job is false “João’s claim that he did the job is false.”

b. *A probabilidade do João de [que tenha feito o trabalho] the probability of.the João of  that has.subj done the job

é alta. is high



“The probability that João did the job is high.”

The contrast in (13) is due to a violation of the θ-Criterion in (13b), as probabilidade “probability” does not have a θ-role to assign to o João.5 The fact that

5.  In response to a reviewer’s question, it is worth pointing out that if one resorts to the θ-Criterion, this does not ensure a commitment to D-Structure, for the θ-Criterion can apply

 Jairo Nunes

the contrast in (13) is replicated in (14) can now be interpreted in terms of the Case Filter. Recall that A-movement out of an embedded finite clause in BP is possible only if the embedded T has just a number feature. If this is the case in (14a), we are led to the conclusion that o João is Case-marked after it moves to the θ-position associated with afirmação “affirmation”. The difference between (14a) and (14b) is now accounted for if the dummy preposition de is a marker of inherent Case. Once inherent Case must be associated with a θ-role (see Chomsky 1986), movement of the embedded subject allows it to comply with the Case Filter in (14a), but not in (14b), as probabilidade does not have a θ-role to assign to the moved DP (cf. (13b)). Thus, the subject of (14a) is to be represented along the lines of (15), with the preposition de being added to o João in the morphological component.6 (15)

[a [afirmaçãok [NPˉ[o João]iˉtk deˉ[CPˉ queˉ[ti fez o trabalho]]]]]  the  affirmation    ˉˉthe João of that did the job

Let us now address the question of why the movement of the embedded subject depicted in (15) is only licit if the CP is also preceded by the dummy preposition de. Building on Stowell (1981), I would like to propose that in BP the presence or absence of de in these constructions respectively signals whether we are dealing with a true complement structure or an appositive of sorts. More specifically, I take de in these constructions in BP to be the realization of the inherent Case assigned to the embedded clause.7 Take the data in (16) below, for example, which do not allow the presence of de. If the preposition encodes a noun complement configuration in virtue of realizing inherent Case, its presence in (16) yields unacceptable results as these sentences involve a predication configuration. (16) a.

A hipótese é (*de) que o João tenha feito isso. the hypothesis is   of that the João has done this “The hypothesis is that João did this.”

perfectly well at LF (even within the GB model). Moreover, the movement theory of control proposed by Hornstein (2001), which I am following here, still keeps the θ-Criterion in the sense that predicates are required to assign their θ-roles and expressions merged within the thematic domain of a given predicate must be assigned a θ-role. What is actually abandoned in Hornstein’s system are the additional assumptions that a given expression cannot bear more than one θ-role and that θ-role assignment is not a licensing condition for movement. 6.  The linear order of (14a) indicates that after o João moves to the relevant θ-position associated with afirmação, the latter moves to a higher position. The nature of such position is orthogonal to our current discussion. 7.  See Picallo (2001, 2002) for arguments within Chomsky’s (2000, 2001) Agree-based system that clauses can participate in Case and ϕ-agreement relations.



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

b. A idéia é (*de) que o João seja o candidato. the idea is   of that the João be.subj the candidate “The idea is that João should be the candidate.” c.

A alegação é (*de) que a Maria viaja muito. the allegation is   of that the Maria travels much “The allegation is that Maria travels too much.”

Assuming that the lack of a preposition in a sentence such as (17) signals that the embedded CP is an adjunct rather than a complement, there still remains the question of why o João cannot move from within the embedded CP. (17) *A afirmação d[o João]i que ti fez o trabalho é falsa. the affirmation of.the João that did the job is false “João’s statement that he did the job is false.”

The question is pressing as Ferreira (2000, 2009) and Rodrigues (2004) have argued, following Hornstein’s (2001) analysis of adjunct control, that A-movement from the subject position of a finite adjunct clause is possible in BP if it takes place prior to the attachment of the will-be adjunct. Take, for instance, the simplified derivation of (18) given in (19)–(22). (18) [O João]i cumprimentou [a Maria]j depois que eci/*j entrou.  the João greeted  the Maria after that entered “João greeted Maria after he got in.” (19) Kˉ=ˉ[[o João]ˉTN:default the João

entrou] entered

Lˉ=ˉ[cumprimentou a Maria]   greeted the Maria (20) Side]ward movement:



K = [[o João]i TN:default entrou] ↓ M = [[o João]i cumprimentou a Maria]

(21) [vP [vP [o João]i cumprimentou a Maria] [depois que the João greeted the Maria  after that [[o João]i TN:default entrou]]]    the João entered (22) [TP [o João]i T[N:default; P:default] [vP [vP [o João]i cumprimentou a Maria] [depois que [[o João]i TN:default entrou]]]]

After the syntactic objects K and L in (19) are built, the computational system makes a copy of o João from K and merges it with L – an instance of sideward movement in

 Jairo Nunes

the sense of Nunes (2001, 2004), yielding M in (20).8 Notice that in (19) o João did not have its Case checked (T has only a number feature) and was therefore active for purposes of A-movement. Note also that at the derivation step depicted in (19)–(20), K is not an adjunct, but a root syntactic object; hence, movement out of K should not yield an (adjunct) island effect (see Nunes & Uriagereka 2000; Hornstein 2001; Nunes 2001, 2004; and Hornstein & Nunes 2002).9 Further computations then yield the structure in (21) with the temporal clause adjoined to the matrix vP. O João then moves again (cf. (22)), landing in the matrix Spec,TP and having its Case valued. (22) finally surfaces as (18) after the lower copies of o João are deleted. Let us now examine why a sideward movement derivation for (17) does not lead to a convergent result. Consider the steps in (23)-(25). (23) a. K= [CP que [o João] TN:default fez o trabalho] that  the João did the job b. L = [NP afirmação] affirmation (24) Sideward movement: a. K= [CP que [o João]i TN:default fez o trabalho]



b. L = [NP afirmação [o João]i]

(25)

DP a the





NP NP

CP

[afirmação [o João]i] affirmation the João

[que [o João]i fez o trabalho] that the João did the job

The steps in (23)–(25) are parallel to those of (19)–(21). So, the difference must show up at later steps. Consider the copies of o João in (22). The higher copy

8.  The adjunct is built before the matrix vP in virtue of the bottom-up nature of the system: The domains that are going to be embedded are built before the embedding domains (see Chomsky 2000 for an implementation in terms of subarrays). For arguments that sideward movement proceeds from the adjunct to the matrix derivational workspace and not the opposite, see Nunes & Uriagereka (2000), Nunes (2001, 2004), and Hornstein & Nunes (2002). 9.  Sideward movement in (20) targets Spec,vP rather than the object position of the main verb, due to Merge-over-Move economy computations (see Hornstein 2001 for discussion). That is, the computational system first exhausts the numeration, plugging a Maria in the matrix object position, and only after that is sideward movement of o João licensed.



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

c-commands each of the lower copies, forming a different chain with each of them. In (25), on the other hand, the two copies are not in a chain configuration as they do not stand in a c-command relation. Assuming that the deletion of copies can only operate with chains, Chain Reduction (see Nunes 2004) can be employed in (22), but not in (25). In turn, failure to delete one of the copies of o João in (25) causes linearization problems as the system gets contradictory instructions: o João should precede and be preceded by que, as well as precede itself (see Nunes 2004 for discussion).10 Note that when de is present, i.e. when we have a true noun complement structure as in (15), the upper copy of o João c-commands and forms a chain with the lower copy, allowing Chain Reduction to apply in the phonological component, delete the lower copy, and circumvent the potential linearization problems. To summarize, dummy prepositions do not play a direct role in the licensing of null subjects inside noun complement structures, as appeared at first sight. Rather, they just disambiguate the potential relation the finite CP can bear with respect to the noun it relates to: a head-complement relation or an adjunction relation. When de is present, we have a head-complement relation and the embedded subject can successively move to the θ-position associated with the selecting noun. When de is absent, we have an adjunct structure and although sideward movement of the subject of the adjunct clause can satisfy Last Resort, the resulting structure cannot be linearized and the derivation crashes. Let us now consider the consequences of this analysis for another type of construction in BP in which the interpretation of a null subject varies depending on whether or not de is present.

10.  The contrast between the derivations of (17) and (18) mirrors the contrast between the derivations of the parasitic gap constructions in (ia) and (iia) below, which under Nunes’s (2001, 2004) analysis involves sideward movement of which paper to the matrix object position, yielding the (simplified) structures in (ib) and (iib). In (ib), sideward movement of which paper to the matrix object position is followed by movement to the matrix Spec,CP. The copy in Spec,CP forms a different chain with each of the lower copies, allowing Chain Reduction to apply and delete the lower copies. By contrast, in (iib) the two copies of which paper do not form a chain and Chain Reduction is inapplicable. The derivation then crashes at PF because it cannot be linearized: without, for instance, should precede and be preceded by which paper in (ib) (see Nunes 2001, 2004 for further discussion). (i)

a. Which paper did you file without reading? b. [[which paper]i did [you [[file [which paper]i] [without reading [which paper]i]]]]

(ii)

a. *Who filed which paper without reading? b. *[who [[filed [which paper]i] [without reading [which paper]i]]]

 Jairo Nunes

4.  Dummy prepositions and adjunct control Modesto (2000) presents sentences such as (26) below as evidence against the proposal that referential null subjects in BP are A-traces. The argument goes as follows: If null subjects in BP are to be analyzed on a par with obligatorily controlled PRO in Hornstein’s (2001) system (i.e. as A-traces), the null subject of (26) should in principle behave like the object control structure in (27) and take the matrix object as its antecedent, contrary to fact. (26) [O João]i convenceu [a Maria]k que eci/*k tinha que sair. the João convinced the Maria that had that leave “João convinced Maria that he/*she had to leave.” (27) [O João]i convenceu [a Maria]k a tk/*i sair. the João convinced the Maria to leave “João convinced Maria to leave.”

However, in order for the argument to be valid, it must be the case that (26) and (27) have exact parallel structures. In particular, the embedded clause should be a complement of convencer “convince” in both cases and the matrix object should also c-command into the embedded clause in both cases. Neither of these tacit assumptions resists further scrutiny, though. As pointed out by Ferreira (2000), the matrix object does not induce a Principle C effect with respect to epithets inside the embedded CP, as illustrated in (28) below, which indicates that the matrix object does not c-command into the embedded clause. In turn, Rodrigues (2004) claims that the embedded clause is actually an adjunct clause, as it induces island effects, as illustrated in (29). (28)

O João convenceu [a Maria]i que [a idiota]i devia assaltar the João convinced the Maria that the idiot should rob

um banco. a bank “João convinced Mariai that [the idiot]i should rob a bank.” (29) a. ??Quemi o João convenceu a Maria [queˉti vem amanhã]? who the João convinced the Maria that comes tomorrow “Who did João convince Maria [will come tomorrow]?” b. ??O quei o João convenceu a Maria [que o Pedro what the João convinced the Maria that the Pedro

precisa comprar ti]? needs buy



“What did João convince Maria that Pedro needs to buy?”

c. *Comoi o João convenceu a Maria [que o Pedro tinha how the João convinced the Maria that the Pedro had



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 



que se vestir para a festaˉti]? that refl dress for the party



“Howi did João convince Mary [that Pedro had to dress for the party ti]?”

In face of data like (28) and (29), Rodrigues proposes that sentences such as (26) should be derived along the lines of what was proposed for sentences such as (18). In other words, (26) is a case of adjunct control into a finite clause, which is derived via sideward movement, as sketched in (30)–(33). (30) K = [[o João] TN:default tinha que sair] the João had that leave L = [convenceu a Maria] convinced the Maria (31) Sideward movement:



K = [[o João]i TN:default tinha que sair] ↓ M = [[o João]i convenceu a Maria]

(32) [vPˉ[vPˉ[o João]i convenceu a Maria] [que [[o João]iˉTN:default the João convinced the Maria  that   the João tinha que sair]]] had that leave (33) [TP [o João]i T[N:default; P:default] [vP [vP [o João]i convenceu a Maria] [que [[o João]i TN:default tinha que sair]]]]

What I would like to bring to this discussion is the observation that if the embedded clause of sentences such as (26) is preceded by the dummy preposition de, we get a new pattern: Extraction out of the embedded clause now becomes acceptable, as shown in (34), and an embedded null subject can take either the matrix subject or the matrix object as its antecedent, as shown in (35). (34) a.

Quemi o João convenceu a Maria de [queˉti vem amanhã]? who the João convinced the Maria of that comes tomorrow “Who did João convince Maria will come tomorrow?”

b. O quei o João convenceu a Maria de [que o Pedro what the João convinced the Maria of that the Pedro

precisa comprar ti]? needs buy



“What did João convince Maria that Pedro needs to buy?”

c.

Comoiˉ o João convenceu a Maria de [que o Pedro tinha how the João convinced the Maria of that the Pedro had



que se vestir para a festa ti]? that refl dress for the party



“Howi did João convince Mary [that Pedro had to dress for the party ti]?”

 Jairo Nunes

(35) [O João]i convenceu [a Maria]k de que eci/k tinha que sair. the João convinced the Maria of that had that leave “João convinced Maria that he/she had to leave.”

The ambiguity found in (35) is arguably what underlies the contrasts in (36) and (37). (36) a.

[O João]i convenceu [a Maria]kˉ que tinha que sei/*k vestir ˉ the João convinced the Maria that had that refl dress



bem para a reunião. well for the meeting



“João convinced Maria that he/*she had to dress well for the meeting.”

b. [O João]i convenceu [a Maria]k de que tinha que sei/k vestir  the João convinced  the Maria of that had that refl dress

bem para a reunião. well for the meeting



“João convinced Maria that he/she had to dress well for the meeting.”

(37) a. *[O João]i convenceu [a Maria]k que tinha que apresentá-loi  the João convinced  the Maria that had that introduce-him

pra Ana. to.the Ana



“*Joãoi convinced Maria that hei had to introduce himi to Ana.”

b. [O João]i convenceu [a Maria] de que tinha que  the João convinced  the Maria of that had that

apresentá-loi pra Ana. introduce-him to.the Ana



“Joãoi convinced Maria that she had to introduce himi to Ana.”

Given that the null subject must be controlled by the matrix subject in the version without de, the reflexive in (36a) must corefer with the matrix subject and not the matrix object. By the same token, (37a) is out because the pronoun is locally bound by the null subject. Conversely, once the null subject may be controlled by the matrix subject or the matrix object in the version with de (cf. (35)), the reflexive in (36b) is ambiguous and the pronoun in (37b) can be coreferential with the matrix subject, thanks to the possibility of object control. However, this general pattern raises a problem. If de always signals that the embedded CP is a complement in these constructions, subject control in the version with de becomes unexpected. Given a Larsonian VP-shell structure for ditransitives, movement of the embedded subject to the matrix Spec,vP should cross the intervening object and yield a minimality effect, as illustrated in (38).



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

(38)

[vP [o João]i [v’ convenceuk+v [VP [a Maria] tk de [CP que ti …]]]



↑-------------------*---------------------------------

This unexpected pattern is arguably spurious, relating to the fact that the sociolinguistic prestige ascribed to de (see Section 3) often leads to instances of hypercorrection. Consider the data in (39) below, for example. Although the verb pensar “to think” does not subcategorize for de in BP, as shown in (39a), it is not uncommon to find cases of hypercorrection with insertion of de in formal speech, as illustrated in (39b), where # stands for hypercorrection. (39) a.

Eu não pensei (*de) nada. I not thought   of nothing “I didn’t think anything.”

b. Eu penso (#de) que é necessário investir mais em educação. I think   of that is necessary invest more in education “I think that it is necessary to invest more in education.”

Although one frequently finds prescriptivist condemnations of hypercorrections of the sort illustrated in (39b), the presence of the preposition in constructions such as (35) is taken to be prescriptively correct, given the availability of structures such as (40), where the preposition is obligatory. (40) Eu nunca consigo convencer a Maria *(de) alguma coisa. I never get convince the Maria   of some thing “I never manage to convince Maria of anything.”

Based on the superficial “stylistic” availability of de-insertion in (39b), I would like to propose that the unexpected subject control reading in (35)/(36b)/(37b) arises from an adjunct structure to which “stylistic” de was added. We can see that the de that appears in these constructions is spurious by examining whether the binding relations allowed in (35)/(36b)/(37) are kept the same when there is wh-movement from the embedded clause. Given that wh-movement can only take place when the embedded CP is a complement, the prediction is that subject control, which is licensed when the embedded CP is an adjunct, should no longer be available in the presence of wh-movement from within the embedded clause. Take the sentence in (41) below, for instance, which contrasts with (36b) in only allowing the object control reading for the embedded null subject. According to the proposal above, the two readings available in (36b) are due to a structural ambiguity. When the embedded CP is a complement (and de is a marker of inherent Case), only the object control reading is possible. Conversely, when the embedded CP is an adjunct (and de is a “stylistic” marker), only the subject control reading is allowed. As predicted, once an adjunct is extracted out of the embedded

 Jairo Nunes

clause in (41) and the complement structure is enforced, the anaphor must take the matrix object as its antecedent. (41)

Comow [o João]i convenceu [a Maria]k de [que tinha que sek/*i how  the João convinced  the Maria of  that had that refl

vestir para a reunião tw]? dress for the meeting “Howw did João convince Maria [that she/*he had to dress for the meeting tw]?”

Similar considerations apply to the contrast between (42a) and (42b) below. Speakers vary in their judgments for (42a) under the reading in which a Maria and a idiota are coreferential. According to the proposal above, this fluctuation in judgments is due to the potential ambiguity of the embedded CP. If CP is interpreted as a complement (with de as a marker of inherent Case), the matrix object c-commands into the embedded clause and coreference with the epithet should induce a Principle C effect. In turn, if CP is interpreted as an adjunct (with the “stylistic” use of de), coreference should be allowed as the object would not c-command the epithet. Once the interpretation of CP as a complement is forced by wh-extraction, as in (42b), judgments now become more streamlined and coreference is not allowed, as predicted. (42) a. %O João convenceu [a Maria]i de [que [a idiota]i devia the João convinced  the Maria of that  the idiot should

se vestir formalmente para a festa]. refl dress formally for the party



“João convinced Mariai that [the idiot]i should dress formally for the party.”

b. *Comok o João convenceu [a Maria]i de [que [a idiota]i how the João convinced  the Maria of  that  the idiot

devia se vestir para a festa tk]? should refl dress for the party

“*Howk did João convince Mariai [that [the idiot]i should dress for the party tk]?”

To sum up, if we use wh-extraction to control for the potential noise due to the “stylistic” use of de in constructions with verbs like convencer “convince”, we again find that de disambiguates the potential complement or adjunct interpretation of a CP. When de is present, we have a complement structure and movement of the embedded subject proceeds upwards; hence the moved subject must land in the matrix object position (Spec,VP) in compliance with minimality requirements. In turn, if de is not present, we have an adjunct structure, and the embedded subject can only reach the matrix domain if it undergoes sideward movement, landing in Spec,vP (see Fn. 9). Wh-movement out of the embedded clause is therefore licit when de is present, but induces an island effect when de is not present.



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

5.  Dummy prepositions and the A-over-A Condition Impersonal constructions constitute another domain where prepositions interact with the licensing of null subjects in BP. As originally noted by Galves (1987), sentences such as (43) in BP are ambiguous in that the DP that appears in the matrix clause may be interpreted as the external or the internal argument of the embedded verb: (43)

O João é difícil de elogiar. the João is difficult of praise.inf Tough-interpretation: “It is hard to praise João.” Raising interpretation: “João rarely praises someone.”

Here I will focus on the raising interpretation. Evidence that the matrix DP in (43) is a subject is the fact that it triggers verbal agreement, as shown in (44) below. Moreover, the fact that these constructions may involve idiom chunks, as illustrated in (45) and (46), indicate that the matrix DPs have raised out of the embedded clause.11 Finally, (47) shows that raising is possible even if the embedded clause has an inflected infinitival. Actually, that raising out of inflected infinitivals is allowed in BP should be no surprise by now, given that raising out of finite clauses is also possible, as seen in previous sections. In other words, Nunes’s (2008) proposal that finite Ts in BP may bear only a number feature in the syntactic component can also be extended to the T head of inflected infinitivals. (44) a.

Eu sou fácil de elogiar alguém. I am easy of praise someone “I easily praise people.”

b. Esses professores são difíceis de elogiar os alunos. these teachers are difficult of praise the students “These teachers rarely praise the students.” (45) a.

Tá fácil do caldo entornar. is easy of.the broth boil.over

b. O caldo tá fácil de entornar. the broth is easy of boil.over “It’s likely that things will go wrong.”

11.  The fact that idiom chunks are allowed in constructions such as (45b) and (46b) also provides a compelling argument against a pro-based approach to null subjects in BP, for idiom chunks cannot be subsumed or doubled by pronouns. For further arguments against a pro-based analysis, see Ferreira (2000, 2009), Rodrigues (2004), Martins & Nunes (2005), and Nunes (2008).

 Jairo Nunes

(46) a.

Tá bem fácil da vaca ir pro brejo. is very easy of.the cow go to.the swamp

b. A vaca tá bem fácil de ir pro brejo. the cow is very easy of go to.the swamp “It’s very likely that things will go wrong.” (47) Esses professores são difíceis de elogiarem alguém. these teachers are difficult of praise.3pl someone “These teachers rarely praise anyone.”

Relevant for our current discussion are the correlations involving the dummy preposition de and the availability of these raising constructions (see Nunes 2008). First, this type of raising is only allowed with predicates that permit the dummy preposition de. As shown in (48) and (49) below, for instance, predicates such as fácil “easy” and difícil “difficult”, which optionally require de, allow raising of the embedded subject; by contrast, predicates such as provável “probable” and lamentável “regrettable”, which do not license de, do not allow raising. (48) a.

É fácil/difícil (d)esses professores elogiarem os alunos. is easy/difficult  of.these teachers praise.3pl the students “It’s easy/hard for these teachers to praise the students.”

b. Esses professores são fáceis/difíceis de elogiarem os alunos. these teachers are easy/difficult of praise.3pl the students “These teachers often/rarely praise the students.” (49) a.

É bem provável/lamentável (*d)os professores terem elogiado is very probable/regrettable   of.the teachers have.3pl praised



o diretor. the director

b. *Os professores são bem prováveis/lamentáveis de terem the teachers are very probable/regrettable of have.3pl

elogiado o diretor. praised the director



“It is very likely/regrettable that the teachers praised the director.”

Second, raising can take place only if the preposition is present, as shown in (50). (50) a. É difícil (d)esses professores elogiarem alguém. is difficult  of.these teachers praise.3pl someone “These teachers rarely praise anyone.” b. Esses professores são difíceis *(de) elogiarem alguém. these teachers are difficult   of praise.3pl someone “These teachers rarely praise anyone.”



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

Finally, although the infinitival clause can move to the subject position or stay in situ, as shown in (51), once it is preceded by de it can no longer move, as shown in (52). (51) a. b.

É difícil esses professores elogiarem alguém. is difficult these teachers praise.3pl someone Esses professores elogiarem alguém é difícil. these teachers praise.3pl someone is difficult “These teachers rarely praise anyone.”

(52) a.

É difícil desses professores elogiarem alguém. is difficult of.these teachers praise.3pl someone

b. *Desses professores elogiarem alguém é difícil. of.these teachers praise.3pl someone is difficult “These teachers rarely praise anyone.”

Nunes (2008) argues that the paradigm in (48)–(52) can be accounted for if de is a marker of inherent Case. In this view, the fact that only some predicates are able to take a de-infinitival (cf. (48a) vs. (49a)) reduces to lexical idiosyncrasies generally involved in inherent Case assignment. In turn, if the infinitival receives inherent Case, it should become inactive for purposes of A-movement; hence, a de-infinitival cannot move to the matrix subject position (cf. (52a) vs. (52b)). Finally, given that both the infinitival clause and the embedded subject can potentially raise to the matrix subject position (cf. (51b) and (50b)), potential movement of the infinitival clause should always block movement of the embedded subject, as they instantiate an A-over-A configuration. In order for the subject to be allowed to move without violating Chomsky’s (1964) A-over-A Condition, the infinitival must be discarded from the competition. This happens when the infinitival receives inherent Case (cf. (52)). We now have an explanation for why movement of the embedded subject requires the presence of de (cf. (50b)): By rendering the infinitival clause immobile, de ends up freeing the embedded subject. Actual subject movement will then depend on whether the infinitival T head bears only number or number and person features. To sum up, in this section we examined another type of construction in BP that at first sight suggests that its null subject must be licensed by the dummy preposition de. Upon close inspection, we found that these impersonal constructions replicate what we had already seen in previous sections, with the only difference being that the movement of the embedded subject this time targets a non-thematic position. More specifically, there is no direct interaction between the dummy preposition and the null subject. The preposition is just a realization of inherent Case assigned to the infinitival clause by the subcategorizing head. However, a by-product of such interaction is that the infinitival clause becomes immobile, allowing the embedded subject to move. The raising constructions discussed in

 Jairo Nunes

this section thus provide additional evidence for the two proposals explored in previous sections: (i) that referential null subjects in BP are A-traces and (ii) the preposition de that precedes clauses is a marker of inherent Case.

6.  Concluding remarks The proposal that referential null subjects in BP are traces of A-movement (see Ferreira 2000, 2009; Rodrigues 2002, 2004; Martins & Nunes 2005; Nunes 2008) provides substantial support for the general approach advocated by Hornstein (2001), according to which construal phenomena should be captured in terms of movement. This paper has provided further evidence to this approach as it showed that the basic proposal can be extended to finite control into noun complement clauses in BP (Section 3) and sharpened an alternative analysis of recalcitrant data involving apparent object control configurations (Section 4). Finally, it provided a new type of argument for a movement approach to referential null subjects in BP (Section 6): As movement of embedded clauses and movement of their subjects may compete with respect to economy considerations minimizing the span of the operation, referential null subjects in BP were shown to be regulated by the A-over-A Condition.

References Chomsky, Noam. 1964. Current Issues in Linguistic Theory. The Hague: Mouton. Chomsky, Noam. 1986. Knowledge of Language: Its nature, origin and use. New York: Praeger. Chomsky, Noam. 2000. “Minimalist Inquiries: The framework”. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik ed. by Roger Martin, David Michaels & Juan Uriagereka, 81–155. Cambridge, Mass.: MIT Press, Chomsky, Noam. 2001. “Derivation by Phase”. Ken Hale: A life in language ed. by Michael Kenstowicz, 1–52. Cambridge, Mass.: MIT Press. Duarte, M. Eugênia. 1995. A Perda do Princípio “Evite Pronome” no Português Brasileiro. Ph.D. dissertation, Universidade Estadual de Campinas. Embick, David. 1997. Voice and the Interfaces of Syntax. Ph.D. dissertation, University of Pennsylvania. Ferreira, Marcelo. 2000. Argumentos Nulos em Português Brasileiro. M.A. thesis, Universidade Estadual de Campinas. Ferreira, Marcelo. 2009. “Null Subjects and Finite Control in Brazilian Portuguese”. Minimalist Essays on Brazilian Portuguese Syntax ed. by Jairo Nunes, 17–49. Amsterdam/Philadelphia: John Benjamins. Galves, Charlotte. 1987. “A Sintaxe do Português Brasileiro”. Ensaios de Lingüística 13. 31–49. Hornstein, Norbert. 2001. Move! A Minimalist Theory of Construal. Oxford: Blackwell.



Dummy prepositions and the licensing of null subjects in Brazilian Portuguese 

Hornstein, Norbert & Jairo Nunes. 2002. “On Asymmetries between Parasitic Gap and Acrossthe-board Constructions”. Syntax 5:1. 26–54. Martins, Ana Maria & Jairo Nunes. 2005. “Raising Issues in Portuguese”. Journal of Portuguese Linguistics 4. 53–77. Martins, Ana Maria & Jairo Nunes. Forthcoming. “Apparent Hyper-raising in Brazilian Portuguese: Base Generated Topics and Agreement across a Finite CP”. Ms., Universidade de Lisboa & Universidade de São Paulo. The Complementiser Phase: Subjects and Wh-dependencies ed. by E. Phoevos Panagiotidis. Oxford: Oxford University Press. Modesto, Marcello. 2000. On The Identification of Null Arguments. Ph.D. dissertation, University of Southern California. Nunes, Jairo. 1990. O Famigerado Se: Uma análise sincrônica e diacrônica das construções com se apassivador e indeterminador. M.A. thesis, Universidade Estadual de Campinas. Nunes, Jairo. 2001. “Sideward Movement”. Linguistic Inquiry 31. 303–344. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge, Mass.: MIT Press. Nunes, Jairo. 2008. “Inherent Case as a Licensing Condition for A-movement: The Case of Hyper-raising Constructions in Brazilian Portuguese”. Journal of Portuguese Linguistics 7. 83–108. Nunes, Jairo & Juan Uriagereka. 2000. “Cyclicity and Extraction Domains”. Syntax 3. 20–43. Picallo, M. Carme. 2001. “Nominalized Clauses, Clausal Arguments and Agreement”. CatWPL 9. 69–84. Picallo, M. Carme. 2002. “Abstract Agreement and Clausal Arguments”. Syntax 5. 116–147. Rodrigues, Cilene. 2002. “Morphology and Null Subjects in Brazilian Portuguese”. Syntactic Effects of Morphological Change ed. by D. Lightfoot, 160–178. Oxford & New York: Oxford University Press. Rodrigues, Cilene. 2004. Impoverished Morphology and A-movement out of Case Domains. Ph.D. dissertation, University of Maryland. Ross, John R. 1982. “Pronoun deletion process in German”. Paper presented at the Annual Meeting of the Linguistics Society of America. San Diego, California. Stowell, Tim. 1981. Origins of Phrase Structure. Ph.D. dissertation, Massachusetts Institute of Technology.Word order

OV sequences in early child Catalan and English Susagna Tubau

Universitat Autònoma de Barcelona In this paper I investigate the nature of preverbal objects in child English and Catalan. OV sequences occur in the children’s productions between approximately 19 months and 2 years of age even though both adult English and Catalan display VO word order. I claim that OV sequences emerge in child English and Catalan after Telicity Phrase and Aspect Phrase have been acquired but are still underspecified, thus not forcing movement of objects and verbs to Tel0 and Asp0 respectively. While V-to-Asp0 without object-to-Tel0 yields VO word order, the opposite yields OV, which disappears as soon as both movements can no longer take place non-simultaneously. In child Catalan, two kinds of OV sequences are distinguished: ‘true’ OV sequences, which occur with non-finite and non-adult inflected verbs, and Focus fronted OV combinations, which are attested with adult-like inflected verbs and whose syntax is different from ‘true’ OVs.

1.  Introduction As reported in Radford (1990)1 for child English and Gavarró (1998), Llinàs-Grau (1998) and Llinàs-Grau & Coll-Alfonso (2001) for child Catalan, most children between approximately 19 months and 2 years of age2 produce OV sequences, i.e. occasionally place the object before the verb. This is unexpected in VO languages, where the canonical position of the (non-focused) object is after the verb. 1.  Braine (1963) already reported the examples other fix (= “fix the other one”) and this do (= “do this”), produced respectively by two children named Andrew and Steven. These kinds of data are defined as ‘groping patterns’ and are simply accounted for as ‘strange’ combinations. Lebeaux (2000:10) assumes Braine’s ‘groping patterns’ to be try-outs to determine the directionality of government. 2.  In the data contributed by Gavarró (1998), Lena, the studied child, produced some OV sequences after 3 years of age. Interestingly, she is an English-Catalan bilingual, who was regularly exposed to both languages at home. The rest of the children considered in this study show a reasonable amount of individual variation with respect to the ages at which OV is first (and last) attested in their transcribed productions (cf. Table 4).

 Susagna Tubau

In the present piece of research, the data of four English monolingual children and four Catalan children with varying degrees of bilingualism with Spanish were considered to study the nature of OV sequences. An anonymous reviewer suggests that comparing the data of English monolingual children to the data of CatalanSpanish bilinguals does not seem fair, for bilingualism may have quantitative and qualitative delay effects on language development. With respect to this issue, I would like to stress that the Catalan subjects chosen for the present study are mainly exposed to Catalan, though Spanish may also be occasionally heard in the household. This is certainly very different from more extreme cases of bilingualism where both languages are used indistinctively by the children’s caretakers. The article is organised as follows: in Section 2, the data are presented; in Section 3, previous analyses of the phenomenon under study are outlined; some theoretical background is introduced in Section 4, and in Section 5, an analysis is put forward that distinguishes true instances of OV from a Focus construction. Finally, Section 6 summarises the main conclusions.

2.  The data The data of the four English monolingual children were taken from the Manchester corpus, available from the CHILDES database (MacWhinney & Snow 1985; MacWhinney 1995). The data of the four Catalan children are also from CHILDES.3 The examples in (1) to (4) have been included to illustrate what OV sequences are. The utterances in (1) and (2) have been produced by Catalan children, while examples in (3) and (4) belong to English children.4

3.  I would like to thank Dr. Montserrat Capdevila for contributing the data of the child named Pere. 4.  Some of the English examples could be treated as N + N combinations instead of OV sequences and others could be given different interpretations as pointed out by an anonymous reviewer. Apart from the examples in (3) and (4) above, examples like the following are found in the data: (i) Baby fit Anne “Anne is fitting the baby”

(Anne 2;0.15)

(ii) Train drive “I’m driving a train”

(Becky 2;0.28)

(iii) Thomas pushing “I’m pushing Thomas”

(Carl 1;11.29)



OV sequences in early child Catalan and English 

(1) Un altr-e pintar one other-m paint.inf “(Let’s) paint another one” (2) El-s peu-s tap-ar aquí the-pl foot-pl cover-inf here “Cover my feet here”

(Pep 2;0)

(Caterina 1;9.15)

(3)

No want snowman draw “I don’t want to draw a snow man”

(Carl 1;11.22)

(4)

One fasten “Fasten (this) one”

(Becky 2;1.24)

In the Catalan examples above, the verbs are infinitives (i.e. uninflected for tense); however, OV sequences are also attested with verbs inflected with a non-adult form which is not appropriate for the context where it occurs. In the case of English, bare forms are also used; those verbs that carry an inflectional mark occur without their corresponding auxiliary and are thus comparable to the Catalan cases of non-adult inflection. The Catalan data in (1) and (2) also co-occur with examples where the object precedes an adult-like inflected verb, as shown in (5) and (6). This phenomenon is unattested in English. (5) A peu destap-o aquí the foot uncover-pres.1sg here “I am uncovering my foot here” (6) Raqueta a mama no vull racket of mum not want.pres.1sg “I don’t want mum’s racket”

(Caterina 1;9.15)

(Pere 2;0.14)

An anonymous reviewer points out that in (i), baby could be the subject instead of the object, as the following tier uttered by the child is Anne’s lap, and although the child’s mother uses the verb fit a lot, she rarely does with the X is fitting Y pattern. However, notice that after Anne’s lap, the mother utters You’re going to fit her, which shows that she interprets Anne as the subject and baby as the object of the verb in (i). With respect to example (ii), I do agree with the anonymous reviewer in that the child’s utterance is ambiguous between an OV pattern and a N + N combination. This particular example has not been counted. Neither has been example (iii): according to the reviewer, it is not clear whether (iii) is an OV sequence, since the child also says Henry is pushing the trucks and Thomas is pushing the trucks. Actually, it is suggested by the reviewer that it may be Carl’s mother who produces a post-verbal subject in reordering sentence (iii) as pushing Thomas right after Carl has produced it.

 Susagna Tubau

In the next sections, I argue that the kind of data illustrated in (5) and (6) cannot receive the same analysis as (1) to (4). More precisely, I claim that while examples like (1) to (4) are true instances of OV sequences, (5) and (6) are derived via movement of the object to a functional category that will be identified as Focus. The lack of examples of the second kind in English is attributed to the fact that Focus fronting is not as readily available in adult English as it is in Catalan. Ferdinand (1996)5 also reports the occurrence of OV sequences in child French. Some of her examples have been included in (7) and (8). (7) Couteau mets knife put.pres.1sg “I put the knife” (8) un-e pomme de terre (donn-e)? a-f potato give-pres.1sg ‘Shall I give a potato?’

(Nathalie, 2;2.2) (Ferdinand 1996:181) (Grégoire, 2;1.25) (Ferdinand 1996:183)

Very much in line with the current proposal, Ferdinand argues that the data in (7) and (8) result from Focus fronting. Her conclusion is strengthened by the data of a child named Philippe, who, as shown in the examples in (9) to (12), not only produces instances of fronted objects, but also of fronted prepositions. (9) la maison on cass-e the.f house man. cl break-pres.3sg “We are breaking THE HOUSE”

(Philippe, 2;1.26)

(10) celui-là on lèv-e, comme ça that-one man. cl lift-pres.3sg like that “We are lifting THAT ONE, like that”

(Philippe, 2;1.26)

(11) sur le lit elle est la dame on the.m bed she be.pres.3sg the.f lady “The lady is ON THE BED” (12) dans le garage je le port-e in the.m garage I it.cl carry-pres.1sg “I carry it INTO THE GARAGE” 

(Philippe, 2;1.19)

(Philippe, 2;2.3) (Ferdinand 1996:187)

From a quantificational point of view, Tables 1 and 2 show the total number of OV sequences in child Catalan and English, respectively. The percentages that these kinds of constructions represent out of the total number of transitive verbs

5.  I thank Elisabeth van der Linden for supplying this reference. The glosses for Ferdinand’s (1996) examples are mine.



OV sequences in early child Catalan and English 

with explicit post-verbal objects produced by the children in the files where OV sequences were attested have also been included. Table 1.  Percentages of OV and VO in child Catalan

Caterina Pere Pep Jordina

OV[+fin]

OV[–fin]

VO[+fin]

VO[–fin]

N

Age range

3 (8.1%) 6 (4.1%) 2 (5.9%) 4 (16.7%)

2 (5.4%) 3 (2%) 5 (14.7%) 4 (16.7%)

28 (75.7%) 116 (78.4%) 21 (61.8%) 14 (58.3%)

4 (10.8%) 23 (15.5%) 6 (17.6%) 2 (8.3%)

  37 148   34   24

1;9.15–1;11.1 1;10.11–2;1.16 1;10.6–2;1.1 1;7.23–1;11.6

Table 2.  Percentages of OV and VO in child English

Anne Becky Carl Joel

OV

VO

N

3 (4.1%) 2 (6.2%) 14 (5.9%) 5 (7.8%)

71 (95.9%) 30 (92.8%) 222 (94.1%) 59 (92.2%)

  71   32 236   64

Age range 2;0.15 2;1.24 1;11.15–2;0.26 2;0.12–2;1.23

As can be seen in the tables above, the production of OV sequences in child English and Catalan is rather scarce, thus making such a phenomenon difficult to observe. Catalan children produce more OV constructions than the English children; in addition, there is also a fair amount of individual variation: some children produce more OV sequences than others, for longer and having started earlier. The eight studied subjects, nonetheless, produced OV sequences at some point within the 19 months-2 years age range. Table 3 shows how the production of OV sequences is distributed across time in the spontaneous speech of Carl and Joel,6 while Table 4 does the same for OV[+fin] and OV[–fin] in child Catalan. Table 3.  Distribution of Carl and Joel’s OVs across time Age Carl

1;11.15 1;11.22 1;11.29 2;0.06 2;0.26

OV 4 7 1 1 1

Age Joel

2;0.12 2;0.26 2;1.03 2;1.10 2;1.23

OV 2 0 0 0 3

6.  Anne and Becky produced the attested OV sequences in one single videotaped session which was transcribed as two separate .cha files.

 Susagna Tubau

As can be observed, in every available transcript between 1;11.15 and 2;0.26 Carl produces OV sequences, with a peak at 1;11.22. By contrast, there are no OV sequences found before 2;0.12 in Joel’s files. After 2;0.12, OVs are not attested again until 2;1.23. After that, no more instances of OV have been found in the corpus. Table 4.  Distribution of OV in child Catalan across time OV[+fin] Caterina Pere

Pep

Jordina

1 2 0 1 0 2 3 1 0 0 1 3 0 0 0 0 0 3

OV[–fin] 2 0 0 1 2 0 0 2 0 3 0 0 0 0 1 0 0 1

Age 1;9.15 1;11.1 2;1.10 1;10.11 1;11.10 2;0.14 2;1.16 1;10.6 1;11.6 2;0 2;1.1 1;7.23 1;8.3 1;8.27 1;9.11 1;9.25 1;10.24 1;11.6

The distribution of different types of OV that emerges in Table 4 suggests that after a peak in the production of OV[–fin], these kinds of constructions disappear, while the type OV[+fin] is more widely attested. This is particularly visible in the data of Caterina and Pere. In addition, for all the studied children both kinds of OV are produced simultaneously at some point. As can be seen, some of the files do not contain any instances of OV at all. However, this does not mean that the child has stopped producing them: in Jordina’s case, for example, both kinds of OV disappear in some of the transcripts and reemerge later. It has to be borne in mind that longitudinal studies do not record the whole of a child’s linguistic productions. To a certain extent, the marginality of the data is also a product of the amount of transcribed material that is available. 3.  Previous accounts of OV sequences in English and Catalan Several analyses of OV sequences in English and Catalan have been put forward in the literature. For English, Radford (1990) reported that these kinds of constructions



OV sequences in early child Catalan and English 

occurred occasionally in early child English and mostly in subjectless sentences. He assumed that objects in OV sequences occupied the subject position, the source of the error being attributed to the child’s exposure to adult constructions such as (13), (14) and (15) in which the agent had been suppressed. (13) The window broke

(anti-causative)

(14) The money has been stolen

(passive)

(15) This sweater cleans well

(middle)

In Powers (2000), the production of OV sequences is related to the existence of two distinct pre-functional stages, with SVO structures emerging in the second one. Powers claims that OV constructions are only attested after SVO word order is found in the child’s grammar: since the subject is missing, the object can occupy its (preverbal) position. According to Llinàs-Grau & Coll-Alfonso (2001), the two proposals outlined above incorrectly predict that OV sequences are unattested with an overt preverbal subject. Examples like (16) and (17) show that Catalan children can produce SOV constructions, thus disconfirming the accounts in Radford (1990) and Powers (2000). (16) Papa aquest columpio compr-ar daddy this swing buy-inf “Dad will buy this swing” (17) Mama pitet pos-a mummy bib put.on-pres.3sg “Mummy is putting the bib on”

(Mireia 1;10) (Llinàs-Grau & Coll-Alfonso 2001:73) (Mireia 1;9) (Llinàs-Grau & Coll-Alfonso 2001:74)

In Gavarró (1998), OV sequences produced by a Catalan/English bilingual child were attributed to the fact that English and Catalan, the two languages being acquired, diverged with respect to the strength of the features of the functional category AGR(eement). While AGR was assumed to be weak in English, thus not forcing verbs to raise overtly, it was taken to be strong in Catalan, thus resulting in the verb always preceding the object, which would stand in Spec,AGRO. Since the child was acquiring two languages with conflicting features for AGR, OV word order was claimed to follow from the optionality of verb movement that the underspecified features of AGR imposed. In Llinàs-Grau (1998), it is assumed that OV sequences are the output of object movement to the Specifier of a functional projection encoding internal aspect of the verb. VO word order, on the other hand, is proposed to be an instance of verb movement to some functional projection, which Llinàs-Grau (1998:21) identifies as “an

 Susagna Tubau

early version of the adult T”. While T in adult grammar triggers obligatory movement of the verb, it does not in child grammar, thus allowing OV to surface occasionally. Llinàs-Grau & Coll-Alfonso (2001) build on Llinàs-Grau’s (1998) observation that OV sequences mostly occur with telic verbs (i.e. delimited predicates with an inherent completion point) and show that a strong correlation between telicity and word order exists in the data of Mireia and Júlia, two Catalan children. Llinàs-Grau & Coll-Alfonso assume that objects of transitive telic verbs must move to the Specifier of AspP, which is available from a very early stage in the process of acquisition. In combination with lack of verb movement to AGR, which is not specified with strong features in the early stages of the acquisition process, movement of the object to Spec,AspP results in OV. It is not clear, however, how to account for those examples of OV that involve atelic verbs: given that AspP only projects with telic predicates, it is predicted that OV word order should not be possible with atelic verbs, contrary to fact. I depart from Llinàs-Grau & Coll-Alfonso (2001) and Gavruseva (2003) in assuming that all objects (and not only those of telic verbs) uniformly move to a functional position above the verb. In particular, Gavruseva (2003:741), following Borer (1994), assumes two different projections to be part of the phrase structure: AktionsartP, merged within the VP shell and with a head that can be [± telic] and AspP, dominating the VP shell. Telic verbs raise to Asp0 to be in a Spec-Head relation with an argument with specified quantity, which yields a telic interpretation (e.g. ‘write a poem’); however, if the argument has an unspecified quantity feature, it will check its accusative case in FP, a functional projection sandwiched between AspP and VP, and the reading will be atelic (e.g. ‘write poetry’). In the case of atelic verbs, AspP is assumed not to project. In the present analysis, what determines OV-VO word order is whether the verb raises to a position higher than the object or not. As shown in the next section, the position where objects raise to is identified as the head of Telicity Phrase, while verbs are assumed to move to Asp0. In Catalan, verbs further raise, if overtly inflected for Tense, to T0.

4.  Theoretical background 4.1  Solà (1996) In order to attempt an explanation for why it is possible for both Catalan and English children to produce OV sequences, some description of the syntax of verbs must be provided. For this purpose, I have adopted Solà’s (1996) account, which was originally formulated to explain the word order patterns attested in Germanic languages.



OV sequences in early child Catalan and English 

The main assumption in Solà (1996) concerns verb movement and its relation with verbal morphology: verb movement to T0 is overt in those languages where the verb is inflected for Tense. Catalan is one of these languages. In the particular case of English, it is well-known that lexical verbs behave differently from auxiliaries. While the latter are normally assumed to undergo head movement from the position where they are base-generated in to T0, the former do not. This observation leads Solà to challenge the idea that English finite lexical verbs are really ‘finite’. He claims that English finite lexical verbs are actually participles that are selected by a T null morpheme specified as [± past]. Within this proposal, the contrast between finite verbs in English and in languages like Catalan and French is straightforward: while these are inflected for Tense in the latter, they are not in the former. Being participles, English verbs do not undergo movement to T0, which is either occupied by a null morpheme or by an auxiliary. English verbs are assumed to be inflected for Aspect7 nonetheless. In other words, English verbs are aspectual participles specified with a [± progressive] feature, which results in the verbal paradigm in Figure 1.

[+progressive]: [–progressive]:

eating eaten ate eat(s) eat

Selected by: be have [+Past] [–Past] modals, to (Solà 1996:235)

Figure 1.  Verbal paradigm for English

In Solà’s analysis, a verbal form such as left contributes the same meaning in (18) and in (19a, b). In both examples, left is a participle that is specified as [–progressive]; whether T0 is occupied by a null morpheme or an overt auxiliary is what makes (18) different from (19a, b). (18) John left (19) a. John has left b. John had left

7.  In Catalan, the forms in Figure 1 are: (estar) menjant, [+progressive] “(be) eating”, (haver) menjat, [–progressive] “(have) eaten”, (anar) menjar, [–progressive] “ate”, menjar, [–progressive] “eat”. Concerning past forms, there is a difference in terms of imperfectivity between vaig menjar “I ate” and menjava [+progressive], “I was eating”.

 Susagna Tubau

The English clausal structure is assumed to be the one in (20). Verbs are basegenerated in VP and head-move to Asp(ect)0, a functional category encoding the [± progressive] property. Objects uniformly move to Tel(icity)P(hrase),8 which, as seen in the tree representation, is merged under AspP. The VO word order necessarily obtains in English. (20)

CP C0

MoodP Mood0

TP T0

ReltP Relt0

AspP Asp0

TelP Tel0

VP



(Solà 1996:228)

In the case of Catalan, verbs overtly move to T0, as they carry overt Tense inflection. V- to-T movement results in VO word order. In the case of non-finite forms, I assume that lacking tense inflection, they do not move to T0. However, they do raise as far as Asp0, thus preceding the object that sits in Tel0 in the same way as inflected verbs do. Given the structure in (20), it is predicted that if verbs are not inflected for Aspect, they would occur to the right of the object, which has raised to Tel0. In the case of Catalan, if verbs are inflected neither for Aspect nor for Tense, they are also predicted to occur after objects. 4.2  On Focus and left-dislocation in Catalan In Catalan, OV word order in adult grammar results from Focus fronting and Left Dislocation structures (Vallduví 1992). These two kinds of constructions are illustrated in (21) and (22) respectively. (21) a.

L-es SABAT-ES vol la Maria The-pl shoe-pl want.3sg the.f Maria “Maria wants the SHOES”

8.  This analysis is similar to the one proposed in Chomsky (2005), where objects are assumed to uniformly raise to Spec,VP, while verbs raise to v0, the functional category immediately dominating the VP.



OV sequences in early child Catalan and English 

b. SABAT-ES port-a la Maria shoe-pl wear-3sg the.f Maria “Maria wears SHOES” (22) a.

La nina la vol la Maria the.f doll it. cl want.3sg the.f Maria “Maria wants the doll”

b. Aquest programa l=ha vist tothom this programme it.cl=have.pres.3sg see.prf everyone “Everyone has seen this programme”

Focus fronting is associated with emphasis and expresses contrastive focus. That is, the constituent that is focussed in (21a, b) is relevant with respect to other elements of the same kind. É. Kiss (1998:245) refers to this kind of focus as ‘identificational’. Identificational focus is different from informational focus, which serves the purpose of presenting new information in a sentence (É. Kiss 1998). Crucially, contrastive focus is assumed to involve movement to the left periphery of the clause, while informational focus does not. In addition, while it is the case that every sentence contains an informational focus, this is not necessarily so for contrastive focus. Left Dislocation, by contrast, can be distinguished from Focus fronting in that the former involves the presence of clitic pronouns, which are absent if the relevant constituent is focussed. In the examples in (22), the constituent la nina (“the doll”) does not receive contrastive intonation and, unlike the examples in (21), cannot be followed by a tag introduced by ‘and not…’. This is shown in the examples in (23). (23) a.

L-es SABAT-ES vol la Maria i no el-s mitjo-ns the-pl shoe-pl want.3sg the.f Maria and not the-pl sock-pl “Maria wants the SHOES and not the socks”

b. Aquest programa l=ha vist tothom * i this.m programme it. cl=have.pres.3sg see.prf everyone and

no aquesta revista not this magazine



“Everyone has seen this programme, *and not this magazine”

As discussed in further sections, the absence of clitics in child OV sequences could be seen as an argument to rule out an analysis of these constructions as Left Dislocated structures. However, as reported in Wexler, Gavarró & Torrens (2002) Catalan children around two years of age tend to omit clitics in obligatory contexts, so the fact that OV sequences are not clitic-doubled is not a conclusive argument in favour of a Focus fronting analysis. By contrast, the fact that some of the examples contain a fronted indefinite is a stronger argument, given the fact that indefinite objects cannot be Left Dislocated in Catalan.9 9.  I thank an anonymous reviewer for pointing this out to me.

 Susagna Tubau

In the analysis I put forward in the next section the possibility that OV is derived via leftward movement to Focus is linked to the presence of overt inflection on the verb. In other words, regardless of whether the object is pronounced with contrastive pitch or not, I take OV sequences to be derived via movement when the verb is finite and inflected for tense in an adult-like manner.

5.  Analysis One of the main claims in the present paper is that OV sequences are possible in child English because insertion of lexical verbs in Asp0, the category that has been assumed to encode [±progressive] distinction, and movement of the object to Tel0, encoding [± telic], are optional around age two10 due to underspecification. Let us assume that, at early stages of acquisition, neither the verb nor the object leave their base-generated position within the VP to check the [± progressive] and the [± telicity] features respectively. VO word order emerges, the verb being in its bare form. Tel0 and Asp0 are underspecified, and movement of the object to Tel0 and V-to-Asp0 are not categorical: if none of the two movements takes place, VO word order is attested; likewise, if both apply, VO results, as well. However, while lack of movement of object to Tel0 in combination with V-to-Asp0 still yields VO word order, object-to-Tel0 in combination of lack of V-to-Asp does not: rather, OV word order emerges. This explains why VO word order is clearly dominant from a very early age in English and OV is so marginal. For child Catalan, non-finite forms are assumed not to be inflected for Tense, which necessarily precludes V-to-T movement. If, in addition to that, the child goes through a stage of Asp0 and Tel0 underspecification, OV word order is attested when the object has moved to Tel0 but the verb has not raised to Asp0. The verb is in the infinitival form or, at most, in a non-adult inflected form (i.e. an apparently inflected form that does not agree with the subject). As shown in (5) and (6) above, nonetheless, OV sequences are also attested with adult-like inflected verbs. This poses a problem for any proposal for OVs which assumes inflected verbs to undergo V-to-T movement, as in being inflected for Tense, the verbs in (5) to (6) should have raised to T0, thus necessarily preceding the object.

10.  Gavruseva (2003) also claims that AspP is underspecified in child French, German and Dutch. In her analysis, underspecification of AspP, a category which is essential for the licensing of Tense chains, results in a particular kind of verbs (i.e. nonpunctual eventives that can be [+telic] or [–telic] depending on the complement the verb takes) surfacing as Root Infinitives.



OV sequences in early child Catalan and English 

There are at least two possible explanations for the data of the kind in (5) and (6). First, it could be argued that V-to-T movement failed to apply in spite of the fact that the verb is inflected for Tense. A second possible explanation involves assuming that (i) verbs do indeed raise to T0 and that (ii) objects raise to a position higher than both Tel0 and T0. The first hypothesis is speculative, as it calls for the assumption that V-to-T movement is optional with adult-like inflected verbs and, more worryingly, unpredictable. The second explanation, by contrast, relies on the existence of a leftperipheral position dominating T0. Given that, as has been shown in Section 3.2, such a position is available in Focus constructions, I take the second hypothesis to be superior to the first. In Section 5.1, I provide evidence in favour of analysing OV sequences with adult-like inflected verbs as instances of Focus fronting, which is a possibility in the target adult grammar being acquired by the children under study. The immediate consequence of my analysis is that two different kinds of OV sequences should be distinguished in child language: ‘true’ OV sequences are instances of objectto-Tel0 and lack of V-to-Asp0 movement; other OV sequences which occur with adult-like inflected verbs are instances of Focus fronting. A further claim I make in this paper is that all OV sequences in child English are of the ‘true’-OV type. That is, OV constructions in child English always result from movement of object to Tel0 in combination with lack of verb movement to Asp0, but never from movement of the object to Foc0. 5.1  Catalan: Evidence for Focus-movement In Tubau (2004), I investigated to what extent it is true that there is no adult input for OV sequences. For that purpose, both the children’s and the adults’ tiers in the available CHILDES files were browsed to find instances of OV word order. The results are reported in Table 5.11 Table 5.  Potential adult input for child OV

Jordina’s files Pere’s files Pep’s files Caterina’s files

Left-dislocation

Focus-fronting

Exclamatives

N

11 (0.3%) 5 (0.1%) 5 (0.1%) 3 (0.2%)

7 (0.2%) 18 (0.5%) 3 (0.06%) 1 (0.06%)

2 (0.06%) 2 (0.05%) 3 (0.06%) 0(0%)

3,495 3,680 4,580 1,730

(adapted from Tubau 2004:160)

11.  In Table 5, N corresponds to the total number of transcribed tiers corresponding to adults (or older children) in the files of every studied child.

 Susagna Tubau

As shown in Table 5, the adults whose utterances were transcribed in the children’s files studied in Tubau (2004) produced very few examples of Focus fronting and Left Dislocated structures. The column ‘exclamatives’ refers to sentences with a fronted wh-element and a deleted complementiser, as shown in the examples in (24) and (25). (24) Quin-s peto-ns que ha don-at! what-pl kiss-pl (that) have.pres.3sg give-prf “What kisses has she given!” (25) Quin vestit més mac-o que port-es avui! what.m dress more nice-m (that) wear-2sg today “What a nice dress you are wearing today!” 

(Tubau 2004:160–1)

Unlike in Spanish, (standard) Catalan exclamative sentences with a fronted whphrase involve the use of the complementiser ‘que’. In the areas around Barcelona, however, where Catalan and Spanish are very much in contact, the complementiser is often dropped in these kind of constructions. Syntactically speaking, this results in the (wh-)object preceding the verb. However, as can be seen in Table 5, the percentage of faulty wh-exclamative is extremely low. The findings in Tubau (2004) are relevant to the present piece of research because they show that there is indeed no input for the ‘true’ kind of OV sequences, which must result from a mechanism that is no longer possible in the adult grammar. The percentages of potential input in the form of Focus-fronting and LeftDislocated structures are very low in the studied files, as well. As suggested by an anonymous reviewer, the possibility that Catalan children produce both ‘true’ and ‘untrue’ OV sequences on the basis of Focus fronting and Left-Dislocated adult data cannot be excluded. However, as shown by examples (26) and (27), ‘true’ OV sequences are different from Focus fronted constructions in that in the latter (i) the verbs are selected by the relevant auxiliary, which is adequately inflected for tense and (ii) the object does not only precede the verb it is an argument of, but also the auxiliary in T0. (26) Un peix-et ha trob-at el Pep a fish-diminutive have.pres.3sg find-prf the.m Pep “Pep has found a little fish” (27) Un cavall ha cur-at (l)a Jordina a horse have.pres.3sg heal-prf the.f Jordina “Jordina has healed a horse”

(Pep 1;10.06)

(Jordina 1;11.06)

In examples (28) and (29), the object is the complement of a non-finite verb which is itself a complement of another verb. Interestingly, instead of occurring to the



OV sequences in early child Catalan and English 

left of the non-finite verb it is a complement of, the object occurs to the left of the matrix inflected verb. (28) E(s)t-o vull sac-ar this-m want.pres.1sg take.out-inf “I want to take this out” (29) Mi(a) e(s)t-o vaig a compr-ar look.imp this-m go.pres.1sg to buy-inf “Look, I am going to buy this”

(Jordina 1;11.26)

(Jordina 1;11.26)

If the OV sequences in (26) to (29) were a result of the non-finite forms (i.e. the past participle in (26) and (27) and the infinitive in (28) and (29)) not having raised from V to Asp0 while the object sits in Tel0, the object would occur to the left of the non-finite verb, but lower down in the structure with respect to the inflected matrix verb. This is precisely what seems to have been the case in an example such as (30), from Gavarró (1998), and in (31), produced by one of the studied children. (30) No vull això f-er not want.pres.1sg this do-inf “I don’t want to do this” [target: No vull fer això] (31) Vull això treu-re want.pres.1sg this do-inf “I want to take this out” [target: Vull treure això]

(Lena 2;10.0) (Gavarró 1998:111) (Pere 1;10.11)

Data like (30) and (31) show that ‘true’ OV sequences are also attested with verbs that are part of more complex verb patterns involving a matrix T0. In (30) and (31), the AspP in the embedded clause behaves differently than the one in the matrix clause. As will be seen in the next section, data comparable to (30) and (31) have been attested in child English, while examples like those in (26) to (29) have not. 5.2  English: ‘True’ OV sequences Given that longitudinal data for the study of language acquisition are generally collected by means of periodic video-taped sessions of variable length, the occurrence of OV word order constructions in child language may go unnoticed. This is why not every existing corpus is suitable for the observation of OV sequences. The Manchester corpus, which is available from the CHILDES database, contains a remarkable number of transcribed files per child. In addition, the recording sessions took place approximately every week for most of the children, which allowed OV sequences to be attested in spite of the fact that these constructions are much less frequent than VO constructions.

 Susagna Tubau

The English data on OV sequences are rather scarce, as shown in Tables 2 and 3 in Section 2: despite the short intervals between recording sessions, only a few examples of OV were attested among the linguistic productions of the studied children. A comparison between the data found in English and the data found in Catalan showed that preverbal objects occurred mostly with bare infinitives in English. I suggested, therefore, that OV sequences in English result from raising of the object to Tel0 and lack of raising of the verb to Asp0. These two conditions occasionally obtain during a period of time in which Tel0 and Asp0 are underspecified thus not forcing objects and verbs to move out of VP. Examples like (30) and (31) have also been attested in child English. However, no instances like the kind of constructions reported in (26) to (29) for child Catalan were found in the English corpora. (32)

No want snowman draw “I don’t want to draw a snowman”

(Carl 1;11.22)

Let us assume that (32) can be analysed as having a simplified syntactic structure that involves no functional projections other than AspP and TelP, which dominate the lexical projection VP. The structure for (32) would be the one in (33), with the two verbs being heads of VPs dominated by AspP and TelP. (33)

NegP Neg′ Neg0 no

AspP Asp0 wantj

VP V0 tj

TelP Tel0

AspP Asp0

TelP

Tel0 snowmani

VP draw ti

The example in (33) is comparable to (30) and (31), which could be given a similar analysis. The child Catalan examples in (26) to (29), conversely, are best analysed as having a structure like (34).



OV sequences in early child Catalan and English 

(34)

FocP object

FocP Foc0

auxiliary/inflected verb T0

TP AspP

past participle/infinitive Asp0

TelP Tel0 tobj

VP V0 tv

tobj

Llinàs-Grau (1998:23) reports an example of OV involving negation, as shown in (35). Unlike the sentence in (33), the object is higher than the sentential negative marker, which dominates TP in Catalan. Such a configuration is expected if, given the finiteness of the verb, which shows that it has moved to T0, the preverbal object is in Focus. In the studied corpora, an example of OV[+fin] involving negation was also found, as shown in (36). (35) Sop-eta no vol-em soup-diminutive not want-pres.1pl “I do not want any soup” (36) Raqueta a mama no vull racket of mum not want.pres.1sg “I do not want mum’s racket”

(Mireia, between 20 and 22 months)

(Pere 2;0.14)

6.  Summary and conclusions In the present paper the nature of an intriguing child language syntactic construction has been investigated. OV sequences, where the object unexpectedly precedes the verb, are occasionally observed in the children’s linguistic productions between 19 months and two years of age, approximately. Although these kinds of constructions are not abundant and disappear shortly after the child starts producing them, they occur, to different extents, in the linguistic productions of every single child whose transcripts have been browsed. Although it may be too strong to claim that OV sequences are universal in the light of the available data, they are certainly commonly observed in the

 Susagna Tubau

speech of children acquiring different languages. In this paper, for instance, OV constructions have been reported to occur in child English, child Catalan, and child French, as well. Further research should determine whether the production of OV sequences can be described as a micro-stage in the process of language acquisition. On the basis of the available data, which is rather scarce due to the limitations imposed by the data-collection methods commonly used in longitudinal studies, I have claimed that OV word order results from a combination of two circumstances: (i) the object has moved to a position outside the VP and (ii) the verb has remained in the position where it has been base-generated instead of raising to Asp0. To formalise my analysis, I have assumed Solà’s (1996) proposal, which takes the verbal projection to be dominated by two functional projections, TelP and AspP, where the object and the verb head-move to, respectively, to check telicity and aspect features. Solà’s account capitalises on the assumption that overt Tense morphology results in movement of the verb to T. To explain the English facts with respect to the syntactic behaviour of lexical verbs, he puts forward the idea that these do not raise further than Asp0, T0 being occupied either by an auxiliary, or by a null morpheme. Child English OV sequences and child Catalan OV[–fin] patterns have been analysed as being a product of a transitional stage: in the early stages of acquisition, objects and verbs do not leave their base-generated position within the VP, and VO word order results; however, when the child acquires TelP and AspP, these are underspecified, not always triggering movement of objects and verbs to Tel0 and Asp0. While V-to-Asp0 movement without object-to-Tel0 results in VO word order, object-to-Tel0 without V-to-Asp0 results in OV. As soon as both object-to-Tel0 and V-to-Asp0 are incorporated into the child’s grammar, OV stops being attested. In child Catalan, by contrast, two distinct kinds of OV sequences seem to exist. Close observation of the available data reveals that, as in child English, child Catalan OV sequences are very often found with non-finite forms or with nonadult inflected forms. Unlike English, however, instances of OV with adult-like inflected verbs were also attested in child Catalan. I have claimed that while child English and Catalan OV sequences with nonfinite verbs can be analysed as resulting from movement of the object to Tel0 and lack of V-to-Asp0 movement co-occurring, cases of OV word order with (adultlike) inflected verbs were best analysed as involving movement to the left-peripheral category Focus. Such a conclusion was in line with Ferdinand (1996), who showed that, around age two, French children produced sentences with Focus fronted constituents.



OV sequences in early child Catalan and English 

References Borer, Hagit. 1994. “The Projection of Arguments”. UMass Occasional Papers in Linguistics. 19–47. Braine, Martin D.S. 1963. “The Ontogeny of the English Phrase Structure: the First Phase”. Language 39. 1–13. Chomsky, Noam. 2005. “On Phases”. Ms., Massachusetts Institute of Technology. É. Kiss, Katalin. 1998. “Identificational Focus versus Information Focus”. Language 74:2. 245–273. Ferdinand, Astrid. 1996. The Development of Functional Categories: the Acquisition of the Subject in French. Ph.D. dissertation, Rijksuniversiteit Leiden. Gavarró, Anna. 1998. “Word Order Alternations and Feature Assignment in Bilingual Catalan Acquisition”. Probus 10. 103–114. Gavruseva, Elena. 2003. “Aktionsart, Aspect, and the Acquisition of Finiteness in Early Child Grammar”. Linguistics 41:4. 723–755. Lebeaux, David. 2000. Language Acquisition and the Form of the Grammar. Philadelphia & Amsterdam: John Benjamins. Llinàs-Grau, Mireia. 1998. “Verb Complement Patterns in Early Catalan”. University of Massachusetts Occasional Papers in Linguistics. New Perspectives on Language Acquisition ed. by Bart Hollebrandse, vol. 22, 15–26. Amherst: GSLA, University of Massachusetts. Llinàs-Grau, Mireia & Mercè Coll-Alfonso. 2001. “Telic Verbs in Early Catalan”. Probus 13.69–79. MacWhinney, Brian. 1995. The CHILDES Project: Tools for Analyzing Talk. Hillsdale: Lawrence Erlbaum. MacWhinney, Brian & Catherine Snow. 1985. “The Child Language Exchange System”. Journal of Child Language 12. 271–296. Powers, Susan. 2000. “Scrambling in the Acquisition of English?”. The Acquisition of Scrambling and Cliticization ed. by Susan Powers and Cornelia Hamann, 95–126. Dordrecht: Kluwer. Radford, Andrew. 1990. Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Solà, Jaume. 1996. “Morphology and Word Order in Germanic Languages”. Minimal Ideas. Syntactic studies in the minimalist framework ed. by Werner Abraham, Samuel D. Epstein, Höskuldur Thráinsson & Jan-Wouter Zwarts, 271–251. Amsterdam & Philadelphia: John Benjamins. Tubau, Susagna. 2004. “Early Catalan OV Sequences: Empirical Evidence for the Poverty of Stimulus Argument”. UCL Working Papers in Linguistics 16. 149–163. Vallduví, Enric. 1992. The Informational Component. New York: Garland. Wexler, Ken, Gavarró, Anna & Vicenç Torrens. 2002. “Feature Checking and Object Clitic Omission in Child Catalan and Spanish”. Romance Languages and Linguistic Theory 2002. Selected Papers from ‘Going Romance’, Groningen 28–30 November 2002 ed. by Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe & Petra Sleeman, 253–268. Amsterdam & Philadelphia: John Benjamins.

Index

A Adjectives  67–85, 156, 161–176 classifying adjectives  73 classifying relational adjectives  75 non-restrictive adjectives  73 restrictive adjectives  73 thematic adjectives  75 transitive adjectives  76 Adjuncts  1–25, 46, 47, 57, 104, 189–190, 253–260 adjunct control  250 Anaphora  109–129, 179–198 anaphoric se  191–193 deep anaphora  111–115 null complement anaphora  109–129 surface anaphora  111–115 A-over-A Condition  253–254 Aspect  145–159 bounded aspect  145–147, 151–159 outer aspect  157–159 unbounded aspect  145–147, 151–159 Atelicity  156–159 Aux-to-Comp  1–12 C Case  243–264 inherent case  243–264 Catalan  267–287 C-Deletion  1, 2, 13–15 Cel  73, 78 fn10, 161–176 Chain Reduction  254 Child language development  51–65, 131–141, 267–287 child Catalan  267–287 child English  267–287 child Hebrew  51–65 child Spanish  51–65 child Palestinian Arabic  51–65

child European Portuguese  51–65 Classifier  145, 157–158 Clitics  9–14, 70, 110–115, 123–126, 164, 168, 234, 277 clitic se  179–198 clitic doubling  61–65, 201–220 clitic left dislocation  7 CLLD  7 Complementizer Deletion  1, 2, 13–15 Complex Event Nominals  145–159 Co-variation  1 Covert operation  87–109, 133, 139, 209, 215 D Definite NP / determiner  8–9, 138–139, 166–175, 201–220 definiteness effect  201–220 Determiner Spreading  73, 77, 161–176 Directionality  25–49 Dislocation  276–287 D-Linking  69 DP  9, 10, 34, 46–47, 67–85, 105 fn11, 114 fn3, 126, 128, 150, 154, 158, 164, 168, 170, 182–196, 298 fn4, 252, 254, 261 discontinuous DP  73 elliptical DPs  67–85 Dummy preposition de  243–264 Dutch  68 fn1, 146 fn1, 183, 226 fn2, 227–228, 232, 237–238 E Ellipsis  67–85 English  5, 13–14, 26, 28–29, 73, 87–88 fn1, 104 fn4, 106, 109, 111, 118, 134, 147,

152, 157, 162, 166–167, 172, 175, 215, 218, 233–234, 236, 267–287 Dissociated features  243, 246 F Focus  6–7, 67–85, 87–109 constrastive focus  68–69 covert focus reconstruction  95 focus extension  93 focus fronting  276–287 focus markers  102 focus propagation  95 identificational focus  87–109 Frankish superstrate  223–238 French  11, 13 fn20, 25–49, 59, 75 fn6–fn7, 80, 82 fn16, 84, 146, 171 fn10, 176, 201–238, 270, 275, 278 fn10 Old French  25–49 Pied Noir French  215 Quebec French  201–220 standard French  207 G Gascon  215 Gerund  148 Gerundive licensing  14–18 Greek  73, 161–176 H Hebrew  51–65 High German  227 I Impersonal se  179–198 Infinitive  145–159 inflected infinitivals  251–253 Spanish nominal infinitive  145–159 Italian  2–3, 5, 14–15, 109, 111–112, 122, 131–141, 146, 180, 186–187, 196–197, 219

  Index L Language acquisition  51 L1 acquisition  51 Left dislocation  276–287 Lexical semantics  25–49 Linearization  51–65, 254 Locative expressions  25–49 M Minimality effects  67–85 Modal operators  131–141 Modal verbs  131–141 N Negation  131–141, 215 Nominal ellipsis  67–85, 165 Noun complement clauses  249 Number  145, 150, 156–159 O OV-sequences  267–287 P Paduan  219 Particles  25–49 Person-less pronouns  184 Pluractionality  153 Plural marking  145–151, 156–159 Pluralization  145–159 Portuguese  1–21, 75, 77, 109–129 Brazilian Portuguese  1–21, 243–264 European Portuguese  1–21, 51–65, 75 fn6, 77 fn8,

European Portuguese dialects  179–198 Prepositional phrases  25–49 Prepositions  25–49 Prosodic typology  224 Pseudoclefts  87–109 Q Questions question particle  201–220 yes/no questions  201–220 R Relative clause  168 reduced relative clause  168 Restructuring  109, 118–129 Romanian  73, 77, 145–159, 161–176 S Scope  131–141 Se  179–198 Semantic subset principle  141 Sideward movement  243, 253–261 Spanish  27, 51–65, 67–85, 87–109, 233, 268, 280 Spanish nominal infinitive  145–159 Stress  223–238 stress timed language  232 expiratory stress  223–238 final stress  223 melodic stress  229 Subjects subject clitic  201–220

expletive subject  179–198 subject doubling  179–198 subject-verb  52–65 null subjects  243–264 Superstrate  223–238 Supine  145–159 Syllable language  223–238 Syllable timed language  232 Syntactic variability  38, 39 T Telicity  151 Trentino  207 Truth Value Judgment Task  135–141 U Unaccusative  51–65 Underspecification  278–279 Unergative  51–65 V V raising  1, 2 Verbal agreement morphology  243–244, 241 Verb-subject  52–65 Vowel deletion  223–238 Vowel reduction  223–238 V-to-C movement  1–18, 59–65 V-to-I movement  59–65 W Wh Movement: Split interrogatives  103 Wh-elements  215 Word language  223–238