Interfaces in Language 3 [1 ed.] 9781443865760, 9781443842556

Citation preview

Interfaces in Language 3

Interfaces in Language 3

Edited by

Marina Kolokonte and Vikki Janke

Interfaces in Language 3, Edited by Marina Kolokonte and Vikki Janke This book first published 2013 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2013 by Marina Kolokonte and Vikki Janke and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-4255-9, ISBN (13): 978-1-4438-4255-6

CONTENTS

Preface ....................................................................................................... vii Acknowledgements .................................................................................... xi Marina Kolokonte and Vikki Janke Part I: Interfaces in Language Acquisition Chapter One ................................................................................................. 3 Definite and Bare Noun Contrasts in Language Acquisition Pedro Guijarro-Fuentes and María Pilar Larrañaga Part II: Syntax-Semantics Chapter Two .............................................................................................. 31 Overt Quantifier Raising of Neg-whQ in Cantonese at the SyntaxSemantics Interface Man-Ki Theodora Lee Part III: Syntax-Morphology Chapter Three ............................................................................................ 57 Two Types of Bound Items and their Interaction with Morphology and Syntax Masaharu Shimada and Akiko Nagano Chapter Four .............................................................................................. 83 Are German Modals Better Verbs than English Modals? An Intra-Verbal Interface John Partridge Part IV: Syntax-Pragmatics Chapter Five ............................................................................................ 105 Yes/No Ellipsis and Polarity Focus in Modern Greek Marina Kolokonte

vi

Contents

Chapter Six .............................................................................................. 121 The Distribution of Non-Obligatory Control and its + Human Interpretation Vikki Janke Part V: Language Policy and Phonology Chapter Seven.......................................................................................... 159 Changing Attitudes: Ways of Phonologically Adapting Proper Names in Archaic Brazilian and European Portuguese Gladis Massini-Cagliari Contributors ............................................................................................. 179 Index ........................................................................................................ 181

PREFACE

The papers in this collection represent a selection from those accepted for presentation at the third Interfaces in Language conference, hosted from 4 – 6 May 2011 at the University of Kent, and organised by the university’s Centre for Language and Linguistics Studies (CLLS). Similar to the previous years, the number of participants was kept low and parallel sessions were avoided in a bid to provide the most fertile environment possible for participant interaction and the forging of new links between language enthusiasts of all stripes. In line with the conference’s title, we invited applications which held true to the interface theme but did not place restrictions on the way in which ‘interface’ was interpreted. The result was a heterogeneous set of talks, interspersed with and complemented by lively discussions, confirming that the interdisciplinary setting we tried to provide was a good way of cultivating discussion between linguists who might otherwise not cross paths. We hope our participants drew as much satisfaction from the event as we did. Subsequent submissions fell loosely into five themes: Language Acquisition; Syntax-Semantics; SyntaxMorphology; Syntax-Pragmatics; Language Policy and Phonology. Starting with Part I, we open with Pedro Guijarro-Fuentes’ and María Pilar Larrañaga’s paper, which discusses acquisition issues raised by the specific and generic interpretation of nominals (bare and definite) in Spanish and English. The licensing conditions on these elements in Spanish and English differ in that bare nouns are only licensed in Spec-IP in the latter language. Gathering data from three children (two monolingual Spanish and one bilingual), the authors predict generic nominals to surface prior to specific nominals. This is not borne out by the data, where it is seen that bare nouns are problematic in all three children. In part II, we move on to Man-Ki Theodora Lee’s paper on neg-whQ constructions in Cantonese. This is a minimalist analysis of quantifier raising, which assumes existence of a NegQP, comprising the negative morpheme mou and the wh-phrase. The unique SOV word order in constructions that have an object neg-whQ is accounted for on this QR account, as are the characteristic dual interpretations (i.e. non-existential and presupposed existential) of neg-whQ constructions. The author extends this analysis to another type of movement, namely neg-raising. If on the right path, this analysis could account for the indefinite interpretation of wh-phrases,

viii

Preface

available when preceded by negation. The topic of Part III straddles the divide between syntax and morphology. We begin with Masaharu Shimada and Akiko Nagano, who adopt a lexeme-based approach of Japanese reflexives, having shown that the properties of pure and near reflexives cannot be captured by the traditional demarcation between bound and free morphemes. Upon a methodical demonstration of the absence of parallelism between the affixal nature of reflexive elements and their interpretative restrictions, they reduce the inventory of reflexives by classifying affixal reflexives in Japanese as surface variants of one underlying lexeme. This new realignment of reflexives maps quite neatly with their syntactic and morphological distributions. Next is John Partridge’s work, which compares morphological, semantic and syntactic properties of German and English modal verbs. With respect to German modals, he finds a stronger morphological paradigm, a greater variety of interpretations (which in English is met by periphrasis), and a number of syntactic parallels with lexical verbs. On this basis, he concludes that German modals behave more like lexical verbs than their English counterparts. Part IV leads us to the bridge between syntax and pragmatics. In this section the authors consider constructions which require syntactic regulation but also pragmatic enrichment. Marina Kolokonte looks at Information Structure requirements of ‘ellipsis in coordination’ in Modern Greek. Based on Kolokonte (2008), she argues that the remnant that is present in the second, elliptical conjunct is displaced to the left periphery of the clause, where it is assigned the Information Structure role of contrastive topic. It is also proposed that this contrastive topicalisation forces narrow focus on the polarity marker. Vikki Janke’s paper on nonobligatory control builds a representation of this relation using the same mechanism that she developed to regulate syntactically mediated control in Janke (2008). Her first aim is to demarcate the point at which pragmatics can take over from syntax, by formulating an LF-rule, which, properly confined, will not over-generate. Her second aim is to derive the +human restriction on these structures. An entirely new theme is explored in the final contribution in Part V from Gladis Massini-Cagliari. The author conducts a diachronic and synchronic investigation of the pronunciation of proper names of foreign origin in medieval (archaic) Portuguese and contemporary European and Brazilian Portuguese. Her diachronic investigation leads her to conclude that the majority of names of foreign origin underwent a phonological adaptation which conforms to the phonological system of the language. Examination of European and Brazilian Portuguese reveals a divide in terms of the degree of phonological adaptation; the process is substantially more permissive in

Interfaces in Language 3

ix

the latter language. This follows from state regulation of name-giving in Portugal but not Brazil. As such we see a range of foreign phonological and stress patterns in Brazil that is absent in Portugal. We thank all of our participants for their contributions, which made the whole enterprise worthwhile. University of Kent, Canterbury August 2012

ACKNOWLEDGEMENTS

Many people have been involved with the Interfaces Series since its inception, all of whom deserve to be acknowledged. But for this volume in particular, we would like to single out Damien Hall, David Hornsby, Michael Hughes, John Partridge and Jeremy Scott from the Centre of Language and Linguistics Studies at the University of Kent, as well as Billy Clarke and Napoleon Katsos, who were our keynote speakers at this year’s conference. We are also most grateful to Cambridge Scholars Publishing for having supported the 3rd publication of this Interfaces in Language series. This enterprise would have neither begun nor continued without John Partridge, so on the occasion of his retirement it seems most apt to dedicate this volume to him. This is for you, John, and comes with our warmest regards.

PART I: INTERFACES IN LANGUAGE ACQUISITION

CHAPTER ONE DEFINITE AND BARE NOUN CONTRASTS IN LANGUAGE ACQUISITION PEDRO GUIJARRO-FUENTES AND MARÍA PILAR LARRAÑAGA

Abstract Germanic and Romance languages differ in the way they allow for bare nouns (BNs) in the specifier position of INFL. Whilst Germanic languages allow for BNs in this position, BNs in the same position in Romance languages are banned. In the present study we analyzed data of two Spanish monolingual children growing up in Spain and a bilingual Spanish/English child living in the UK. The results show that all three children go through a stage in which target-deviant BNs are used. One monolingual child, Irene, overcomes this stage very early on, whilst the other child still makes errors in the very last recording. We argue that this untypical development is due to the not yet fully fledged implementation of INFL in Magín’s grammar. The bilingual child makes the same type of errors as Magín, but as verbal inflections are target like, we argue that the preverbal target-deviant BNs in subject position are due to cross-linguistic influence from English. The target-deviant BNs in preposition phrases are attributed to the high number of chunks children have to learn and which have to be learned separately and whose meaning cannot be inferred from the meaning of the nouns at stake.

1. Introduction Recent theoretical linguistic approaches to language cognition have made strong theoretical predictions for close associations between language acquisition/learning and the appearance of complex properties that characterize and distinguish human languages across space and time. Examination of sub-modules of grammatical systems in isolation (e.g.,

4

Chapter One

phonology, syntax, semantics, morphology and pragmatics) cannot provide us with an adequate account of how language acquisition is possible. Modern linguistic descriptions indicate that sub-modules of linguistic systems, by definition, are not independent of each other, rather, they work concurrently to make the interaction between sound, structure and meaning plausible. Only an incorporated approach involving the study of the interaction between different levels of linguistic knowledge - known as linguistic interfaces - has the assurance of enlightening us on fundamental questions. More particularly within the generative framework, which is the framework adopted for our purposes in this paper, grammar consists of the computational system or narrow syntax, the output of which feeds into phonetic form (PF) and logical form (LF) at the interfaces between different levels (Bos, Hollebrandse & Sleeman 2004; Chomsky 2007). Therefore, the enquiry of linguistic interfaces has the potential to make a crucial contribution towards a tenable, integrated model of the acquisition, constitution and organization of human mental linguistic systems in general, and more particularly, for language acquisition. Largely, but not solely, within contemporary generative acquisition theory on first and bilingual language acquisition, empirical research is motivated by the hypothesis that acquisition delays in children (as well as L1 attrition and variability in adult L2 acquisition) are often related to the inherent complexity of acquiring interface-conditioned properties (e.g. Müller & Hulk 2001; Paradis & Navarro 2003; Platzack 2001; Serratrice et al. 2004; Sorace 2004, 2005, 2011 and references therein; Tsimpli et. al. 2004, amongst others). More specifically, several studies in monolingual and bilingual L1 acquisition have reported that in L1 acquisition, narrow syntactical phenomena and interface knowledge do not go hand in hand (Kramer 2000). Some of these studies have focused on the acquisition of a relatively narrow set of structures, e.g., overt vs. null pronominal subjects and object expressions in different languages and by different learners acquiring a wider range of languages. Let us take a closer look at the acquisition of overt vs. null pronominal subjects which relates to the pro-drop parameter involving the syntax-information structure interface. In pro-drop languages, such as Greek, Italian and Spanish, overt subjects are strongly favoured when they introduce new information or when a contrast is established or a focus is required. On the contrary, null subjects are preferred when there is no switch in reference in a series of utterances in discourse, and when there is no need for focus and contrast. Serratrice et al. (2004) investigated the acquisition of overt vs. null pronominal subjects in an Italian/English bilingual child compared to two groups of MLUw-matched monolinguals.

Definite and Bare Noun Contrasts in Language Acquisition

5

This study showed that in Italian the bilingual child produced overt pronominal subjects in contexts where monolinguals prefer a null subject (see Schmitz (2007) for comparable results in German-Italian bilingual children, Patuto (2012) German-Italian and German-Spanish children). In addition, the bilingual child used post-verbal strong object pronouns rather than preverbal weak pronominal clitics. Similar results were obtained in adult L2 acquisition. Adult L2 learners of null subject languages whose first language (L1) does not allow null subjects seem to use null subjects correctly in obligatory contexts, but overuse overt subjects in contexts requiring a null pronoun (Serratrice et al. 2004). These results indicate a cross-linguistic influence in specific contexts at the syntax-pragmatics interface. In light of these findings, it is important to investigate different interface phenomena with a precise analysis of what is entailed by the various interface properties. This article discusses original research from the syntax-semantic angle, addressing interface issues in Spanish first and bilingual language acquisition (defined broadly as language(s) learned in childhood) and aims to contribute to this general debate by studying the syntax and semantics of bare nouns in argument positions. The paper is organized as follows. The next section discusses the syntax and semantics of bare nouns in the adult target grammar, i.e., Spanish. Section three presents our predictions and research questions after reviewing some relevant studies for our study. Next, section four presents our own study in which we describe the participants, methodology and results. The main part of this article finishes with a discussion and general conclusions.

2. Linguistic Phenomena: Spanish Under current minimalist proposals (Chomsky 1995, 2007), parametric differences among grammars are associated with properties of the lexicon. The lexicon, for example, includes functional categories related to number (Num), agreement (Agr), tense (T). These functional categories can be different from one language to another as to their configuration of associated formal features and feature values. Moreover, not all formal features have the same status whereby some are un-interpretable and others interpretable at the level of L(ogical) F(orm). Un-interpretable features encode only syntactic/formal properties relevant to the syntactic component, whereas interpretable features carry meaning and have semantic import. With this in mind, languages can vary (a) as to the functional categories instantiated in their grammar and (b) the features associated with particular functional categories. In this sense, let us present in what follows the

6

Chapter One

functional categories involved and the features associated to them that constrain the grammar of Bare Nouns and DPs in subject and object position. (1) Bare plurals in Romance (Spanish) a. Subjects: *Leones viven en el Sahara.1 Viven leones en el Sahara. ‘Lions live in the Sahara.’ b. Objects: Busco juguetes. (Generic in Spanish) ‘I’m looking for toys.’ c. Prepositions: 1. Está en casa. ‘(He/she) is at home.’ 2. Va con amigos. ‘(He/she) goes with friends.’ First we will focus on the syntactic constraints to end up with the description of the semantic intricacies of the distribution of BNs and phrases with determiners. Comparing (1a) and (1b) which holds across in two different syntactic positions, one can detect clear asymmetries. Namely, whereas BNs are disallowed in Romance languages (e.g., Spanish) in preverbal subject position, BNs are permitted in object position. Moreover, bare nouns are allowed in prepositional phrases as (1c1) demonstrates. In short, a combination of lexical/syntactic features constrains the appearance of bare nouns, namely, lexically governed positions (O, P) allow for bare nouns, whereas BNs are banned from positions which are not lexically governed such as the specifier of INFL. Let us turn to the semantic import of BNs. Following Diesing (1992) amongst many others on the semantic literature, the bare nouns in subject position hold an existential or kind reference meaning depending on the predicate. As regards bare nouns in object position, there it is a less clear-cut issue since they can hold an existential meaning with some extensional verbs (e.g., ‘John eats bananas’ with the implication that there are some bananas that were eaten by John), whereas they do not hold such an existential meaning with other intentional verbs (e.g. ‘the child needs shoes’). The same holds for BNs in prepositional phrases (1c2). Bare nouns in prepositional phrases denote, however, in a restricted number of cases other meanings which are not directly linked to the meaning of the noun at stake as (1c1) illustrates, where the bare noun denotes ‘home’. In explaining the distribution and interpretation of the generic interpretation of plural NPs, we are dealing with a semantic parameter (Chierchia 1998): whether a phrase like the lions can have both a

Definite and Bare Noun Contrasts in Language Acquisition

7

maximality and a generic interpretation, or the maximality interpretation is the only option available. Based on the syntax/semantic interface set in (2) that conceives that syntactic categories are mapped onto semantic types: (2)

NP …….. DP ……. e

Chierchia (1998) proposes the Nominal Mapping Parameter (NMP), which accounts for typological variation amongst languages regarding the supply and semantic interpretation of bare nouns. Languages differ depending on whether the nominals are mapped directly as semantic arguments (type e) or semantic predicates (type ). Within the NMP three types of languages are distinguished. In Type I languages like Chinese and other classifier languages which do not possess determiners and number, NPs are categorized as [-pred,+arg], where every lexical noun is a mass noun and therefore NPs are mapped directly as arguments. In Type II languages (i.e., Germanic languages), NPs are mapped freely as either semantic arguments (names of kinds, regularities that occur in nature) or as predicates yielding a [+arg, +pred] feature specification. Under the [+arg, +pred] setting of the parameter, bare nominals are permitted with the caveat that individual lexical items are associated with one feature value or the other. That is, NPs are restricted by the count/mass and singular/plural properties of the noun. Under Chierchia’s NMP, singular mass nouns are allowed in Germanic languages because they can directly map as e, if they are marked in the lexicon as [+arg]. Bare plurals are also permitted because count nouns are of type ). A summary of these presented language types so far is summarized under (3). (3)

The Nominal Mapping Parameter

a.

[+arg, -pred] feature specification (e.g. Chinese Type) • generalised bare arguments • all nouns are mass nouns • no plural morphology • generalised classifier system

b.

[+arg,+pred] feature specification (e.g. Germanic Type) • bare mass nouns and plurals in argument position • no bare singular count nouns • plural morphology

Chapter One

8

c.

[-arg,-pred] feature specification (non-existent language type)

Lastly, Type III languages (Romance languages including Spanish) are characterized by the generalized use of determiners as NPs and its specification is the following [-arg, +pred] (3d). d.

[-arg,+pred] feature specification (e.g. Romance Type) • no bare nominals in argument position • count/mass distinction • morphological plural

Thus, every noun is mapped as a predicate only in a Romance language and is therefore non-argumental since predicates cannot occur in argument position (see (5a) below). Note, however, that proper nouns which are per se BNs are not bound to this restriction (4a-b). This fact can be easily accounted for if we assume that the proper nouns have a D layer with a zero D. (4) a. Alaitz compró libros. b. [DP į [NP libros] c. Libros compró Alaitz ‘Alaitz bought books’ In (5a), bare nominal subjects with generic interpretation are banned because Spanish (Romance) Nouns are ) and NPs cannot be arguments unless D3 is fully projected (5b). This restriction involves that a DP layer renders the nominal into a semantic argument with a zero-D as in lexically governed positions as previously shown in (1b)-(1c) and with an overt D in functionally governed positions (5b). That is, the grammaticality is warranted if a determiner is used to render the sentence felicitous, as shown in (5b): (5) a. *Gatos son amistosos. cats are friendly ‘Cats are very friendly.’ b. Los gatos son amistosos. [+generic, +specific] the cats are friendly ‘Cats are very friendly.’ c. *Alaitz fue a comprar tornillo Alaitz went buy screw ‘Alaitz went to buy a screw’

Definite and Bare Noun Contrasts in Language Acquisition

9

Note, however, that D cannot be projected at all if the noun is count and singular as in (5c). Let us now mention one important semantic property of determined nouns. In (5b), the definite plural can have both kind (generic) and subset-denoting (specific) readings. Making the long story short, grammatical properties are determined by interpretable features (such as the use of articles to mark definiteness in Germanic languages, or the use of plural NPs with definite articles to express generic meanings). That is, Germanic languages choose the ‘maximality’ value on the definiteness scale, but not the ‘generic’ value (Dayal 2004). More importantly, Germanic (e.g., German and English) and Romance languages (e.g., Spanish) differ regarding the possibility of bare nouns and their semantic interpretation. Spanish is a language with bare plurals in post-verbal object position, but not in the subject position, while Germanic languages allow bare plurals in both positions. This leads to clear asymmetries across language families which are constrained by the semantic interpretable features [genericity, specificity] (Chierchia 1998; Dayal 2004). Recall in Chierchia's paradigm, a broad typology of language types could be made based on the choice of values for binary features [±arg] for argument and [± pred] for predicate. In sum, nouns in Spanish have the specification [-arg, +pred] and need a DP layer in order to be able to emerge as arguments. Bare nouns in both subject and object syntactic positions comprise a null Do which is licensed by a lexical or a functional head. Hence, children acquiring Spanish need to know that the D-layer is always projected wherever a BN occurs and as a consequence: •BNs are banned from the preverbal subject position •Singular count BNs are banned from any position Moreover, they need to know that there is no bi-univocal relationship between kind reading and the status of BNs since determined nouns may be interpreted as generic.

3. Learnability Task and Predictions Current studies on the acquisition of determiners across languages point to the fact that determiners emerge early in languages where determiner insertion is generalized: Chierchia et al. 1999 for Italian; Guasti et al. 2004; Guasti & Gavarró 2003 for Catalan; Marinis 2003 for Greek; Kupisch 2006 in bilingual German-French and German-Italian children, to name but a few. In particular, Chierchia et al. (1999) argue that differences in the acquisition rate are due to syntactic/semantic factors. Based on corpus data analysis it

10

Chapter One

was shown that definite articles start to be present earlier in French and in Italian than in English and Swedish. Moreover, Guasti and Gavarró (2003) and Guasti et al. (2004) also report the appearance of articles earlier in Catalan than in Dutch; and although the omission rate is very similar at the first stage of development, the omission rate stays high in Dutch and drops quite considerably in Catalan and Italian. The results reported also by Marinis (2003) for Greek seem to go in the same direction: Greek children behave as their Romance counterparts with early and widespread use of articles. Taking together the main findings from the studies reviewed, some of the predictions made by the Nominal Mapping Parameter are held true, that is, earlier surfacing of D in Romance languages than in Germanic languages. Other studies with direct implications to ours are the ones by Pérez-Leroux and Roeper (1999) who studied 36 school and pre-school children and 33 college students whose mother tongue was English. According to their results, children at a very early age show knowledge of the specific environment in which a noun is licensed, and, more importantly, children seem to possess the semantic distinctions from very early on in their language development. In a more recent study which also considers the interpretation of BN/NPs distinctions, Gavarró et al. (2006) consider the extent of the children’s mastery of the semantic meanings by comparing the acquisition of Catalan definite DPs vs. Bare Nouns (BN) in object and subject positions. Thirty-three children (3-5 years old) were examined for comprehension of bare noun and definite contrast in Catalan. Catalan, like Spanish (the language under investigation in the present paper) allows bare nouns in object position, but not in subject position. Interestingly, Catalan objects match English subjects in terms of their semantic interpretation. Results indicate that children at the age of 3 are not able to discriminate between bare nouns and definite use of articles in object position, although by the age of 4 children showed some degree of sensibility. The authors argue that bare nouns possess a default generic reading that can swing into an ‘instance-of-a-kind’ semantic interpretation. More importantly, the acquisition patterns are motivated by a principle of semantic contrast that seems to confine the initial overuse of bare nouns to specific readings. Taking into account the findings from previous studies on the acquisition of bare nouns in different syntactical positions and the predictions made by the Interface hypothesis, we predict that the semantic features interact with their respective syntactic positions and thus represent a challenging learning task for monolingual and bilingual children at very early stages, due to the interface character of this domain. That is, does the child figure out that a noun can hold a generic and specific reading depending on the specific syntactic position (subject vs.

Definite and Bare Noun Contrasts in Language Acquisition

11

object/prepositional phrase)? In Romance languages (Spanish in our case) there is not a morpheme which directly encodes genericity. Previous literature on semantic acquisition (e.g., Hollander et al. 2002; Gelman, Star & Flukes 2003) indicates that children attain generic semantic interpretation rapidly and effortlessly. Hence, we would expect children to acquire the generic objects in Spanish with ease. Children should be able to acquire the semantic interpretation of the BN/DPs together with the acquisition of the specific categories of the DP phrase. However, following Chierchia et al’s (1999) main conclusions, we would therefore expect children to initially start mapping nominals with generic interpretation, and as a second step in the acquisition process children would induce nominals with specific interpretation when they have a fully integrated semantic and pragmatic linguistic system. Furthermore, following Dayal’s (2004) wide spectrum of cross-linguistic variation on how languages represent semantic meanings (i.e., generic vs. specific), we would expect differences between individuals depending on the language they are exposed to, but not with regard to the syntactic position in which bare nouns/DPs can appear. Bearing in mind the predictions of the Subset/Superset relationship of the L1 (and L2 grammars) for the present semantic domain (e.g., Wexler & Manzini 1987), the Nominal Mapping Parameter system predicts a very specific implementation order of the category determiner. The child would start with the most restrictive, economical grammar, i.e., with BNs with generic interpretation, and later on would obtain the knowledge of characterizing generic from specific interpretations, by implementing the language specific forms that introduce disambiguation when the grammar expands. In the null learning hypothesis, the most economical one retains the default meaning. That is, restrictive in the sense that children should start with the feature specification that excludes the most, so that they alter their hypothesis on the basis of positive evidence. Consider (6) below. (6)

Chinese Type

Romance Type

German Type

12

Chapter One

In (6) above, the most restrictive feature specification within the Nominal Mapping Parameter is [+arg, -pred], which relates to the Chinese setting. This is based on the notion that in languages of this type: (i) nouns occur without determiners, (ii) the extension of nouns is mass, (iii) there is no plural marking and (iv) a classifier system is operative. Properties (i) to (iv) are also encompassed in the Romance and Germanic type of languages. Upon noticing definite articles, plural marking and quantifiers joined directly with nouns in the input, children should change the initial setting to [-arg, +pred] (which corresponds to the Romance setting). The value of the Romance type languages is the next most restrictive feature matrix, because in Romance type languages: (i) the extension of nouns is mass or count, (ii) nouns occur with determiners, (iii) there is plural marking on count nouns and (iv) a classifier system is operative. No nouns without the D-layer are licit in the second setting. That is, children are expected to always project a DP-layer when the noun phrase is in an argument position. Nevertheless, children have to work out which nouns are mass and which ones are count, together as to when a null Do is licit or illicit. Lastly, unrestricted appearance of bare mass nouns in argument positions should direct the children to change the value of the NMP to the setting of the Germanic type languages, [+arg, +pred]. The Germanic type value is the least restrictive feature matrix, because: (i) the extension of nouns is mass or count, (ii) nouns may occur with or without determiner, (iii) there is plural marking on count nouns and (iv) a classifier system is operative with mass nouns. Since it is estimated that all children start with the Chinese setting regardless of the language they are acquiring, we predict an initial stage of article-drop (i.e., children will only use bare nouns). Children acquiring Spanish, after having discovered definite articles, plural morphology and/or numerical quantifiers conjoining directly with nouns, should switch to the Romance setting. This predicts that at a second stage children will project a

Definite and Bare Noun Contrasts in Language Acquisition

13

DP-layer. Because input to Spanish children contains bare nouns in argument positions, children have to uncover in which position bare nouns are licit or illicit (e.g., uncovering the distinction between argument and non-argument; identifying the subject and object position and that Spanish displays a subject and object asymmetry). Indecision on the child’s part would imply an optional use of the definite articles. To sum up, the main research questions of the present paper are to test: a) Whether the bare noun stage which has been attested to in the literature can be found in two monolingual and one English/Spanish bilingual child. b) Are target-deviant BNs equally possible in all syntactic positions in monolingual and bilingual children?

4. Study In the present study, we analyze the early performance in three longitudinal corpora. All corpora are available in the Childes database (http://childes.psy.cmu.edu/) (McWhinney & Snow 1985; McWhinney 1995). Magín belongs to the Aguirre corpus, Irene to the Irene corpus and M to the Deuchar corpus.

4.1 Participants The utterances of two monolingual children living in Spain4, the male Magín and the female Irene as well as a bilingual Spanish/English child living in the UK have been taken into account. Table 1-1: Children under investigation Name

Language

Irene Magín M

Monolingual Sp. Monolingual Sp. Bilingual En/Sp.

Investigation period by age 1;7-2;10,24 (29 files) 0;11,01-2;11,27 (57 files) 1;03,04-2;06,02 (9 files)

4.2 Methodology For the present paper, we counted all nominal expressions occurring in subject and object position. The nominals were distributed in 6 categories N[+def], N[-def], target deviant N[det], target-like and target-deviant BNs5

14

Chapter One

(namely, bare nouns). Proper nouns were counted for the subject option separately from the remaining BNs since proper nouns are allowed in all positions in Spanish as opposed to BNs which are banned from the preverbal position. Proper nouns are not very frequent with objects or prepositional phrases and were ignored in these categories. In order to test whether the acquisition of the DP is constrained by the position in which they are located, we divided the whole corpus in pre- and post-verbal subjects, objects, prepositional phrases and predicates. Moreover, the verbs were divided into four classes: transitive/unergative, unaccusative, copulative (ser ‘to be’), copulative (estar ‘to be’). The nominals occurring in isolation, imitations and repetitions were ignored.

4.3 Results Figure 1 illustrates the overall distribution of target-like, target-deviant BNs and target-like DPs as well as proper nouns in early Spanish. The first observation that comes to mind is that, generally speaking, DPs and BNs are far more frequent with objects than with subjects in all three participants. Noteworthy is the fact that a negligible number of subjects appears as proper nouns. Not surprisingly, Irene has more data than Magín and the latter more than M, a trend that can be easily explained by the number of files considered in this study. Target-like BNs are attested to in all children, although they are more frequent with objects than with subjects. Interestingly, all three subjects use target deviant BNs at some point of their development. However, the incidence of target-deviant BNs overall is not very high and some interesting usage patterns can be attested to. As expected, the monolingual child Irene does not use many target-deviant BNs. As opposed to Irene, both the monolingual Magín and the English/Spanish bilingual M make extensive use of target-deviant BNs, an issue which will be analyzed in detail below. The relationship between nominals with determiners and the target deviant counterparts is depicted in Figure 2. This figure shows that both monolinguals use a large number of nominals with determiners with all verb classes except for the copula ser. The bilingual M has only a couple of examples with determiners. As to the target deviant BNs, Irene’s use of these nominals stands well below the critical level of 5% and is only found with one subject of a transitive verb. As opposed to Irene, Magín’s data show plenty of target-deviant BNs distributed across subject positions of the transitive and unaccusative verbs. As opposed to the data in the monolingual subjects, the bilingual M’s target-deviant BNs with subjects outnumber the nominals with determiners. And, interestingly, the only subjects with

Definite and Bare Noun Contrasts in Language Acquisition

15

target-deviant BNs in M are found with unaccusative verbs. It should be noted, that target-deviant BNs are not used with copula. The target-deviant BNs are not the only errors the subjects make as Figure 3 demonstrates.

Figure 1-1: overall distribution of subject BNs and DPs (tokens) in child Spanish

Figure 1-2: N[det] compared to target-deviant BNs (tokens) in subject position according to four verb classes

16

Chapter One

Figure 1-3: Target-like BNs as compared to target-deviant N[det] (tokens) in subject position across verb classes.

Target-like bare nouns are not very numerous in subject position in any of the children. The bilingual child M has neither target like BNs nor target-deviant N[det] in subject position. Interestingly, both monolingual children use N[det] in contexts in which a BN is expected. BNs are allowed in all structural positions (VS, VO, OV and PP) except for SV in adult Spanish. Figure 4 depicts all the target-deviant BNs according to position SV or VS. Figure 4 shows that both monolinguals and the bilingual child use target-deviant BNs in postverbal position more frequently than in preverbal position. The bilingual child has no target-deviant BNs in postverbal position. The monolingual child Irene has no examples of BNs in preverbal position. Target-deviant BNs are attested to in preverbal position by the monolingual child Magín and the bilingual M. Target-deviant BNs with objects and predicates are shown in Figure 5.

Definite and Bare Noun Contrasts in Language Acquisition

17

Figure 1- 4: Target deviant BNs (tokens) in preverbal and postverbal position across verb classes.

Figure 1-5: Nominals with Objects and predicates (tokens).

Chapter One

18

The proportions of target-like BNs, target-like N[det], target-deviant BN and target-deviant N[det] are depicted in Figure 5. Target-deviant BNs are not numerous in the case of Irene as it was the case with the subjects. Overall, Magín uses more target-deviant BNs than Irene with objects and predicates. The bilingual child M does not make any errors. Since all subjects make errors of omission of determiners to a more or lesser extent, it should be investigated whether there is a period in which the errors are more frequent and disappear at a later stage. Table 1-2: Total number of target-deviant BNs(%) until age 2;0 and from 2;0 onwards. Until 2;0 From 2;0 onwards

Irene 13(10,24%) 2(0,24%)

Magín 45(17,11%) 20(3,3%)

M 3(75%) 1(1,42%)

Since the age of two is cross-linguistically a turning point in which many grammatical features are implemented in child grammar and the NMP predicts illicit BNs at early stages, we made an ad hoc cut at this age in order to ascertain, whether target-deviant BNs are a phenomenon that should be ascribed to early developmental stages. Table 2 shows the results for the three subjects. The monolingual child Irene makes the bulk of errors until the age of 2;0. The two remaining errors after 2;0 are, because of the magnitude of the data, negligible. Once again, the monolingual Spanish child Magín differs from Irene and M in at least two ways. The monolingual child Magín, moreover, makes far more errors than both Irene and M together and his errors prevail in the entire studied period. M only makes three errors in the whole study and they are located before the age of 2. In sum, target-deviant bare nouns are attested to in all three children, but to a different extent in early and later stages. In order to exclude the possibility that illicit BNs are used in positions in which BNs are lexically licensed, let us examine the use of target-deviant BNs in lexically governed (V) and ungoverned (SpecIP) and preverbal and postverbal positions.

Definite and Bare Noun Contrasts in Language Acquisition

19

Figure 1-6: Target-deviant BNs (tokens) according to syntactic position.

The position in which BNs occur varies strongly according to subject. The monolingual child Irene makes errors with both subjects and objects, but, interestingly, all errors occur in post-verbal position (VS, VO). As opposed to Irene´s data, Magin’s corpus shows errors with both subjects and objects in preverbal and post-verbal position. Note that a large number of errors is attested to in SV position. By contrast, the bilingual child M makes only a few errors. Recall, however, that the data in this child are not numerous. It is noteworthy, however, that two of the three errors occur in preverbal subject position. As regards prepositional phrases, all three children drop determiners in prepositional phrases for a long period of time.

5. Discussion and general conclusions Our main results can be summarized as follows: -The two monolingual and the English/Spanish bilingual subjects go through a stage in which determiners are dropped in Spanish. -As expected for Spanish, the monolingual child Irene reaches an almost 100% correct performance at around the age of two. -The monolingual child Magín very much differs in terms of

20

Chapter One

performance as compared to the monolingual Irene. Some of the target-deviant utterances in Magín´s data very much resemble the data of the bilingual English/Spanish child M. The first finding is related to the early appearance of determiners in child language in Spanish by monolinguals and bilinguals which goes in line with previous studies reported in this paper (Chierchia et al. 1999 for Italian; Guasti et al. 2004; Guasti & Gavarró 2003 for Catalan; Marinis 2003 for Greek; Kupisch 2006 for bilingual children). The main issue to be discussed in this paper was whether the syntactic position in which noun phrases appear has any bearing in the implementation of the category Det. The data we have examined clearly show that the syntactic position of the noun phrases at stake is most important when implementing Det. The data we have scrutinized show that all children make errors in the postverbal predicate position. The same holds for the postverbal subject, object position and when the nominals occur under P. Errors in the postverbal position with subjects, predicates and objects practically cease to occur after the critical age of two in the subjects Irene and M. With regard to prepositions, Irene makes errors with prepositional phrases in the entire study as do Magín and M. The common denominator to nominals occurring under P, O or VS is that they are located in complement position. If the structural position is the same, why are DPs in the object and subject position error free and why do errors persist until the very last recording with prepositions? In the case of objects, the nominals used by the children in fact refer to kinds as all children use target-like BNs very early. Moreover, children use indefinites and nominals with definite determiners which refer to specific entities in a target-like fashion very early. The stage in which target-deviant BNs in object and postverbal subject position, that is BNs which are intended as definite or indefinite DPs, lasts for a short period of time in the case of Irene and M. Even in this period of time in which BNs are target-deviant, they are licensed by V and, hence, syntactically allowed. If the omission errors disappear, it is because the children have implemented definiteness to their grammars. But, what about the prepositional phrases? Indeed, some of the nominals located under P refer to specific entities or are indefinite. However, all children produce a few target-like BNs in prepositional phrases which eventually denote kinds, but they crucially also use target-like BNs in chunks as in (7). (7) a. Van en tren. (M, 2;05,05) ‘(They) go by train.’

Definite and Bare Noun Contrasts in Language Acquisition

21

b. Y no lo tengo en casa, cómprame uno! (Magín, 2;06,10) ‘And (I) do not have it at home, buy me one.’ c. De cristal (Irene, 2;09,11) ‘In glass’ The fact that the children use determiners very early with nominals under P in a target-like fashion reveals that they are aware of the semantic properties of nominals with and without determiners. However, the BNs under (7) cannot be said to denote kinds per se. The meaning of P + N is not the sum of the meaning of P and N, but has a meaning as a chunk as (7b) shows. The preposition en has a local interpretation whereas casa refers to a building. However, en casa cannot be interpreted as ‘in the building’ but only as ‘at home’. The local meaning of at remains in the background and home does not necessarily denote a building in the construction at home. Children must learn the meanings of the chunks in the same way they learn other lexical items and may eventually get confused in such a way that they may interpret the BNs in chunks as denoting kinds. This is the more so likely if we look at utterances with target-deviant BNs with prepositions (8). (8) a. *Aquí en brazo (Irene 2;08,14) ‘Here in arm’ b. *No cabe en cama (M, 2;02,05) ‘(It) does not fit in the bed’ c. *la casita de perro (Magín 2;4,25) ‘The little house of the dog’ In examples in (8), Irene should have used a definite article (8a) because she was referring to her own arm, and M meant the bed (8b) and Magín referred to the dog (8c), but they omitted the determiner. Note that Spanish knows an almost identical chunks en brazos ‘in someone’s arms’, en cama ‘in bed’. It is plausible to assume that the high number of chunks with prepositions in Spanish hinders the otherwise rapid implementation of NMP in VO and VS positions. Nevertheless, the chunks are not the whole story. If we look at the use and appearance of bare nouns, we can make the observation that target-like bare nouns with a kind reading appear very early on and, crucially, simultaneously with objects (9) as do nouns with determiners (10) with a definite reference. (9) a. Echo agua (Irene, 1;08,09) ‘(I) throw water’

22

Chapter One

b. Be(ber) aba(=agua) (Magín, 1;7) ‘Drink water’ (10) a. (N)o está (l)a f(l)or (Magín, 1;7) ‘The flower is not (there)’ b. Ya (.) quita el viyeo (.) papa. (Irene 1;7,22) ‘Well (.) turn off the video (.) Daddy’ However, omission errors with objects remain until the age of around 2 in all children. After this age, omission errors drop dramatically in the case of Irene and M which is in line with our expectations for a Romance language. Magín’s performance very much differs from the performance of Irene and M and from the cases reported in the literature and deserves, therefore, a detailed analysis. (11) a. *Ha hecho chistá (=chiste) (Magín 2;05) ‘(I) made a joke’ b. *Yo quiero limpiar popo (polvo) (Magín, 2;07, 26) ‘I want to clean dust’ c. *Toca la música (Magín 2;05) ‘(He) plays music’ d. *Te disparo que echa el humo (2;06,20) ‘(I) shoot so that it throws smoke’ e. *Redondo es un sol (2;06,10) ‘Round is a sun’ M continues making omission errors after the age of 2;0 as (11a) shows where an indefinite determiner is expected because new information is introduced. New information is also introduced in (11b) but only a definite article gives a felicitous utterance although it has a kind reading. Examples (11c) and (11d) have a kind reading as well but disallow determiners. The sun is unique and as such has a specific reading and, therefore, requires a definite article and not an indefinite one. All these examples show that there is no biunique relationship between kind readings and bare nouns, specific readings and definite articles. Moreover, encyclopaedic knowledge is required in order to be able to use the word sun correctly which might not yet be available at the age of 2;06 and can, hence, explain the error. If we turn now to the subjects, Magín makes a number of errors with kinship names and animal names. It must be noted that papá ‘Daddy’ and mamá ‘Mummy’ are usually analyzed as proper nouns because they are unique and can, consequently, appear in preverbal subject position as bare nouns.

Definite and Bare Noun Contrasts in Language Acquisition

23

However, the remaining kinship names are common nouns and hence require a determiner if they are to be used as subjects in preverbal position. Examples in (12) clearly demonstrate that some common nouns belonging to the class of kinship terms are analyzed as proper nouns by Magín. (12) a. Tú te llamas una nena ‘Your name is a child’ b. Ab(u)e(l)a se llama ‘Her name is grandmother’ c. Ab(u)ela se ha roto (1;09,27) ‘Grandmother broke’ d. Es nene (1;10) ‘(It) is child’ e. Se llama, niña. (MOT: cómo se llama tu amiguita?) ‘(Her) name is child’ (Mot: What is the name of your little friend?) It is worth mentioning in this respect that some of the target-deviant BNs in preverbal position belong to the semantic class of animals (13). (13) a. *Lorito van (1;09;01) ‘Parrots go’ b. *Mosquito baja (2;0) ‘(The) mosquito comes down’ c. Me [*] ha mordido perro. (2;11,20) ‘(The) dog bite me.’ It is not quite clear whether animals are misanalysed as proper nouns by Magín, but it is not quite implausible in light of (13a-b). However, other common nouns appear in preverbal position as target-deviant BNs as well so that a semantic misinterpretation of the animal names does not adequately explain all the errors. An observation about (13a) is in order here. This example contains an agreement error of a special kind because the third person plural is used with a singular subject. One could argue that the plural marking in the noun is missing, which is not implausible, but other examples with agreement errors reveal that Magín has massive verb agreement problems (14) that persist until late in his linguistic development which require a closer scrutiny. (14) a. *Se van señora (1;10) ‘Go lady’

24

Chapter One

b. *El pepino pi(n)chan [: pincha] (1;10) ‘The cucumber punctures’ c. *Y la nube se van [*] . (1;10) ‘The cloud go’ e. Sí, son pez, en lacuario [: el acuario] . (2;05) ‘Yes, are fish in the aquarium) f. Se van, van xxx sol, redondo sol . (2;6,20) ‘Go, go xxx sun, round sun’ g. No las cangrejas está [*], no . (2;6,20) ‘No, the crab are, no’ h. *Dónde estoy [*] cogiendo una piedra, me ha [*] caí(d)o . (2;6,20) ‘Where am taking a stone, has fallen down’ In almost all errors listed under (14), a mismatch in number agreement (14a-g) is found, but not only number (14h). Let us look at the former more closely first. The subject stands in the singular and the verb in the plural (14a-f) or vice-versa the subject stands in the plural and the verb in singular (14g). In light of the example (14e) one is tempted to assume that the source of these errors is a mere morpho-phonological problem in the implementation of the plural morpheme –s in the nominals, since it is assumed that more than one fish is in the aquarium. But example (14f) disconfirms this hypothesis since, sol sun is a unique referent and can build the plural only in metaphorical contexts, probably not yet known to the child. That is, the number of the noun is target-like in this case. Moreover, the child uses the plural correctly in (14h) and, more importantly, uses closed syllables with –s in numerous non-morphological contexts. Note that all these examples were chosen as to show that the child under study has a major problem at an age well beyond the age of 2;0. And, finally, as example (14h) shows, the problem is not only phonological, since the first person is used for the third. The errors discussed so far show that the implementation of INFL is not fully fledged in Magín in the entire period under consideration, a finding that very much contrasts with the acquisition literature available for Spanish monolinguals and bilingual children (Ezeizabarrena, 1996). Our interpretation of these facts is only tentative, but Magín seems to treat INFL as a lexical category which licenses BNs in the specifier position. This would explain the high number of target-deviant BNs in preverbal position and the agreement errors in both positions. The bilingual subject M makes identical errors (only 2) in preverbal subject position, but does not make any agreement errors. Consequently, the mentioned errors must have another explanation. Bare nouns are allowed in the preverbal subject

Definite and Bare Noun Contrasts in Language Acquisition

25

position in English, the dominant language in M, since the child is living in an English speaking environment. Note, however, that this type of error disappears after the age of two in 2;0. Target-deviant BNs in preverbal position must be explained in a syntactic framework, but the remaining errors in the lexical positions, which represent the majority, require another type of explanation. The data scrutinized have shown that all children have enormous difficulties to represent kinds and referential entities in subject, object and predicate positions in the appropriate manner at an early stage and over the entire age span studied in this paper with prepositions. This goes in line with Gavarró´s et al. (2006) results who made the observation that three year olds had difficulties understanding semantic features of nominals. Consequently, our prediction that children would first use nominals which refer to kinds, in a second step, use nominals that refer to specific entities could not be confirmed, since all children use target-like bare nouns and specific nominals at the around same age. All the subjects in this study encounter problems with the semantics of bare nouns as opposed to nouns with determiners. The syntactical constraints of the different lexical positions are acquired early with the exception of Magín, who has persistent agreement problems whose origin must be studied in the future. Apart from this peculiarity, it can be concluded that syntactic and semantic features are acquired separately, and semantic properties seem to pose a greater problem than the syntactic ones.

Notes 1. The standard grammars of Spanish state that bare nouns are banned from the preverbal position. However, a quick glance at newspapers, especially the online versions, reveals that bare nouns in preverbal position are rather frequent, in particular in the headlines. To our knowledge, this phenomenon has not been studied to date. 2. Note that preverbal BNs in subject position are ungrammatical, an issue that will be discussed in more detail below. 3. In Spanish pre-verbal NP subjects are possible with an existential reading only. Even in such cases, BNs must be modified (i) or in conjoined structures (ii) such as the following examples taken from Laca (1999: 908): (i) Eléctricas letras verdes intermitentes anunciaron la llegada del vuelo. (ii) Fotógrafos y cámaras de la televisión llegaban con la obsesión puesta en los ojos y en los codos. 4. A natural step in our research programme is to investigate parents’ performance too in relation to the linguistic phenomena under investigation. 5. The term target deviant BN has to be understood in the sense that a bare noun is used in a context where a noun with determiner is expected.

26

Chapter One

References Bos, P., Hollebrandse, B. & P. Sleeman. 2004. The pragmatics-syntax and the semantics-syntax interface in acquisition. IRAL 42, 101-110. Chierchia, G. 1998. References to kinds across languages. Natural Language Semantics, 6, 339–405. Chierchia, G., Guasti, M. T. & A. Gualmini. 1999. Nouns and articles in child grammar and the syntax/semantics map. Ms. U. of Milan, U. of Sienaand U. of Maryland. Chomsky, N. 1995. The Minimalist Program. MIT Press, Cambridge, MA. —. 2007. Approaching UG from Below. In U. Sauerland & M. Gaertner (eds.), Interfaces + Recursion = Language? Chomsky's Minimalism and the View from Syntax-Semantics, pp. 1-30. Berlin: Mouton de Gruyter, Dayal, V. 2004. Number marking and (in)definiteness in kind terms. Linguistics and Philosophy, 27, 393-450. Diesing, M. 1992. Bare Plural Subjects and the Derivation of Logical Representations. Linguistic Inquiry 23, 353-380. Ezeizabarrena, M. J. 1996. Adquisición de la morfología verbal en euskera y castellano por niños bilingües. Bilbao: Universidad del País Vasco/Euskal Herriko Universitatea. Gavarró, A., Pérez-Leroux, A.T. & T. Roeper. 2006. Definite and bare noun contrasts in child Catalan. In. Torrens, V. & L. Escobar (eds.), The Acquisition of Syntax in Romance Languages, pp. 51-68. Amsterdam: John Benjamins. Gelman, S., Star, J. & J. Flukes. 2003. Children’s use of generics in inductive inferences. Journal of Cognition and Development 3, 179–199. Guasti, M., de Lange, T. J., Gavarró, A. & C. Caprin. 2004. Article omission: across child languages and across special registers. In J. van Kampen and S. Baauw (eds.), Proceedings of GALA 2003, pp. 199–210, Utrecht: LOT. Guasti, M. T. & A. Gavarró. 2003. Catalan as a test for hypothesis concerning article omission. In B. Beachley, A. Brown and F. Conlin (eds.), Proceedings of the 27th Annual Boston University Conference on Language Development, pp. 288–298, Somerville: Cascadilla Press. Hollander, M., Gelman, S. & J. Star. 2002. Children's interpretation of generic noun phrases. Developmental Psychology 38, 883–894. Kramer, I. 2000. Interpreting Indefinites. An Experimental Study of Children's Language Comprehension. PhD dissertation, University of Utrecht. Kupisch, T. 2006. The acquisition of determiners in bilingual German-Italian

Definite and Bare Noun Contrasts in Language Acquisition

27

German-French children. Munchen: Lincom Europa. Laca, B. 1999. Presencia y ausencia de determinante. In I. Bosque & V. Demonte (Eds.), Gramática descriptiva de la lengua española, pp 891–928. Madrid: Espasa Calpe. Longobardi, G. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25, 609-665. Marinis, T. 2003. The Acquisition of the DP in Modern Greek. Amsterdam: John Benjamins. MacWhinney, B. 1995. The CHILDES project: tools for analyzing talk. Hillsdale, NJ: Lawrence Erlbaum Associates. MacWhinney, B. & C. E. Snow. 1985. The Child Language Data Exchange System. Journal of Child Language, 12, 271-296. Müller, N. & A. Hulk. 2001. Cross-linguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 4, 1-21. Paradis, J. & S. Navarro. 2003. The use of subjects by a Spanish-English bilingual child: Cross-linguistic influence or the influence of the input? Journal of Child Language 30, 371-393. Patuto, M. 2012. Der Erwerb des Subjekts in (Nicht-)Nullsubjektsprachen: Die Rolle des Spracheneinflusses und der Sprachdominanz bei bilingual deutsch-italienisch, deutsch-spanisch und französisch-italienisch aufwachsenden Kindern. PhD dissertation, Bergische Universität Wuppertal. Pérez-Leroux, A. T. & T. Roeper. 1999. Scope and the structure of bare nominals: evidence from child language. Linguistics 37, 927–960. Platzack, C. 2001. The vulnerable C-domain. Brain and Language 77, 364-377. Serratrice, L., Sorace, A. & S. Paoli. 2004. Transfer at the syntax-pragmatics interface: subjects and objects in Italian-English bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7, 183-205. Schmitz, K. 2007. L'interface syntaxe - pragmatique: Le sujet chez des enfants franco-allemands et italo-allemands. AILE [Acquisition et Interaction en Langue Etrangère] 25, 9-43. Sorace, A. 2004. Native language attrition and developmental instability at the syntax-discourse interface: Data, interpretations and methods. Bilingualism: Language and Cognition 7, 143-145. —. 2005. Selective optionality in language development. In L. Cornips and K.P. Corrigan (eds), Syntax and Variation: Reconciling the Biological and the Social, pp. 55-80. Amsterdam: John Benjamins. —. 2011. Pinning down the concept of “interface” in bilingualism.

28

Chapter One

Linguistic Approaches to Bilingualism. 1, 1-33. Tsimpli, I.-M., Sorace, A., Heycock, C. & F. Filiaci. 2004. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8, 257-277. Wexler, K. & R. Manzini. 1987. Parameters and learnability in binding theory. In T. Roeper and E. Williams (ed.), Parameter Setting, pp. 41-76. Dordrecht: Reidel.

PART II: SYNTAX-SEMANTICS

CHAPTER TWO OVERT QUANTIFIER RAISING OF NEG-WHQ* IN CANTONESE AT THE SYNTAX-SEMANTICS INTERFACE MAN-KI THEODORA LEE Abstract This paper reports a NegQP (Lee 2011) structure for negative wh-quantifiers (Neg-whQ) in Cantonese, in the combination of a negator mou and a wh-phrase (any DP for other non-existential quantifiers) where the former precedes the later, and discusses the unique overt quantifier raising phenomenon with such kind of strong quantifiers (Diesing 1992) at the syntax-semantics interface. The overt quantifier raising is claimed to be triggered by the uninterpretable features [quant] (Chomsky 1995) and [neg] to account for the sole SOV order of constructions with strong quantifiers in Cantonese. The overt quantifier raising accounts for the dual interpretations of constructions with a neg-whQ, having the non-existential interpretation as well as the possible presupposed existential or sentential negation interpretation.

1. Introduction Among the many languages having overt quantifier raising (QR), such as Hungarian, French and Icelandic etc., Cantonese negative wh-quantifiers (Neg-whQ) appear to be one of the kind. Neg-whQ, the negative wh-quantifier, refers to any non-existential quantifiers in the combination of the negator mou and any ordinary wh-phrases, for example mou-bingo ‘nobody’, mou-matje ‘nothing’ and mou-bindou ‘nowhere’. NegQP proposed in my previous paper (Lee 2011), which has the negator mou at its spec position, an invisible quantifier operator Ø as its head and any wh-phrase (any NP for other non-existential quantifiers) as its complement, is unique in Cantonese. Apart from the same syntactic properties that Cantonese and Mandarin Chinese share in general, object neg-whQs

32

Chapter Two

undergo obligatory and overt QR in the former whereas neg-whQs can only appear as subject in the latter. This paper looks at Cantonese neg-whQ at a syntax-semantic interface and follows Chomsky’s (1995, 2000, 2001) approaches that quantifiers raise to satisfy uninterpretable features in order to receive the correct interpretation. By proposing the NegQP, I aim to provide a unified account for all movements related to negative quantifiers and even strong quantifiers in Cantonese. I include data to support the idea that the [quant] and [neg] features inherited from the head and spec positions of NegQP trigger movements of the object neg-whQ into [spec, vP] and [spec, NegP] positions. The overt quantifier raising of NegQP allows for the two possible interpretations from the neg-whQ itself, which are the original non-existential/ sentential negation interpretation and an additional (presupposed) existential interpretation. In section 2, I define neg-whQ and present my original data with overt quantifier raising phenomenon in Cantonese. Section 3 reports my proposal of NegQP and the two possible overt movements derived from it, which are the neg-whQ raising (QR) and the neg-raising. The neg-whQ raising represents the movement of neg-whQs from base object positions as one constituent, while the neg-raising represents the movement of only the negative morpheme out of the NegQP leaving the wh-phrase in situ. In section 4, I present empirical evidence (from other languages) and data on Cantonese to support my claim and section 5 explains how NegQP accounts for the data on NPI licensing and WCO cancelling. Last but not least, section 6 concludes on the account of overt QR of neg-whQ in Cantonese at the syntax-semantic interface.

2. Neg-whQ and the Overt Quantifier Raising Phenomenon in Cantonese The morphology of Cantonese Neg-whQ is the composition of a negative morpheme mou with any wh-phrase. In general, mou-bingo (no-who), mou-matje (no-what) and mou-bindou (no-where) are the common neg-whQs. They can appear either as the subject as in (1) or the object as in (2). Strictly speaking, they are not totally equivalent to the English non-existential ‘nobody’, ‘nothing’ and ‘nowhere’ in semantics. Neg-whQ constructions give not only a non-existential interpretation or sentential negation interpretation as in (2a), but also a presupposed existential interpretation as in (2b). (1)

Mou-bingo zungji ngo (jaa). NO-WHO LIKE ME SP ‘Nobody likes me. ’ / ‘Not anyone (someone) likes me.’

Overt Quantifier Raising of Neg-whQ in Cantonese

(2)

33

Ngo mou-matje sik-guo (wo). I NO-WHAT EAT-PVF SP a. ‘I ate nothing.’ / ‘I didn’t eat anything (something).’ b. ‘I ate only a few things.’

In certain contexts (with different tones of the sentence final particle), (1) can be possibly interpreted as “Not many like me” or “Only a few like me” and (2) can possibly be interpreted as “I didn’t eat a lot” or “I ate only a few”. In contrast with its most equivalent English counterparts, the non-existential quantifier in (3) does not seem to give a distinct interpretation from the negator and N(egative)P(olarity)I(tem) 1 combination in (4), and also neither of them give existential interpretation. (3)

Nobody likes me.

(4)

I don’t like anybody.

Neg-whQs are unique among other (strong) non-existential quantifiers, as they give the exceptional existential or sentential negation interpretation. As illustrated in (5)-(6), the ordinary non-existential quantifiers (e.g. moujan ‘nobody’ and mouje ‘nothing’) in the combination of the negative morpheme mou and an ordinary NP give only the non-existential interpretation. (5)

Moujan zungji ngo. NOBODY LIKE ME ‘Nobody likes me.’

(6)

Ngo mouje sik-guo. I NOTHING EAT-PFV ‘I ate nothing.’

The SVO word order is the canonical structure for Cantonese. Even so, noun phrases are likely to be dislocated resulting in OSV structure where the object is topicalized and VOS where the subject is right dislocated (look at Cheng’s2 work for more details on dislocation constructions). However, the unique SOV structure is only observed with strong (object) quantifiers while it is prohibited with any other ordinary NPs as in (7). Mou-bingo in (8) must precede the verb whereas (9), where it appears in-situ, is ungrammatical.

Chapter Two

34

(7)

*Ngo nei zungji. I YOU LIKE ‘I like you.’

(8)

Ngo mou-bingo zungji (jie). I NO-WHO LIKE SP a. ‘I like nobody.”/ “I don’t like anyone.’ b. ‘I only like a few people.’

(9)

*Ngo zungji mou-bingo. I LIKE NO-WHO ‘I like nobody.’

Cantonese is a dialect of Chinese sharing similarities with Mandarin Chinese (MC) in their grammar. Their wh-phrases can be licensed as indefinites as widely stated in the literature 3 and they both observe the unique SOV structure. On a par with MC, wh-phrases in Cantonese do not inherit internal quantificational force but need triggers. They not only can be interrogative words but also can be licensed as negative polarity items (e.g. in negated context, in conditional clause, etc.) and universal quantifiers. The observed overt QR to object quantifiers is only restricted to strong quantifiers, therefore suoyou dungxi ‘everything’ and the wh-phrase shenme ‘what’ in Mandarin Chinese in (10) and sojau-je ‘everything’ and the wh-phrase matje ‘what’ in Cantonese in (11) are licensed to be strong quantifiers with ‘dou-quantification’ 4 and must appear in preverbal positions. (10)

Mary suoyou dungxi / shenme dou xihuang chi. MARY ALL THING / WHAT ALSO LIKE EAT ‘Mary likes to eat everything.’

(11)

Mary sojau-je / matje dou soeng maai. MARY ALL THING / WHAT ALSO WANT BUY ‘Mary wants to buy everything.’

Regardless of the identical grammar the two languages have, there is no identical neg-whQ in MC. Meiyou-shei ‘nobody’ can appear in the subject position as in (12). However, as the object, it does not appear to undergo such overt quantifier raising and renders ungrammaticality in (13). In order to get the non-existential interpretation, the only option is to have the

Overt Quantifier Raising of Neg-whQ in Cantonese

35

wh-phrase in-situ and licensed as NPI as in (14). These examples suggest that the combination of a negative morpheme and a wh-phrase does not form one constituent in the same way as it does in Cantonese. The neg-whQ as an object in Cantonese, where the negator mou immediately precedes the wh-phrase and undergoes overt QR to a preverbal position, is therefore unique in Cantonese. (12)

Meiyou-shei xihuan ni. NO-WHO LIKE YOU ‘Nobody likes you.’

(13)

*Wo meiyou-shei xihuan. I NO-WHO LIKE ‘I like nobody.’

(14)

Wo meiyou xihuan shei. I NO LIKE WHO ‘I don’t like anybody.’ / ‘I like nobody.’

3. The Proposal I follow my previous proposal of an NegQP5 that Cantonese neg-whQ forms one constituent. Unlike the other strong quantifiers that require an external licenser as discussed above (e.g. dou-quantification), such neg-wh-quantifiers are licensed as neg-whQ in NegQP with an internal licensor mou and an invisible operator Ø. Neg-whQs undergo obligatory and overt movements and the kind of movements trigger alternate interpretations between the non-existential and presupposed existential/sentential negation interpretations. The NegQP accounts for the two possible overt movements, namely the neg-whQ raising (as QR) and the neg-raising.

3.1 The Structure of Neg-whQ as One Constituent The proposed NegQP gives the structure of any neg-whQs in Cantonese, where the negator mou immediately precedes the wh-phrase (e.g. bingo ‘who’, matje ‘what’ and bindou ‘where’). It has the negator mou at its spec position carrying the [neg] feature, an unpronounced quantifier operator Ø at its head position carrying the [quant] feature and any wh-phrase as its complement. It accounts for all negative quantifiers, which inherit both [neg] and [quant] features. This structure explains how the quantifier operator

36

Chapter Two

licenses wh-phrases as indefinites, and thus how neg-whQs can be decomposed into negation and an existential/indefinite element6. Both the non-existential and implied existential presupposition interpretations are allowed where neg-whQ moves to preverbal position as one constituent. According to Diesing7, ‘strong quantifiers’ must undergo overt quantifier raising (QR). I would later on consolidate my argument that neg-whQs form one constituent rather than simply mou in NegP attracting the wh-phrase to land at preverbal position, by giving data on neg-whQ constructions relating to a gap under long distance raising. Neg-whQs therefore belong to a kind of strong quantifiers, and movements (the neg-whQ raising) of such strong quantifiers to preverbal position involved are claimed to be overt QR. (15)

The NegQP can be further extended to a QuantP as in (16) to represent all quantifiers (including wh-phrases as indefinites) in Cantonese according to the compositions of quantifiers as explained in (17). Any classifiers, for example sojau ‘every’, jamho ‘any’, appear in the spec position (x in the below structure) or else it can be left empty; the unpronounced quantifier operator Ø remains as the head; and any DPs including the wh-phrases appear as the complement of QuantP. This can possibly explain the cases where wh-phrases can be raised or stay in-situ to be licensed as indefinites and remain in situ in interrogatives. It can also represent existential quantifier (e.g. jaujan) as strong quantifier that also observe overt QR. Due to the limitation of scope, this paper focuses only on Cantonese neg-whQs.

Overt Quantifier Raising of Neg-whQ in Cantonese

37

(16)

(17)

Cantonese quantifiers Cantonese quantifiers are formed in the combination of X and Y, where X is the classifier that determines whether the quantifier is universal (e.g. sojau ‘every-’) or non-existential (e.g. mou ‘no-’) or existential (e.g. mau ‘some-’), and Y can be any DPs modified by X or any wh-phrases.

3.2 The Overt Movements (Driven by Uninterpretable Features) The proposed NegQP accounts for the two possible overt movements. They include the neg-whQ raising (as one constituent) where dual interpretations (the non-existential “nothing” and the existential presupposition “only a few”) are preserved, and the neg-raising where only the non-existential (sentential negation) interpretation is preserved. Adopting the recent minimalist approach8, I hereby claim that the overt movements (Move) are motivated by matching and deleting the uninterpretable features in the probes and interpretable features in the goals under Agree. In order to get the Full Interpretation9, movements must apply to eliminate all uninterpretable features at interface. Following Begehelli and Stowell’s10 location in the functional structure of the clause of the five QP-types in (18), the landing site of the neg-whQ is Spec of NegP because it “checks [+Neg] in Spec of NegP, under agreement with the Neg-operator in Neg0” (p.76).

38

Chapter Two

(18)

As illustrated in the structure in (19), the neg-whQ in NegQP inherits both the interpretable [Neg] and [Quant] features from the negative morpheme mou and the invisible quantifier operator Ø. They match with the two uninterpretable features [uNeg] of the probe Neg0 and [uQuant] of the probe v. Since the goal and the probes Match these feature sets prompt Agree (downward right angled arrows indicate probing throughout this paper). The neg-whQ then is required to undergo successive cyclic movements in order to satisfy the unvalued features. It cannot simply be referred as Chomsky’s object shift11, since such raising is restricted only to and even obligatory to object quantifiers in Cantonese creating SOV order. The overt quantifier raising observed in Cantonese is on a par with Kiss’ proposal12 that quantified NP in Hungarian raises.

Overt Quantifier Raising of Neg-whQ in Cantonese

39

(19)

Under Agree, the two possible movements, namely the neg-whQ raising and the neg-raising, are triggered in order to account for the dual interpretations.

3.3 Neg-whQ Raising One possible overt movement by NegQP, which is neg-whQ raising, leads to the overt QR phenomenon in Cantonese where the neg-whQ raises as a whole to a preverbal position, as in (20). Notice that such raising is obligatory, because the neg-whQ staying in base generated object position leads to ungrammaticality in (21). (20)

Ngo mou-bingoi zungji ti (wo13). I NO-WHO LIKE SP a. ‘I like nobody.’ / ‘I don’t like anybody.’ b. ‘I like only a few people.’

40

(21)

Chapter Two

*Ngo zungji mou-bingo.

The derivation with the observing movements (upward curved arrows indicate movement throughout this paper) is represented in (22) as the following. (22)

The neg-whQ mou-bingo appears in preverbal position and makes (20) grammatical and yields the two possible interpretations. I follow Chomsky’s approaches14 that neg-whQs raise to satisfy uninterpretable and EPP features in order to receive the correct interpretation and the movement is governed by the Last Resort15 (LR). In the representation in (22), the [uQuant] in vP and [uNeg] in NegP match with the interpretable features in NegQP and force the neg-whQ to move. The neg-whQ mou-bingo inherited both [neg] and [quant] features as a whole undergoes successive movements from its base generated object position. First, it raises to [spec, vP] and values [uQuant] and EPP features. Then it undergoes successive cyclic movement to [spec, NegP] to further value [uNeg] and EPP features. At this point, all features are valued for the probes and they become inactive and drive no further syntactic operation. This structure can account for both OSV (the object topicalization) and the unique SOV word orders, depending on whether or not the subject raising from V-to-T takes place.

Overt Quantifier Raising of Neg-whQ in Cantonese

41

Note that the negator is base generated in the Neg0 head in negated context. Therefore this structure can also represent neg-whQ constructions in negated contexts as in (24). (23)

Last Resort Move F raises F to target K only if F enters into a checking relation with a sublabel of K.

(24)

Ngo mou-bingoi m zungji ti (wo16). I NO-WHO NOT LIKE SP a. ‘I don’t like nobody.’ / ‘I don’t dislike anybody.’ (Lit. ‘There is nobody, who I don’t like.’) b. ‘I don’t like only a few people.’

3.4 Neg-Raising The NegQP gives rise to another possible overt movement, which is the neg-raising, and explains how wh-phrases in Cantonese can be licensed as indefinites (NPI) by the preceding negation. A definite negative interpretation is obtained from this movement, where only one interpretation is preserved as in (25). (25)

Ngo moui zungji [NegQP ti bingo]. I NO LIKE WHO ‘I don’t like anybody.’

Before any movement, the probes v and Neg0 are active carrying [uQuant] and [uNeg] respectively and the EPP features. The probe v Agrees with the invisible operator Ø in the goal NegQP and the probe Neg0 Agrees with the negative morpheme mou in it, therefore movements are driven to eliminate all uninterpretable features. The probes search down their c-command domain and attract the closest features under Agree. First, the [uQuant] and EPP features are valued with the unpronounced quantifier operator Ø raising out of NegQP to lands at [spec, vP], then the [uNeg] and EPP features are valued with the negative morpheme mou raising out of NegQP to lands at [spec, NegP]. All uninterpretable features are valued and therefore the wh-phrase bingo needs not to raise and can be licensed as NPI in-situ by the c-commanding mou in [spec, NegP]. These operations in respective sequence obey both Attract F17 and Minimal Link Condition18 constraints. The syntactic operations involved are represented as follows:

42

Chapter Two

(26)

(27)

Attract F: K attracts F if F is the closest feature that can enter into a checking relation with a sublabel of K.

(28)

Minimal Link Condition: K attracts Į only if there is no ȕ, ȕ closer to K than Į, such that K attracts ȕ.

The negative morpheme mou moves out of NegQP to [spec, NegP] position and the invisible quantifier operator Ø moves out of NegQP to [spec, vP] in (26), allowing the wh-phrase to stay in-situ within the NegQP. This creates such licenser-licensee relationship of mou and the wh-phrase bingo, and therefore the latter is licensed as NPI. In such constructions, the sentential negation interpretation is forced and only the non-existential interpretation remains.

Overt Quantifier Raising of Neg-whQ in Cantonese

43

4. It is Quantifier Raising (QR) Quantifier Raising is parameterized between being covert (e.g. English) and overt (e.g. Hungarian, French and Scandinavian languages).Tthis paper argues that the unique SOV structure observed with object neg-whQ in Cantonese is a kind of overt QR as suggested in different literatures19. In other languages observing overt QR, the negative or quantified objects land at a position preceding the vP domain. Such overt QR can be optional with quantified objects in Icelandic to obtain the sentential negation interpretation as in (29) 20 and some cases (e.g. tout ‘all’ and beaucoup ‘many’) in French, and obligatory in Hungarian as in (30)21 and some cases (e.g rien ‘nothing’) as in (31)22 in French. (29)

Ég hef [NegP [engine stig ]i [vP fengiõ ti]] I HAVE NO POINTS RECEIVED a. Zero-quantification: ‘I scored zero points.’ b. Sentential negation: ‘I haven’t got any point yet/I haven’t been judged yet.’

(30)

János [minden diákot]i [VP szeretne [ha meghívná ei ]]. JOHN EVERY STUDENT WOULD:LIKE IF INVITED:WE ‘John would like if we invited every student.’

(31) a. *Pierre n’a [vP mangé rien] b. Pierre n’a [rieni [vP mangé ti]] PIERRE NEG-HAS NOTHING EATEN ‘Pierre didn’t eat anything.’ With reference to Chomsky and Huang, a syntactic movement must be involved if subjacency is observed in relating an element with a trace embedded in an island23. I hereby provide data where ungrammaticality is observed relating the raised neg-whQ with its trace to support my claim that the unique SOV structure is a consequence of syntactic movements but not simply dislocation. The following examples (32)-(33) illustrate how the optional long-distance movements are banned by the Subjacency Condition. (32)

Ngo zungji [IP nei mou-matjiei m sik [NP ti]]. I LIKE YOU NO-WHAT NOT EAT a. (Lit. ‘I like that there is nothing that you don’t eat.’) b. (Lit. ‘I like that there is only a few things that you don’t eat.’)

44

(33)

Chapter Two

*Ngo zunji mou-matjiei [IP nei ti m sik [NP ti]].

It is unproblematic for the neg-whQ to stay in the subordinate clause as in (32). However, moving it out of the subordinate clause renders ungrammaticality, as in (33). According to Chomsky’s Subjacency Condition, an element cannot move to a position where there are two bounding nodes (including DP and IP) in between the target landing site and its base position24. There is a subjacency effect in (33) because the neg-whQ is crossing the two bounding nodes (NP and IP). If neg-whQ is base generated in the preverbal position, it would just be crossing one bounding node and the ungrammaticality cannot be explained. There are other cases as in (34) showing that optional long distance movements to a further higher position are not allowed where the embedded clause is overtly marked finite and a strong islandhood25 is observed. (34)

*Ngo mou-bingoi gotdak [IP ti nei ti gin-guo ti]. I NO-WHO THINK YOU MEET-ASP ‘Nobodyi, I think you met himi.’

(35)

Ngo mou-bingoi gotdak [IP ti nei ti gin-guo kuii]. I NO-WHO THINK YOU MEET-ASP HIM

This is in accordance with May that quantifier raising is clause bound26 and explains the ungrammaticality of (34) due to the QR constraints. It is not certain whether such further raising is QR or simply left dislocation after the obligatory QR to the preverbal position in the embedded clause. Even so, this can be explained by the resumptive pronoun rendering grammaticality in (35). If such further raising is simply left dislocation, the moved NP should be able to relate back to the resumptive pronoun in its base generated position. The ungrammaticality in (36) cannot be explained if the raised neg-whQ is base-generated in the preverbal position, because grammaticality should be expected by replacing its trace by a resumptive pronoun. (36)

*Ngo mou-bingoi gotdak [IP ti nei kuii gin-guo].

5. Accounting for the Data In this section, I will include data on NPI licensing and WCO cancellation in dative and infinitival constructions.

Overt Quantifier Raising of Neg-whQ in Cantonese

45

5.1. Licensing NPI in Dative Constructions Several scholars27 discuss the licensing of wh-phrases as polarity items in Mandarin Chinese, and that Cantonese wh-phrases are licensed by negations as NPI, as in the following (as discussed above). (37)

Koei mou fatjin matje HE NO DISCOVER WHAT ‘He doesn’t discover anything.’

The proposed overt QR of the NegQP licenses another wh-phrase as NPI in dative constructions, where the neg-whQ is the indirect object undergone raising to preverbal position and the licensed wh-phrase as NPI is the direct object. Grammaticality is preserved after the overt and obligatory QR as in (39). (38)

(39)

*Ngo gaaisiu bingo bei [mou-bingo]. I INTRODUCE WHO TO NO-WHO ?’Who do I introduce to nobody?’ Ngo [mou-bingo] gaaisiu bingo bei ti I NO-WHO INTRODUCE WHO TO a. ‘Nobody, I introduce anyone to.’ b. ‘I do not introduce anyone to anyone.’

The overt QR not only saves (38) syntactically since mou-bingo cannot stay in-situ, but also semantically as (39) gives the possible reading where the direct object bingo is licensed as NPI. The structure in (40) explains how the licensor-licensee relationship is preserved by overt QR. The neg-whQ, which originally stays inside the PP, does not c-command the wh-phrase as a direct obect, and (38) can only be interpreted as an interrogative interpretation that makes no sense. In the representation (39), the raised NegQP mou-bingo not only satisfies the [uQuant], [uNeg] and EPP features and provides grammaticality, but also creates the c-commanding relationship with the direct object wh-phrase bingo and licenses it as NPI “anything” giving rise to the possible interpretations in (39a, b).

46

Chapter Two

(40)

5.2. Cancelling WCO in Infinitival Constructions The proposed overt QR of NegQP also cancels the WCO effect, where a neg-whQ is the complement of the verb within the infinitival and the anaphor is within the direct object of the main verb. (41)

(42)

*Ngo daai [kuii aa ma] hui gin [mou-bingo]i I BRING HIS MOTHER TO MEET NO-WHO ‘I bring his*i/j mother to meet nobodyi.’ Ngo [mou-bingo]i daai [kuii aa ma] hui gin ti I NO-WHO BRING HIS MOTHER TO MEET a. ‘Nobodyi, I bring hisi/j mother to meet.’ b. ‘There is only a few peoplei, who I bring theiri mother to meet.’

Overt Quantifier Raising of Neg-whQ in Cantonese

47

The construction in (41) is ungrammatical with the neg-whQ in-situ. Also, it observes the WCO effect when mou-bingo is co-indexed with the anaphor kui. However, the overt QR provides grammaticality in (42) cancelling WCO where mou-bingo now successfully binds the anaphor kui as illustrated by the representation (43). (43)

The neg-whQ mou-bingo as the object in the infinitival undergoes overt QR, first raising to [spec, vP] to value [uQuant] and satisfy the EPP features, then raising successively to [spec, NegP] to satisfy both [uNeg] and EPP features. As long as it c-commands the anaphor kui within the DP as the object of the matrix verb, the WCO effect is cancelled, rendering (42) grammatical.

5.3. Licensing Wh-phrase as NPI in Infinitival Constructions Again, the proposed overt QR to neg-whQ licenses the object wh-phrase of the matrix verb as NPI and accounts for the declarative interpretation.

48

(44)

(45)

Chapter Two

*Ngo daai bingo hui gin [mou-bingo] I BRING WHO TO MEET NO-WHO ?’Who do I bring to meet nobody?’ Ngo [mou-bingo]i daai bingo hui gin ti I NO-WHO BRING WHO TO MEET a. ‘Who do I bring to meet only a few people?’ b. ‘Nobody, I bring anyone to meet.’

Example (44) is ungrammatical with the neg-whQ in-situ at the object position in the infinitival and allows only an interrogative interpretation where the wh-phrase is not licensed as NPI. The overt QR renders grammaticality in (45) and allows both the interrogative reading in (45a) and the interpretation where the direct object bingo is now licensed as NPI as in (45b). Moving the neg-whQ as a whole allows dual interpretations (nonexistential and existential presupposition) and therefore the interrogative interpretation of (45) is preserved. Or else, the interrogative reading does not make any sense (?’Who do I bring to meet nobody?’) if not , the neg-whQ gives the property of dual interpretations. The structure for (44) is similar to the one in (40) and it is not repeated again. As I proposed, the structure could account for the case where only the negator mou raises to preverbal position in (46). (46)

Ngo moui daai bingo hui gin [ ti bingo] I NO BRING WHO TO MEET WHO a. ‘I don’t bring anyone to meet anyone’ b. ‘Who do I not bring anyone to meet?’ c. ‘Who do I not bring to meet anyone?’

The representation (47) explains the ambiguity of (46) due to the overt raising of mou from NegQP. With only the negator mou raising out of the NegQP, it satisfies [uNeg] and the EPP feature at [spec, NegP]. The quantifier operator Ø raises to [spec, vP] and satisfies [uQuant] and the EPP feature. The wh-phrase bingo needs not to raise and remains in NegQP and is licensed as NPI. In representation (46), the negator mou c-commands both wh-phrases bingo in VP1 and in NegQP. This gives rise to the declarative interpretation that both bingo phrases are licensed as NPIs, and the interrogative interpretations that either bingo is licensed as NPI while the other remains to be interrogative.

Overt Quantifier Raising of Neg-whQ in Cantonese

49

(47)

5.4. Summary Under the proposed structure, a neg-whQ undergoing obligatory and overt QR licenses NPI in dative and infinitival constructions, where the neg-whQ mou-bingo being the indirect object is c-commanding the direct object bingo (another wh-phrase) in the former construction; and the object neg-whQ in the infinitival raises and licenses the other wh-phrase as NPI in the later construction. In addition, it also cancels WCO in infinitival construction, where the object neg-whQ in the infinitival raises and binds the object bingo (another wh-phrase) in the matrix clause after the overt QR. This section explains how both the raising options account for grammaticality and give rise to possible interpretations (given that the raised neg-whQ gives both non-existential and existential presuppositional interpretations).

50

Chapter Two

6. Conclusion This paper explains how the structure NegQP can account for the dual interpretations (the non-existential and the existential presuppositional interpretations) regarding neg-whQ constructions in Cantonese at the syntax-semantic interface. NegQP gives the correct structure for neg-whQs where the negative morpheme mou precedes the wh-phrase forming negative quantifiers, and explains the overt movement that is driven by the uninterpretable features resulting in the unique SOV word order with an object neg-whQ in Cantonese. In addition, it also accounts for the two possible movements related, namely the neg-whQ raising and the neg-raising. Following the recent minimalist approach, movements (QR) are driven by [uQuant], [uNeg] and EPP features in order to receive Full Interpretation. The NegQP inherits the [Neg] feature from mou in its specifier position and [Quant] feature from the invisible quantifier operator Ø in its head position, and trigger overt quantifier raising under Agree. I also attempted to explain the structure of all quantifiers and wh-phrases as indefinites under the QuantP. I conclude that neg-whQ is a kind of strong quantifier in Cantonese and is required to undergo overt QR to render grammaticality as well as the possible interpretations.

Notes 1

Negative Polarity Items (NPI) requires a negation-licensing context and therefore has to be c-commanded by negation. 2 Cheung, Lawrence Y. L. ‘Obligatory XP-raising in Chinese*’, Toronto Working Papers in Linguistics, 28 (2008), 15-28. 3 Cheng, Lisa Lai-Shen. ‘On Dou-quantification. Journal of East Asian Linguistics 4 (1995), 197-234; Huang, C.-T. James. Logical Relations in Chinese and the Theory of Grammar. PhD thesis, MIT, (1982); Li, Yen-Hui Audrey. ‘Indefinite Wh in Mandarin Chinese.’, Journal of East Asian Linguistics, 1.2. Li, C. and S. A. Thompson, 1992; Cheng, Lisa Lai-Shen. ‘Wh-words as Polarity Items.’, Chinese Languages and Linguistics II, Symposium Series of Institute of History and Philology, Academia Sinica, Taiwan, (1994), 615-640; Lin, Jo-Wang. Polarity Licensing and Wh-Phrase Quantification in Chinese, PhD Dissertation, University of Massachusetts at Amherst, 1996; Yuan, Boping. ‘Behaviours of wh-words in English speakers’ L2 Chinese wh-questions: evidence of no variability, temporary variability and per- sistent variability in L2 grammars.’, Bilingualism: Language and Cognition 10, (2007b), 277–98; Yuan, Boping. ‘Non-permanent representational deficit and apparent target-likeness in second language: evidence from wh-words used as universal quantifiers in English and Japanese speakers’ L2 Chinese.’, In Snape, N., Leung, Y.-K.L. and Sharwood-Smith, M., editors, Representational Deficits in SLA: In honour of Roger Hawkins. Amsterdam: John Benjamins, (2009),

Overt Quantifier Raising of Neg-whQ in Cantonese

51

69–103; Yuan, Boping.‘Domain-wide or variable-dependent vulnerability of the semantics—syntax interface in L2 acquisition? Evidence from wh-words used as existential polarity words in L2 Chinese grammars.’, Second Language Research 26, (2010), 219-260. 4 Cheng, Lisa Lai-Shen. ‘On Dou-quantification.’, Journal of East Asian Linguistics, 4 (1995), 197-234. 5 Lee, Man-ki Theodora. ‘Overt quantifier raising of Negative-wh-quantifiers in Cantonese.’ In Chris Cummins, Chi-Hé Elder, Thomas Godard, Morgan Macleod, Elaine Schmidt & George Walkden (eds.). 2011. Proceedings of the Sixth Cambridge Postgraduate Conference in Language Research. Cambridge: Cambridge Institute of Language Research, (2011), 92-107. 6 Kratzer, Angelika. ‘Stage-level and individual-level predicates.’, In G. Carlson, & F. J. Pelletier (Eds.), The generic book. Chicago: University of Chicago Press, 1995, 125-175; Potts, Christopher. ‘When even no’s neg is splitsville’. URL: http://ling.ucsc.edu/Jorge/potts.html, 2000; Penka, Doris and von Stechow, Arnim. ‘Negativen indefinita unter modalverben.’, In R. Muller, & M. Reis (Eds.), Modalitat und modalverben im deutschen, Sonderheft 9 edition, 2001. 7 Diesing, Molly. Indefinites. Cambridge, MA: MIT Press, 1992. 8 Chomsky, Noam. The minimalist program, Cambridge, Mass.: MIT Press, 1995b; Chomsky, Noam. ‘Minimalist Inquiries: The Framework.’, In Roger Martin, David Michaels and Juan Uriagereka (eds), Step by Step: Essays on Minimalis Syntax in Honour of Howard Lasnik. Cambridge, Massachussetts: MIT Press, 2000; Chomsky, Noam. ‘Derivation by Phrase.’, In Kenstowicz, Michael (ed.), Ken Hale. A Lifein Language, Cambridge, Mass.: MIT Press, 2001, 1-52. 9 Chomsky, Noam. Barriers. Cambridge, Mass.: MIT Press, 1986. 10 Beghelli, Fillippo, and Tim Stowell. ‘Distributivity and negation.’, In A. Szabolcsi (ed.), Ways of Scope taking, 1996, 71-107. 11 Chomsky, Noam. ‘Derivation by Phrase.’ In Kenstowicz, Michael (ed.), Ken Hale. A Lifein Language, Cambridge, Mass.: MIT Press, 2001, 1-52. 12 Kiss, Katalin É. ‘NP Movement, Operator Movement, and Scrambling in Hungarian.’ In K. Kiss, ed., Discourse Configurational Languages, Oxford University Press, 1995, 207-243. 13 Since the colloquial constructions in Cantonese are discussed in this paper, many examples appear to be more natural with the final particle at the end of the sentence. 14 Chomsky, Noam. The minimalist program. Cambridge, Mass.: MIT Press, 1995b; Chomsky, Noam. ‘Minimalist Inquiries: The Framework.’ In Roger Martin, David Michaels and Juan Uriagereka (eds), Step by Step: Essays on Minimalis Syntax in Honour of Howard Lasnik. Cambridge, Massachussetts: MIT Press, 2000; Chomsky, Noam. ‘Derivation by Phrase.’ In Kenstowicz, Michael (ed.), Ken Hale. A Lifein Language, Cambridge, Mass.: MIT Press, 2001, 1-52. 15 Chomsky, Noam. The minimalist program, Cambridge, Mass.: MIT Press, 1995b, 280. 16 Since the colloquial constructions in Cantonese are discussed in this paper, many examples appear to be more natural with the final particle at the end of the sentence. 17 Chomsky, Noam. The minimalist program, Cambridge, Mass.: MIT Press, 1995b, 297.

52 18

Chapter Two

Chomsky, Noam. The minimalist program, Cambridge, Mass.: MIT Press, 1995b, 311. 19 Christensen, Ken Ramshøj. 2003. NEG-shift in the Scandinavian Languages and English. Paper presented at Grammatik I Fokus [Grammar in Focus], Lung University, (February 6. 2003); Christensen, Ken Ramshøj. Quantifier Movement and Derivation by Phase-Now You See It, Now You Don’t, in Workshop on Comparative and Theoretical Linguistics: “When and why do constituents move?”, Dept. of English, University of Aarhus, 2004; Confais, Jean Paul. Grammaire Explicative. Schwerpunkte der französischen Grammatif für Leistungskurs und Studium, München: Max Hueber Verlag, 1978; Haegeman, Liliane. The Syntax of Negation, Cambridge University Press, 1995; Rizzi, Luigi. Relativized Minimality, Cambridge, Mass.: MIT Press, 1990; Kiss, Katalin É.‘NP Movement, Operator Movement, and Scrambling in Hungarian.’ In K. Kiss, ed., Discourse Configurational Languages, Oxford University Press, 1995, 207-243; Rögnvaldsson, Eiríkur. ‘OV Word Order in Icelandic’ in Allan, R. D. S. & M. P. Barnes (eds.), Proceedings of the Seventh Biennial Coference of Teachers of Scandinavian Studies in Great Britain and Northern Ireland, London: University College, 1987, 33-49. 20 Christensen, Ken Ramshøj. Quantifier Movement and Derivation by Phase-Now You See It, Now You Don’t, in Workshop on Comparative and Theoretical Linguistics: “When and why do constituents move?”, Dept. of English, University of Aarhus, 2004, 6 (17b). 21 Kiss, Katalin É.‘NP Movement, Operator Movement, and Scrambling in Hungarian.’ In K. Kiss, ed., Discourse Configurational Languages, Oxford University Press, 1995, 226. 22 Nølke, Henning. Fransk Grammatik og Sprogproduktion, København: Kaleidoscope, 1997, 234. 23 Chomsky, Noam. ‘On WH Movement.’, In Akmajian et al (eds.), Formal Syntax. Academic Press, New York, 1977; Huang, C.-T. James. Logical Relations in Chinese and the Theory of Grammar. PhD thesis, MIT, 1982. 24 Chomsky, Noam. ‘On WH Movement.’, In Akmajian et al (eds.), Formal Syntax. Academic Press, New York, 1977. 25 Chomsky, Noam. Barriers. Cambridge, Mass.: MIT Press, 1986; Ross, John R. Constraints on variables in syntax. Doctoral dissertation, MIT, 1967. 26 May, Robert. Logical Form: Its Structure and Derivation. Cambridge, Mass.: MIT Press, 1985. 27 Cheng, Lisa Lai-Shen. ‘Wh-words as Polarity Items.’ Chinese Languages and Linguistics II, Symposium Series of Institute of History and Philology, Academia Sinica, Taiwan, 1994, 615-640; Cheng, Lisa Lai-Shen. ‘On Dou-quantification. Journal of East Asian Linguistics 4 (1995), 197-234; Cheng, Lisa Lai-Shen and Rooryck, John. ‘Licensing Wh-in-situ.’ SYNTAX 3.1, 2000, 1-19; Li, Yen-Hui Audrey. ‘Indefinite Wh in Mandarin Chinese.’ Journal of East Asian Linguistics. 1.2. Li, C. and S. A. Thompson, (1992); Lin, Jo-Wang. Polarity Licensing and Wh-Phrase Quantification in Chinese, PhD Dissertation, University of Massachusetts at Amherst, 1996.

Overt Quantifier Raising of Neg-whQ in Cantonese

53

References Aoun, J., Choueiri, L. and N. Hornstein. 2001. Resumption, movement, and derivational economy. Linguistic Inquiry 32, 371-403. Cheng, L. 1994. Wh-words as Polarity Items. Chinese Languages and Linguistics II, Symposium Series of Institute of History and Philology, Academia Sinica, Taiwan, 615-640. —. 1995. On Dou-quantification. Journal of East Asian Linguistics 4, 197-234. Cheng, L. & J. Rooryck 2000. Licensing Wh-in-situ. Syntax 3, 1-19. http://www.lisacheng.nl/pdf/Wh-Polarity.pdf Cheung, L.Y. L. 2008. Obligatory XP-raising in Chinese*, Toronto Working Papers in Linguistics, 28,15-28. Christensen, K. R. 2003. NEG-shift in the Scandinavian Languages and English, paper presented at Grammatik I Fokus [Grammar in Focus], Lung University, (February 6. 2003). —. 2004. Quantifier Movement and Derivation by Phase-Now You See It, Now You Don’t, in Workshop on Comparative and Theoretical Linguistics: “When and why do constituents move?”, Dept. of English, University of Aarhus. Chomsky, N. 1977. On WH Movement. In P.Culicover, T.Wasow and A. Akmajian (eds), Formal syntax. New York: Academic Press, —. 1986. Barriers. Cambridge, Mass.: MIT Press. —. 1995b. The minimalist program. Cambridge, Mass.: MIT Press. —. 2000. Minimalist Inquiries: The Framework. In R. Martin, D. Michaels and J. Uriagereka (eds), Step by Step: Essays on Minimalis Syntax in Honour of Howard Lasnik. Cambridge, Mass: MIT Press. —. 2001. Derivation by Phrase. In M. Kenstowicz (ed.), Ken Hale: A Life in Language, pp. 1-52.Cambridge, Mass.: MIT Press. Confais, J. P. 1978. Grammaire Explicative. Schwerpunkte der französischen Grammatif für Leistungskurs und Studium. München: Max Hueber Verlag. Diesing, M. 1992. Indefinites. Cambridge, MA: MIT Press. Haegeman, L. 1995. The Syntax of Negation. Cambridge: Cambridge University Press. Huang, C.-T. J. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD dissertation, MIT. Kayne, R. S. 1998. Overt vs. Covert Movement. Syntax 1, 128-191. Kiss, K. É. 1995. NP Movement, Operator Movement, and Scrambling in Hungarian. In K. Kiss (ed), Discourse Configurational Languages, pp. 207-243. Oxford: Oxford University Press.

54

Chapter Two

Kratzer, A.. 1995. Stage-level and individual-level predicates. In G. Carlson, & F. J. Pelletier (eds), The generic book, pp.125-175. Chicago: University of Chicago Press. Lee, M. T. 2011. Overt quantifier raising of Negative -wh-quantifiers in Cantonese. In C. Cummins, C.H. Elder, T. Godard, M. Macleod, E.Schmidt & G. Walkden (eds), Proceedings of the Sixth Cambridge Postgraduate Conference in Language Research. Cambridge: Cambridge Institute of Language Research. Li, Y-H A. 1992. Indefinite Wh in Mandarin Chinese. Journal of East Asian Linguistics 1, 125-155. Lin, J-W.1996. Polarity Licensing and Wh-Phrase Quantification in Chinese. PhD dissertation, University of Massachusetts at Amherst. May, R. 1985. Logical Form: Its Structure and Derivation. Cambridge, Mass.: MIT Press. Mahajan, A. 1990. The A-/A-bar distinction and movement theory. PhD dissertation, MITWPL. Nølke, H. 1997. Fransk Grammatik og Sprogproduktion, København: Kaleidoscope. Rizzi, L.1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Ross, J. R. 1967. Constraints on variables in syntax. PhD dissertation, MIT. Rögnvaldsson, E. 1987. OV Word Order in Icelandic. In R. D. S. Allan & M. P. Barnes (eds), Proceedings of the Seventh Biennial Coference of Teachers of Scandinavian Studies in Great Britain and Northern Ireland, pp. 33-49. London: University College,. Penka, D. & A. von Stechow. 2001. Negativen indefinita unter modalverben. In R. Muller, & M. Reis (eds), Modalitat und modalverben im deutschen, Sonderheft 9 edition. Potts, C. 2000. When even no’s neg is splitsville. URL: http://ling.ucsc.edu/Jorge/potts.html [accessed 2000]. Saito, M. 1992. Long Distance Scrambling in Japanese*. Journal of East Asian Languages 1, 69-118. Soh, H. L. 2005. Wh-in-situ in Mandarin Chinese. Linguistic Inquiry 36, 143-155. Svenonius, P. 2000b. Quantifier movement in Icelandic. In P. Svenonius, (ed.) The Derivation of VO and OV, pp. 255-292. Amsterdam: John Benjamins.

PART III: SYNTAX-MORPHOLOGY

CHAPTER THREE TWO TYPES OF BOUND ITEMS AND THEIR INTERACTION WITH MORPHOLOGY AND SYNTAX MASAHARU SHIMADA AND AKIKO NAGANO

Abstract There are languages that have both bound and free types of anaphors. Kishida (2009) claims that bound and free anaphors in Japanese are necessarily pure and near reflexives, respectively, in the sense of Lidz (2001). Kishida’s suggestion is that the morphological boundness of anaphoric expressions reflects their syntactic and semantic characteristics. However, close examination of Japanese anaphors reveals that boundness is not a crucial factor for their syntax and semantics, leading us to support a lexeme-based analysis of the lexicon of human language (Aronoff 1994; Beard 1995, among others). The analysis derived from the lexeme-based morphology gives us important insights into the interactions between lexicon, morphology and other components of grammar. Specifically, it classifies bound forms into two types, one of which functions as a lexeme, not as an affix.

1. Introduction Morphemes are defined as minimal linguistic units bearing meanings and are generally classified into two types, free morphemes and bound morphemes. Lexical words are realised as free morphemes, as in man, love and tall, and affixes are realised as bound morphemes, as in -s in hats, un- in unhappy and -ise in formalise. Kishida (2009) offers an explanation of anaphoric expressions that is quite consistent with this traditional morpheme-based perspective. She explores the idea that morphological boundness and semantic properties have a one-to-one correspondence, claiming that free-morpheme anaphors and bound-morpheme anaphors

58

Chapter Three

behave differently in the way they refer to antecedents. Objecting to Kishida’s view, Shimada and Nagano (2011) provide empirical evidence suggesting that morphological boundness and semantic properties are not related. They argue that morphological boundness is not a distinctive feature of affixes and that contentful lexical elements as well as affixes take bound forms. This situation might be explained from a morpheme-based perspective in such a way that there are two types of bound morphemes. Shimada and Nagano (2011), however, show that the lexicon organised this way is successfully captured in an analysis based on a lexeme-based theory of morphology, developed by Matthews (1991), Anderson (1992), Aronoff (1994), Beard (1995) and Fradin (2003), among others, but not a morpheme-based theory of morphology. The aim of this paper is to examine the consequences of Shimada and Nagano’s (2011) analysis, focusing on the classification of bound forms and the possibility of eliminating the notion of a morpheme to further support a lexeme-based approach. This paper is organised as follows. In section two, we summarise Kishida’s analysis of Japanese anaphors and introduce Kishida’s generalisation of the relationship between morphological boundness and anaphoric functions. In section three, empirical evidence is provided against Kishida’s generalisation. The behaviour of anaphors is not predicted according to morphological boundness. In section 4, we develop a lexeme-based analysis to explain the data observed. In section 5, we discuss some consequences. In particular, we address derivational morphology, showing that bound forms should be divided into two types. In section 6, we summarise the discussions.

2. Close Relationship between Morphological Boundness and Semantics The interaction between morphological boundness and semantics has been examined in the literature. This section is devoted to a brief review of previous studies arguing for their close relationship in terms of anaphoric binding We are mainly concerned with Kishida’s (2009) observation on anaphoric binding in Japanese. In section 2.1, we give data discussed in Kishida (2009), and in section 2.2, we introduce the generalisation that she offers based on her observation.

2.1. Affixal Anaphors in Japanese The Japanese anaphors zibun ‘self’ and zibunzisin ‘self,’ given in (1),

Two Types of Bound Items and their Interaction

59

have been widely examined in research on generative linguistics. As shown below, they are morphologically free nouns and act as hosts of Case particles:1 (1) a.

John-ga zibun-o hihan-si-ta. John-NOM self-ACC criticism-do-PAST ‘John1 criticised himself1.’ b. John-ga zibunzishin-o hihan-si-ta. John-NOM self-ACC criticism-do-PAST ‘John1 criticised himself1.’

In addition to these morphologically free anaphors, Japanese has morphologically bound anaphors, often called affixal anaphors or bound-morpheme anaphors, as discussed in a limited body of literature that includes Takezawa (1991), Aikawa (1993), Tsujimura and Aikawa (1999), Kishida (2009), and Shimada and Nagano (2011). Zi- and ziko-, which are written in Chinese characters as ‘⮬’ and ‘⮬ᕫ,’ respectively, are instances of Japanese bound anaphors. Japanese does not have only native but also Sino-Japanese morphemes. Bound-morpheme anaphors combine with Sino-Japanese morphemes, often forming Sino-Japanese verbal nouns. For example, zi- attaches to a Sino-Japanese bound morpheme such as -satsu (ẅ) ‘killing’, forming a verbal noun zisatsu (⮬ ẅ ) ‘suicide’. The verbal noun can be further combined with a light verb -suru ‘do’ to form a Sino-Japanese verb zisatsu-suru ‘kill oneself’, which is often called a zi-verb. Ziko- also attaches to Sino-Japanese free morphemes such as hihan (ᢈุ) ‘criticism’ and form a verbal noun zikohihan ‘self-criticism’. Again, it becomes a verb called a ziko-verb with the attachment of -suru, as in zikohihan-suru ‘criticise oneself.’ Typical examples of sentences containing zi-verbs or ziko-verbs are illustrated in (2): (2) a.

John-ga zi-satu-si-ta. John-NOM self-killing-do-PAST ‘John1 killed himself1.’ b. John-ga ziko-hihan-si-ta. John-NOM self-criticism-do-PAST ‘John1 killed himself1.’

Although zisatsu-suru and zikohihan-suru cannot take a syntactic object, some zi-verbs and ziko-verbs can take an object, as illustrated below:

Chapter Three

60

(3) a.

John-ga musuko-o zi-man-si-ta. John-NOM son-ACC self-boast-do-PAST ‘John1 boasted about his1 son.’ b. John-ga taizyuu-o ziko-kanri-si-ta. John-NOM weight-ACC self-control-do-PAST ‘John1 controlled his1 weight.’

Additional examples of zi-verbs and ziko-verbs are given in (4). The zi-verbs and ziko-verbs in (4a) are of the objectless type and those in (4b) are of the object-taking type: (4) a.

zi-satu-suru ‘kill oneself,’ zi-ai-suru ‘take care of oneself,’ zi-ritu-suru ‘establish oneself,’zi-ten-suru ‘revolve,’ zi-kai-suru ‘collapse,’ zi-baku-suru ‘explode,’ zi-syu-suru ‘turn oneself in,’ziko-syookai-suru ‘introduce oneself,’ ziko-hihan-suru ‘criticise oneself’ b. zi-man-suru ‘boast about oneself,’ zi-kyoo-suru ‘confess oneself guilty,’ziko-kanri-suru‘self-administer,’ ziko-sinkoku-suru ‘report by oneself’

As shown in (2) and (3), zi- and ziko- semantically refer to local subjects. Thus, it can be stated that these morphologically bound elements function as anaphoric expressions.

2.2. Kishida’s Generalisation In Kishida (2009), zi- and ziko- are grouped together as affixal anaphors. Kishida claims, however, that although they are anaphoric expressions, they differ from free-morpheme anaphors such as zibun in reflexivity. According to Kishida, a so-called statue reading is not acceptable for affixal anaphors. Specifically, this claim means that the referent of ziko-, for example, is necessarily identical to its antecedent, but zibun can refer to something related to its antecedent. The statue reading was originally discussed by Jackendoff (1992) with the following example: (5)

Ringo started undressing himself. (himself = Ringo, a statue of Ringo)

In addition to the situation in which Ringo is removing the clothes that he is wearing, (5) can describe the situation in which Ringo is removing the

Two Types of Bound Items and their Interaction

61

clothes that the statue depicting him is wearing. This additional interpretation is called a statue interpretation, and the English anaphor himself is taken as an anaphoric expression that allows it. Based on the observation that zibun can refer to a statue of John in (1a), but ziko- in (2b) cannot, Kishida claims that ziko- and zibun are different in that the former resists the statue interpretation. Kishida (2009) refers to the classification of anaphors given in Lidz (2001). Lidz classifies anaphors into two types: pure reflexives and near reflexives. Pure reflexives must be completely identical to their antecedents, whereas near reflexives can refer to something related or similar, but not necessarily identical, to their antecedents. In fact, a statue reading is only possible for near reflexives. Kishida (2009) proposes that zi- and ziko-, requiring a strict semantic identity with their antecedents, are pure reflexives and that zibun, allowing for a statue reading, is a near reflexive. Kishida (2009) then draws the generalisation that affixal anaphors function as pure reflexives and that non-affixal anaphors function as near reflexives, as follows: (6) a. bound-morpheme anaphor = pure reflexive anaphor b. free-morpheme anaphor = near reflexive anaphor Kishida (2009) observes further differences between affixal anaphors and non-affixal anaphors. First, zibun allows for non-local binding as well as local binding, but ziko- allows for the latter only: (7) a. Mary2-wa [John1-ga ziko1/*2-hihan-si-ta] to omot-ta. Mary-TOP John-NOM self-criticism-do-PAST C think-PAST ‘Mary2 thought that John1 criticised {himself1/*her2}.’ b. Mary2-wa [John1-ga zibun1/2-o hihan-si-ta] . Mary-TOP John-NOM self-ACC criticism-do-PAST to omot-ta C think-PAST ‘Mary2 thought that John1 criticised {himself1/her2}.’ Thus, (7) shows that the non-affixal anaphor zibun can refer to both the local subject John and the matrix subject Mary, but the affixal anaphor zikocan only refer to the local subject. Second, ziko- and zibun differ in the possibility of the so-called non-sloppy reading in comparative deletion constructions. Kishida makes the following observation:

62

Chapter Three

(8) a. Mary-ga John yorimo hageshiku ziko-hihan-si-ta. Mary-NOM John than severely self-criticism-do-PAST ‘Mary criticised herself more severely than John criticised himself.’(sloppy identity reading) *‘Mary criticised herself more severely than John criticised her.’ (*non-sloppy identity reading) b. Mary-ga John yorimo hageshiku zibun-o hihan-si-ta. Mary-NOM John than severely self-ACC criticism-do-PAST ‘Mary criticised herself more severely than John criticised himself.’(sloppy identity reading) ‘Mary criticised herself more severely than John criticised her.’ (non-sloppy identity reading) Thus, (8) shows that non-sloppy identity reading is prohibited with the affixal anaphor ziko- in comparative deletion constructions, whereas it is allowed for the free anaphor zibun. Again, these examples show that the affixal anaphor ziko- requires a strict identity with its antecedent, whereas the non-affixal anaphor zibun does not. In summary, Kishida asserts that the semantic property of anaphoric expressions depends on their surface morphological forms and that boundness and reflexivity are related to each other. She proposes that the relationship between boundness and reflexivity can be generalised as stated in (6) and claims that this generalisation is valid cross-linguistically. It should be noted that Kishida’s generalisation implies that morphological boundness and semantic properties are related. If Kishida’s view is correct, it can be concluded that Universal Grammar (UG) is constructed in such a way that a systematic correspondence is guaranteed between a morphological form and semantics.

3. Problems with Kishida’s Generalisation The generalisation in (6) and its implication that morphological boundness determines semantic properties are simple and attractive. However, as shown in Shimada and Nagano (2011), a close examination of Japanese anaphors immediately reveals that the generalisation in (6) is not valid. In fact, some non-affixal anaphors behave as pure reflexives rather than near reflexives, contrary to the generalisation in (6). In section 3.1, we briefly review Shimada and Nagano’s observations and give some empirical data suggesting that the generalisation in (6) does not hold. In section 3.2 and section 3.3, we provide further arguments against a direct connection between morphological boundness and semantics.

Two Types of Bound Items and their Interaction

63

3.1. Non-affixal Anaphors as Pure Reflexives Kishida attributes the pure reflexive property of zi- and ziko- to their morphological property of boundness. However, Shimada and Nagano (2011) argue against her idea, observing that some non-affixal anaphors function as pure reflexives, contrary to the generalisation in (6). Shimada and Nagano consider Japanese anaphors other than zibun, especially mizukara (⮬ࡽ) and onore (ᕫ). As the following sentence shows, they occur as free morphemes and allow for local binding, suggesting that they belong to non-affixal or free-morpheme anaphors: (9)

John-ga mizukara/onore-o hihan-si-ta. John-NOM self-ACC   criticism-do-PAST ‘John1 criticised himself1.’

Note also that even ziko, which Kishida considers to be exclusively affixal, can be non-affixal; it can occur on its own and be used as a free morpheme. Consider the following example, in which the accusative Case particle -o is attached to ziko: (10)

John-ga ziko-o hihan-si-ta. John-NOM self-ACC criticism-do-PAST ‘John1 criticised himself1.’

As the object of hihansita, ziko appears as a free determiner phrase (DP) in (10), suggesting that it can be a non-affixal or free morpheme anaphor. What is most important here is that these free anaphors behave as pure reflexives rather than near reflexives, as Shimada and Nagano (2011) observe. First, consider the following examples: (11) a. John-ga zibun-o arat/sibat/nade-ta. John-NOM self-ACC wash/bind/stroke-PAST ‘John1 washed/bound/stroked himself1.’ (zibun = John, a statue of John) b. John-ga mizukara/onore/ziko-o arat/sibat/nade-ta. John-NOM self-ACC wash/bind/stroke-PAST ‘John1 washed/bound/stroked himself1.’ (mizukara, onore, ziko= John, *a statue of John) Zibun allows for a statue interpretation, as noted earlier. (11a) can describe the situation in which John washed a statue depicting himself. In

64

Chapter Three

contrast, (11b) (where mizukara, onore, or ziko occurs instead of zibun) resists this interpretation. Although the anaphors mizukara, onore, and ziko are morphologically non-affixal like zibun, they cannot refer to a statue of John, requiring a completely identical interpretation in (11b). In this sense, mizukara, onore, and ziko should be classified as pure reflexives. The generalisation in (6) does not correctly capture their behaviour. Next, we move on to the possibility of non-local binding: (12)

Mary2-wa [John1-ga mizukara/onore/ziko1/*2-o hihan-si-ta] Mary-TOP John-NOM self-ACC criticism-do-PAST to omot-ta. C think-PAST ‘Mary2 thought that John1 criticised {himself1/*her2}.’

(12) shows that mizukara, onore, and ziko are necessarily locally bound. Therefore, the non-affixal anaphors mizukara, onore, and ziko, like zi-/ziko-, are pure reflexives, contrary to the generalisation in (6). Finally, consider how mizukara, onore, and ziko behave in a comparative deletion construction. If (6) is a correct generalisation, then they should allow not only a sloppy identity interpretation but also a non-sloppy identity interpretation in the following example. Unfortunately, this is not the case:

(13)

Mary-ga John yorimo hageshiku mizukara/onore/ziko-o Mary-NOM John than severely self-ACC hihan-si-ta. criticism-do-PAST ‘Mary criticised herself more severely than John criticised himself.’(sloppy identity reading) ‘Mary criticised herself more severely than John criticised her.’ (*non-sloppy identity reading)

As observed in (13), mizukara, onore, and ziko do not allow for a non-sloppy interpretation. This diagnosis further indicates their pure reflexivity. The classification of Japanese anaphors is thus summarised as follows.

Two Types of Bound Items and their Interaction

65

Table 3-1: Classification of Japanese anaphors

Affixal anaphors Non-affixal anaphors

Pure reflexives zi-, zikomizukara, onore, ziko

Near reflexives Zibun

Shimada and Nagano’s observation that Japanese non-affixal anaphors, such as mizukara, onore, and ziko, function as pure reflexives suggests that the reflexivity of anaphors is not determined based on boundness, posing a serious problem for the generalisation in (6), which states that affixal anaphors and non-affixal anaphors correspond to pure reflexives and near reflexives, respectively. In fact, morphological boundness and semantic properties do not necessarily have a one-to-one correspondence. Note that Kishida (2009) herself observes a similar fact in Korean. Korean has an affixal anaphor ca- ‘self,’ which behaves like a pure reflexive, and a non-affixal anaphor caki ‘self,’ which behaves like a near reflexive, as shown in Kang (2001): (14) a. John-nun ca-phokhaystta. John-NOM self-blew-up ‘John protected himself.’ (ca- = John, *a statue of John) b. John-nun caki-lul pholhayssta. John-NOM self-ACC blew-up ‘John protected himself.’ (caki = John, a statue of John) However, Kishida’s (2009) example in (15) shows that Korean also has a non-affixal anaphor that behaves like a pure reflexive. Another non-affixal anaphor casin ‘self,’ contrasting with caki, does not seem to license statue readings:2 (15)

John-nun casin-lul pholhayssta. John-NOM self-ACC blew-up ‘John protected himself.’ (casin = John, ?a statue of John)

According to the generalisation in (6), the non-affixal anaphor casin should be classified as a near reflexive, but it actually behaves like a pure anaphor. The data in Korean also suggest that the idea of a systematic correspondence between morphological boundness and semantic properties is problematic.

66

Chapter Three

3.2. Contraction Contraction phenomena reveal that morphological boundness and semantics can be independent of each other. Consider pairs of non-reduced and reduced forms of auxiliaries. For example, the free auxiliaries would and will have their bound counterparts -‘d and -‘ll, respectively. The following is cited from Carstairs-McCarthy (2010: 37): (16) a. In those days {I would/I’d} go to Brighton every weekend. b. {Tom will/Tom’ll} eat all the muffins. As Anderson (2005) and Carstairs-McCarthy (2010) correctly suggest, the difference between an auxiliary of free form and its reduced counterpart is phonological, but not syntactic or semantic. Would and -‘d, for example, share the same syntactic and semantic properties and differ only in the phonological aspect. Will and -‘ll only compete with each other phonologically. Contraction phenomena also show that there is no systematic correlation between boundness and semantics.

3.3. Quantifiers The behaviour of Japanese quantifiers also seems to be problematic for an analysis assuming a systematic connection between morphological boundness and semantics. Let us begin by considering a quantificational expression subete ‘every’, which functions as a universal quantifier in Japanese. The following is a typical example of sentences containing subete: (17)

subete-no gakusei-ga hahaoya-o hihan-si-ta. all-GEN student-NOM mother-ACC criticism-do-PAST ‘Every student1 criticised his1 mother.’

As is well known, Japanese allows the occurrence of pronominal elements that are phonetically null, unlike English. In (17), the empty possessive pronoun exists and modifies the object hahaoya-o, just as his in the English translation does. Here, we signify the empty pronominal pro, as in the following: (18)

subete-no gakusei-ga [pro hahaoya-o] hihan-si-ta. all-GEN student-NOM PRON mother-ACC criticism-do-PAST ‘Every student1 criticised his1 mother.’

Two Types of Bound Items and their Interaction

67

We can immediately confirm that subete in (18) functions as a universal quantifier because the empty pronominal is interpreted as a semantically bound variable, as is his in the English translation. (18) has an interpretation in which x criticised x’s mother, with the value of x determined according to a selection of an entity from the set consisting of all students. In other words, (18) can be interpreted in such a way that Tom, one of the students, criticised Tom’s mother, Bill, one of the students, criticised Bill’s mother, and so on. This indicates that subete, like every, can bind a pronoun as a quantifier. Turning to the morphological status of subete, it is evident that it is morphologically free and can occur on its own. Thus, subete is a universal quantifier in a free form. Interestingly, Japanese has a morphologically bound counterpart of subete, zen-. Whereas subete constitutes a quantificational DP with a genitive Case marker -no, as in subete-no gakusei ‘every student’, zenattaches to a nominal expression as if it were an affix, forming a quantificational DP. The expression using zen- corresponding to subete-no gakusei ‘every student’ is thus zen-gakusei ‘every student.’ It can also be an antecedent of pronominals and induce bound variable interpretations, as can subete-no gakusei, as shown in the following: (19)

zen-gakusei-ga pro hahaoya-o hihan-si-ta. all-GEN student-NOM PRON mother-ACC criticism-do-PAST ‘Every student1 criticised his1 mother.’

Note that the same parallelism is observed in long-distance binding. The empty pronominal pro can be bound by subete-no gakusei and zen-gakusei in the next clause up: (20) a. subete-no gakusei-ga [Yamada-sensei-ga pro hahaoya-o all-GEN student-NOM Yamada-Prof.-NOM PRON mother-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘Every student1 thought that Prof. Yamada criticised his1 mother.’ b. zen-gakusei-ga [Yamada-sensei-ga pro hahaoya-o all-student-NOM Yamada-Prof.-NOM PRON mother-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘Every student1 thought that Prof. Yamada criticised his1 mother.’

68

Chapter Three

The behaviour of zen-gakusei and subete-no gakusei in pronominal binding thus leads us to conclude that zen-gakusei, together with subete-no gakusei, can be treated as a quantificational DP: Subete and zen-, as quantifiers, show parallel syntactic and semantic behaviour in other respects, although they contrast with each other in morphological boundness. Let us examine their parallelisms. First, consider weak crossover (WCO) effects. It is known that pronouns must be c-commanded by their quantificational antecedent to be construed as bound variables: (21) a. [TP Everyone1 [VP loves his1 mother]]. b. * [TP His1 mother [VP loves everyone1]]. In (21a), his is c-commanded by everyone, successfully obtaining a bound variable interpretation. In (21b), in contrast, his cannot be interpreted as a bound variable, because it is not c-commanded by everyone. Thus, (21b) is an example that is excluded as a WCO violation. As discussed in the literature, such as in Saito and Hoji (1983), Saito (1985), and Hoji (1985), among others, WCO effects are observed in Japanese as well, as shown in the following: (22) a. daremo-ga  pro hahaoya-o aisitei-ru Everyone-NOM PRON mother-ACC love-PRES ‘Everyone1 loves his1 mother.’ b. *pro hahaoya-ga daremo-o aisitei-ru PRON mother-NOM everyone-ACC love-PRES ‘His1 mother loves everyone1.’ In (22a), the empty pronominal is bound by the quantifier co-commanding it. (22a) is interpreted in such a way that x loves x’s mother. In (22b), in contrast, pro is not c-commanded by the quantifier, inducing a WCO violation, and the intended bound variable interpretation is impossible. The same pattern is observed in the case of the quantifiers subete-no and zen-. Compare (18) and (19), repeated here as (23a, b), with (24a) and (24b): (23) a. subete-no gakusei-ga [pro hahaoya-o] all-GEN student-NOM PRON mother-ACC hihan-si-ta. criticism-do-PAST ‘Every student1 criticised his1 mother.’

Two Types of Bound Items and their Interaction

69

b. zen-gakusei-ga pro hahaoya-o hihan-si-ta. all-GEN student-NOM PRON mother-ACC criticism-do-PAST ‘Every student1 criticised his1 mother.’ (24) a. *pro hahaoya-ga subete-no gakusei -o PRON mother-NOM all-GEN student-ACC hihan-si-ta. criticism-do-PAST ‘*His1 mother criticised every student1.’ b. *pro hahaoya-ga zen-gakusei-o hihan-si-ta. PRON mother-NOM all-student-ACC criticism-do-PAST ‘*His1 mother criticised every student1.’ In contrast to (23), pro in (24) is not c-commanded by subete-no hahaoya or zen-hahaoya, resulting in ungrammaticality. This indicates that both (24a) and (24b) are excluded as WCO violations. WCO effects are also observed in the environment of long-distance binding. Again, compare (20), repeated below, with (25): (20) a. subete-no gakusei-ga [Yamada-sensei-ga pro hahaoya-o all-GEN student-NOM Yamada-Prof.-NOM PRON mother-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘Every student1 thought that Prof. Yamada criticised his1 mother.’ b. zen-gakusei-ga [Yamada-sensei-ga pro hahaoya-o All-student-NOM Yamada-Prof.-NOM PRON mother-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘Every student1 thought that Prof. Yamada criticised his1 mother.’ (25) a. *pro hahaoya-ga [Yamada-sensei-ga subete-no gakusei-o PRON mother-NOM Yamada-Prof.-NOM all-GEN student-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1.’ b. *pro hahaoya-ga [Yamada-sensei-ga PRON mother-NOM Yamada-Prof.-NOM zen-gakusei-o hihan-si-ta] to omot-ta. all-student-ACC criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1.’

Chapter Three

70

The empty pronominals in (25) are not c-commanded by subete-no gakusei or zen-gakusei, and the sentences in (25) exhibit WCO effects. From the data on WCO effects observed thus far, it can be said that zen-gakusei, like subete-no gakusei, is a quantificational expression. Thus, zen- and subete can be understood as having the same semantic property, although they differ in morphological boundness. Further parallel behaviour of subete and zen- is observed concerning WCO effects. It is well known that Japanese allows scrambling, a kind of movement that changes surface word orders. For example, in addition to basic SOV orders, OSV orders are also observed in Japanese: (26) a. John-ga Mary-ni at-ta John -NOM Mary-DAT see-PAST ‘John saw Mary.’ b. Mary-ni John-ga at-ta Mary-DAT John-NOM see-PAST ‘John saw Mary.’ It is generally assumed that OSV orders are derived through scrambling, that is, the movement of an object to a sentence-initial position. In (26b), Mary-ni occurs in a sentence initial position as a result of scrambling, leaving a trace, signified as t, in the original object position. This is illustrated in (27): (27)

[Mary-ni]1 John-ga t1 at-ta Mary-DAT John-NOM see-PAST ‘John saw Mary.’

Long-distance scrambling is also possible, as in the following: (28) a. Bill-ga [John-ga Mary-ni at-ta] to omot-ta. Bill-NOM John-NOM Mary-DAT see-PAST C think-PAST ‘Bill thought that John saw Mary.’ b. [Mary-ni]1 Bill-ga [John-ga t1 at-ta] to omot-ta. Mary-DAT Bill-NOM John-NOM see-PAST C think-PAST ‘Bill thought that John saw Mary.’ Saito (1992) and Tada (1990, 1993) observe some asymmetries between short scrambling and long-distance scrambling in Japanese; short scrambling, but not long-distance scrambling, can suspend WCO effects. The following are typical examples exhibiting WCO violations:

Two Types of Bound Items and their Interaction

71

(29) a. *pro hahaoya-ga daremo-o sikat-ta PRON mother-NOM everyone-ACC scold-PAST ‘*His1 mother scolded everyone1.’ b. *pro hahaoya-ga [Yamada-sensei-ga daremo-o PRON mother-NOM Yamada-Prof.-NOM everyone-ACC sikat-ta] to omot-ta scold-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada scolded everyone1.’ Interestingly, daremo is able to bind pro after short scrambling, whereas long-distance scrambling makes no difference in grammaticality: (30) a. [daremo-o]1 pro hahaoya-ga t1 sikat-ta everyone-ACC PRON mother-NOM scold-PAST ‘*His1 mother scolded everyone1.’ b. * [daremo-o]1 pro hahaoya-ga [Yamada-sensei-ga t1 everyone-ACC PRON mother-NOM Yamada-Prof.-NOM sikat-ta] to omot-ta scold-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada scolded everyone1.’ We can confirm the parallelism between subete and zen- based on the short scrambling/long-distance scrambling asymmetry shown above. The same pattern holds for the quantificational expressions containing subete and zen-. As mentioned earlier, subete-no gakusei and zen-gakusei in object positions cannot bind pro contained in subjects, whether the subjects reside in the same clause, as in (24), or in the upper clause, as in (25). The relevant examples are reproduced below for the reader’s convenience: (24) a. *pro hahaoya-ga subete-no gakusei-o PRON mother-NOM all-GEN student-ACC hihan-si-ta. criticism-do-PAST ‘*His1 mother criticised every student1.’ b. *pro hahaoya-ga zen-gakusei-o hihan-si-ta. PRON mother-NOM all-student-ACC criticism-do-PAST ‘*His1 mother criticised every student1.’

Chapter Three

72

(25) a. *pro hahaoya-ga [Yamada-sensei-ga subete-no gakusei-o PRON mother-NOM Yamada-Prof.-NOM all-GEN student-ACC hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1.’ b. *pro hahaoya-ga [Yamada-sensei-ga PRON mother-NOM Yamada-Prof.-NOM zen-gakusei-o hihan-si-ta] to omot-ta. all- student-ACC criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1. ’ .

Short scrambling of subete-no gakusei and zen-gakusei improves the grammaticality of these examples, as shown in (31), although long-distance scrambling does not, as shown in (32): (31) a. [subete-no gakusei-o]1 hahaoya-ga t1 hihan-si-ta. all-GEN student-ACC mother-NOM criticism-do-PAST ‘*His1 mother criticised every student1.’ b. [zen-gakusei-o]1 hahaoya -ga t1 hihan-si-ta. all-GEN student-ACC mother-NOM criticism-do-PAST ‘*His1 mother criticised every student1.’ (32) a. *[subete-no gakusei-o]1 hahaoya-ga [Yamada-sensei-ga t1 all-GEN student-ACC mother-NOM Yamada-Prof.-NOM hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1.’ b. *[zen-gakusei-o]1 hahaoya-ga [Yamada-sensei-ga t1 all-student-ACC mother-NOM Yamada-Prof.-NOM hihan-si-ta] to omot-ta. criticism-do-PAST C think-PAST ‘*His1 mother thought that Prof. Yamada criticised every student1.’ The grammatical patterns in (31) and (32) strongly suggest that subete and zen- both function as quantificational expressions in a similar way despite their differing morphological boundness. Scope interaction phenomena also indicate that subete and zen- are equally involved in universal quantification. The literature, such as Hoji

Two Types of Bound Items and their Interaction

73

(1985), notes that subject QPs always take a wide scope over object QPs in SOV orders, whereas a scope ambiguity arises in OSV orders induced by scrambling: (33) a. daremo-ga dareka-o hihan-si-ta everyone-NOM someone-ACC criticism-do-PAST ‘Everyone criticised someone.’  (Ѧ㸼ѩ, *ѩ㸼Ѧ) b. Dareka-o daremo-ga hihan-si-ta someone-ACC everyone-NOM criticism-do-PAST ‘Everyone criticised someone.’  (Ѧ㸼ѩ, ѩ㸼Ѧ) Daremo unambiguously takes a wide scope in (33a). In contrast, (33b) has an additional interpretation in which dareka takes a wide scope. The same paradigm is observed even if daremo is replaced with subete-no gakusei or zen-gakusei: (34) a. subete-no gakusei-ga dareka-o hihan-si-ta all-GEN student-NOM someone-ACC criticism-do-PAST ‘Every student criticised someone.’  (Ѧ㸼ѩ, *ѩ㸼Ѧ) b. zen-gakusei-ga dareka-o hihan-si-ta all-student-NOM someone-ACC criticism-do-PAST ‘Every student criticised someone.’  (Ѧ㸼ѩ, *ѩ㸼Ѧ) (35) a. Dareka-o subete-no gakusei-ga hihan-si-ta someone-ACC all-GEN student-NOM criticism-do-PAST ‘Every student criticised someone.’  (Ѧ㸼ѩ, ѩ㸼Ѧ) b. Dareka-o zen-gakusei-ga hihan-si-ta someone-ACC all-GEN student-NOM criticism-do-PAST ‘Every student criticised someone.’  (Ѧ㸼ѩ, ѩ㸼Ѧ) (34) and (35) also suggest that the semantic functions of the free item subete and the bound item zen- are the same.

3.4. Summary In this section, we have shown that Kishida’s generalisation in (6) is not valid and that the connection between morphological boundness and semantics is not strict and systematic. The difference in morphological boundness between linguistic elements does not mean that they are semantically different. Morphological boundness is not a reflection of semantic properties.

74

Chapter Three

4. A Lexeme-based Analysis of Japanese Anaphors and its Extension In section3, we observed that the generalisation in (6) is not tenable. In fact, free items and bound items can have identical semantic properties. For example, the Japanese anaphors mizukara and zi- are both pure reflexives, the auxiliary would has a bound counterpart -‘d, and the Japanese quantifiers subete and zen- are equally involved in universal quantification. Shimada and Nagano (2011), examining the reflexive properties of Japanese anaphors, developed a lexeme-based analysis to explain their classification, presented in Table 1 in section 3.1. In this section, we briefly review Shimada and Nagano’s (2011) analysis of Japanese anaphors and extend and apply this analysis to the contraction and quantification data given in section 3.2 and section 3.3. Shimada and Nagano (2011) attempt to explain how free and bound anaphors in Japanese behave similarly, using the framework of a lexeme-based morphology (Aronoff 1976, 1994; Anderson 1992; Fradin 2003). Remember that the free anaphor mizukara and the bound anaphor zishare the same semantic properties, both functioning as pure reflexives. In other words, mizukara and zi- are semantically identical, although their surface forms are quite different. Shimada and Nagano’s idea is that mizukara and zi-are different surface forms of the same anaphor, that is, the same lexeme. A lexeme is a lexical unit at an abstract level, as introduced in Matthews (1991). Aronoff (1994) states that a lexeme is an abstract morphological entity binding sound form, meaning, and syntax together. According to Fradin (2003), a lexeme consists of graphemic (G), phonological (F), syntactical (SX), morphological (M) and semantic (S) information. Morphological boundness is described in sound form in Aronoff’s term or in G or M information in Fradin’s term. Semantic information is independent of, and not conveyed in, surface forms. Lexeme-based approaches thus presuppose the separation of semantics and surface forms, as suggested in Aronoff (1994). It is also assumed that a lexeme has several variants for its surface realisation, namely, allomorphs. In English, for example, a noun has singular and plural forms, as illustrated by the pair man and men (Carstairs-McCarthy 2002: 23). A verb also changes its surface form according to person, number, and tense. However, the syntactic categories and lexical meanings of nouns and verbs are invariable regardless of the surface form they take. Although they are distinct in surface form, both man and men are nouns and have the same lexical meaning. The lexical unit

Two Types of Bound Items and their Interaction

75

common to man and men that exists at the abstract level is the lexeme. Man and men are thus allomorphs of the same lexeme MAN and share the same lexical meaning. Adopting a lexeme-based morphological theory, Shimada and Nagano regard mizukara and zi- as different surface forms of the same lexeme, that is, MIZUKARA. They are identical in semantics because they are surface forms of the same lexeme. One might wonder whether mizukara and zi- are allomorphs of the same lexeme because they look totally unrelated in surface form (see also Carstairs-McCarthy 2002: 32-33 on this point). They have no mora in common and they differ in the number of mora: mizukara has four moras, and zi- one mora. They seem to be different in every phonological aspect of their surface forms. However, Shimada and Nagano provide evidence from the Japanese writing systems which suggests that they are allomorphs of the same lexeme. Chinese characters are used as one of the writing systems in Japanese, and it is important to note that Chinese characters are read in two ways in Japanese: in Sino-Japanese pronunciation and in native Japanese pronunciation. Considering mizukara and zi- in terms of this writing system, they can be grouped together because they are written with the same Chinese character ‘⮬’; zi- is its Sino-Japanese pronunciation, and mizukara is its native Japanese pronunciation. Assuming that the graphemic (G) information of a lexeme is relevant to its orthography, the fact that zi- and mizukara are written with the same Chinese character strongly suggests that they belong to the same lexeme. The lexeme MIZUKARA can be realised as the bound morpheme zi- or as the free morpheme mizukara and shows the same semantic properties. MIZUKARA [+pure reflexivity] (lexeme) zi(bound form)

mizukara (surface forms) (free form)

Figure 3-1. The lexeme ‘MIZUKARA’

The two ways of pronouncing Chinese characters in Japanese reflect this lexeme-based organisation of the lexicon (cf. Morioka 2004, Part 1). As mentioned previously, the lexeme-based morphological view implies that surface forms and semantic properties are separated in the lexicon. Remember the generalisation in (6), which means that morphological boundness and semantics are closely related to each other. In short, the lexeme-based perspective and the generalisation in (6), or Kishida’s view,

76

Chapter Three

are inconsistent with each other. The behaviour of Japanese anaphoric expressions provides empirical support for the lexeme-based approach. The relationship between a free form of an auxiliary and its contracted bound form can be explained in the same way. It is possible to take would and ‘d as the surface forms of the same lexeme, say, WOULD, and the lack of any difference in syntax and semantics between them immediately follows. Similarly, will and ‘ll can be taken as the allomorphs of the same lexeme. Turning to subete and zen-, their parallel behaviour is consistent with the lexeme-based perspective. Assuming that the lexeme SUBETE exists in the lexicon and that subete and zen- are both its surface forms, we can explain their distributions and grammatical behaviour. One might object again that subete and zen- are too different in surface form to be regarded as the allomorphs of the same lexeme; the former is a free form and the latter is a bound form, and they have nothing phonological in common. However, the empirical evidence for the lexeme-based treatment is again obtained from the writing systems. The Chinese character ‘඲’ is used not only for subete but also for zen-. Subete is its Japanese reading and zen- is its Chinese reading. Thus the lexeme-based analysis is applicable to the contraction phenomena and the parallel behaviour of the quantificational expressions subete and zen-. It can be concluded that the lexicon of human language has a lexeme-based organisation.

5. Some Consequences Thus far, we have demonstrated that lexeme-based approaches are promising for explaining grammatical phenomena concerning lexicon. In this section, we discuss some consequences of using a lexeme-based theory rather than morpheme-based theory. In particular, we are concerned with the notion of morpheme and the classification of bound forms. We show that the lexeme-based approach leads us to divide bound items into two types, lexical elements and functional elements.

5.1. No Notion of Morpheme In a morpheme-based theory, morphemes exist in the lexicon as fundamental and substantial units underlying words. According to this view, the surface forms mizukara and zi- are characterized as morphemes and regarded as grammatical elements having a psychological reality. They bear semantic information, as Kishida assumes. The surface form is the substance. In contrast, in a lexeme-based theory, lexemes are substantial lexical elements, with no notion of morpheme. As discussed in section 4, lexemes

Two Types of Bound Items and their Interaction

77

are abstract lexical units. Unlike morphemes, lexemes cannot be identified as audible or visible surface forms, which are just their phonological, or superficial, manifestations. The surface forms mizukara and zi-, for example, are not really lexical items: they bear no substantial information about the lexical content and have no psychological reality. They are superficial realisations of the lexeme MIZUKARA. Though invisible, MIZUKARA is an element we can visualise for interpretation and calculation in our mind. Paradoxically and interestingly, visible or audible parts are not substantial, whereas the invisible, inaudible or abstract entities called lexemes are real elements in the lexicon.

5.2. Two Types of Bound Forms Given that there is no notion of morpheme in a lexeme-based framework, the terms free morpheme and bound morpheme are cover terms to be used only for the convenience of discussion. It is assumed that content words, or lexical categories, are realised as free morphemes, whereas affixes, or functional categories, are realised as bound morphemes. Remember that Kishida calls the Japanese anaphor zi- an affixal anaphor. This terminology would be natural in a morpheme-based approach, with zi- taken as a bound morpheme. Mizukara, on the other hand, should be a lexical category because it is classified as a free morpheme. In the lexeme-based approach developed thus far, however, zi- is a surface form but not a morpheme. Its substance lies in the lexeme MIZUKARA. Thus, although it is a bound form, it should belong to a lexical category because it is only a surface manifestation of the lexeme MIZUKARA. Lexemes, as lexical categories, are different in nature from affixes, which lack the semantic content necessary for lexical categories and function like functional categories rather than lexical categories. A lexeme-based theory thus predicts that there are bound forms that are lexical in nature. Put differently, bound forms are divided into two types, lexical and functional. The former is a surface form of a lexeme, and the latter is an affix, which we tentatively consider to be a phonological realisation of a functional feature (Beard 1995). The Japanese anaphor zi- is shown to be a lexical element but not an affix, contrary to the classification by Kishida. It should be noted that we can determine whether a bound form is a surface form of a lexeme or an affix. Bound forms, whether lexical or affixal, must combine with other elements; they cannot occur on their own. Affixes participate in the word formation process of affixation. Given that bound realisations of lexemes are lexical in nature, the word formation process in which that they are involved should be compounding, a typical

Chapter Three

78

process observed in lexical categories or lexemes, as in: (36)

housewife

The compound in (36) consists of the nouns house and wife. If a lexeme-based approach is correct and lexical properties are determined at an abstract level rather than a surface level, bound forms and free forms should be the same for the possibility of compound formation. As Scalise (1984) correctly and persuasively points out, there is a difference between affixation and compounding in that affixation shows positional restriction. Prefixes only occur to the left, while suffixes only occur to the right. For example, the suffix -ise only occurs to the right, as in hospitalise, and the form ise-hospital is not allowed. This is not the case for compounding. For example, the lexeme HOUSE comes to the left in the compound housewife and to the right in greenhouse. Let us observe that the Japanese anaphor zi- can occur both leftward and rightward. The following examples are cited from Shimada and Nagano (2011): (37) a. zi-sitsu (‘⮬ᐊ’) self-room ‘one’s own room’ b. kaku-zi (‘ྛ⮬’) each-self ‘each’ In (37a), zi- attaches to -shitsu from the left. In (37b), in contrast, it attaches to kaku- from the right. Thus, the examples in (37) show that zi- is involved in compounding. This means that the bound form zi- is not an affix but a lexical element, that is, a phonological realisation of a lexeme. It is thus demonstrated that there are bound forms other than affixes and that they are lexical in nature. Bound forms should be grouped into lexeme types and affix types. In fact, the words containing zi- ‘⮬,’ such as ziman-suru ‘boast oneself,’ are compounds in which affixation is not involved at all. The existence of bound forms of a lexical type is not limited to Japanese. English bound forms such as psych- and -logy appearing in psychology, for example, are often called combining forms (CFs), and CFs are widely shared by European languages (Baeskow 2004). There have been controversies in the literature about what CFs are and how they differ from affixes. Shimada and Nagano (2011) argue that bound forms like psych- and -logy are neither CFs nor suffixes but are bound allomorphs of lexemes that

Two Types of Bound Items and their Interaction

79

are involved in forming compounds like psychology. Psych- and -logy have free counterparts, mind and study, respectively, belonging to the same lexemes with them. They do not show any positional restriction, and it is possible to create a new compound with -logy posited to the left and phobia-, another CF meaning ‘fear,’ posited to the right, as in logophobia. CFs are freely able to be involved in compounding as bound forms of lexemes. Shimada and Nagano’s (2011) lexeme-based analysis eliminates the notion of CF in UG.3

6. Summary In this paper, following Shimada and Nagano’s (2011) idea that the bound anaphor zi- and the free anaphor mizukara, written with the same Chinese character, belong to the same lexeme and have the same semantic reflexive property, we have explained related phenomena, such as contraction and quantification. Using a lexeme-based analysis, we have distinguished between two types of bound forms. Lexical types are surface forms of lexemes and functional types are affixes. This classification is observed cross-linguistically, which strongly suggests that the lexicon of human language conforms to a lexeme-based organisation.

Notes 1

Japanese has a writing system that utilises Chinese characters. Zibun and zibunzisin are written as ‘⮬ศ’ and ‘⮬ศ⮬㌟,’ for example. In this paper, we provide Chinese scripts for Japanese when necessary. 2 Kishida suggests that casin, compared with caki, is special as an anaphor in many ways, including its near reflexive usage. 3 See Shimada and Nagano (2011) for Hungarian CFs. See also Amiot and Dal (2007) for a similar approach to French CFs.

References Aikawa, T. 1993. Reflexivity in Japanese and LF-analysis of Zibun-binding. PhD.dissertation, Ohio State University, Columbus. Amiot, D. & D. Georgette. 2007. Integrating Neoclassical Combining Forms into a Lexeme-based Morphology. In G. Booij, L. Ducceschi, B.Fradin, E. Guevara, A. Ralli, and S.Scalise (eds), On-line Proceedings of the Fifth Mediterranean Morphology Meeting, pp. 323-336. http://mmm.lingue.unibo.it/. Anderson, S. R. 1992. A-morphous Morphology. Cambridge: Cambridge

80

Chapter Three

University Press. —. 2005. Aspects of the Theory of Clitics. Oxford: Oxford University Press. Aronoff, M. 1976. Word Formation in Generative Grammar. Cambridge, MA: MIT Press. —. 1994. Morphology by Itself: Stems and Inflectional Classes. Cambridge, MA: MIT Press. Baeskow, H. 2004. Lexical Properties of Selected Non-native Morphemes of English. Tubingen: Gunter Narr. Beard, R. 1995. Lexeme-Morpheme Base Morphology. New York: SUNY Press. Carstairs-McCarthy, A. 2002. An Introduction to English Morphology. Edinburgh: Edinburgh University Press. —. 2010. The Evolution of Morphology. Oxford: Oxford University Press. Fradin, B. 2003. Nouvelles Approches en morphologie. Paris: Universitaires de France. Hoji, H. 1985. Logical Form Constraints and Configurational Structures in Japanese. PhD.dissertation, University of Washington. Jackendoff, R. 1992. Mme Tussaud Meets the Binding Theory. Natural Language and Linguistic Theory 10, 1-31. Kang, B.M. 2001. The grammar and use of Korean reflexives. International Journal of Corpus Linguistics 6, 134-150. Kishida, M. 2009. Classification of Reflexives. Proceedings of the 11th Seoul International Conference on Generative Grammar 229-245. Lidz, J. 2001. Condition R. Linguistic Inquiry 32, 123-40. Mahajan, A. 1990. The A/A-bar Distinction and Movement Theory. PhD.dissertation, MIT. Matthews, P. H. 1991. Morphology, 2nd ed. Cambridge: Cambridge University Press. Morioka, K. 2004. Nihongo to kanji [Japanese and Chinese characters]. Tokyo: Meiji Shoin. Saito, M. 1985. Some Asymmetries in Japanese and Their Theoretical Implications. PhD Dissertation, MIT. —. 1992. Long-distance Scrambling in Japanese. Journal of East Asian Linguistics 1, 69-118. Saito, M. & H. Hoji. 1983. Weak Crossover and Move Į in Japanese. Natural Language and Linguistic Theory 1, 245-259. Scalise, S. 1984. Generative Morphology. Dordrecht: Foris. Shimada, M. & A. Nagano. 2011 Zi-nouns in Japanese and Related Issues, Studies in Language and Literature 59, 75-106. Tada, H. 1990. Scrambling(s). unpublished manuscript. MIT. —. 1993. A/A-bar Partition in Derivation. PhD Dissertation, MIT.

Two Types of Bound Items and their Interaction

81

Takezawa, K. 1991. Zyudobun, nokakubun, bunnrifukanosyoyukobun to ‘teiru’ no kaishaku [On the Interpretation of Passive, Ergative, Inalienable Possesion and ‘teiru’]. In Y. Nitta (ed) Nihongo no Boisu to Tadosei [Voice and Transitivity in Japanese], pp. 59-81. Tsujimura, N. & T. Aikawa. 1999. Two Types of Zi-verbs in Japanese. The Journal of Association of Teachers of Japanese 33, 26-43.

CHAPTER FOUR ARE GERMAN MODALS BETTER VERBS THAN ENGLISH MODALS? AN INTRA-VERBAL INTERFACE1 JOHN PARTRIDGE

Abstract German modal verbs cannot be seamlessly equated with their English ‘counterparts’, not only on the basis of their semantics and pragmatics but also because they have a larger inventory of members and forms and a more extensive range of grammatical usage. It could perhaps be thought that they are better ‘real verbs’ or maybe ‘verbier’ than English ones, hence that there are interfaces within the category ‘verb’ which go beyond the purely semantic, and do not accord from one language to another. This paper examines the adequacy of this assumption on the basis of an account of some of the formal properties of German and English modal verbs.

1. Introduction: Terminological Matters 1.1 The 2005 Duden-Grammatik Describes Modal Verbs Thus: ‘Ein Modalverb ist ein Verb, das in Verbindung mit dem Infinitiv eines anderen dessen Inhalt modifiziert‚ (Duden-Grammatik 2005: 1266) [‘A modal verb is a verb which modifies the content of another when it occurs in combination with its infinitive’ My translation: JGP] This pithy description sets up a canonical structural description: modal verbs catenate directly with infinitives without the intervention of the particle zu in German or to in English:

Chapter Four

84

1)

Ohne Brille kann ich kaum sehen

parallelled in German by 2)

I can hardly see anything without my glasses

Although there is an attractive morphological similarity between those verbs traditionally perceived on this basis as modal (though by no means constituting the whole inventory) in English (can, may, must, shall, will) and German (dürfen, können, mögen, müssen, sollen, wollen), on closer examination it becomes quite rapidly clear that this similarity does not extend directly to their semantics and pragmatics, more to their morphosyntax2. The Duden definition falls far short of explaining the semantic and pragmatic relation between the modal and the catenated lexical verb: rather, the relationship which the modal mediates is between the subject and the verb, in this case one of ability. A much more satisfactory summation is that given by Zifonun, Hoffmann and Strecker (1997: 1253): ‘Als MODALVERBEN zählen die Verben, für die gilt; [1] Sie regieren den reinen Infinitiv, dann jedoch außerdem kein akkusativisches Teilkomplement. [2] Sie werden nicht im Imperativ gebraucht. [3] Sie haben ein volles Paradigma der Tempusformen. [4] Sie haben keinen eigenen Valenzrahmen, sondern transportieren den des Verbs, auf dem sie operieren, weiter. [5] Sie werden dazu verwendet, Sachverhaltsentwürfe auf der Folie von Redehintergründen, z.B. situativen Umständen, Normen oder Wissensvoraussetzungen, einzuordnen.’ German modal verbs ‘[1] [2] [3] [4] [5]

govern the pure infinitive, but no other accusative part of the complement do not occur in the imperative invert subject and verb in interrogative sentences. (This applies to all verbs in German.) have a full range of tense forms (infinitive, present, preterite, past participle) do not have a valency pattern of their own but continue those of

Are German Modals Better Verbs than English Modals?

[6]

85

the verbs they operate on are used to organise content conceptually according to the discourse background, e.g. the situational context, norms or presupposed knowledge’ (adapted from Zifonun, Hoffmann, Strecker 1997: 1253. My translation: JGP)

1.2 German Modal Verbs: An Inventory On the basis of the first five of these formal criteria the following provisional inventory can be established: müssen (must) sollen (be to, should) dürfen (be allowed to, may) mögen/möchte (may, like would like(to))

followed by ‘pure‘ infinitive (i.e. no intervening zu (see above)

wollen (want to, be about to) können (can) e.g. 3) 4) 5) 6) 7) 8)

Du musst (you may) Du sollst (you are to Du darfst (you may) Du möchtest (would like to) Du willst (want to) Du kannst (can)

(nicht) brauchen (not) need) haben (have) sein (be)

auf den Ball gehen go to the ball

followed by zuinfinitive

(occasionally bleiben, stehen) 9) 10)

Du brauchst (nicht) auf den Ball zu gehen (You (don’t have to go to the ball)) Du hast auf den Ball zu gehen (You have to go to the ball)

Chapter Four

86

11) 12)

Du warst auf dem Ball zu sehen (You could be seen at the ball) gehören followed by past participle3 (deserve/ought to be) Er gehört erschossen (he deserves to be shot)

‘Half-Modals’ Verbs such as pflegen (care for, attend to, modally tend habitually, use to) and drohen (threaten) are verbs which have lexical significance of their own and can therefore be classed as full lexical verbs, but also have a modal, attitudinal, force mediating the relationship between utterer and proposition. Scheinen (seem, appear [followed by a zu-infinitive]) fits neatly between a vaguely epistemic modal and a lexical interpretation. scheinen (appear, seem) 13) Er pflegte unanständige Witze zu erzählen (He used to tell dirty jokes) 14) Sie schien betrunken zu sein (She seemed to be drunk) 15) Er drohte seine Frau zu erschlagen (He threatened to kill his wife)

1.3. Modal Verbs in English Adopting an adapted version of Zifonun, Hoffmann and Strecker’s criteria we might postulate a characterisation of English modal verbs as below: English modal verbs [1] [2] [3] [4] [5]

govern the pure infinitive, but no other accusative part of the complement do not occur in the imperative invert subject und verb in interrogative sentences. (Applies to all auxiliary verbs in English.) have a defective range of tense forms present, past, but no infinitive, present or past participle do not have a valency pattern of their own but continue those of the verbs they operate on

Are German Modals Better Verbs than English Modals?

[6]

87

are used to organise content conceptually according to the discourse background, e.g. the situational context, norms or presupposed knowledge (JGP, following Zifonun, Hoffmann, Strecker (1997 (adapted)))

1.3.1 A Provisional Inventory can may must shall will

followed by ‘pure‘ infinitive, i.e. no to

dare need

followed by to-infinitive, however occasionally ‘pure’ infinitive

ought

followed by to-infinitive (following Palmer (1976))

2. Morphological Contrast Initially German and English modal verbs look seductively similar in their morphology (Table 1): Table 4-1: German and English modal base forms Base form German English dürfen no cognate können can mögen may müssen must sollen shall wollen will However, dürfen stands out from the original list in that it has no cognate English form and that English must use may or the pragmasemantic periphrasis be allowed to in order to express both epistemic and deontic senses.

Chapter Four

88

Additionally, there is no semantic or pragmatic correspondence between the obviously cognate English will and the German wollen, as will is used to express a prediction of futurity: 16)

It will rain tomorrow. Wollen, on the other hand, is used to indicate a certain volition.

17)

Ich will meine Mutti (sehen)

and its meaning is more comparable with that of may, morphologically cognate with the German möchte4, which can be better rendered by the periphrasis want to/would like to. The future-oriented characteristics of English will are taken over in German by werden, just as the intentionality of German wollen is taken over in English by the periphrasis want to: 16࢝) 17࢝)

Morgen wird es regnen. I want (to see) my Mummy.

Zifonun et al’s list indicates that werden is not a modal verb, but a future-referential auxiliary verb when catenated with a full verb, and serves a copular function when linked with a following nominal phrase. Under no circumstances may it stand alone, however. The German half-modals pflegen, scheinen and drohen and similarly English dare, need and ought form a half-way house: pflegen, scheinen and drohen demand a zu-infinitive: 18) 19) 20)

Er pflegte seine Freundin mit Erdbeeren zu bewerfen Sie schien sich damit abgefunden zu haben Er drohte sie zu heiraten. Similarly we may observe the to-infinitive in English

18࢝) 19࢝) 20࢝)

He used to throw strawberries at his girlfriend She seemed to have accepted it He threatened to marry her,

with pflegen, drohen and threaten clearly exercising full verb function 21) 22)

Er pflegte sorgfältig seinen Hund He took meticulous care of his dog

Are German Modals Better Verbs than English Modals?

23) 24)

89

Er drohte ihr mit dem Tod He threatened her with death,

and (modal) scheinen, seem and use(d) to each serve a copular or temporal function. In the case of the English half-modals dare, need and ought, dare and need(n’t) are verbs with a full verbal morphology, and can be unreservedly regarded as full lexical verbs but can sometimes be used modally without a to-infinitive: 25) 26) 27) 28)

Don’t you dare do that again How dare you speak to me like that He daren’t show his face in this house You needn’t/don’t have to come to the funeral if you don’t want to

The German semantic equivalent of dare, wagen, is clearly non-modal from a morphosyntactic point of view: it employs zu-concatenation and overall displays full verb behaviour. Direct equivalents for (25) are hard to find, but (27) may be paralleled by (29) and (30) 29)

Er wagte es, meine Frau mit ‘Liebchen’ anzusprechen

29࢝) 29) 30)

He dared to call my wife ‘darling’ Er wagt es nicht mehr, sich bei uns sehen zu lassen He no longer dares to show his face in our house

English need is paralleled in German by German brauchen, with need requiring to-catenation, and brauchen zu-catenation. When in the negative need obligatorily drops to-catenation, whilst brauchen may optionally retain zu-catenation. 31) 32)

You needn’t /don’t have to keep on whingeing like that Du brauchst nicht dauernd (zu) meckern However, need not is also covered by the German modal nicht müssen:

28࢝)

Du musst nicht zur Beerdigung kommen, wenn du nicht möchtest/willst

Chapter Four

90

but the apparently cognate must not (i.e. be forbidden to/not be allowed to) is rendered by nicht dürfen or nicht sollen. Ought, the third of the putative semi-modals, which has a restricted morphology, lacking the otherwise obligatory 3rd person singular -s-ending, but can only be catenated using the to-infinitive is without exception pragmatically modal. 33)

You ought to wash more often

2.1. Periphrasis Before proceeding to a full comparison of the contrastive morphology of English and German modal verbs, a further lexical-semantic peculiarity of English modal verbs must be pointed out: periphrasis. German modals thereby reveal themselves to be richer in content and more flexible than English ones. Although German modal verbs can be periphrastically circumscribed, this is practically obligatory in English, if certain interpretations are to be made apparent. This is shown primarily in the infinitive. As, unlike their German counterparts, English modal verbs lack an infinitive this has to be performed by periphrasis (Table 2): Table 4-2: German infinitives and English infinitival periphrases German dürfen können mögen mussen sollen wollen

English no cognate *to can *to may *to must *to shall/should *to will

English Periphrasis to be allowed to to be able to to be allowed to, to be compelled to, have to to be (obliged) to/ be said/ supposed to to be about to/ claim to/ want to

This also comes to light in the formation of the gerund: the German infinitive can be transformed into a gerund by taking on the syntax of a neuter noun. A similar process occurs in nominalization (Table 3), although it must be admitted that such expressions rarely occur.

Are German Modals Better Verbs than English Modals?

91

Table 4-3: German gerunds and nominalisations Infinitive

Gerund

Nominalisation

dürfen können mögen müssen sollen

das Dürfen das Können das Mögen das Müssen das Sollen

der Bedarf

wollen

das Wollen

die Macht das Soll (debt/ deficit) der Wille(n)

In English a gerund cannot be formed, owing to the lack of an infinitive, and so the periphrasis has to be nominalised (Table 4): Table 4-4: English periphrastic gerunds and nominalisations Stem can may

Gerund *canning *maying

must

*musting

should will

*shoulding *woulding

Periphrasis being able to being allowed/permitted to being compelled to, having to being obliged to being about to, insisting on , being in the habit of

Nominalisation ability/possibility possibility, allowance/permission compulsion obligation ?immediate futurity, insistence, habit

It should also be noted that the pragmasemantic polysemy of German modals is lacking in English modal periphrases – in the various differing periphrases of a single modal verb deontic, epistemic and alethic modalities for example are kept strictly apart. 34) 34࢝) 34࢝࢝)

You may go to the toilet You are allowed to go to the toilet Du darfst auf die Toilette gehen

deontic

33) 33࢝) 33࢝࢝)

You may have been infected It is possible that you have been infected Es kann sein, dass du infiziert worden bist

epistemic

Chapter Four

92

34) 34࢝) 34࢝࢝)

A triangle may not have more than three sides It is not possible for a triangle to have more than three sides

alethic

Ein Dreieck darf nicht mehr als drei Seiten haben

Also, implicative interpretation (cf. Karttunen (1971)) is lost: 35)

Nach dem Unfall konnte sich das Unglückspaar gerade noch retten

35࢝)

After the accident the unfortunate pair were just about able to escape (and did),

and an alternative interpretation 35࢝࢝)

?After the accident the unfortunate pair could just about escape (but couldn’t be bothered, and died horribly)

can hardly be entertained.

2.2. Morphological Patterns in English and German Modal Verbs 2.2.1 German Inflected Forms In the main German modal verbs follow the ‘weak’ verb pattern, but are to some extent defective (Table 5): a) b) c)

In the first and third persons singular the endings -e und -t are absent In the singular of wollen the stem vowel -o is replaced by -i Umlauted vowels in the stem are not retained in the present singular, but reappear in the past subjunctive.

Are German Modals Better Verbs than English Modals?

93

Table 4-5: German inflected modals; morphology Infinitive

Pres. sing.

Pres. pl.

Pres. Subj

Preterite

dürfen können mögen müssen sollen wollen

darf, -st kann, -st mag, -st muss, -t soll, -st will, -st

dürfen können mögen müssen sollen wollen

dürfe könne möge müsse solle wolle

durfte konnte mochte musste sollte wollte

Past Subj. dürfte könnte möchte müsste sollte wollte

2.2.2 German Uninflected Forms All German modal verbs lack a present particle, gerundive and an adjectival form (derivational exceptions are perhaps vermögend (well-off) and ((wohl)wollend) (well-)wishing)) All German modal verbs have a weakly formed free-standing perfect participle, except that the umlaut of the stem form disappears. When catenated with other verbs the infinitive is used (Table 6). Table 4-6: German modals: participles Infinitive dürfen können mögen müssen sollen wollen

Present Participle/ Gerundive *dürfend *könnend *mögend *müssend *sollend ? wollend

Deviations/ Derivations? vermögend wohlwollend

Past Participle gedurft/ dürfen gekonnt/ können gemocht/ mögen gemusst /müssen (?gesollt)/ sollen gewollt /wollen

2.3 English: Inflected Forms English modal verbs are formed according to idiosyncratic patterns (Table 7). a)

There is no 3rd person singular –s. All present forms are identical.

Chapter Four

94

Table 4-7: English inflected modals; present tense morphology Person 1

Pronoun I

2

you

3

he/she/it

Verb singular can/ may/must/ shall/will can/ may/must/ shall/will can/ may/must/ shall/will

Pronoun we you they

Verb plural can/ may/must/ shall/will can/ may/must/ shall/will can/ may/must/ shall/will

b)

English verbs in general have hardly any subjunctive forms. At the very most the 3rd person singular –s-ending is omitted or in some cases have and be are used instead of has/had and is /was. Modals have no subjunctive forms.

36) 37) 38) 39)

I recommend he serve at least 30 years’ imprisonment If I were a carpenter and you were a lady… The president demands they have their heads cut off I recommend they be shot at dawn

c)

The preterite is formed entirely idiosyncratically – for example the present and preterite of must are identical; however, can, shall and will form similar preterites in a way which cannot be explained by their stem forms (Table 8)

Table 4-8: English inflected modals; present and preterite morphology Stem form

Circumscription

can may must

[be able to] [be allowed to] [be obliged to, have (got) to5, need to] [be going /about to] [be going to/ about to/ tend to/]

shall will

Present sing. & pl. can may must

Present subj.

Preterite

Past subj.

-

could might must

-

shall

-

should

-

will

-

would

-

Are German Modals Better Verbs than English Modals?

95

2.4 English Uninflected Forms English modal verbs essentially lack uninflected forms and have to resort to periphrastic forms6 (Tables 9 & 10): Table 4-9: English gerundive periphrasis Stemform can may must shall should will

Present participle /gerundive *canning *maying *musting *shalling *shoulding *woulding

Circumscription/ Periphrasis being able to being allowed to being compelled to being about to, tending to being obliged to being about to

Table 4-10: English infinitival gerundive periphrasis Infinitive *to can *to may

Circumscription/ Periphrasis to be able to to be allowed to

*to must

to be compelled to, have (got) to, need to

*to shall

to be going to/about to to

*to should *to will

to be going/about to/ tend to/

Present Participle being able to being allowed to being compelled to, having (got) to, needing to being going to/about to being obliged to being going/about to/ tending to/

Past Participle been able to been allowed to been obliged to, had to, needed to been going to/about to been obliged to been going/about to/ tend to/

Chapter Four

96

2.5 Shared Characteristics and Differences in Morphology and Syntax In the present singular English and German modal verbs display morphological deficiencies. English and German modal verbs lack gerundive and a present participle. Neither English nor German modal verbs have a continuous (or progressive) form. This applies to all German verbs, however. English modal verbs lack an infinitive, a past participle and any form of the subjunctive. German modal verbs display all non-finite forms and two subjunctives. English modal verbs can only be nominalised in their circumscribed/periphrastic form. The gerund is formed in a similar fashion.

3. Syntactic Patterns 3.1 Simplex Sentences 3.1.1 Inversion In interrogative sentences subject-verb inversion takes place in German modal verbs, as is the case with all German verbs. Only in English auxiliary verbs (including modals) does subject-verb inversion take place in interrogative sentences. The auxiliary verb do must be added (with subsequent inversion) to English full verbs. As is the case with all German verbs German modals are negated by the simple addition of nicht, but no inversion takes place. 3.1.2 Negation Only in English auxiliary verbs (including modals) is negation performed by the simple addition of not. If there is no auxiliary verb present a form of do must be added to English full verbs. 3.1.3 Pronominalisation and Ellipsis The sentential complements of German modal verbs can be pronominalised by an accusative es-object.

Are German Modals Better Verbs than English Modals?

40)

Ich

muss I (must) soll I (am (supposed) to) darf (may/ am allowed to) will (want to) mag/möchte (may/would like to) kann (can)

97

auf den Ball gehen / es go to the ball / *it

The sentential complement of an English modal verb cannot be replaced by the object pronoun so which would be expected for propositional pronominalisation. Instead, the sentential complement undergoes progressive ellipsis (cf. Ross (1969)):7 41) 42) 43) 44) 45)

He can’t [have [been [doing [it]]]]. He can’t have been doing He can’t have been He can’t have He can’t

3.1.4 Negative Contraction When English auxiliary verbs (including modals) are negated the negating not can be contracted to n’t. Full verbs must have do added to be negated, with possible contraction of not. 46) 47) 48) 49)

* You gettn’t a quart into a pint pot You can’t get a quart into a pint pot. *You wearn’t too much make-up You don’t wear too much make-up.

The negation of German verbs, either lexical or auxiliary allows no contraction of the invariant negative particle nicht. 50) 51)

Das kann doch nicht Ihr Ernst sein (You just can’t be serious) Er läuft nicht so schnell wie Usain Bolt (He doesn’t run as fast as Usain Bolt)

Chapter Four

98

3.2 Tagging Question-tagging is generally 8 performed in English by copying the modal (or supplying a form of do in the absence of a modal) from the main clause, reversing its polarity and following it by its subject. 52) 53) 54) 55)

You can’t get a quart into a pint pot, can you? You can get a quart into a pint pot, can’t you? You don’t wear too much make-up, do you? You wear too much make-up, don’t you?

A rhetorical, usually sarcastic or doubting use of tagging may occur when polarity is not reversed: 56)

So you think you’re Elvis, do you?

Tagging with a reverse-polarity inverted modal does not occur in German, which simply adds on the invariant particle phrase nicht (wahr), or a regional expression such as ne?, ned?, net?, nich?, nit?, gell?, gelle?, gelt?, oder?, wa(s)?, woll? Indeed there is no copying of either the auxiliary or the lexical verb. Interestingly, however, nicht wahr can be seen to be verbally derived from ‘Ist das nicht wahr?’ parallel to the invariant French ‘n’est-ce pas?.’ [The invariant verbally-derived tag ‘innit?’, putatively derived in predominantly Asian areas of the British Midlands from ‘isn’t it?’ is spreading virulently, particularly amongst young people, and seems likely to establish itself as part of everyday youth vernacular. It is certainly widely parodied as such.]

3.3 Catenation German modal verbs do not allow ‘zu’ before the full verb. 57) 58) 59) 60) 61) 62)

Du musst Du sollst Du darfst Du möchtest Du willst Du kannst

auf den Ball *zu gehen

English modal verbs do not allow ‘to’ before the full verb.

Are German Modals Better Verbs than English Modals?

63) 64) 65) 66) 67) 68)

You can You should You may You must You shall You will

99

*to go to the ball

German modal verbs may be catenated with one another at will, admittedly at times with some rather bizarre semantic consequences: 69) 69‫)ޖ‬

Ich hätte nicht gehen wollen müssen/brauchen I should not have needed to want to go

English modal verbs may not be catenated with one another unless circumscribed by a periphrasis9: 70)

He must have been able to do it

4. Concluding Remarks In view of the foregoing examples and considerations let us come back to the original question: Are German modal verbs better verbs than English ones? This formulation may sound rather naïve, but I believe to have shown that although English modal verbs have more differentiated and specific applications than their German counterparts, German modal verbs are more polysemous and more versatile than English ones, have a considerably wider inventory of inflections and display a syntactic behaviour which corresponds more to that of a full verb. In this sense I believe I have shown that German modal verbs are indeed ‘better’ or ‘verbier’ verbs than English ones and can maintain that there is within the class of verbs in both languages an interface which resolves itself if not totally canonically or idiosyncratically, but at least round roughly central parameters. In this respect the features which Zifonun et al. list for German modal verbs as well as the representation of English modal verbs derived from them should be correspondingly expanded to account better for their propositional and attitudinal reflexes and for German modals’ capacity to take an accusative object, usually the neuter pronoun, and those of the 2005 Duden comprehensively re-worked on both morphosyntactic and pragmasemantic principles. Can’t say better than that, innit?

100

Chapter Four

Notes 1 This is a revision and expansion of a paper originally given in German at the IVG (Internationale Vereinigung für Germanistik) XII conference in Warsaw in August 2010. A considerably condensed version is due to appear shortly in the conference proceedings. 2 Can does not necessarily correspond directly to können, or may to mögen, must to müssen, shall to sollen, will to wollen. There is no English morphological equivalent for dürfen and the intentionality of wollen’s semantics disqualify its English semantic counterpart want to from membership of the class of modals: English will and would are standardly used to signal actual futurity or conditionality, also exasperation and vexation at habitual obstinacy. Bouma (1977) considers epistemic modals in German and English almost mathematically on the basis of gradable possibility or probability. 3 It might, however, also be thought that the verbs (nicht) brauchen (not need to), haben (have), sein (be) , bleiben (stay, remain), stehen and gehören (belong, with following past participle deserve to be) are half- rather than full modals (see 1.2.1). The Internet grammar www.canoo.net also lists ‘verstehen’ und ‘wissen’ (roughly, to understand how to or know how to). However, this question is not further investigated here.ġ 4 Will can however be used modally in English if it indicates a) an annoying tendency to obstinacy He will/would keep on wearing that dirty old flasher’s mac; or b) a repeated activity: He will/would stand there for hours 5 Have (got) to occupies a special place within modal periphrases, in that it is not restricted to one particular modality, and must be seen as a parallel term to must, e.g. epistemic You’ve got to be joking (J. McEnroe), deontic You have (got) to have a wash, and alethic Triangles must have three sides. As a periphrasis it allows both gerund formation and nominalisation of the form having to. My thanks to Damien Hall for raising this point at the Interfaces 3 conference. 6 It might be tempting, as has been suggested, to combine these tables into one. However, while some forms may be identical, their functional category is not 7 At the 2010 IVG conference Elke Ronneberg-Siebold pointed out that ellipsis also takes place with the complements of German modals, but to a much more restricted, formulaic extent, for example in the sexual boast “Ich kann/will immer (bumsen)”, lavatorial distress “Ich muss mal (auf die Toilette gehen) or assuring the possibly sceptical that one is allowed to do something which might appear to be taboo “Ich darf”. However the systematic breakdown of constituents seen in English does not appear to figure here, although it may be developing. 8 Cases in which the original polarity is retained, but where sincerity appears not to be doubted, e.g. “You’ve got backache, have you?” or “You can speak German, can you?” lend themselves to analysis as echo-questions rather than as “true” tag-questions. 9 Some dialects, I believe primarily Black American, will however allow this, e.g. I might coulda done it.

Are German Modals Better Verbs than English Modals?

101

References Bouma, L. 1973. The semantics of the modal auxiliaries in contemporary German. Janua linguarum, series practica, 146. The Hague: Mouton. Crystal, D. 2008. A Dictionary of Linguistics and Phonetics, 6th edition. Oxford: Blackwell. Dudenredaktion. 2005. Die Grammatik. Mannheim: Dudenverlag. Huddleston, R. & G. K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP Karttunen, L. 1971. Implicative verbs. Language 47, S. 340-358. Palmer, F.R. 1974. The English Verb. London: Longmans. Ross, J.R. 1969. Auxiliaries as main verbs. In W.R. Todd (ed.), Studies in Philosophical Linguistics, Series One. Carbondale. Illinois: Great Expectations Press. Zifonun, G., Hoffmann, L. & B. Strecker. 1997. Grammatik der Deutschen Sprache, Band 2. Berlin: Walter de Gruyter.

PART IV: SYNTAX-PRAGMATICS

CHAPTER FIVE YES/NO ELLIPSIS AND POLARITY FOCUS IN MODERN GREEK MARINA KOLOKONTE Abstract The aim of this paper is to put forward a tentative proposal with respect to the Information Structure role of the negative polarity particle in the elliptical construction dubbed Yes/No Ellipsis. Building on an earlier suggestion by the author that yes/no ellipsis in Modern Greek involves contrastive topicalisation of the remnant that is present in the elliptical conjunct, in this paper, it is proposed that this contrastive topicalisation induces narrow focus on the negative polarity marker that follows the remnant. This proposal is in line with independent proposals for the existence of an association between contrastive topicalisation and narrow focus.

1. Introduction Yes/no ellipsis, exemplified in (1), is an elliptical construction attested in many languages, including Modern Greek (henceforth MG). It is predominantly found in coordinated environments, as seen in (1) and its prototypical feature is that the second conjunct of the coordinated clause has been stripped away of its elements except of one constituent and a polarity marker. Furthermore, the constituent that is left behind is contrasted with a constituent in the first conjunct. Regarding terminology, I will refer to the first clause as the antecedent clause and the second the elided clause. The pronounced element of the elided clause, I will call the remnant and the constituent of the antecedent clause that the remnant corresponds to I will call the correlate.

Chapter Five

106

(1)

O Petros latrevi ti thalasa ala i Maria the Petros-NOM loves-3SG the sea-ACC but the Maria-NOM oxi. not ‘Peter loves the sea but Maria doesn’t.’

In the ellipsis literature, yes/no ellipsis has been standardly analysed on a par with verb-phrase ellipsis (VPE) due to certain similarities that the two constructions seem to have in common; for example, both seem to be allowed in subordinate environments, as in (2), and they are able to have a linguistic antecedent in a separate utterance in the discourse, as in (3). Due to this reason, the construction is also dubbed TP-Ellipsis (Laka 1990). (2)

Ton Petro ton proskalesan, ala nomizw pws the Peter-ACC him-CL invite-3PL-PAST but (I) think that ton Pavlo oxi. the Paul- ACC not ‘As for Peter, they have invited him, but as for Paul, I think they didn't.’

(3)

A: B:

I Katerina to agorase to vivlio. the Katerina-NOM it-CL buy-3SG-PAST the book ‘Katerina bought the book.’ O Petros oxi. the Peter-NOM not ‘Peter didn’t.’

On the other hand, it has been recently acknowledged (Winkler 2005) that yes/no ellipsis as defined above is remarkably similar to another elliptical construction, stripping, which is illustrated in (4). (4)

O Yorgos proskalese ti Maria, the Yorgos-NOM invite-3SG-PAST the Maria-ACC ala oxi tin Eleni. but not the Eleni-ACC ‘George invited Maria, but not Helen.’

Comparing (4) to (1), we observe that, at a superficial level, the differentiating characteristic between the two constructions, yes/no ellipsis and stripping, is the order of the elements in the elided clause; in (4), the stripping case, the polarity marker precedes the remnant rather than

Yes/No Ellipsis and Polarity Focus in Modern Greek

107

following it. Winkler, investigating both phenomena in German, argues that yes/no ellipsis and stripping can be classified under the same type of ellipsis, which she calls Sentence-bound ellipsis. Leaving the details of Winkler’s analysis aside, one of her main claims in that work is that in Sentence-bound ellipsis the remnant in the second clause is displaced to the left periphery of the clause for Information Structure requirements. Winkler's claim about the Information Structure requirements of stripping is rather uncontroversial. Many studies on stripping in various languages (e.g., Depiante 2000; Busquets 2006) have put forward the idea that stripping is constrained by focus requirements; more specifically, it is argued that the remnant in the elided clause is a focused constituent which constitutes an alternative to a focused element in the antecedent clause. Therefore, for (4), repeated below as (5), the remnant tin Eleni has as its correlate ti Maria in the antecedent clause. Since the latter is the constituent that provides the new information in the antecedent clause, it must be focused. Due to a parallelism requirement between correlate and remnant, the remnant must be focused as well. (5)

O Yorgos proskalese ti Maria, the Yorgos-NOM invite-3SG-PAST the Maria-ACC ala oxi tin Eleni. but not the Eleni-ACC ‘George invited Maria, but not Helen.’

The aforementioned claim is more controversial for yes/no ellipsis. Due to the tendency to consider yes/no ellipsis as the equivalent of VPE in the languages that do not instantiate the latter and because VPE is said not to be subject to such Information Structure constraints, there is no wide consensus on the Information Structure of a yes/no ellipsis sentence. With regard to MG, in earlier work (Kolokonte 2008), I offered evidence that supports Winkler’s proposal. I showed that on a closer inspection, the alleged similarities between yes/no ellipsis and VPE (as mentioned briefly above) do not hold. On the contrary, I argued that, at least for MG, yes/no ellipsis and stripping exhibit a good number of common characteristics to be considered subtypes of the same type of ellipsis. Most importantly for the present discussion, I claimed that both elliptical constructions are constrained by Information Structure, their major difference being the Information Structure role of the remnant; I suggested that the remnant in stripping should be taken as a focused constituent whereas the remnant in yes/no ellipsis should be considered a contrastive topic. Taking this idea further, in this paper, it is proposed that contrastive topicalisation of the

108

Chapter Five

remnant forces narrow focus on the negative polarity marker. In section 2, I review briefly the earlier suggestion that the remnant in yes/no ellipsis is a contrastive topic. In section 3, I provide an overview of the literature that draws attention to the association between contrastive topicalisation and narrow focus. Particular attention is given to Umbach’s (2005) analysis of coordinated constructions with but. Building on this literature, I put forward the claim that in yes/no ellipsis the presence of the negative marker is a requirement of but-coordination and that contrastive topicalisation of the remnant forces a narrow focus reading of the negation.

2. Contrastive Topics and Yes/No Ellipsis 2.1 Contrastive Topic versus Contrastive Focus In the literature of Information Structure, there is often lack of clarity with regard to the definitions of contrastive focus and contrastive topic (see Erteschik-Shir 2007 for a comprehensive overview). According to one influential view (Molnár 2001), contrast is best seen as an autonomous phenomenon of Information structure that can either be combined with focus, in which case we have contrastive focus or with topichood, where we have the case of contrastive topics. However, the interpretive properties of contrastive focus and topic are quite distinct. Molnár (2001) argues that contrastive focus is about the exclusion of all the alternatives (which she calls ‘exhaustive exclusion’ or ‘strong exclusion’) whereas the contrastive topics exclude some of the relevant alternatives (‘non exhaustive exclusion’ or ‘weak exclusion’). She provides the following data from Hungarian as evidence for this distinction. (6)

[FPÉTER] jár Lundban egyetemre. Peter is going Lund-in university-to ‘It is Peter who is studying in Lund.’ (Molnár 2001: (20B))

(7)

[CTPÉTER] [LUNDBAN]F jár egyetemre. Peter Lund-in is going university-to As for Peter, he is studying in Lund.’ (Molnár 2001: (21))

According to Molnár, contrastive focus on the subject Péter in (6) excludes all the other individuals of the possible set that could have gone to Lund University. On the other hand, the contrastive topic in (7), realised

Yes/No Ellipsis and Polarity Focus in Modern Greek

109

with a fall-rise intonation, indicates that there is at least one member of the set, Péter, for which the predication holds. Büring's Contrastive Topic theory (Büring, 1999, 2005, 2003) is in line with Molnár’s view above. Büring proposes that the function of contrastive topics is to mark deviance from the question, in the sense that they answer a sub-question to the question asked. The CT marking in (8) indicates two things: the deviance from a complete answer and also the fact that the question about what male pop stars were wearing is left open. (8)

A: B:

What did the pop stars wear? The [female]CT pop stars wore [caftans]F.

(Büring 2005: (23))

Umbach (2005), following Vallduví and Vilkuna (1998) distinguishes between rheme/theme on one hand and focus/background on the other. The theme of a sentence contains contextually bound information including a topic whereas the rheme provides information about the topic. A focus, however, comes with a set of alternatives which are contrasted and which have certain constraints, such as that they have to be semantically independent but both included as members of a set specified in the preceding discourse. (9)

[Peter is member of the basketball team. The team was expected to attend a charity event] Only Peter was able to join us.

Following Rooth (1992), focus on Peter puts a contrast with the other alternatives that could potentially hold; these alternatives are restricted to the members of the basketball team provided in the preceding context. According to this demarcation, it is possible that a focus can occur both in the rheme and the theme part of the sentence; when focus is in the theme part, it represents a contrastive topic and when it is found in the rheme, it corresponds to narrow focus. (10) A: What museums did you and Mike visit in Paris? B: [CTI] visited [Fthe Louvre] and [CT Mike] visited [F the Musée d'Orsay]. In the first conjunct of B's reply in (10), there is a focus accent on the subject I which is the theme part of the clause and also on the rheme part, the Louvre that constitutes new information (since it is this part of the clause

110

Chapter Five

which provides the constituent that corresponds to the wh-word). The answer given by speaker B in the first conjunct is only partial with regard to the question asked which has two parts: which museums speaker B visited and which ones Mike did. It thus triggers the expectation that other topics are to be addressed. This is in line with Molnár's and Büring's views on the semantics of contrastive topics. In what follows, I will assume that the function of a contrastive topic is as described above.

2.2. Contrastive Topics and the Left Periphery In certain languages, word order seems to be governed by Information Structure in the sense that focused and topicalised constituents need to occupy a designated position in the clause structure; this position is often associated with the left periphery of the clause. Regarding topichood, cross-linguistically, there are various syntactic configurations that seem to be motivated by the need for topic marking, such as Contrastive Left Dislocation in Germanic Languages and Clitic Left Dislocation (henceforth CLLD) in Romance languages and Modern Greek (Anagnostopoulou 1997). Studies of information structure in Italian and Spanish suggest that of the many dislocation configurations, CLLD marks contrastive topics. Benincà and Poletto (2004), for example, show that in Italian, a clitic-doubled element dislocated at the left periphery of the clause signals that it may have been included in a set introduced in the context. CLLD gives rise to what they call 'List Interpretation' of topics, which suggests that discussion about other topics will follow. (11)

[Context: a farm producing a set of goods that are known to the people involved in the conversation] La frutta la regaliamo, la verdure la vendiamo. the fruit it give-1PL for free the vegetables it sell-1PL ‘We give fruit for free, while we sell the vegetables.’ (Benincà and Poletto 2004: (47a))

In MG, contrastive topics are also shown to be associated with CLLD (Alexopoulou 1996, Baltazani 2006). This is illustrated in (12).

Yes/No Ellipsis and Polarity Focus in Modern Greek

111

(12) Q: Pjos agorase ta vivlia ? who-NOM bought-3SG the books-ACC ‘Who bought the books?’ ǹ1: ȉȠ vivlio ton mathimatikon to agorase the book-ACC of mathematics CL.NEUT.3SG.ACC bought-3SG o Nikos. the Nikos-NOM ‘The maths book, Nikos bought it.’ A2:#O Nikos to agorase the Nikos-NOM CL.NEUT.3SG.ACC bought-3SG to vivlio ton mathimatikon. the book-ACC of mathematics A3: #To agorase to vivlio ton mathimatikon CL.NEUT.3SG.ACC bought-3SG the book-ACC of mathematics o Nikos. the Nikos-NOM In (12A1), the left dislocated NP to vivlio ton mathimatikon forms a subset of the super-set the books, introduced in the question. (12A2), where the same constituent is dislocated to the right is infelicitous because a right dislocated constituent is permissible, only if it has been explicitly mentioned in the preceding linguistic context (Valiouli 1994). The answer in (12A3) is inappropriate for the same reason. (12A1) is the only appropriate answer: the object to vivlio ton mathimatikon, is dislocated to the left periphery and functions as a contrastive topic, thus creating the expectation that the speaker is following ‘a book by book’ strategy of answering the question in (12) and that her/his answer implies that there are other relevant books in the discourse.

2.3 Contrastive Topicalisation and Yes/No Ellipsis Based on the semantics and syntax of contrastive topics, as these were discussed briefly above, in earlier work (Kolokonte 2008), I proposed that Greek yes/no ellipsis involves contrastive topicalisation of the remnant. Some of the arguments in support of this idea are presented below. As is well known, focalised constituents can cooccur with focusing adverbs. If the remnant in yes/no ellipsis were focused, it should be possible to be accompanied by a focusing adverb. However, (13) shows that this is impossible.

Chapter Five

112

(13)

*Oloi simfonisan, mono oi goneis sou oxi. all-3PL agreed-3PL, only the parents-NOM yours not Intended meaning: Everybody agreed, not only your parents.

Furthermore, the order of the polarity marker makes a difference to the interpretation of the sentence. (14)

Simfonisan oli? agreed-3PL all ‘Did everybody agree?’ A1: Oi kopeles oxi The girls not ‘As for the girls, they didn’t agree.’ A2: Oxi oi kopeles. not the girls ‘Everyone agreed except the girls.’

From the translation of the examples, it is obvious that in yes/no ellipsis in (A1) the remnant is taken as a topic and thus translated with the ‘as for...’ phrase. Furthermore, in MG there are certain word orders which restrict the focus to certain elements in the clause; in the OVS order for example, focus must be carried by the object of the clause (Tsiplakou 1998); in the example (15), the occurrence of the direct object to forema in the leftmost position, without this being clitic-doubled, suggests that it is contrastively focused. (15)

[FTo forema] agorase i Sophia. the dress-ACC bought-3SING-PAST the Sophia-NOM ‘It is the dress that Sophia bought (not the shoes).’

In this context, yes/no ellipsis is not permissible, as shown in (16) (16)

*[FTo forema] agorase i Sophia ala ta papoutsia the dress bought-3SG the Sophia-NOM but the shoes ohi. no.

Yes/No Ellipsis and Polarity Focus in Modern Greek

113

However, if the object is doubled by the clitic, the yes/no ellipsis sentence becomes grammatical, as in (17). As the presence of the clitic is a method of defocusing an element (Holton et al. 1997), this forces the correlate in (17) ‘the dress’ to function as a topic. Due to the parallelism Constraint (Chomsky 1995) that dictates that the correlate and the remnant must have a parallel interpretation, it naturally follows that the remnant in yes/no ellipsis must also be considered a topicalised constituent and not a focused one. (17)

To forema to agorase i Sophia ala ta papoutsia The dress itCL bought-3SG the Sophia-NOM but the shoes ohi. not ‘As for the dress, Sophia has bought it but as for the shoes, she hasn’t.’

Regarding the derivation of the elliptical conjunct in yes/no ellipsis, Kolokonte (2008) followed a deletion analysis in the spirit of Merchant (2003, 2004). It was suggested that the remnant starts in its thematic position within the IP and from that position it dislocates to the left periphery through CLLD. Assuming Rizzi’s (1997) hierarchy of functional projections, the remnant merges with TopP, a functional projection in the Comp domain, designated for topics. Movement of the remnant is obligatory in order to acquire a contrastive interpretation. However, this earlier work did not expand on the precise role of the negative polarity marker. My tentative proposal in this paper, the details of which remain to be explored fully in subsequent work, is that contrastive topicalisation forces narrow focus on the negative polarity marker. In the next section, I provide a brief overview of the association of contrastive topicalisation and narrow focus, before moving to the formulation of my proposal.

3. Contrastive Topicalisation and Narrow Focus in Yes/No Ellipsis 3.1 Contrastive Topicalisation and Narrow Focus In the literature of contrastive topics, it is often argued that the presence of a contrastive topic in a sentence induces narrow focus on another constituent (Vallduví and Vilkuna 1998; Büring 1999; Molnár 2001; Gyuris 2004).

114

(18)

Chapter Five

[CT Reykjavikiin] Pekka lensi [F Icelandairilla]. ‘To REYKJAVIK, Pekka flew by ICELANDAIR.’

In her analysis of the Finnish example in (18), Molnár (2001) shows that the contrastive topic Reykjavikiin imposes an additional narrow focus on Icelandairilla. The presence of the contrastive topic indicates that there are things other than the one referred to by the contrastive topic about which the same question could be asked, and it is possible that the answer to those questions would be different. Similar facts seem to hold in Romance languages as well as MG. With regard to the former, Arregi (2003), for example, shows that CLLD, which he also takes to involve contrastive topicalization, forces narrow focus on some other part of the sentence, as illustrated in (19), where Juan being a contrastive topic forces narrow focus on la moto. (19)

[CT A Juan], le di [F la moto]. to Juan him I-gave the bike ‘Juan, I gave the bike.’ (Arregi 2003: (5))

Similarly, Alexopoulou (1996) argues that, in MG, when a contrastive topic is present, a contrast is usually implied in the focused part of the sentence as well. The relevant data is presented in (20). (20)

[CT I Ana] pige stin Kriti [F me to plio] ke the Ana-NOM went-3SG to the Crete with the boat and [CT o Yorgos] pige stin Kriti [F me to aeroplano] the Yorgos-NOM went-3SG to the Crete with the aeroplane. ‘Ana travelled to Crete by boat and George travelled tȠ Crete by plane.’

The above requirement of contrastive topics, that an additional element in the sentence bears narrow focus, seems to be supported by intonational facts as well; contrastive topics, which have a fall-rise accent, are typically followed by a constituent which carries the second intonational peak of the sentence (Gyuris 2004) and which is associated with a high tone accent, typical of focused elements (Büring 1999). However, at this point it must be noted that the focus induced by the contrastive topic on the constituents in (19) and (20) need not, but can be, contrastive focus. The constituents la moto in (19) and me to plio and me to aeroplano in (20) are not contrastively focused in the strict definition of the

Yes/No Ellipsis and Polarity Focus in Modern Greek

115

term, according to which contrastive focus signals a correction by a speaker to a previous statement in the preceding context (Drubig 2003), as shown in the Spanish example in (21) below. (21)

A María, la he invitado [CF yo], no Juan. A María, CL-3SG-FEM have-1SG invited I, not Juan. ‘It was me that invited María, not Juan.’

In the next section, I turn to Umbach’s analysis of but-coordinated structures.

3.2 Umbach's (2004, 2005) Approach to But-coordination Following the brief overview of the literature on the Information Structure of a sentence, which includes notions such as contrastive focus and contrastive topic, I will present some aspects of Umbach’s approach (Umbach 2004, 2005) to coordinated structures with but that are relevant to my proposal. Standard accounts of but-coordination point out two important properties; first, that a contrast by but involves conjuncts that are similar in some respects and dissimilar in others ('Mary is tall but Susan is short') and secondly, that but-contrast expresses a denial-of-expectation ('It rains but we will go for a walk'), where the second conjunct of the coordination denies an inference triggered by the first conjunct ('When it rains, people stay at home'). Crucially, this inference is based on our existing world-knowledge. Umbach shows that the former property is not restricted to butcoordination but it is a general characteristic of coordination. Following Lang (1984, 1991), she argues that coordination involves syntactic, semantic and prosodic parallelism between the two conjuncts. With respect to semantic parallelism, Umbach assumes that coordination entails two specific constraints; firstly, that neither of the two conjuncts subsumes each other and secondly, that there must be a common concept subsuming both conjuncts (Umbach 2004). (22) (a) *Susan had a drink, and/but Maria had a vodka. (b) Susan ordered a vodka and/but Maria ordered a martini. (22a) is ungrammatical, because the notion of vodka is subsumed under the notion of drinks; (22a), therefore, violates the first constraint of semantic parallelism. What (22) shows is that two coordinated elements

116

Chapter Five

have to be alternatives with respect to each other; this entails that they must be semantically independent, while at the same time being included in the same semantic superset. With regard to the latter property, denial of expectation, Umbach refutes the idea that the inference triggered by the first conjunct is based on world knowledge. She proposes that this expectation is induced by the use of but in conjunction to the implicit or explicit question which a but-coordination sentence would be an appropriate answer to. Her argument is based on two novel observations about but-coordination. The first is that but-coordination is focus sensitive; depending on the focus of the but-conjunct, different contrasts can be established; (23), for example, where the verb-phrase is focused, raises the expectation that the VP of the antecedent clause will be focused whereas (24), where focus is on the subject 'Mary', raises the expectation that the subject of the antecedent clause will be focused. (23)

...but Mary has [F read ‘Pride and Prejudice’].

(24)

...but [F Mary] has read ‘Pride and Prejudice’.

Therefore, according to Umbach (2005), an analysis of but needs to be based on the Information Structure of the conjuncts. Secondly, Umbach shows that a but-coordination is an appropriate answer to an implicit question, which will give rise to 'a yes... but no' reply. This means that, in contrast to and-coordination, if but-coordination is to be an appropriate answer to a question with two conjuncts, one needs to be a confirmation and the other a denial. Umbach calls this the deny-confirm condition. (25) (a) (b) (c) (d) (e) (f)

Did John clean up his room and wash the dishes? YESJohn cleaned up his room and YESwashed the dishes #YESJohn cleaned up his room, but YEShe washed the dishes #NOJohn didn’t clean up his room, but NOhe didn’t wash the dishes YESJohn cleaned up his room but NOhe didn’t wash the dishes. YESJohn cleaned up his room but NOhe skipped the dishes. NOJohn didn’t clean up his room, butYES he did wash the dishes

The sentences in (25d - f) are felicitous, because this deny-confirm condition is respected; on the other hand, (25b, c) which violate this constraint are not acceptable. What the above examples show is that a conjunct introduced by but first adds an alternative to those under

Yes/No Ellipsis and Polarity Focus in Modern Greek

117

discussion (the expectation of washing the dishes) and then, this alternative results in a false proposition when combined with the background, therefore requiring negation. Bringing together the aforementioned observations, Umbach's basic claim is that due to its focus sensitivity, but is associated with a focused constituent in the second conjunct which provides the expected alternative. This constituent corresponds to a sister alternative in the first conjunct. The two constituents form the set of alternatives under discussion referred to in the question. The meaning of but consists in forcing a denial condition: the proposition resulting from replacing the sister alternative in the first conjunct with the expected alternative is false. In the next section, I turn to yes/no ellipsis and show how Umbach's proposal can be implemented in a yes/no ellipsis structure.

3.3 Contrast in Yes/No Ellipsis Based on what was discussed in the previous section, but is focus-sensitive and it can create different types of contrasted pairs, depending on which constituent in the second conjunct bears the focus accent. The different possibilities are illustrated in (26) below. (26) (a) Mary [F went to Paris] but she [F skipped London]. (b) [F John] visited the Louvre but [F Mike] didn't. (c) John [F did] visit the Louvre but Mike [F did not]. (27) [F I] visited [F the Louvre] and/but [F Mike] visited [F the Musée d'Orsay] In (26a) the verb phrases are contrasted, in (26b) we have a contrast of individuals whereas in (26c) the contrast is on verum elements (focus of the verb phrase/predicate). With respect to the notions of rheme and theme, we can see that the focus of the second conjunct doesn't need to be on the new information (the rheme part of the sentence); in (27), whereas there is a double contrast, the themes (I and Mike) are also focused. Turning to the yes/no ellipsis data, my suggestion is that Umbach's analysis about the meaning of but can be applied to the Greek coordinator ala. I argue that in a yes/no ellipsis sentence, the pair of contrasted alternatives involves the themes of the two conjuncts.

118

(28)

Chapter Five

To cd to thimithika, ala ta vivlia oxi. the cd itCL remembered- 1SG, but the books not 'As for the cd, I remembered it, but as for the books, I didn't.'

The ala-sentence in (28) would be an appropriate answer to either an explicit or implicit question along the lines of (29). (29)

Did you remember to bring the cd? And what about the books? Did you remember to bring those as well?

Consistent with Umbach's approach, the use of ala forces the denial condition; ala adds an expected alternative to those under consideration (I remembered to bring the books) but when this alternative is combined with the background, it results to a false proposition. This demands the presence of an explicit negation in the second conjunct. Observe that a yes/no ellipsis with the presence of and ('ke') would be ungrammatical, as in (30). (30)

To cd to thimithika, ke ta vivlia oxi. the cd itCL remembered- 1SG, and the books not 'As for the cd, I remembered it, and as for the books, I didn't.'

Finally, based on the idea that contrastive topics impose a second semantic focus in the sentence, as discussed in section 3.1., I argue that the negative particle will also bear a semantic focus (verum focus). This is consistent with Gyuris’s (2004) proposal that the semantic focus imposed by the contrastive topic can be either identificational focus (Kiss 1998), a quantifier position in the preverbal field, a verum focus or a negative particle. Summarising then, what I have tentatively suggested in this section is that yes/no ellipsis is a standard case of but-coordination, where the second conjunct first introduces an alternative with respect to a background and then this alternative is denied with respect to the first conjunct (exclusion of additional alternatives). The two alternatives bear the Information Structure role of contrastive topic. However, the full implementation of this proposal is the subject of further investigation.

Yes/No Ellipsis and Polarity Focus in Modern Greek

119

References Alexopoulou, T. 1996. Contrastive focus and contrastive topics in Greek. Paper presented at Proceedings of the Edinburgh Linguistics Department Conference '96, Available at: . Anagnostopoulou, E. 1997. Clitic Left Dislocation and Contrastive Left Dislocation. In E. Anagnostopoulou, H. Riemsdijk van and F. Zwarts (eds), Materials in Left Dislocation, pp. 151-192. Amsterdam: John Benjamins. Baltazani, M. 2006. Intonation and pragmatic interpretation of negation in Greek. Journal of Pragmatics 38, 1658-1676. Benincà, P., & C. Poletto. 2004. Topic, focus and V2: Defining the CP sublayers. In L. Rizzi (ed) The structure of CP and IP: The Cartography of Syntactic Structures, vol. 2. Oxford: Oxford University Press. Büring, D. 1999. Topic. In P. Bosch and R. van der Sandt (eds), Focus-Linguistic, Cognitive, and Computational Perspectives, pp. 142-165. Cambridge: Cambridge University Press. —. 2005. Intonation, Semantics and Information Structure. Available at: . —. 2003. On D-trees, Beans, and B-Accents. Linguistics & Philosophy 26, 511-545. Busquets, J. 2006. Stripping vs. VP-Ellipsis in Catalan: What is deleted and when? Probus 18, 159-187. Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Depiante, M.. 2000. The syntax of deep and surface anaphora: A study of Null Complement Anaphora and Stripping/Bare Argument Ellipsis, PhD dissertation, University of Connecticut. Drubig, H. B. 2003. Towards a typology of focus and focus constructions. Linguistics 41, 1-50. Erteschik-Shir, N. 2007. Information Structure. Oxford: Oxford University Press. Gyuris, B. 2004. A New Approach to the Scope of Contrastive Topics. ZAS Papers in Linguistics 35, 133-156. Holton, D., Mackridge, P., & I. Philippaki-Warburton. 1997. Greek: a comprehensive grammar of the modern language. London: Routledge. Kiss, K. É. 1998. Identificational versus Information focus. Language 74, 245-273. Kolokonte, M. 2008. Bare Argument Ellipsis and Information Structure, PhD dissertation, University of Newcastle.

120

Chapter Five

Laka, I. 1990. Negation in syntax: On the nature of functional categories and projections. PhD dissertation, MIT. Merchant, J. 2003. Remarks on Stripping. Ms. University of Chicago. —. 2004. Fragments and ellipsis. Linguistics and Philosophy 27, 661-738. Molnár, V. 2001. Contrast from a Contrastive Perspective. In ESSLLI 2001, Workshop on Information Structure, Discourse Structure and Discourse Semantics. Available at: . Rizzi, L. 1997. On the fine structure of the left periphery. In L. Haegeman (ed), Elements of Grammar, pp. 281-337. Dordrecht: Kluwer. Rooth, M. 1992. A Theory of Focus Interpretation. Natural Language Semantics 1, 75-116 Tsiplakou, S. 1998. Focus in Greek: Its structure and interpretation. PhD dissertation, University of London. Umbach, C. 2004. On the notion of contrast in information structure and discourse structure. Journal of Semantics 21, 155-175. —. 2005. Contrast and information structure: a focus-based analysis of but. Linguistics 43, 207-232. Valiouli, M. 1994. Anaphora, Agreement, and Right Dislocation in Greek. Journal of Semantics 11, 55-82. Vallduví, E., & M. Vilkuna. 1998. On rheme and kontrast. In P. W. Culicover and L. McNally (eds), Syntax and semantics, vol. 29: The limits of syntax, pp. 79-108. San Diego: Academic Press. Winkler, S. 2005. Ellipsis and Focus in Generative Grammar. Berlin/ New York: Mouton de Gruyter.

CHAPTER SIX THE DISTRIBUTION OF NON-OBLIGATORY CONTROL AND ITS + HUMAN RESTRICTION VICTORIA JANKE

1. Introduction A central aim in Janke (2007) was to represent obligatory control (OC) without recourse to PRO. In this respect, it followed a line of works that have demonstrated purposefully (see Bresnan 1978, 1982; Brame 1977; Chierchia 1984; Evers 1988; Manzini and Roussou 2000) or incidentally (Borer 1989; Landau 2000), the irrelevance of PRO to the control relation. Another aim was to reduce OC to a modified form of binding which would retain the essence of the binding relation without invoking a direct dependency between the lexical elements involved. It was shown that the differences between PRO and lexical anaphors (as documented in Lasnik (1992)) did not translate necessarily into differences between control and binding per se, but referred to the elements involved in these relations, rather than the relations themselves. Dispensing with PRO but retaining the relation avoided problems associated with reduction in this direction. This reduction was implemented by analysing control as a binding relation between ș-roles. Rather than positing an interpretative link between a PRO-subject and matrix argument, as in (1a), it linked properties of ș-roles by unifying the interpretative properties of the infinitive verb’s ș-role with those of the ș-role assigned to the matrix verb’s controlling argument, as in (1b). This allowed the subject properties of the controlled infinitival to be retained by a path created by the external ș-role introduced by the infinitive verb. In this way, PRO was dispensed with, without losing the generalisations for which it was introduced.

122

Chapter Six

(1)

Since interpretation in Non-Obligatory Control (NOC) structures is not pinned to a syntactic argument, NOC cannot be reduced to binding. But the redundancy of PRO to OC extends to NOC, so this paper will develop the PRO-free approach introduced for OC for application to NOC. The system of ș-role percolation devised in Neeleman and van de Koot (2002), and developed for OC in Janke (2007, 2008), will be supplemented by two extra-syntactic rules, which provide the external ș-role of NOC-clauses with the appropriate amount of semantic content for both their generic and specific interpretations. The present paper takes the motivation for a PRO-free analysis of control for granted1. It starts by delineating the properties which distinguish OC from NOC, before moving on to two types of interpretation possible in NOC, namely generic and discourse-regulated ones. Section 4 sets out the main points of the system of ș-role copying devised in Neeleman and van de Koot (2002), and developed for control in Janke (2007, 2008). With the copying mechanism clear, section 5 introduces a structural restriction on this mechanism, thereby preventing it from over-generating. The focus of section 6 is the licensing of genericity. Two interpretative rules are formulated. One suffices for generically interpreted clauses and captures the +human restriction evident in NOC. The other, regulated by Accessibility (as developed in Ariel (1988)), is a discourse metric, which steers the interpreter to NOC’s specific interpretations. Finally, these interpretive rules are tested on independent structures, namely the linked reading effect (Lebeaux 1984) and null generic objects (Rizzi 1986).

2. Distinguishing NOC from OC 2.1 Properties of OC The understood subject of OC clauses requires a theta-marked argument as its antecedent, which must be local, c-commanding and unique: (2) a. Billi tried [PROi to organise himself] b. *Iti was tried [PROi to organise himself] c. *Billi thinks it was tried [PROi to organise himself]

The Distribution of Non-Obligatory Control

123

d. *Bill’si aunt tried [PROi to organise himself] e. *Billi asked Benj [PROi&j to kiss Bobby behind the bike shed] Example (2b) shows that the antecedent must be a theta-role-bearing argument, whilst (c) demonstrates locality. Example (d) establishes c-command is operative and (e) shows the ban on split antecedents.2

2.2 Properties of NOC The restrictions operative in OC do not regulate NOC. NOC relations may conform to some of them, but they do not exhibit all and in some instances lack them entirely (see Williams 1980). These criteria give us the following candidates for NOC: infinitival subject clauses, as in (3a), controlled interrogative complements, shown in (b), verbal gerunds as in (c), control with implicit arguments, displayed in (d and e) and long-distance control, in example (f). (3) a . b. c. d. e. f.

PRO To go to the lecture drunk wasn’t one of your best ideas Peter knows how PRO to fix the head gasket PRO Walking back home yesterday, a brick fell on my head It is fun PRO to dance (It is fun for x, for x to dance) PRO To finish off one sentence in peace would be nice (for x) Peter said that PRO to get there on time would be very difficult

Infinitival subjects, verbal gerunds and implicit control constructions have no structurally represented antecedent. The interrogative complement has a non-local argument in the super-ordinate clause, but this is not the antecedent for the implicit subject, which carries a generic interpretation. Long-distance control breaks locality, but also tolerates split antecedents, separating it further from OC: (4)

Peter said to Rita that PRO to get there on time would be very difficult for them

The first of our aims is to account for the distributional differences between OC and NOC: I.

Our theory should provide an account for why OC is subject to four conditions which NOC is not: its antecedent is obligatory, unique, local and must c-command the understood subject.

124

Chapter Six

In contrast to the absence of restrictions that mainly characterise NOC, this relation is subject to a semantic restriction that OC is not. As noted in Manzini (1983), in NOC, the understood subject must be interpreted as human. But in OC it is possible to have non-human subjects acting as antecedents: (5) a. This booki promises PROi to be a great read b. The filmi tries PROi to convey a humane side to the dictator. c. The farmer needed the cropi PROi to exceed last year’s quota. NOC does not share this option. In (6) the verb used does not itself force a human interpretation of the infinitival subject, but it must nevertheless be human. Despite the inclusion of a potential antecedent (the book) and that no human is mentioned, the implicit subject is interpreted as a ‘generic human’, where its most accurate paraphrase is (6b), not (c): (6) a. This book is a great example of how PRO to entertain children. b. This book is a great example of how one can entertain children c. This book is a great example of how books can entertain children The semantic criterion that NOC’s understood subjects must meet then, says the following:3 (7)

Understood subjects of NOC must be +human

The +human characteristic of NOC constructions is something a theory of control must derive, and hence forms our second aim: II.

Our theory should account for why the understood subject in NOC is human.

In addition to this semantic restriction on the understood subject in NOC, such clauses carry independent tense: (8) a. [Going to the lecture drunk today] will upset your mother next week b . [Walking home yesterday] will guarantee you a lift today c. Peter said yesterday that [to get there on time today] will be very difficult for him d. John1 knew already what [to buy him1 in London tomorrow]

The Distribution of Non-Obligatory Control

125

Independent tense becomes important when we formulate the LF-rule operative on these clauses’ understood subjects. But first some space is dedicated to the different types of interpretations available in NOC, which will anticipate the interpretative rules introduced at the end of the section.

3. Generic and Discourse-Specific Interpretations in NOC There are two different types of interpretation possible in NOC constructions, generic and discourse-specific. Each will be regulated by a different rule: one where the understood subject is interpreted generically and the other when it has a specific interpretation. Returning to (3b and d), we can see that the interpretative subject is understood generically. Generic infinitival sentences do not express properties of any specific events or individuals but refer to generalisations over events, and to custom-like regularities: (9) (10)

It is fun to dance Dancing the tango is fun

For these generic interpretations if the understood subject were attributed with a +human specification by an LF-interpretative rule, this would provide the minimum semantic content for it to function as an argument. A (first) formulation of this rule is given in (11) below. (11)

LF-Rule: The external argument in NOC clauses is interpreted as [+human] (Formulation I)

In both the sentences in (9) and (10), the interpretation is such that ‘dancing is fun for people in general’. But adding a specific time reference steers away from the generic interpretation, and creates an expectation for a conversational cue to guide us to a referent for the dancer(s): (12) a. It was fun to dance yesterday b. Dancing the tango on Saturday night was fun The interrogative structure in (3b) repeated below as (13a) also has a generic interpretation that moves towards a referential reading with the addition of an episodic marker, as in (13b). And the addition of a reflexive, as in (14), goes further, by forcing an interpretation in which the understood subject is equated with the matrix subject.

126

Chapter Six

(13)

a. Peter knew how to fix the head gasket =Peter knew how one could/should fix the head gasket b. Peter knew how to fix the head gasket yesterday

(14)

John wondered whether to talk to himself in public

For both implicit and interrogative examples, then, a more specific rule than that in (11) must be made available, as +human does not do the possible interpretations justice. This is also so for examples which omit a generic reading, and require a specific antecedent, whether this is syntactically represented, but not locally, as in (3f) or simply inferred, as in (3a,c & e). For these, it is argued that referential candidacy is determined by the degree of saliency of prospective antecedents. The notion of Accessibility, as developed in Ariel (1988), determines how the +human reference supplied by the LF-interpretative rule is supplemented. Briefly, a separate discourse-governed rule takes the most accessible antecedent (where ‘accessible’ is defined below) in the discourse and link its reference to that of the +human argument: (15)

Discourse-Rule: An underspecified [+human] argument can only be linked to a highly accessible antecedent

Sections 5 and 6 develop the application of and restriction on these two extra-syntactic rules so that they cover all the NOC examples, whilst being blocked from over-generating into OC cases. In the immediate section the theta-based mechanism developed for OC in Janke (2007) is set out, which will provide the bridge to demonstrate the point at which the interpretative rules take over from the syntactic mechanism. It is in this way that the relations’ differing distributions will be sourced.

4. Syntactically Regulated Control 4.1 Non-atomic ș-roles On the present view of control, itself a development of Samek-Lodovici (2003), a ș-role is a non-atomic construct, whose components can be distinguished on the basis of their differing contributions to argumenthood. These include an argument component, whose task is to ensure that formal requirements such as structural position are met, and an interpretative component, whose task is to ensure that interpretative relations are secured. I label these A and B respectively.

The Distribution of Non-Obligatory Control

(16)

A: B:

127

Formal Licensing Component Interpretative Component

A full representation of a ș-role includes both of these components, and in most circumstances their sub-parts go unnoticed, as they remain together as a complex. (17)

Ĭ-role = [ș A, B]

Ĭ-roles are assigned to arguments via a system of recursive upward copying and one-step downward application, as developed in Neeleman and van de Koot (2002). A transitive verb’s requirement for a subject and an object is encoded by its two ș-roles (one external, the other internal),4 which percolate to the node immediately above an argument with the relevant property. Under immediate domination, the verb’s requirements have been met. The ș-role assigned to the object in (18) stops percolating at VP, where it immediately dominates the direct object, whilst the external role continues until TP, where it immediately dominates the subject: (18)

If ș-roles are non-atomic, it should be possible to distinguish the differing roles of A and B, which collectively comprise ș-role assignment. One might also expect to find instances of A and B operating independently. Although a full explication exceeds the space restrictions of this paper, the examples in (19) and (20), illustrate their differing functions. Component A is concerned with a verb’s adicity, whereas B is a syntactic representation of the predicate’s argument variable. So when B’s requirements are met, the argument variable representing the predicate’s semantics receives a value. In (19), B is linked to the argument variable x and applied to the DP, with the result that the DP is interpreted as the argument of the verb. Component A,

128

Chapter Six

however, ensures that the verb’s adicity and the number of DP-arguments projected in the structure correspond. Its application to an argument morphologically marked as such licenses that argument position5: (19)

(20)

4.2 Control as Binding But it is the B component which regulates control, being concerned with interpretation as it is. The example (22a) demonstrates how this component regulates relations independently of A, by looking at the asymmetric relation between a reflexive and its antecedent. A first thing to note is that a reflexive lacks referential properties itself, requiring an antecedent that enables it to be interpreted (a quantifier or referential category), and since this is a syntactically regulated requirement, this must be registered on the reflexive. By introducing just the selectional requirement, B, which essentially says ‘bind me’, the lexical encoding of the variable of the anaphor is represented. Via the same mechanism of percolation, the reflexive links to its antecedent: B percolates upwards from the anaphor that introduces it until it immediately dominates an argument that provides the variable with a value. Note that this is independent of regular ș-role assignment, also in operation, which for ease of exposition is represented on a separate tree in (b).

The Distribution of Non-Obligatory Control

129

(22)

Similar to reflexives, OC infinitival clauses lack the referential properties necessary for interpretation. Their implicit subject’s reference is determined by a designated argument in the matrix clause, either a subject or an object: (23) (a) Billi hoped to wini

(b)

Bill ordered Beni to dancei

But unlike reflexives, there is no audible construct, which begs the question of where this selectional requirement should be encoded. The argument developed in Janke (2007) is that the external ș-role of infinitival complement’s verb encodes this requirement, and is the source of the subject properties apparent in control complements. And within this external ș-role, it is specifically the B component that forges the interpretative dependency between the infinitival clause and the matrix antecedent: the control verb selects for an infinitival CP of the control type (where ‘type’ is defined below), and this ‘pull’ from the control verb licenses the B component of the external ș-role to detach from A and percolate in isolation beyond CP to an argument that can provide it with a value. Creating a thematic path, B percolates to a node where it immediately dominates its controller in the matrix clause. Under this configuration, OC results:

130

Chapter Six

(24)

Note that A is not licensed to leave CP, allowing us to capitalise on a distinction between predication on the one hand and OC and reflexive binding on the other. Unlike OC and reflexive binding, predication cannot cross CP boundaries: (25) a. *Bill seemed that Ben met Mary drunk yesterday b. John arranged to win c. I arranged for myself to win ((18, 19 & 22) in Janke 2008) With A and B as separable components, the problematic status of (25a) lies with A: the A component of the ș-role cannot cross CP, thereby prohibiting predication across CPs. In contrast, the B component can cross CP, with the result that both binding and control are permitted across this boundary. Their both escaping this restriction lends support to the mechanism adopted here, which regulates both relations with the same component.

4.3. Regulation of ș-role Decomposition (25) shows that predication is more restricted than OC and binding. The tighter restriction on predication becomes relevant to any regulation implemented for ș-role decomposition, which must prohibit detachment of ș-role components in the former environments but not the latter. Kiparsky’s Elsewhere Principle (Kiparsky 1973), which gives precedence to a more specific relation - predication in this instance -, will ensure that the dislocation of B from A (i.e. binding and OC) only occurs when complete

The Distribution of Non-Obligatory Control

131

ș-role percolation (namely predication) is barred: (26)

Given two competing rules, R1 and R2, which operate in two domains of application, D1 and D2, such that the D1 forms a sub-set of D2, then R1 will block R2 from applying in D1.

Ĭ-role decomposition makes it possible to distinguish two key components that contribute to argumenthood: interpretation and case. It is the interpretation of the infinitival subject that is relevant to the control relation, not its case, and ș-role decomposition makes this representation possible, by retaining the thematic path of the external ș-role and linking its interpretative component to the designated argument of the control verb (thereby distinguishing this approach empirically from one based on raising as in Hornstein 2001). In the next section, I return to NOC, demonstrating where the syntactic analysis is superseded by the non-syntactic one, which in turn, helps to derive the differing distributions of these relations, as per Aim I.

5. Structural Restrictions on Copying: Deriving the OC/NOC Divide The syntactic mechanism developed for OC does not extend to NOC, since in NOC there is (often) no structurally represented antecedent at all. The example of implicit control below shows that however far the ș-role of the verb percolates 6 , there is nothing in the structure to function as a semantic argument, rendering the copy procedure futile. As such the ș-role remains unassigned. (27)

132

Chapter Six

But NOC and OC do have their introduction of an external ș-role in common, and the principle of Full Interpretation requires the interpretative component of this ș-role (B) to have the content of a semantic argument. The two control types differ in how this is achieved. In the OC-relation, saturation of the interpretative component occurs syntactically via the copying mechanism, where B separates from A and percolates to an antecedent in the matrix clause. In the NOC-relation, this is achieved by semantic means. B is specified as [+human] at the LF-interface, a specification which can be supplemented by a discourse rule directing it towards a highly accessible antecedent (detailed below). There are two ways then, of specifying how the external ș-role is semantically interpreted: (28)

(I) Specify B as human (II) Copy B to (an antecedent in) the super-ordinate clause

(II) is the mechanism of percolation at work as discussed above. (I) needs regulation. At this point, I assume that tensed TP forms a choice point at which (I) above becomes available, with a view to refining this later. The option is recursive, becoming available at every tensed TP. (29)

At tensed TP: B can be specified as + human (First Formulation)

5.1 Deriving the OC Distribution We have seen where option (I) in (28) becomes available, but the circumstances governing its availability must be restricted to prevent any over-generation. OC resulted when the B separated from A at CP. Separation was regulated by the Elsewhere Principle. Through its preference for whole ș-role percolation, this principle ensures that copying of B in isolation only occurs when whole ș-role percolation is impossible, such as across infinitival CPs.

The Distribution of Non-Obligatory Control

133

(30)

But although Elsewhere makes separation of B from A possible, something extra needs to be said about how and why the ș-role goes beyond TP in the first place. For Elsewhere to kick in, it must be that (29) does not occur in OC structures. Given that OC constructions contain control verbs, which demand a certain type of complement, we might expect the divergence in terms of percolation beyond TP to fall out from this difference. The claim here is that the OC head syntactically selects for a CP with an unassigned B. (31)

OC-type head selects for a CP with an unassigned B

The structures for which such syntactic selection occurs include subject-control structures, object-control structures and some headed by adjuncts: (32) (33) (34)

Paul hoped to sleep well that night Peter persuaded Paul to get a good night’s sleep Peter crossed the road while speaking to Paul

In examples (32) and (33), the matrix verb selects for an unassigned B in its CP-complement, whereas in example (34), it is the head of the adjunct that selects for the unassigned B. This selection by the controlling head in the super-ordinate clause for a B in its sister node provides the pull for the copying of the unassigned B out of the embedded clause. The tree below demonstrates the point at which Elsewhere allows for the separation of B from A, and its application to the DP.

134

Chapter Six

(35)

If the option to specify B as human, were available at the TP* in (35), the saturated ș-role would not percolate further. This would ignore the selectional requirement of the controlling head for a CP with an unassigned B: (36)

(29) does not require all constructions with verbs that select complements to be OC structures. In (37) below, the infinitival clause is a complement selected by suggest, but this is not an OC structure, as the absence of any structurally represented antecedent attests: (37)

John suggested to leave/leaving at once

The Distribution of Non-Obligatory Control

135

In the right context, this can either mean that I leave, that you leave or that we leave, but John need not necessarily be included.

5.2 Deriving NOC Having carved out the distribution of OC, we can now set this against NOC, which in the main, characterises those instances of control which are not selected by the matrix control verb. The only exception is the class of interrogatives, whose complements are selected, but it will be argued later that copying in the sense of (31) above is absent. Essentially, we will see that the complement of a wh-expression must be an open proposition. Aside from interrogatives, NOC structures are not complements, so their B-components are not selected for by control heads. Deferring discussion of interrogatives until last, these structures included infinitival subject clauses (38), verbal gerunds (39), control by implicit arguments (40) and long-distance control (41): (38) (39) (40) (41)

[To go to the lecture drunk] wasn’t one of your best ideas [Walking back home yesterday], a brick fell on my head It is fun [to dance] Paul said that [to get there on time] would be very difficult for him]]

In all these examples, the ș-role receives an interpretation within its clause via the option in (29). On economical grounds, this option available in tensed TPs should be taken. Copying derives a more complex structure, where what determines complexity includes the number of copies of ș-roles in a given structure. A structure with fewer ș-roles is more economical, making (I) the preferred option. Examples (38) and (39) are of a clause-initial infinitival subject and gerundive subject respectively, so their NOC characteristics follow straightforwardly. In both cases there is no pull from a control verb to extract the B-component from of its clause, since the verb does not select for a clause with an unassigned B. In the absence of selection, the more economical option is preferred and the external ș-role of each clause is subject to the LF-interpretative rule that specifies the argument variable as [+human]. 7 Examples (40) and (41) also involve subject-clauses. The former is an example of extra-position, the latter that of an embedded subject within a complement, so again the fact that their external role is interpreted within the clause follows straightforwardly. I now return to interrogatives and how to reconcile them with the rule in (29). We saw in (31) that OC verbs select for a CP with an unassigned

136

Chapter Six

B-component. In virtue of the unassigned B component copying to the super-ordinate clause, the complements of OC verbs are predicates. But the infinitival verb in NOC has an external ș-role with a +human specification. Absence of copying, and assignment of a +human specification makes the NOC infinitival clause a proposition. But what of interrogative complements? These are selected for by their matrix verbs, but the wh-feature introduced by interrogatives must have a proposition with only one open position in its scope, namely that of the bound variable: (42)

+wh scopes over a proposition that contains a variable bound by the wh

Interrogative complements must be propositions since questions are only formed from propositions. A good precedent for this assumption can be found in Karttunen (1977). In this work, a question is formed in two steps. The first is that of constructing a "proto-question", which has an operator attached to the front of a proposition, whereas in the second, this proto-question is manipulated to make a real question. But crucial for present purposes, is that one can only make a proto-question out of a proposition, so everything which ends up as a question starts out as a proposition (see especially Hamblin (1973) and Karttunen (1977)). Even in Groenendijk and Stokhof (1989) for whom interrogatives do not always denote propositions (not always being of type t), interrogatives are always of a type made from s and t, that is, nothing which indicates any missing arguments. So there is a conflict of interest in terms of what a control verb selects for syntactically and what it requires semantically. On the one hand it demands a CP with an unassigned B, which makes the complement a predicate, and on the other, the wh-feature of the interrogative requires its complement be a proposition. To attribute C’/TP with both would result in a contradiction, since the infinitival cannot be simultaneously a predicate and a proposition: (43)

* [ unassigned B, +wh] ĺ [predicate and proposition]

The alternative pursued here is that verbs with interrogative complements, such as know in (13a), select a CP, but not with an unassigned B. Without the pull from the matrix verb, the B-component of the external role is not copied beyond TP, but specified as +human, which fits in well with their generic character and also falls in place with their NOC classification.

The Distribution of Non-Obligatory Control

137

There is a crucial respect in which OC and NOC differ. OC comprises the controlling predicate’s syntactic selection of a CP with an unassigned B, and the copying of this unassigned B to the controlling argument in the super-ordinate clause. This binds the infinitival to the designated controller. It is the absence of this first component in NOC that is responsible for its different distribution: Infinitival subjects and gerunds have no higher predicate that select for a B-component and interrogative complements do not make a syntactic selection for a CP with an unassigned B. With this we arrive at the first of our aims set out in section 2. I

Our theory should provide an account for why OC is subject to four conditions, which NOC is not: its antecedent is obligatory, must be unique, local and must c-command its understood subject.

In the next section, I turn to the implementation of the LF-interpretative rule introduced in (15), which applies in all NOC environments, including those that have a generic interpretation, those whose interpretation comes from an antecedent in the structure, and also those whose reference is determined by a specific inferred argument. Starting with generically interpreted clauses first, we will see that although the +human specification afforded by the rule is exactly the right amount of information for their interpretation, these complements’ genericity must also be licensed at the semantic level. I look first at interrogative complements, and then at implicit constructions.

6. From Generic Readings to Specific Readings 6.1 Interrogatives: Syntactic vs. Semantic Selection In the main, the understood subjects of interrogative complements receive a generic interpretation, their nearest overt counterpart being one. But in addition to this understood subject having a reference distinct from the overt argument in the matrix, interrogatives have a modal feel to them, as demonstrated in Bhatt and Izvorski (1998), where they are matched to their nearest paraphrases: (Bhatt and Izvorski’s 36 and 37): (44) a. a’. b. b’.

Matt knows how to solve the problem Matt knows how one could/should solve the problem Daniel knows what to do with one’s life8 Daniel knows what one should/could do with one’s life

138

Chapter Six

Falling under NOC, the representation of interrogative complements proceeds as shown in (45). Their external ș-role is copied up to TP, and in virtue of the LF-rule in (22), it is attributed with a +human interpretation. (45)

Through this +human specification, a generic interpretation becomes possible. Having been specified, no further copying of the ș-role is motivated so it remains in the infinitival complement. Failure to copy the ș-role further, makes this complement a proposition, as required by the wh-feature. The generic operator, introduced by the hidden deontic modal, is of semantic consequence: (46)

Peter knows [how to fix the head-gasket] ĺ scope of genericity

Despite the infinitival being a complement of the interrogative verb, the selection is not of the OC type. That is these verbs do select for a CP, but not a CP with an unassigned B. Such syntactic selection would render the complement a predicate, which would contravene the semantic requirement for the complement of the wh to be proposition.

6.2 Licensing Implicit Readings I return now to the example of ‘implicit’ control, in which an inferred generic argument is felt:

The Distribution of Non-Obligatory Control

(47)

139

It is fun to dance

Bhatt and Izvorski (1998) would claim that fun has an implicit argument, itself generic and it is this that controls, and hence determines, the interpretation of the implicit subject. In fact, they make the stronger claim that an arbitrarily interpreted PRO is always dependent on a generic implicit argument. On this account, two factors predict whether PRO can be interpreted generically: an implicit antecedent and that antecedent’s genericity, neither on its own being sufficient. It is not clear whether the implicit argument on which PRO is dependent for its interpretation in their proposal is posited as a structurally represented one: (48)

PROarb ļ Generic implicit argument in the immediately higher predicate (Bhatt and Izvorski’s (19))

But this formulation appears to be too strong, whether or not it refers to an implicit generic argument at the syntactic or the semantic level: (49)

Bill: Ben: Bill:

The headmaster just phoned What did he say? He said to introduce yourself to the class before he arrives. (‘He said to me that you should introduce yourself to the class before he arrives’)

If in the above example, the matrix verb said has an implicit argument, this argument’s reference must be the speaker, Bill, and Bill is not the antecedent of the subject of the infinitival clause, which is understood as the addressee, namely Ben. So the claim that arbitrary PRO is always referentially dependent on an implicit controller in the immediately higher predicate is too strong. A weaker statement that replaces the bi-conditional with a conditional can perhaps be upheld: (50)

If there is an implicit argument then PRO co-refers with that argument

Thus amended, it accounts for the interpretation in (51). (51)

To dance is fun

140

Chapter Six

Note that examples such as (52) below, set against the indexical predicate in (51), do not prove that PROarb always has an implicit controller: (52) *To dance is certain/likely/sure Certain and such predicates express the extent to which something is true or false, whereas dance expresses an activity. Since activities don’t have truth values it is expected that the latter cannot be predicated of the former, making the example orthogonal to the issue of implicit arguments. If we use an adjective which says nothing about truth, or a DP, that predicate’s combination with to dance is felicitous: (53) a. To dance in tap shoes is noisy b. To dance drunk is a messy affair The adjective noisy also has no implicit argument, yet the sentence still carries a generic reading. So an account of generically understood subjects in infinitivals does not necessarily depend on the presence of a generic implicit argument. The weaker claim here is that the ș-role’s +human specification enables the generic interpretation. And both of the examples below allow for a generic interpretation: (54) a. It is fun to dance [For people in general] [it is in general fun to dance] b. Peter knew how to fix the head gasket Peter knew how one should fix the head gasket. So the generic interpretation is not necessarily due to another generic implicit argument, but does it depend on anything else? If we look again at the example in (54a), we can see a double layer of genericity, as indicated by the brackets. The existence of two different sources of genericity is revealed by examining its possible interpretations. First, note that generics differ from universals in that the former allow exceptions, where as the latter do not: (55) a. It is fun to dance b. It is always fun to dance Example (a) above is a true statement even if we can think of an occasion when it isn’t fun to dance. The universal added in (b) however, would not be true if we could think of an occasion on which dancing would be less than pleasant. To test whether (54a), repeated here as (56a), has two

The Distribution of Non-Obligatory Control

141

sources of genericity, we can see if insertion of a universal before people, as in (b), or fun, as in (c), is tolerated. Exceptions are indeed possible in two places, since the interpretation of the sentence in (a) does not entail either (b), (c) or (d) below: (56) a. b. c. d.

For people in general it is in general fun to dance For all people it is in general fun to dance For people in general it is always fun to dance For all people it is always fun to dance

These examples illustrate two independent sources of genericity in such an infinitival: the unassigned ș-role that is attributed with a +human specification and a generic operator, operating at sentence level. But although we have seen that the generic reading of the +human ș-role exists independently, there is reason to believe that this interpretation is still in need of licensing by a generic operator. There are no examples of a generic subject, for example, when the sentence in which it sits is not generic as a whole. Recall that insertion of an episodic marker makes a generic reading of the sentence impossible: (57)

It was fun to dance yesterday  ‘It was fun for people in general to dance yesterday’

What this effectively does is remove the matrix generic operator and with that the availability of a generically understood subject. So it seems that when the whole structure is within the scope of a generic operator, the structure licenses the genericity of the understood subject if that subject’s specification is minimal. The +human specification that follows from the LF-rule, repeated below in (58), qualifies as such. It still remains to show that this rule operates independently of control constructions, an issue to which we return last. (58)

LF-Rule: An unassigned ș-role is interpreted as +human

The external role of the infinitival verb percolates to TP, the point at which the +human specification option becomes available. With this minimal specification, the generic interpretation is made available, but through the generic operator at sentence level, the interpretation is licensed.

142

Chapter Six

(59)

The sparse nature of its specification makes it compatible with a generic reading, but will allow it to be enriched when a particular antecedent suggests itself, for example when an episodic marker such as yesterday is added. The addition of such a marker removes the possibility of a generic reading of the understood subject. If the ș-role’s specification were any more detailed, this specification might clash with those of the particular antecedent with which it might be subsequently linked. This argument is developed further in the next sub-section, where I turn to the discourse metric that fleshes out the interpretation of +human ș-roles in non-generic contexts.

6.3 Discourse-Regulated Antecedents A +human specification renders a ș-role compatible with generic readings, where minimal specification is required, but something more is needed for those instances in which the reference of the inferred subject is concrete: (60) (61) (62)

Peter said that [to get there on time] would be very difficult. Walking back home yesterday, a brick fell on my head To finish off one sentence in peace would be nice

A pragmatically governed rule, which directs an underspecified argument towards a highly salient conversational cue for interpretation, will supplement this minimal specification: (63)

Discourse-Rule: An underspecified [+human] argument can only be co-referential with a highly accessible antecedent

The Distribution of Non-Obligatory Control

143

What makes an antecedent highly accessible can be determined using a metric along the lines of that developed in Ariel (1988), in which it is shown that the weaker a referential dependent, the more salient its antecedent must be. I begin by motivating an extra-syntactic mechanism, before introducing Ariel’s metric. Syntactically speaking, non-anaphoric expressions are free, but in practice something in the discourse restricts them: (64)

As for Billy’s girlfriend, her mother can’t stand h’r

Syntax allows us to construe the reduced pronoun h’r as distinct from the DP Billy’s girlfriend, but the fact that they are understood as co-referential points to something extra-syntactic regulating our use of referential expressions during discourse. Accessibility, as developed in Ariel (1988), in which it is argued that the form/content of a (pro)noun instructs us as to how to link it to an antecedent, allows us to represent this. The basic idea is that the more information a (pro)noun encodes, the less obvious its antecedent needs to be and using these criteria, (pro)nouns are ordered on a hierarchical scale. At one end lie low accessibility markers, such as proper nouns, used by a speaker who is unable assume that their addressee has any prior knowledge about who the speaker is referring to. The more lexical material an accessibility marker includes, the lower on the scale it is, which makes sense given that each addition contributes more information. At the opposite end of the scale, lie high accessibility markers, which are those expressions used when a speaker can assume their addressee has immediate or easy access to their referent. In terms of saliency, 1st and 2nd person make more accessible antecedents than 3rd person, since they refer to people directly involved in the discourse. Zero pronouns are the highest accessibility markers, always requiring highly accessible antecedents. The omitted subject of the diary-drop style of speech, as in (65) is strongly biased towards an interpretation in which it is understood to be the 1st person. Since the speaker of the discourse is the most accessible antecedent, any other interpretations, although compatible with the form of the verb, would be at odds with the contextual cues available and communication would fail. (65) a. Went to the lecture yesterday. Fell asleep. Got a D for the essay. b. #John went to the lecture yesterday. Sue fell asleep. You got a D for your essay.

144

Chapter Six

Given their absence of lexical material, control subjects must be at the top of the accessibility hierarchy. But whereas those of OC pattern with reflexives in that their referential sources are sought syntactically, the inferred subjects of NOC tally better with the restrictions on pronouns: there may or may not be a linguistic antecedent, but if present it need not be local or in any particular configuration. With these distinctions in mind, the following Ariel-based hierarchy of accessibility can be constructed, where (1) indicates markers of high accessibility and (5) low: (66)

1) null pronouns 2) clitics/weak pronouns 3) strong pronouns 4) definite descriptions 5) proper nouns

Factors which grade the referential forms above according to how accessible their antecedents must be include saliency, for example if their antecedent is marked as a topic, competition between referential candidates and distance between a (pro)noun and its referential source: (67) a. Saliency - topicality b. Competition - presence/absence c. Distance -local/inter-sentential/intra-paragraph Application of the principles behind Accessibility can guide interpretation of non-obligatorily controlled clauses in the following way. Recall first that the discourse rule, repeated below, restricts their interpretation to an antecedent that is highly accessible: (68)

Discourse-Rule: An underspecified [+human] argument can only be co-referential with a highly accessible antecedent

NOC-subjects lie at the top end of the scale: they demand their referential sources to be highly salient, free from competitors and if present in the structure, closer than other potential candidates (67). The residual NOC structures for which we must still account are long-distance control, verbal gerunds and non-generic implicit control cases: (69) (70) (71)

Peter said that to get there on time would be very difficult. Walking back home yesterday, a brick fell on my head To finish off one sentence in peace would be nice

The Distribution of Non-Obligatory Control

145

Taking long-distance control first, the understood subject does have one potential antecedent in the super-ordinate clause, marking it as the most accessible antecedent vacuously. There are no other competitors and Peter is the topic of the conversation, so a construal under which the understood subject is interpreted as a third unmentioned party, say Paul, is barred. If, however, Paul is mentioned in the discourse, then distance and salience regain significance: (72)

Paul was confident that he could attend the concert and still make it to the viva on time. Peter said that to get there on time would be very difficult.

Despite there being two competitors for the understood subject, and Peter being closer, the most salient and therefore most obvious antecedent choice is Paul, the topic of the conversation. But the sparse information provided in (72) does not absolutely rule out an interpretation in which both Peter and Paul for example, are the antecedents of the understood subject and this is exactly what is required from a pragmatically driven rule, which should guide rather than direct absolutely. Paul remains the preferred option, but both Peter and Paul together remain possible antecedents. The verbal gerund in (70) can be accounted for similarly. Again, there is an understood subject with no visible lexical features, making 1st or 2nd person, both of whom are present in the discourse to hand, most accessible and hence preferred antecedents. The 1st person possessive pronoun in the following clause, however, provides the crucial cue and we interpret the understood subject as 1st person (73a) rather than 2nd (73b). (73) a. Whilst I was walking back home yesterday, a brick fell on my head b. #Whilst you were walking back home yesterday, a brick fell on my head Lastly, the case of implicit control, where the implicit argument refers to someone specific, also follows from Accessibility Theory without complications: (74)

To finish off one sentence in peace would be nice

The indexical predicate nice has an inferred argument which may or may not be represented structurally, but for present purposes it is enough that it is the only conceivable antecedent for the infinitival’s understood

146

Chapter Six

subject, there being no other competitors. Having illustrated how the combination of the discourse rule in (68) and Ariel’s notion of Accessibility can regulate the interpretation of the unassigned external ș-role, this section ends with evidence for the extra-syntactic nature of this rule. Firstly, recall that the OC relation comprises two key parts: the control verb’s specification for a designated controller and the copying mechanism which carries out this direction, by linking the interpretative components of the two ș-roles. I have argued that NOC lacks this first component entirely, in that there is never a higher predicate that selects an NOC clause in the way that occurs for OC. In the absence of such selection, the discourse rule comes into play, by filling the reference of the underspecified external ș-role. But this discourse-rule should not be able to interfere with the interpretation of OC-subjects, whose reference is regulated syntactically. (77) suggests that is so. First, observe the function of the topic marker, as for, in example (75a) where it marks Peter as the topic. Topic-hood makes Peter a highly salient prospective antecedent so the external ș-role of the infinitival is directed towards it for its reference. If demoted from this salient topic position, however, Peter cannot be the antecedent of the infinitival’s understood subject, as indicated in (b): (75) a. As for Peter, the boss suggested to align himself with the union. b. *As for Peter’s sister, she suggested to align himself with the union. Note that demoting Peter from topic position does not have any effect on an overt pronoun. This is expected as overt pronouns do not require an equally high accessible antecedent, so they do not need to link back to a topic: (76)

As for Peter’s sister, she suggested he align himself with the union.

Topic-hood also has no impact on OC examples, showing that this extra-syntactic referential aid cannot interfere with a relation that is regulated syntactically. In (77), despite John being marked as the topic, it cannot be the antecedent of the infinitival. The control verb’s specification that its complement be predicated of its external argument cannot be overridden: Mary must be the infinitival’s interpreted subject. (77)

*As for Johni, Mary promised him Ĭi to leave

The Distribution of Non-Obligatory Control

147

This section and the one preceding have offered two rules, which collectively have covered the interpretations in NOC structures. In all of these structures, the external role of the infinitival clause is subject to the LF-interpretative rule through which the ș-role receives a +human specification. Generically interpreted clauses, among which are those with implicit arguments not modified by an episodic marker and interrogative complements whose subjects can be paraphrased as one, are subject to this rule alone. Long-distance control cases and implicit control cases that are interpreted specifically, are subject to the additional discourse rule, which supplements the +human specification. The rule, guided by Ariel’s Accessibility, steers interpretation to a highly accessible referential source. In the next section, these rules are extended to two different phenomena. The first is the so-called linked reading effect (see Lebeaux (1984)), and the second is the generic null-object in Italian, as described in Rizzi (1986). This null-object is of potential significance to the LF-rule introduced for understood subjects, since it implies that this rule operates on objects, too. The widening of this rule’s applicability beyond that of understood subjects, takes it in the direction of a general grammatical principle.

7. The Linked Reading Effect and Null Objects 7.1 The Linked-Reading Effect When two NOC clauses combine in the same sentence, their understood subject must usually co-refer. Lebeaux (1984) provides some examples that demonstrate this linked-reading effect:9 (78) a. Ĭ1 making a large profit as a slum landlord requires Ĭ1/*2 exploiting the tenants b. Ĭ1 becoming a movie star involves Ĭ1/*2 being recognised by everyone c. Ĭ1 to know him is Ĭ1/*2 to love him In each of the examples above, the understood subject of embedded verb must be interpreted as identical to the understood subject of the uppermost verb. Despite the strong preference for shared reference, Rizzi (1986) offers the following example in which two arbitrary subjects may refer to different entities: (79)

It is difficult Ĭ1 to hope that Ĭ1/2 winning the race will be easy (Rizzi’s 24d)

Chapter Six

148

It is not easy to produce such examples, in fact, once we factor out those verbs that centre around competitions, such as win, lose etc, the preference for a linked reading is very strong: (80) a. #It’s difficult Ĭ1 to expect that Ĭ2 reading the book will be easy b. #It’s unwise Ĭ1 to hope that Ĭ2 bringing the train in on time will be likely It could be that the example in (79) is more akin to that provided in Williams (1980), where for some speakers there need not be strict identity between the understood subject and its antecedent; it is sufficient that the antecedent include it: (81)

I1 want Ĭ1+ to meet at 6.

(Adapted from Williams’ 68)

Such constructions have more recently been labelled as partial control in Landau (2000). If one thinks of the way in which spectators of sport often include themselves when expressing how their team fared, the absence of an obligatorily linked reading in (79) might only be an apparent one: (82)

Bill: Ben:

How did Man-U do on Saturday? Ah, we were robbed. The referee had it in for us from the start…

Noteworthy in Lebeaux’s original examples, is their generic character. When paraphrased, the nearest overt counterpart to each of the null-subjects is the indefinite pronoun one: (83) a. For one to make a large profit as a slum landlord requires one to exploit the tenants b. For one to become a film star involves one being recognised by everyone c. ?For one to know him is for one to love him The overt generic pronouns above are also unable to refer to different entities, which must question whether the linked-reading effect is an issue peculiar to control. However, application of the LF-rule in (58) to the unassigned external role of each infinitival in Lebeaux’s original examples in (78), will restrict the interpretation of the argument variable represented

The Distribution of Non-Obligatory Control

149

by each ș-role to +human. At this point, with two minimally specified ș-roles, we need only recall the weaker version of Bhatt and Izvorski’s generalisation of implicit control, repeated from (50). (84)

If there is an implicit argument than PRO co-refers with that argument.

Translated into current terms, this would mean that the unassigned ș-role, which has been attributed with the +human specification, will share its reference with another unassigned ș-role if present: (85)

If we provide a reference for the understood subject in the embedded clause in the examples in (80), the generalisation in (84) is no longer relevant. The reference of the understood subject is supplemented, as guided by Accessibility, with the result that it no longer depends on the interpretative subject in the matrix clause for its reference. (86) a. It’s difficult Ĭ1 to expect that Ĭ2 reading the book will be easy for Billy. b. It’s unwise Ĭ1 to hope that Ĭ2 bringing the train in on time will be likely for British Rail. But does this leave Rizzi’s exception to the linked reading effect without an account? Aside from the English example, Rizzi (1986) shows that in Italian it is also possible to have two instances of arbitrary interpretations, whose references are distinct from each other. The example in (87) (a) has the possible interpretation given in (b)10

150

Chapter Six

(87) a. È difficile [ PROarb’ sperare [che il governo possa ‘It is difficult[ PRO to hope [that the government can autorizzare ___arb’’ a PRO vivere così authorise___ [PRO to live like that]]]’ b. ‘It is difficult for x to hope that the government can authorise y to live like that’. (Rizzi 1986 his 25b) But again, note that a paraphrase of (87), which uses the overt counterpart of the null generic pronoun, la gente, may also have an interpretation such that the two pronouns are understood as being distinct from one another. The sentence in (88a) can take (b) as one of its possible interpretations: (88) a. È difficile per la gente sperare che il governo possa is difficult for the people to-hope that the government could autorizzare la gente a vivere cosi' authorize the people at live like-that. 'It is difficult for people to hope that the government could authorise people to live like that.’ b. ‘It is difficult for x to hope that the government could authorise y to live like that’. Marco Tamburelli (pc) provides the following context in which such an interpretation becomes available: we are discussing how birds, unlike humans, are free because they can fly from one country to another without having to carry a passport. When I suggest that voters should only elect a government that includes this as part of their manifesto, you utter the sentence in (a) above, which could mean: It is difficult for people (i.e. voters in general) to hope that the government could authorise people (i.e. human beings) to live like that". The availability of this referential dissociation is important because it again points to the linked reference between the understood subject and object being a problem that is orthogonal to control. Syntactically speaking, the linked-reading effect is accidental and given that the rules of discourse guide us as to their interpretation, the existence of a minority of counter examples is exactly what is expected and desired from a such a rule, which should be set to prefer the norm, yet be sufficiently pliable not to bar the limited amount of deviance from this norm.

The Distribution of Non-Obligatory Control

151

7.2 Null Objects The interface rules proposed here for unassigned external ș-roles have no subject/object asymmetry intrinsic to them, so should be applicable to unassigned internal ș-roles. Rizzi (1986) documents a series of constructions in Italian that look very much like unassigned internal ș-roles. These null-objects exhibit syntactic effects, making it clear that their ș-roles are projected. Implicit objects are found in Italian and English: (89) a. This leads (people) to the following conclusion b. Questo conduce (la gente) all seguente conclusion (90) a. This sign cautions (people) against avalanches b. Questo cartello mette in guardia (la gente) contro le valanghe (Rizzi 1986 his 1 & 2) Rizzi (1986) provides five counts of evidence that the generic null pronoun of Italian is syntactically active, unlike its English counterpart. Here I concentrate on two. Contrary to English, Italian null objects can control (91) and bind (92). (91) a. Questo conduce ___ all seguente conclusione a’. Questo conduce ___ a [PRO concludere quanto segue] ‘This leads ___ to conclude what follows’ b. This leads ___ to the following conclusion b’. *This leads ___[PRO to conclude what follows] (Rizzi’s 6b &d and 8b & d) (92) a. La buona musica riconcilia ___ con se stessi b. *Good music reconciles ___ with oneself (Rizzi’s 11) In Rizzi (1986) it is argued that Italian hosts a null pronoun, which receives ș-role, whereas the null pronoun of English is not projected and the relevant ș-role is saturated lexically. In line with rest of this paper, I formulate the contrast in terms of the ș-role: in Italian the ș-role projects and in English it does not. In the Italian null-object construction, the object properties are interpreted in terms of the internal ș-role of the matrix verb. This internal role controls the reference of the understood generic subject in (91a), and the antecedent of the generic reflexive in (92).

152

Chapter Six

(93)

No null-object projects, so the internal ș-role remains unassigned. This allows the ș-role to percolates to VP, its maximal extension. To be the antecedent of the understood generic subject in (91a) and of the generic reflexive in (92), the ș-role must be minimally specified, which is what the LF-rule introduced for the external ș-role of infinitivals can provide: (94)

LF-Rule: Unassigned ș-role is interpreted as +human

If on the right track, a reformulation of the choice point at which the rule above becomes available is in order. In section 5 the point was assumed to be TP (see (29)), on the basis of CP blocking ș-role percolation. Since the maximal extension of an internal ș-role is VP, we can reformulate rule in (94), so that the option to interpret a ș-role as +human becomes available when that ș-role reaches its maximal extension: (95)

When a ș-role, [ș A, B], reaches its maximal extension, B can be specified as + human

And with this, the second of our aims is met: II

Our theory should account for why the understood subject in NOC must be human.

The proposal for internal ș-roles is a tentative one, but it would be interesting to find out whether this +human internal ș-role can also become specific. If so, the discourse rule that applies to the understood subject of infinitivals generalises to that of understood objects. Note there is no OC of objects, since this relation is licensed through copying, made possible by Elsewhere. If a ș-role can be assigned, separation of B from A is not be permitted. For the same reason, null-subjects in finite clauses are ruled out: (96)

*John1 promised Mary Ĭ1 would leave

The Distribution of Non-Obligatory Control

153

8. Conclusions This paper developed an extra-syntactic analysis of NOC, such that all external ș-roles of NOC infinitivals were subject to an LF-rule that would ensure they be interpreted as arguments in the semantics. Specification was minimal (+human), which catered for generically interpreted NOC structures. A discourse rule was added for implicit subjects with specific interpretations. Distributional differences between OC and NOC were captured by the control verb’s s- selection of a CP with an unassigned B. Their predicate status fell out from copying the unassigned B beyond the embedded clause. Application of the two rules introduced for NOC was extended to the linked-reading effect, although it was noted that this effect is not peculiar to control. Finally, I introduced a possible approach to the null-objects of Rizzi (1986), which were also reinterpreted in terms of the internal ș-role, thereby suggesting an analysis of the phenonemon that need not depend on a structurally represented null object.

Acknowledgements For useful comments and enjoyable discussions, my thanks to Ad Neeleman, Marco Tamburelli and Rob Truswell.

Notes 1

Details can be read in Janke (2007). These last two restrictions have more recently come into question in Landau (2000) for reasons the present account rejects. Space considerations prevent their consideration but I refer the interested reader to Janke (2007). 3 The formulation of the semantic criterion above may seem too strong in light of the example below: a) Field mice are easy feed for the nocturnal owl. To go out at night, therefore, is a dangerous undertaking. But such sentences are more indicative of our tendency to refer to non-human entities as though they had peculiarly human characteristics (such as our reasoning capacity) and in light of this they do not constitute counter examples to the +human requirement. Another potential counter example is (b) below. But there is reason to believe that this is a nominal gerund as opposed to a verbal gerund. Verbal gerunds, for example, do not take determiners, contra (c), and permit adverbs as modifiers, as opposed to APs, contra (d). b) Melting at room temperature is typical of ice c) The melting of ice is expected at room temperature d) Unexpected/*unexpectedly melting of ice at room temperature… 4 For an explanation of how external and internal ș-roles are distinguished in this 2

154

Chapter Six

theory, I refer the reader to the original (Neeleman and van de Koot 2002). For expository purposes only, I represent the external ș-role as underlined. 5 Marking is either case-marking on the category itself or via agreement with the verb. 6 A restriction on ș-role percolation is formulated later, ensuring that no overgeneration occurs. 7 We return to the discourse metric that fleshes this +human specification out below. 8 This sentence is not available in my grammar, but another attributed to Chomsky in Lebeaux (1984), which makes the same point, is: John asked Bill how PROarb to behave oneself under such circumstances 9 With the current PRO-free analysis of control in mind, I use the ș-role notation (Ĭ) to indicate that it is the external ș-role that is relevant in this relation. 10 For evidence that this interpretation is one of real ambiguity, I refer the reader to the paper itself, in which Rizzi induces a Principle B violation when the two null-categories share governing categories.

References Ariel, M. 1988. Referring and Accessibility. Journal of Linguistics 24, 65–87. Bhatt, R. & R. Izvorski. 1998. Genericity, Implicit Arguments and Control. Proceedings of Student Conference in Linguistics (SCIL) 7. Distributed by MITWPL. Bhatt, R. & R. Pancheva. 2006. Implicit arguments. In: M. Everaert & H. van Riemsdijk (eds.), Blackwell Companion to Syntax vol. II, pp. 554-584. Oxford: Blackwell. Borer, H. 1989. Anaporic Agr. In: O. Jaeggli & K. Safir (eds.), The Null-Subject Parameter. Dordrecht: Kluwer Academic Publishers. Brame, M.K. 1977. Alternatives to the Tensed S and Specified Subject Conditions. In: E. Sturm & F. Weerman (eds.), Generatieve Syntaxis p137- 411. Leiden: Martinus Nijhof. Bresnan, J. 1982. Control and Complementation. Linguistic Inquiry 13, 343-434. Chierchia, G. 1984. Topics in the Syntax and Semantics of Infinitives and Gerunds. PhD. dissertation, University of Massachussetts. Evers, A. 1988. Non-finite verb forms and subject è-role assignment. In: M. Everaert, A. Evers, R. Huybregts & M. Trommelen (eds.), Morphology and Modularity: Publications in Language Sciences, vol. 29. Dordrecht: Foris Publications. Groenendijk, J. & M. Stokhof. 1989. Type-Shifting Rules and the Semantics of Interrogatives. In G. Chierchia, B. Partee & R. Turner (eds.), Properties, Types and Meanings, Semantic Issues, vol 2 pp. 21-68. Dordrecht: Kluwer.

The Distribution of Non-Obligatory Control

155

Hale, K. & S. Keyser. 1993. On Argument Structure and the Lexical Expression of Syntactic Relations. In: K. Hale & S. Keyser (eds.), The View from Building 20, pp. 53-109. Cambridge, MA: MIT Press. —. 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press. Hamblin, C. 1973. Questions in Montague English. Foundations of Language 10, 41-53. Hornstein, N. 2001. Move! A minimalist theory of construal. Oxford: Blackwell Publishers. Janke, V. 2007. Control without PRO. PhD dissertation: University College London. —. 2008. Control without a Subject. Lingua 118 (1), 82 -118. Karttunen, L. 1977. Syntax and Semantics of Questions. Linguistics and Philosophy 1, 3-44. Kiparsky, P. 1973. ‘Elsewhere’ in phonology. In S. Anderson & P. Kiparsky (eds.), A Festshrift for Morris Halle, pp. 93-106. New York: Holt, Rinehart and Winston. Landau, I. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions. Studies in Natural Language and Linguistic Theory. Dordrecht: Kluwer Academic Publishers. Lasnik, H. 1992. Two notes on control and binding. In: R. Larson, S. Iatriadou, U. Lahiri, J. Higginbotham (eds.), Control and Grammar, pp. 235-251. Amsterdam: Kluwer Academic Publishers. Lebeaux, D. 1984. Anaphoric Binding and the Definition of PRO. In: C. Jones & P. Sells (eds.), NELS 14 GLSA Publications, pp. 253-274. Amherst, MA: UMASS. Lebeaux, D. 1985. Locality and Anaphoric Binding. The Linguistic Review 4, 343-363. Manzini, R. 1983. On control and control theory. Linguistic Inquiry 14, 421-46. Manzini, R. & A. Roussou. 2000. A minimalist theory of A-movement and control. Lingua 110 (6), 409-447. Marantz, A. 1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT Press. Neeleman, A. & H. van de Koot. 2002. The Configurational Matrix. Linguistic Inquiry 33, 529-574. Neeleman, A. & K. SzendrĘi. 2007. Radical pro drop and the morphology of pronouns. Linguistic Inquiry 38(4), 671-714. Rizzi, L. 1986. Null Objects in Italian and the Theory of pro. Linguistic Inquiry 17, 501-557. Samek-Lodovici, V. 2003. The internal structure of arguments and its role in

156

Chapter Six

complex predicate formation. Natural Language and Linguistic Theory 21, 835-881. Williams, E. 1980. Predication. Linguistic Inquiry 11, 203-238.

PART V: LANGUAGE POLICY AND PHONOLOGY

CHAPTER SEVEN CHANGING ATTITUDES: WAYS OF PHONOLOGICALLY ADAPTING PROPER NAMES IN ARCHAIC, BRAZILIAN AND EUROPEAN PORTUGUESE* GLADIS MASSINI-CAGLIARI

Abstract This paper focuses on the pronunciation of proper names of foreign origin when used in Portuguese, in a historical approach. The analysis makes a comparison between the way medieval and contemporary Portuguese incorporate foreign names into the language. For the medieval Portuguese, the names come from the Cantigas de Santa Maria (King Alfonso X – 1121-1284). These chants gathered miracles and names of people from several parts of Europe (and not only from Iberian Peninsula). By observing how the verses can be divided into poetic syllables and how rhymes are constructed, it is possible to make reasonably good hypotheses on how they were pronounced and adapted to Portuguese, looking at the spelling. Taking into account our knowledge of medieval Portuguese phonology, not all names underwent a phonological adaptation. In Portugal today there are specific laws which determine the names that can be used by the Portuguese people from their tradition. In this respect, they are not allowed to incorporate foreign names. On the other hand, in Brazil people are free to choose names for their children. This fact leads to the situation in which many people choose foreign names that are not phonologically adapted to the Brazilian pronunciation. These foreign names used in Brazil were randomly collected from different media and from all over the country (TV, newspaper, magazine, school registers, personal information, etc.). From the data analysis of the three situations it is concluded that an important change in the attitude concerning the phonological adaptation of foreign names in the language can be observed. The medieval tendency is

160

Chapter Seven

no longer used. Portugal has retained only traditional names and Brazil presents an interesting process of incorporating foreign names to the phonological system of the language.

Introduction It is well known that, in many ways, sometimes the role of proper nouns in the general context of the language is considerably marginal, once compared to common nouns. When first names of foreign origin given to children by their parents and used in different historical periods of the language are compared, concerning the ways they do or do not undergo phonological adaptation (especially in prosodic terms), their grade of adaptation can vary through the history of Portuguese. In spite of this fact, in all the three situations analysed here, a certain amount of marginality (or, in other words, formal characteristics that result from violating the phonological principles of the language) can be observed since the first written registers in Galician-Portuguese (13th century) till nowadays. One of the most ancient and famous Portuguese grammarians, Jeronymo Soares Barbosa (1822, 116), in his Grammatica philosophica da lingua portugueza ou principios de grammatica geral applicados à nossa linguagem [‘Philosophical Grammar of Portuguese Language or Principles of General Grammar Applied to Our Language’], has already recognised this ‘marginal’ character of proper nouns in the context of language grammars and dictionaries, assuring that: ‘These Names [...] do not properly belong to the considered languages as common analytical Methods, and because of this, they do not have place in their Vocabularies, but only in Historical and Arts Dictionaries, to which they belong’.1 According to Correia (2009: 27), proper nouns, contrarily to common nouns, do not appear in dictionaries because they do not carry linguistic meaning: they are strictly referential words, capable of establishing relation with individual entities of reality. Maybe because they are empty, in terms of lexical meaning, the main focus of a proper name rests on its sound, which, freed from the barriers of literal meaning, can also be liberated from the structural barriers of Phonology. Because of this, relating to Portuguese, in archaic and contemporary periods, the adaptation of proper nouns of foreign origin deserves a more accurate study, since this phenomenon deals with a limit situation, concerning phonological adaptation. On the one hand, several previous studies (among them, M. Freitas 1992; T. Freitas, Ramilo & Soalheiro 2003; M. Freitas & Neiva 2006; Assis 2007) have shown that the phonetic realization of proper nouns of foreign

Changing Attitudes

161

origin is sensitive to the application of contemporary Portuguese phonological processes, when these words are inserted in Portuguese phrases – the same for Brazilian (henceforth, BP) or European Portuguese (from now on, EP). In contrast with common nouns, proper nouns do not always fit the phonological patterns of the language, for the speakers tend to carry over into Portuguese the foreign pronunciation or spelling of the names. On the other hand, there are very few studies about the phonetic realization of proper nouns of foreign origin in past periods of Portuguese.2 Relating to the archaic period of the language, the Cantigas de Santa Maria (‘Chants of the Holy Mary’, henceforth, CSM), written in GalicianPortuguese, although they have been compiled in Castile, are particularly revealing to this respect, since the 420 religious chants composed by order of King Alfonso X the Wise (1121-1284) reunite miracles collected in several different parts of Europe (and not only in the Iberian Peninsula). Because of this, names of persons and places of different origins are present in the poetic narratives of the Holy Mary wonders. D’Azevedo (1900: 47) worries that the study of Portuguese onomastics has not received the needed attention, although it is one of the most interesting chapters of the origins of Portugal, since successive races of people left lasting vestiges of their passage in this word category. Silva Neto (1970[1957]: 308) also recognises in the anthroponymy of the language a less conservative character, when they are compared to other kinds of names, including proper nouns such as toponyms. He argues that names of persons, following the fashion, were romanised soon in the first periods of the colonisation; however, this process did not happen to names of places, which were nominated following the ancient tradition. To the author, toponyms are highly conservative because very often they refer to geographical peculiarities of the regions. This is the reason why A. F. Carvalho (1950: 151) considers the study of proper nouns origin much more complicated than the study of the etymology of common nouns. Their mobility results from the influence of factors of different nature, such as social, historical and psychic. In contrast, the constancy of the names for places constitutes the fundamental characteristic of the toponyms. Câmara Jr. (1985[1975]: 208) considered the frequent foreign form of first and family names as a proof of the persistent integration of foreigners in Portuguese social life in the Middle Age and in the Renaissance period. To this respect, he affirms that the richness of contemporary Brazilian proper names started to be historically configured by borrowing foreign words, since the very beginning. Contrarily to this tendency, in Portugal, the choice of first names continues to reside mainly on religious motivations or

162

Chapter Seven

continuity in the family context. Regarding the differences in incorporating foreign proper names in Brazil and in Portugal, that results in different ways of treating their pronunciation (and, consequently, their orthographic and phonological forms), this paper aims to focus on persons’ first names in the most important Portuguese-speaking geographic areas (Brazil and Portugal) nowadays. The following sections will be dedicated to the analysis of the way proper names of foreign origin are or are not graphically and phonetically adapted to Portuguese, in three moments of Portuguese historical and geographical continuum: Archaic Portuguese (AP), contemporary European Portuguese (EP) and contemporary Brazilian Portuguese (BP).3 Concerning AP, the corpus is composed by King Alfonso X’s Cantigas de Santa Maria (based on the reasons previously exposed in this paper). Relating to EP, the laws which regulate the incorporation of foreign names in the official onomastics of the country are considered. In what concerns BP, data previously randomly collected in several different media by this researcher and by members of my research team will be analysed, as well as registers of names of personal acquaintances to this researcher. But, previously to the presentation of the phonological characteristics of proper nouns in AP, BP and EP, we will provide a brief overview of the most important properties on (Brazilian) Portuguese phonology and stress, in order to facilitate following the comparison of the adaptation that is taken place. In onset position, the inventory of BP shows an amount of 19 distinctive consonants: /p, b, t, d, k, g, f, v, s, z, ‫ݕ‬, ‫ݤ‬, m, n, ݄, l, ‫ݠ‬, ‫ݐ‬, h4/. When there is only one consonant preceding the nuclear vowel, this consonant can be any one of the 19 consonantal phonemes. In word initial position, / ݄, ‫ݠ‬/ occur only in a very few loanwords (such as nhoque [‘݄‫ܧ‬ki], ‘gnocchi’, a kind of Italian pasta, and lhama [‘‫ܣݠ‬m Ѻ a], ‘llama’) and / ‫ݐ‬/ never occurs. Syllables that present /݄, ‫ݠ‬, ‫ݐ‬/ in initial position must not be preceded by a nasal vowel (Silva 1999: 156). There are also restrictions concerning syllables with two prevocalic consonants. When C1 and C2 occur, the first consonant is always an obstruent (category which includes plosives and labial /f,v/ fricatives) and the second consonant must be a liquid (/l, ‫ݐ‬/). Although, /dl/ never occurs and /vl/ occurs only in a very restrict group of proper nouns of foreign origin (Silva 1999: 157). The majority of syllables in Portuguese are open, but closed syllables are allowed. When only one consonant occurs after the nuclear vowel, this must be one of the following: /S/, /R/, /L/, /N/5 - archiphonemes,

Changing Attitudes

163

represented by capital letters, because they correspond to underspecified fricatives, rhotic, lateral and nasal segments, respectively. Generally, only one consonant in coda position is allowed in Portuguese, but, when a second consonant appears in coda (in words such as circunstância ‘circumstance’ and solstício ‘solstice’), it must be a fricative /S/ (Cagliari 1997: 34). There are seven distinctive vowels in Portuguese: /i, e, ‫ܭ‬, a, ‫ܧ‬, o, u/. There are neutralization processes that reduce this inventory to 5 vowels, in pretonic positions: /i, e, a, o, u/ (medial vowels /e,o/ in this position, can be respectively realised as [i, e, ‫ ]ܭ‬and [u, o, ‫ ;)]ܧ‬and to 3, in word final position: /i, a, u/, phonetically realized as [ܼ, a, ‫( ]ݜ‬Câmara Jr. 1985[1970]). Phonetically nasal vowels and diphthongs are phonologically interpreted as a sequence of an oral vowel and a nasal consonant, in coda position (e. g. [‫]ܣ‬Ѻ = /aN/) (Câmara Jr. 1985[1970]). The BP default stress patterns are paroxytone words ending in a light syllable (cása ‘house’; gáto, ‘cat’; léite, ‘milk’) or oxytone words ended by a heavy syllable – i.e. a closed syllable containing rhotic, lateral, fricatives or nasal consonants in coda position (pomár ‘area where fruit trees are grown’, papél ‘paper’, rapáz ‘boy’, jardím ‘garden’). Marginal exceptional stress patterns are oxytone words ended by light open syllables (café ‘coffee’; sofá ‘sofa’; urubú ‘vulture’), proparoxytone words (árvore ‘tree’; fonética ‘Phonetics’; lâmpada ‘lamp’) and paroxytone words ending in heavy syllables (âmbar ‘amber’; túnel ‘tunnel’; jóvem ‘young’; órgão ‘organ’; pônei ‘pony’).6

Foreign Proper Names in the CSM As we have already pointed out, for the analysis of the proper names in AP, the 420 Cantigas de Santa Maria, compiled by Alfonso X will be considered as research corpus. This vast collection of poems reunites chants in honour of Virgin Mary, with musical notation, compiled by order of the King of Castile in the second half of the 13th century. They have survived in four codices: the Toledo MS (To), the smallest and the most ancient one, now in the Biblioteca Nacional de Madrid; the códice rico [‘rich codex’] (T), in the Escorial Library, Spain, the richer one in artistic content, which was completed by the Florence MS (F), the códice de las histórias [‘the stories codex’], in the Biblioteca Nazionale Centrale, Florence, Italy; and the códice de los músicos [‘musicians codex’] (E), also in the Escorial Library, the most complete.7 The recourse to a poetic corpus, in order to realise researches of this nature, is an imperative, since no oral registers of the language in that period have survived. Because of this, we have searched for clues that could point

164

Chapter Seven

out for the sounds represented by the written letters, which, in one way or another, can reveal phonetic clues, both in segmental and (mainly) prosodic levels (Massini-Cagliari 1995, 1999, 2005). Being particularly revealing of the phenomenon focused in by this article, the CSM constitute an excellent material for this kind of research. Firstly, because they contain narratives of miracles collected in several different places in Europe, which imply the textual necessity of portraying people and places (surely referred to by their names) of other origins than Galician-Portuguese; secondly, because they constitute metre-based and rhymed poems, which bring clues to the phonetic realization of these names, departing from the observation of the verse division in poetic syllables and of the rhyme matching. The data was collected from Mettmann’s (1972) Glossário, which compiled all the words that appear in the 420 CSM, including anthroponyms, toponyms and other kinds of proper nouns, beyond the other usual lexical entries. For this study, we have mapped all the persons’ names in the Glossário. As it can be seen in table 1, these names were adapted to the AP phonological patterns in their vast majority, taking into consideration mainly word internal phonological phenomena, especially syllabification and lexical stress.8 Table 7-1: Phonological adaptation of people names of foreign origin in the CSM. People names of foreign origin in the CSM Cases in which the names fit AP phonological patterns Cases in which the names do not fit AP phonological patterns Cases in which it was not possible to determine stress position TOTAL

Quantity (percent) 100 (73%) 31 (22.6%) 6 (4.4%) 137 (100%)

Among the cases in which it was not possible to determine stress position, the major problem was the lack of sufficient data for the analysis. That was the case of the following first names: Abiron, Libano, Nicolas, Ruben, Siagrio and Yprocras, that occurred only once in the corpus and away from the rhyme position, which caused the impossibility of determining with certainty the stress positioning. However, all these names fit AP syllabification patterns.

Changing Attitudes

165

The research focused mainly on the comparison between those names in the corpus which appear as not conforming to the phonological patterns of the language and those which do. This procedure shows which phonetic contexts were considered ‘problematical’ (in the sense that they do not fit the patterns of Galician-Portuguese Phonology) by native speakers in those days, and which ones could be considered ‘acceptable’ (passing to AP without any change). Among the contexts which do not trigger phonetic changing, i.e., figuring in names that were considered as already ‘adapted’ to the phonology of the language in those days, there are oxytone words ended by nasals (Aben Mafon, Abirron, Abran, Aragon, Merlin, Octavian, Reymon, Salomon), rhotics (Ander - from Santander [Saint Ander] -, Artur, Bondoudar, Vitor), laterals (Marçal), and fricatives (Bonifaz, Fiiz, Tomás). Following the patterns which occur till present days in BP and EP, AP default stress pattern corresponds to paroxytones ended by one light syllable (amíga ‘female friend’, cása ‘house’) or oxytones ended by one heavy syllable, i.e., a syllable ended by a rhotic, a lateral, a fricative or a nasal consonantal archiphoneme (amór ‘love’, anél ‘ring’, soláz ‘pleasure’, sazón ‘time, occasion, station’).9 Exceptional marginal patterns (such as oxytones ended by open syllables or proparoxytones or paroxytones ended by heavy syllables) are extremely rare in that moment of the temporal continuum of the language.10 It was possible to map two names in which the stress positioning was altered, maybe in order to fit the rhyme: Brutus and Colistanus. These Latin names, in origin, are not oxytones, but paroxytones. Although, there is evidence that they were realised as oxytones, at least in the context of the cantiga in which they appear, because these two words perfectly rhyme with the tonic monosyllable chus ‘more’. (1)

Dun mercador que avia | per nome Colistanus, que os levass' a Bretanna, | a que pobrou rei Brutus; e entrou y tanta gente | que non cabian y chus, e mui ricos mercadores | que levavan grand' aver. O que a Santa Maria der algo ou prometer... (CSM 35, 8th stanza, in Mettmann’s edition, 1986: 146)11

Although in its vast majority the people names mapped in the CSM are already adapted to AP Phonology, there could be found a very few names that do not fit the phonological patterns of the language in those days and that do not undergo phonological adaptation process, remaining

166

Chapter Seven

exceptional, in what concerns the prosodic structure. In this case, there are names such as Elisabet and Octavian, whose syllabic structure presents obstruent consonants in coda position – this pattern was already unacceptable in medieval AP. There could be also found non-adapted names, concerning stress position, such as: Jaymes, Lucas, Marcos (the last two names remain unaltered – and, henceforth, exceptional, in what concerns the stress positioning – until nowadays). Oxytone names ended by open syllable were also mapped: Aleixi, Davi, Salomé, Tomé. This stress pattern was extremely rare in those days, being found only in a few adverbs (such as aqui ‘here’, ali ‘there’, cf. Massini-Cagliari 1999) and substantive nouns (rubi ‘ruby’, cf. Costa 2006).

The Official Legislation on the Incorporation of Foreign Names in EP Relating to Portugal, Castro (2003: 15) informs that, in the case of attributing names to persons, ‘the weight of official norms makes itself heavily felt’.12 It is necessary that the State makes explicit, by means of a process of civil register, the agreement relating not only to the chosen name, but also to the way the name is written and pronounced. The Portuguese Code of Civil Register also restricts the extension of the citizens’ names, in the way that it establishes that the full name must be composed by six grammatical words at the maximum, of which only two must correspond to first names; the other four are surnames. The counting of the names is based on the concept of ‘grammatical compound word’ [‘vocábulo gramatical composto’], which is defined as ‘a word constituted by two or more simple words, which bear an autonomous meaning, sometimes dissociated from the meaning of its components’ (Instituto dos Registos e Notariados 2009: 1).13 In what concerns its origin, the names must be Portuguese, i.e., they must appear in the ‘national onomastics’, represented in the official catalogue of proper names.14 Only foreigners born in Portugal can choose non-Portuguese names (Castro 2003: 16-17). But the official Institute of Registers (Instituto dos Registos e Notariados 2009: 1) considers that, when the graphic and phonetic adaptation to the Portuguese language is made, this is equivalent to make Portuguese a name originally foreign. As Castro (2003: 23) points out, even the spelling of a proper name must be officially approved, and graphic variations for the same name are not allowed. As an example, the author says that, in Brazil, there can be ladies called Rosemary, Rosemeire, Resemere, Rosemery, Rosimeire, Rosimere, Rosimeri, Rozemeire, all names derived from the same English form

Changing Attitudes

167

Rosemary. In contrast, in Portugal, all of these ladies have to be called Rosa Maria; not even Rosamaria would be accepted as a Portuguese name. As shown by Monteiro (2008: 45-46), although rigid, the rules used in Portugal for the composition of the names for persons are recent: they have had it about eighty years, since they were established by the Code of Civil Register [Código do Registo Civil] on April 10th 1928. In the ancient regime (before 1834), there were no legal norms concerning this subject, and in the first Portuguese Civil Code, dated from 1867, there were no rules about these matters. However, Nunes (1989[1919]) and Câmara Jr. (1985[1975]) point out that the main legal changes relating to the composition of proper names in Portugal are relative to the combination of first and surnames and to the number of allowed names; there was no change concerning the use of foreign names. In this sense, concerning the choice of first names, historical studies have shown that the presence of foreign origin words always existed, and there were moments in which their entry to the language was more striking. Following Mattos e Silva (2005), in the formative period of the Portuguese language (5th to 7th centuries), there was the invasion of Germanic people in the Iberian Peninsula, who have brought to Portugal a series of Germanic names (such as Fernando, Orlando, Rolando) which are very popular till nowadays. In other words, the resistance to foreign names in the Portuguese legislation is the most important distinctive element between Portugal and Brazil in present days.

The Graphic Form and the Sound of the Foreign Names in Brazil 15 Contrarily to Portuguese people, Brazilian citizens are free to choose the names of their children. According to Calaça (2001:31), there are only two prohibitions regarding the choice of proper names in Brazilian legislation: they cannot be ridiculous16 or immoral17. The parents are free to choose any spelling they prefer for their children’s names. The freedom concerning the choice of a name in BP extends to its spelling. Hence, many times the Brazilian speaker appreciation of foreign origin names reflects on the chosen graphic form for names of vernacular origin. In this sense, their writing can be different from the usual one, presenting double letters, consonant + groupings and predominance of . Cagliari (1998: 387), based on the orthographic system that was official at that time in Brazil, points out that the letters K, W and Y were only used

168

Chapter Seven

in foreign words, in abbreviations, in proper names and to represent logical and mathematical calculus. He adds that the usage of foreign words brings novelties to the writing system of the language, causing new relations between sounds and letters to emerge. The possibility of graphic variation concerning persons’ names in Brazil is the most important factor that causes native or very common names in Portuguese to appear ‘foreign’ or ‘strange’. A few examples are: Nathalya, Andreya, Anthonio, Cellya, Christyna, Edwuardo, Filippe, Helenna, Henriky, Jhullio, Kamylla, Karlla, Karolyna, Maryo, Markus, Maryana, Mauryccio, Phelipe, Regyna, Rickardo, Sonnya,18 etc. This tendency extends even to native names, borrowed from Brazilian indigenous languages, such as Jhanaína. In spite of their deviant orthographic pattern, these names sound exactly as their native correspondents in usual spelling. However, many of the most usual names in BP are in fact of foreign origin, mainly English. According to Carvalho (2009: 68-69): In no other lusophone country is there such an unrestricted adoption of English proper names as in Brazil, mainly in lower social strata. João, Manuel, Maria, Severina, Francisco are being substituted by Magaiver, Kelly, Marilyn, Kennedy, Tyronne, Daiane and many more, even more strange, which can be found in the list of students’ names in state schools. It seems that the choice is based on the paraphrase ‘the more bizarre, the better’.19

This paper does not aim to discuss the reasons why Brazilian parents prefer to choose an English name instead of names from other foreign languages, since these reasons extrapolate the limits of strictly linguistic factors; otherwise, we intend to discuss the consequences of the dense entrance of English first names, with their phonological and graphic specific native characteristics, for the phonological identity of contemporary BP. In Brazil, there are no official laws which prohibit the adoption of a non-standard Portuguese spelling or the maintenance of the original spelling. Being so, it is equally possible to maintain the original foreign orthographic pattern or transform it – in this case, not exactly in direction of the Brazilian Portuguese orthographic patterns. Thus, sometimes the spelling adopted in Brazil is not the original foreign one, but an attempt to represent the original sound of the name, following its phonetic realization in the foreign language. To exemplify, we can quote names such as: Quéli [‘k‫ܭ‬lܼ] (from Kelly), Sanflauer [s‫ܣ‬Ѻ’fla‫࡬ݜ‬e‫( ]݋‬from Sun Flower), Rérisson [‘x‫ݐܭ‬isõ‫࡬ݜ‬Ѻ] (from Harrison), and Taison [‘taܼ࡬ sõ‫࡬ݜ‬Ѻ] (from Tyson).

Changing Attitudes

169

Other times, attempting to guarantee a phonetic realisation in BP resembling the original pronunciation of the name, the English graphemes / and , that in this language can correspond to the English diphthongs /ay/ and /ey/, are substituted by letters that represent in BP the sounds the person who chooses the name wants to reach as target. Examples: Brayan [‘b‫ݐ‬aܼ࡬ ‫ܣ‬Ѻ] (from Bryan), Carolaini [ka‫ݐ‬o’laܼ࡬ nܼ] (from Caroline), Daiana [daܼ࡬ ’‫ܣ‬Ѻna] (from Diana), Maycon [‘maܼ࡬ kõ‫࡬ݜ‬Ѻ] (from Michael), Greicy [‘ܳ‫ݐ‬eܼ࡬ sܼ] (from Grace), Jeymes [‘d‫ݤ‬eܼ࡬ mܼs] (from James), and Deivid [‘deܼ࡬ vܼd‫( ]ܼݤ‬from David).20 Another example of the process of spelling adaptation that targets a graphic representation for the original sounds of the English name occurs when uncommon consonantal groups to BP are used. This is the case of the attempt of representing the English affricate consonant /d‫ݤ‬/, usually graphically represented in English by and . In BP context, graphic groupings such as and can be used as a tentative of representing the affricate, in names such as Dheniffer [‘d‫ݤ‬enife‫( ]݋‬from Jennifer), Dhéssica [‘d‫ܭݤ‬sika] (from Jessica) and Djerson [‘d‫݋ܭݤ‬sõ‫࡬ݜ‬Ѻ] (from Jerson).21 To achieve the same goal, the sequence can be used, in names such as Diônas [‘d‫ݤ‬onas] (from Jonas) and Diully [‘d‫ݤ‬ulܼ] (from Julie),22 since in the BP variety spoken in the geographic region in which the data was collected there occurs the phonological process of palatalisation of /d/ before /i/, which results in phonetic realizations to the first syllables of those two names respectively as [d‫ ࡬ܼݤ‬o] or [d‫ݤ‬o] and [d‫ ࡬ܼݤ‬u] or [d‫ݤ‬u]. The pronunciation of names that preserve the English original graphic form is occasionally regulated by the rules of orthographic deciphering in BP and not in English. In other words, the relations between letters and sounds are proper of BP and not of English. This phenomenon can be observed in names such as Walter [‘va‫࡬ݜ‬te‫]݋‬, realized with an initial [v], Jefferson [‘‫ܭݤ‬fe‫݋‬sõ‫࡬ݜ‬Ѻ], Jackson [‘‫ݤ‬akisõ‫࡬ݜ‬Ѻ] and Gerald [‘‫ݐܭݤ‬a‫࡬ݜ‬d‫]ܼݤ‬, phonetically realized with [‫ ]ݤ‬in the initial positions of the first syllable, and Charles [‘‫ݕ‬a‫݋‬lܼs], pronounced with [‫ ]ݕ‬in the initial syllable. However, some foreign proper nouns whose use is quite common in Brazil today do not fit the phonological parameters of BP, in contrast to what happens to foreign borrowed common nouns. Occasionally, BP native speakers bring to their own native language traces of the original pronunciation of foreign names, transferring unusual characteristics to the system, chiefly at prosodic levels. Examples of proper nouns with a syllable pattern that is not usual in BP are Wilson [‘‫࡬ݜ‬i‫࡬ݜ‬sõ‫࡬ݜ‬Ѻ], William [‘‫࡬ݜ‬i‫]ܣݠ‬Ѻ , Washington [‘‫ݕܧ࡬ݜ‬ƭtõ‫࡬ݜ‬Ѻ], Wellington [‘‫ܭ࡬ݜ‬lƭtõ‫࡬ݜ‬Ѻ], Wesley [‘‫ܭ࡬ݜ‬sleܼ࡬ ] and Wendel [‘‫࡬ݜ‬Ӂde‫]࡬ݜ‬, that contain a pre-nuclear glide /w/, which occurs in onset position in BP only after velar stops /k, ܳ/.

170

Chapter Seven

Irregular syllable patterns in BP can also be found in names in which there are obstruent consonants in coda position, such as Klebson and Judson. In this case, the anomalous syllable structure is usually ‘solved’ by repositioning the stop consonant in syllabic onset with an inserted epenthetic nuclear vowel after the consonant: kle.b(i).son [‘kl‫ܭ‬bܼsõ‫࡬ݜ‬Ѻ], Ju.d(i).son [‘‫ݤ‬ud‫ݤ‬isõ‫࡬ݜ‬Ѻ]. Non-allowed consonantal clusters in BP, even when they occur in names whose spelling corresponds to the original English spelling, use to have their anomalous syllable structure resolved by an epenthesis, as in Stewart [is.’tu.a‫݋‬.t‫]ܼݕ‬. However, the most common irregularity (at the phonological level) related to the use of foreign names or names that are supposedly of foreign origin in BP is concerned with stress positioning. In the case of names imported from English, non-canonical stress positioning is much more frequent than the adoption of default stress patterns, i.e., proparoxytones and paroxytones ended by closed syllables are common. The English names formed with the use of specific endings (such as –ton and –son) fit this pattern:23 Kleiton, Washington, Wellington/ Welinton/Uélinton, Anderson, Robinson, Nelson, Jeferson, Everson). Other nasals endings also cause the names to be fitted to this pattern: Alan, Cristian, Helen/Hellen, Sheron. There are also phonetic proparoxytones in which there is an insertion of an epenthetic vowel after a consonant originally in coda position: Vagner/Wagner [‘va.ܳi.ne‫ ]݋‬and Ingrid [‘ƭ.ܳ‫ܼݐ‬.d‫]ܼݤ‬. In this case, there also occur paroxytones or proparoxytones ended by rhotics (Éder [‘‫ܭ‬de‫]݋‬, Kleber [‘kl‫ܭ‬be‫]݋‬, Wander [‘v‫ܣ‬Ѻde‫ )]݋‬or fricatives (Gladis [‘ܳlad‫ܼݤ‬s], James [‘d‫ݤ‬eܼ࡬ mܼs], Nicholas [‘nikolas], Deives [‘deܼ࡬ vܼs]). In all these examples (and in the penultimate position, in the case of proparoxytone words) there appear closed syllables in unstressed word-final position. This structure is not usual in BP, although syllables closed by rhotic consonants (revólver ‘revolver’) and containing nasal vowels (which are phonologically interpreted as a sequence of an oral vowel followed by a nasal consonant, e.g. jóvem ‘young’) can be found marginally in this position. In general, imported proper names demonstrate a preponderance of exceptional stress patterns (proparoxytone words and paroxytone words ended by heavy syllables), even when the reference is the target language (BP) and not the original language (English). In this way, it is possible to say that the ‘foreign’ or ‘alien’ character ascribed to foreign names by BP speakers remains chiefly in the exceptional prosodic pattern they assume. In fact, this ‘foreign’ character (of the spelling, by the one hand, and of the prosody, on the other hand) attributed to borrowed names can be

Changing Attitudes

171

demonstrated when we analyse recent creations of first names in BP, which intend to be ‘innovative’ and ‘different’ and that take as their model foreign proper nouns previously borrowed by BP. As an example of ‘innovative’ spelling, we can quote the names of two twin sisters, Kauany and Keroly, created from possible syllabic and stress patterns in BP, but graphically represented by unusual orthographic forms. All people names which present alternative spelling to vernacular names, based on the adoption of the letters and in consonantal and vocalic combinations which are not usual in Portuguese official spelling, can also be quoted as examples. Good examples of how a marginal prosodic pattern can bring to names a ‘foreign’ character are recent creations, which are based in English names – but, although they seem to be English to BP speakers, they are not properly English: Keirrison, Kimarrison, Richarlyson, Gelson, Clerison, Jandison, Silgleison, Madson, etc. A curious example is the adaptation that suffered one of the most popular first names in Brazil today: Máicon (also spelled Maycon, Maikon, Maykon, etc.), adapted from Michael, used mainly to pay homage to the singer Michael Jackson. The name Michael presents a lateral consonant in coda position at the end of the word; if the phonological processes that apply to foreign common nouns to adapt them to BP were applied to this name, the result would be Máicou [‘maܼ࡬ ko‫]࡬ݜ‬. However, this form does not present a nasal or a rhotic consonant in the unstressed word ending position – elements that seem to be crucial for the recognition of an English proper noun by BP speakers. On the other hand, the form Máicon perfectly corresponds to this expectation. Following this model of creation, we can also quote the name Livian,24 which was formed by the junction of the mother’s name, Lilian, with the traditional name Lívia (which, by its turn, is already marginal, in terms of stress positioning). Compared to Lívia, Lívian is even more marginal in prosodic terms, with the presence of a nasal consonant closing the last syllable.

Conclusion The study of proper nouns is a highly promising field, in terms of investigating the phonological identity of a specific language, in diachronic or comparative perspectives. Specifically considering the names analysed in this paper, it is possible to see that there was a much more acute grade of adaptation of foreign first names in AP medieval times than it occurs today, in BP.

172

Chapter Seven

N. Carvalho (2009: 30) asserts that ‘to be recognised as terms of Portuguese language, loans adopt patterns created by popular terms, concerning phonological system, as much as syllabic typology and morphological structure’.25 In this way, because they adopt a deviant prosodic pattern, names of real or supposed foreign origin are not recognised as ‘Portuguese’, maintaining its ‘marginal’ character. Maybe it is because of this that the category of proper nouns is considered as marginal to dictionaries of the language since Jeronymo Soares Barbosa (1822), even more when the personal first names are in fact of foreign origin, because they constitute a class apart, inclusively in what respect Phonology, field in which they present a series of ‘irregularities’, when compared to other (common) nouns of the language. Analogously, in terms of spelling, they do not fit the official orthographic patterns, since they ‘allow’ several ‘concessions’, in terms of graphic creativity and letters combination. In this way, despite the fact that proper names can be phonologically marginal, they do not represent a menace to the phonological identity of the language. Although the majority of parents who choose foreign or foreign-like names do not speak English or other foreign language, many BP speakers consider foreign first names to be much more ‘elegant’ than Portuguese proper nouns. Other parents choose a foreign word to name their children because of their different phonetic characteristics. Other parents choose a foreign name only because they do not want a popular or a very commonly used name for their children, believing that uncommon and unique names represent people with a special personality. Other parents only want to pay homage to a public personality who they admire for some reason (Cunha 2007, 60). Moreover, the motivation for this behaviour is the hope on the part of the parents that their children will have better chances and opportunities in life; in other words, they wish for a change in the status quo. This explains why the choice of a foreign name (or a name that seems to be foreign) is so common in less privileged social classes (although the use of foreign names is not exclusive to them). By this reason, they choose ‘exotic’ first names (from the point of view of pronunciation), that appear to them to be ‘new’ or ‘different’, or not to be ‘Brazilian’. In this sense, the choice of an exotic name, in phonological terms, is a conscious way of denying the sound identity of the language. Castro (2003, 21) compares the usage of foreign proper nouns in BP and in EP and suggests that the main difference, in what concerns the acceptance of these names, resides on the safety (in terms of menace to cultural identity) which multiculturalism of Brazilian society brings in. According to Castro, Portuguese society is beginning to be multicultural, following the

Changing Attitudes

173

Brazilian society that always has been and has never stopped being. As such, Brazil can welcome foreign words without reducing them to the patterns of the vernacular language. Castro claims that the way Portugal and Brazil protect themselves from foreign ‘threats’ reflects an important difference between the two countries. In this sense, Brazilian citizens, being confronted with the difference brought by the foreign linguistic elements, do not feel menaced, but, on the contrary, desire for themselves the alterity the foreign element brings (Garcez and Zilles 2002: 22-23). In summary, this paper has shown that vernacular forms have always cohabited with foreign words in the language, especially in the field of proper names. In the registers of AP analysed here, there is a heavy tendency of adapting the foreign forms into vernacular ones, including persons’ names. This tendency remains in EP, by the force of official laws, but it is considerably different from what occurs today in Brazil, where foreign names carry out a ‘magic’ attraction used to transfer to the owner some of the social status those names bring. In doing so, the linguistic consequence is to house ‘strange’ foreign phonological patterns into the language. This strangeness, in phonological terms, resides almost always on the prosodic level, concerning specifically stress positioning, since the syllable pattern is currently adapted. From this point of view, the presence of foreign or foreign-like prosodic patterns within the language does not menace its phonological identity, as long as they are stylistically used precisely to bring into a proper name a ‘strange’, marginal characteristic.

Notes * This research was funded by FAPESP (2010/06386-0) and CNPq (302222/2009-0). 1 Estes Nomes [...] não pertencem propriamente ás Línguas consideradas como Methodos vulgares analyticos, e por isso, não costumão ter lugar nos Vocabulários das mesmas; mas so nos Diccionarios Históricos, e das Artes, aos quaes pertencem. [All translations that appear in this paper are my own.] Gonçalves (2008) discusses the concept of ‘proper nouns’ in 18th century Portuguese grammar historiography, from which Jeronymo Soares Barbosa is one of the most prominent figures. 2 One of the few studies on this respect is Cunha (1956), about the pronunciation of the troubadour surname Codax. 3 Of course, in medieval times, Galician-Portuguese was not spoken in the contemporary Brazilian territory. 4 Or /x/, in several varieties. 5 In coda position, nasal consonants are phonetically realised as open syllables containing nasal vowels. 6 See Massini-Cagliari (1999) and references therein for details of the BP stress pattern.

174

Chapter Seven

7 This research recurred to facsimile editions of To (Afonso X O Sabio 2003) and E (Anglés 1943) and to microfilms of T and F. 8 For a detailed presentation of AP phonological patterns, see Mattos e Silva (1989, 1991) and Massini-Cagliari (2005). 9 In this case, the nasal consonant could be phonetically realised in an open syllable containing a nasalised vowel. 10 On AP stress patterns, see Massini-Cagliari (1999, 2005) and Costa (2006). 11 ‘of a Merchant called Colistanus so that He might take them to England, colonized by King Brutus. A great many rich merchants carrying great wealth boarded the ship until it could hold no more.’ ‘He who gives or promises something to Holy Mary...’ (Translated by Kulp-Hill 2000, 46). 12 O peso da norma faz-se sentir gravemente. 13 um vocábulo constituído por dois ou mais vocábulos simples que possui um significado autónomo, muitas vezes dissociado dos significados dos seus componentes. The particles (such as prepositions) are not considered in counting. 14 The official catalogue of Portuguese onomastics can be consulted on the web; see Portal do Cidadão (2007). 15 Part of the ideas in this section already appeared in the paper presented at the Conference Interfaces in language 2, Canterbury, University of Kent, Centre for Language and Linguistic Studies, May 2009, entitled: Loans and foreign first names as clues to phonological identity in Brazilian Portuguese. The present paper constitutes a development of the paper presented in the previous Interfaces conference. 16 It is not clear what Brazilian legislation considers to be a ‘ridiculous’ name. It probably refers to embarrassing names, such as the ones presented in the TV Program Domingão do Faustão, in April 17, 2011 (which were officially registered, in spite of the prohibition): Xerox (‘Photocopy’), Fotocópia (‘Photocopy’), Autenticada (‘Proved Authentic’), Carimbo (‘Stamp’, ‘Seal’), Epílogo (‘Epilogue’), Delícia Caldas (‘Delicious Syrups’), and the twin sisters Minúscula (‘Minuscule’) and Maiúscula (‘Capital Letter’). 17 A few combinations of names and surnames can be considered immoral, in this sense. As an example, we can quote a boy named Caio, with a surname Pinto, an embarrassing name that in BP sounds exactly as ‘cai o pinto’ (‘the penis falls’). 18 All the names, excepting the first, are extracted from Souza (2011), a Ph.D. thesis developed under my supervision. 19 Em nenhum país lusófono há uma adoção indiscriminada de nomes próprios em inglês como no Brasil, sobretudo nos baixos estratos sociais. João, Manuel, Maria, Severina, Francisco vão sendo substituídos por Magaiver, Kelly, Marilyn, Kennedy, Tyronne, Daiane e muitos outros, mais estranhos, que constam da lista de chamada das escolas públicas. Parece que a escolha é baseada na paráfrase ‘quanto mais estranho, melhor’. 20 Examples extracted from Souza (2011). 21 Examples extracted from Souza (2011). 22 Examples extracted from Souza (2011). 23 In the graphic representation for the examples, the underline patterns correspond to stressed syllables.

Changing Attitudes

175

24 First name of the daughter of the famous Brazilian comedian Renato Aragão, who is also an actress. 25 Os empréstimos, para serem reconhecidos como termos da língua portuguesa, adotam padrões criados pelos termos populares, quanto ao sistema fonológico, quanto à tipologia silábica e quanto à estrutura morfológica.

References Afonso X o Sabio. 2003. Cantigas de Santa María: edición facsímile do Códice de Toledo (To). Biblioteca Nacional de Madrid (Ms. 10.069). Vigo: Consello da Cultura Galega, Galáxia. Anglés, H. 1943. La Música de las Cantigas de Santa María del Rey Alfonso el Sabio – Facsímil, transcripción y estudio critico por Higinio Anglés. Barcelona: Diputación Provincial de Barcelona; Biblioteca Central; Publicaciones de la Sección de Música. Volume II – Transcripción Musical. Assis, A. B. G. 2007. Adaptações fonológicas na pronúncia de estrangeirismos do Inglês por falantes de Português Brasileiro. MA Dissertation; Araraquara: FCL/UNESP. http://portal.fclar.unesp.br/poslinpor/teses/ana_beatriz_goncalves_de_a ssis.pdf (accessed October 27, 2011). d’Azevedo, P. A. 1900. Nomes de pessoas e nomes de logares. Revista Lusitana. Lisboa, 6(1), 47-52. Barbosa, J. S. 1822. Grammatica philosophica da lingua portugueza ou principios de grammatica geral applicados à nossa linguagem. Lisboa: Academia Real das Sciencias. http://purl.pt/128 (accessed December 27, 2010). Cagliari, L. C. 1997. Fonologia do Português. Análise pela Geometria de traços. Campinas: edição do autor. —. 1998. Alfabetizando sem o bá-bé-bi-bó-bu. Sao Paulo: Scipione. Calaça, I. Z. P. 2001. Nomes próprios estrangeiros no português brasileiro. A Página, 108, 10, December 2001: 31. http://apagina.pt/arquivo/Artigo.asp?ID=1644 (accessed March 28, 2008). Câmara Jr., J. M. 1985. Estrutura da Língua Portuguesa. 15. ed. Petrópolis: Vozes. [1st ed. 1970] —. 1985. História e estrutura da língua portuguesa. 4. ed. Rio de Janeiro: Padrão, 1985. [1st ed. 1975] Carvalho, A. F. 1950. Contribuição para o estudo da antroponímia portuguesa. Três épocas nos nomes de habitantes dum concelho da Beira – Besteiros, hoje Tondela. XIII Congresso Luso-Espanhol para o Progresso das Ciências. Lisboa. Tomo VIII: 151-165.

176

Chapter Seven

Carvalho, N. 2009. Empréstimos lingüísticos na língua portuguesa. Sao Paulo: Cortez. Castro, I. 2003. O lingüista e a fixação da forma. Actas do XVIII Encontro Nacional da Associação Portuguesa de Lingüística. Porto 2002. Lisboa: Associação Portuguesa de Lingüística, 11-24. Correia, M. 2009. Os Dicionários Portugueses. Lisboa: Caminho. Costa, D. S. 2006. Estudo do acento lexical em Português Arcaico por meio das Cantigas de Santa Maria. MA Dissertation; Araraquara: Faculdade de Ciências e Letras, UNESP. http://portal.fclar.unesp.br/poslinpor/teses/daniel_soares_da_costa.pdf (accessed October 27, 2011). Cunha, C. F. 1956. O cancioneiro de Martin Codax. Rio de Janeiro: Imprensa Nacional. Cunha, L. J. 2007. O processo discursivo de designação de pessoas: a determinação histórico-social do nome próprio. In: BOLOGNINI, C. Z. (Ed.) A língua inglesa na escola. Campinas: Mercado de Letras. 49-52. Freitas, M. A. 1992. Empréstimos, teoria auto-segmental e abertura vocálica. Cadernos de Estudos Lingüísticos 23: 71-81. Freitas, M. A. and A. M. S. Neiva. 2007. Estruturação silábica e processos fonológicos no inglês e no português: empréstimos e aquisição. Revista Virtual de Estudos da Linguagem (ReVEL) 7, 1-27. http://paginas.terra.com.br/educacao/revel/. (accessed January 30, 2007). Freitas, T.; M. C. Ramilo and E. Soalheiro. 2003. Processo de integração dos estrangeirismos no português europeu. In Actas do XVIII Encontro Nacional da Associação Portuguesa de Lingüística. Lisboa, Portugal. http://www.iltec.pt/pdf/wpapers/2003-redip-estrangeirismos.pdf (accessed January 30, 2007). Garcez, P. M. and A. M. S. Zilles. 2002. Estrangeirismos – Desejos e ameaças. In Faraco, C. A. (Ed.) Estrangeirismos – Guerras em torno da língua. 2nd ed. Sao Paulo: Parábola. 15-36. Gonçalves, M. F. 2008. Un repaso al concepto de ‘nombre proprio’ en la gramaticografía portuguesa del setecientos. Gramma-Temas 3: España e Portugal en la tradición gramatical: 105-124. Instituto dos Registos e Notariado. 2009. Composição do nome. Lisboa: Ministério da Justiça. http://www.irn.mj.pt/sections/irn/a_registral/ registo-civil/docs-do-civil/dar-o-nome/ (access December 27, 2010). Kulp-Hill, K. 2000. Songs of Holy Mary of Alfonso X, the Wise. A Translation of the Cantigas de Santa Maria. Tempe, Arizona: Arizona Center for Medieval and Renaissance Studies. Massini-Cagliari, G. 1995. Cantigas de amigo: do ritmo poético ao

Changing Attitudes

177

lingüístico. Um estudo do percurso histórico da acentuação em Português. Ph.D. Thesis; Campinas: IEL/UNICAMP. —. 1999. Do poético ao lingüístico no ritmo dos trovadores: três momentos da história do acento. Araraquara: FCL, Laboratório Editorial, UNESP; Sao Paulo: Cultura Acadêmica. —. 2005. A música da fala dos trovadores: Estudos de prosódia do Português Arcaico, a partir das cantigas profanas e religiosas. Habilitation Thesis; Araraquara: Faculdade de Ciências e Letras, UNESP. Mattos e Silva, R. V. 1989. Estruturas trecentistas: elementos para uma gramática do português arcaico. Lisboa: Imprensa Nacional - Casa da Moeda. —. 1991. O Português arcaico: fonologia. São Paulo: Contexto. —. 2005. Aspectos do léxico da língua portuguesa do período formativo à primeira metade do século XVI. In: Telles, C. M. and R. B. Souza (eds.) V Encontro Internacional de Estudos Medievais. Salvador: Quarteto. 112-117. Mettmann, W. 1972. Glossário. In: Afonso X, O Sábio. Cantigas de Santa Maria. Coimbra: Universidade. v. IV: Glossário. —. (Ed.) 1986. Cantigas de Santa María (cantigas 1 a 100): Alfonso X, el Sabio. Madrid: Castalia. Monteiro, N. G. 2008. Os nomes de família em Portugal: uma breve perspectiva histórica. Etnográfica 12(1), 45-58. Obata, R. 1986. O livro dos nomes. Sao Paulo: Círculo do Livro. Portal do Cidadão. 2009. Atribuição do Nome a um Recém Nascido. Disponível em http://www.portaldocidadao.pt/PORTAL/entidades/MJ/IRN/pt/SER_atr ibuicao+do+nome+a+um+recem+nascido.htm (accesses March 28, 2009). Nunes, J. J. 1989. Compêndio de gramática histórica portuguesa: Fonética e Morfologia. 9th ed. Lisboa: Livraria Clássica Editora. [1st ed. 1919] Silva, T. C. 1999. Fonética e Fonologia do Português. Sao Paulo: Contexto. Silva Neto, S. 1970. História da língua portuguesa. 2nd ed. Rio de Janeiro: Livros de Portugal. [1st ed.1957] Souza, S. M. L. S. 2011. Antropônimos de origem inglesa: adaptações ortográficas e fonético-fonológicas realizadas por falantes do português brasileiro de São Luís-MA. Ph.D. Thesis. Araraquara: UNESP.

CONTRIBUTORS

Pedro Guijarro-Fuentes is Reader in Spanish at the University of Plymouth. His main research interests are in the interdisciplinary field of Spanish applied linguistics, sociolinguistics, psycholinguistics, and bilingualism. Current research projects include the study of spatial prepositions, object marking and word order by a variety of learners. [email protected] Victoria Janke completed her PhD in Linguistics at University College London, where she focused on the syntactic representation of control. She is currently Lecturer in English Language & Linguistics at the University of Kent. Her main research areas are syntax and first & second language acquisition in both typical and atypical development. [email protected] Marina Kolokonte gained her PhD in Linguistics at Newcastle University, specialising on ellipsis in Modern Greek and Romance languages. She is currently Lecturer in English Language & Linguistics at the University of Kent. Her research interests include syntax, pragmatics, information structure and second language processing. [email protected] María Pilar Larrañaga is an assistant Professor in Spanish at the University of Plymouth. Her main research interests are Spanish and Basque syntax as well as the acquisition of grammar in bilingual children and L2 learners of Spanish. [email protected] Man-Ki Theodora Lee is currently a PhD student of the Department of Language and Linguistic Science at the University of York. Her main research areas are syntax and second language acquisition. [email protected] Gladis Massini-Cagliari is a Professor at Unesp, Sao Paulo State University, Brazil. She obtained her PhD in Linguistics from Campinas State University, Brazil, after which she completed a year’s postdoctoral research at the University of Oxford. Her research interests lie in historical linguistics and phonology, with particular reference to contemporary Brazilian and medieval Portuguese. [email protected]

180

Contributors

Akiko Nagano is Associate Professor of Human-Social Information Sciences at Tohoku University (Japan). She completed her PhD in Linguistics at Tsuda College (Japan). Her research interests are English morphology (derivation, compounding, diachronic changes) and morphological theory. [email protected] John Partridge John Partridge took his BA and Cert. Ed. In Modern Languages at Liverpool and his Dr. phil. in Theoretical Linguistics and German Philology at Regensburg before returning from Germany to teach German Linguistics at the University of Kent at Canterbury. He is a former secretary of the Forum for Germanic Language Studies and founded and directed the Kent Centre for Language and Linguistic Studies (CLLS) until his recent retirement. He has published works on pragmatics, the linguistics of crosswords, contrastive English-German linguistics and Information Technology in language teaching, and maintains an active interest in all these areas. Masaharu Shimada is Associate Professor of Foreign Language Center at the University of Tsukuba, Japan. He completed his PhD in Linguistics at the University of Tsukuba. His main research areas are syntax and its interfaces with morphology. [email protected]

INDEX accessibility, 122, 143, 146 affixal anaphors, 60 affixation, 78 affixes, 57, 77 alethic, 92 anaphoric binding. See binding anaphors, 57 pure reflexives, 61 anthroponymy, 161 bare nouns, 6, 13 bare nouns in object position, 6 bare nouns in subject position, 6 bare plurals. See bare nouns binding, 61, 121 local binding, 63 non-local, 64 reflexive binding, 130 BNs. See bare nouns Büring's Contrastive Topic Theory, 109 but-coordination, 115 Cantigas de Santa Maria, 161 Cantonese, 32 catenation, 83 c-command, 123 Childes database, 13 circumscription, 94 Clitic Left Dislocation, 110 compounding, 78 contraction, 66 contrast, 4, 10, 108 contrastive focus, 108 Contrastive Left Dislocation, 110 contrastive topics, 107 deny-confirm condition, 116 deontic modality, 91 derivational morphology, 58 development, 3 untypical, 3 discourse-specific intepretation, 125 Elsewhere. See Kiparsky's

Elsewhere Principle epistemic modality, 86, 91 EPP features, 40, 41, 45 existential meaning, 6 formal features interpretable, 5, 9, 37 uninterpretable, 5, 32, 41 Full Interpretation, 37, 132 functional categories, 5 generic interpretations. See genericity genericity, 11, 122, 140 German modal verbs, 83 Germanic languages, 3, 7, 9 half-modals, 88 Information Structure, 107 interpretable features, 37 Japanese (language), 58 Japanese quantifiers, 66 Kiparsky’s Elsewhere Principle, 130 L2 acquisition, 5 Last Resort, 40 left periphery, 113 lexeme-based analysis, 57 lexeme-based morphology, 74 lexemes, 57, 74 lexicon, 5, 7, 57 linguistic interfaces, 4 syntax-pragmatics, 5 syntax-semantic, 32 linked reading effect, 147 loanwords, 162 logical form (LF), 4 maximality interpretation, 7 Minimal Link Condition, 41, 42 Modern Greek (language), 105 morphemes bound, 57 free, 57 morphological boundness, 57, 58, 62 narrow focus, 108

182 narrow syntax, 4 negative polarity, 34, 108 neg-raising, 35, 41 neg-whQs, 32 object, 31 neg-wh-quantifiers, 35 neutralization, 163 new information, 4 Nominal Mapping Parameter (NMP), 7, 10, 11, 21 non-atomic ș-roles, 126 non-canonical stress positioning, 170 non-existential interpretation, 42 non-obligatory control (NOC), 123 non-sloppy reading, 61 nouns count, 12 mass, 12 proper, 14 NPI. See negative polarity NPI licensing, 32, 44 null subject. See pronominal subject object quantifiers, 34 obligatory control (OC), 121, 122 onomastics, 161 overt QR. See overt quantifier raising overt quantifier raising, 31, 34, 36, 38 parameters, 5 periphrasis, 90 phonetic form (PF), 4 phonological adaptation, 159 Portuguese (language), 159 Archaic Portuguese (AP), 162 Brazilian Portuguese (BP), 162 European Portuguese (EP), 166 Portuguese phonology, 159 predicates, 8 PRO, 121 pro-drop parameter, 4 pronominal clitics, 5 pronominal subject, 4 null, 4

Index overt, 4 proper nouns, 160 proper nouns of foreign origin, 160 quantificational DPs, 68 quantifiers existential, 36 non-existential, 33 object, 34 strong, 32, 34, 36 question-tagging, 98 reflexivity, 62, 65 rheme, 109 Romance languages, 3, 9 saliency, 144 scrambling, 70 sentential negation interpretation, 42 sloppy identity, 64 spelling adaptation, 169 split antecedents, 123 statue interpretation, 61 stress patterns, 163 stripping, 106 strong quantifiers, 36 subjacency, 43 subject position, 16 successive cyclic movement, 38 theme, 109 toponymy, 161 uninterpretable features, 37 Universal Grammar (UG), 62 universal quantifiers, 67 verbal nouns, 59 verb-phrase ellipsis, 106 WCO cancellation, 32, 44 WCO effects, 47, 68, 70 wh-quantifiers, 31 word order OSV, 40 SOV, 40, 43 Yes/no ellipsis, 105 ș-role, 121 ș-role copying, 122 ș-role percolation, 122, 128, 131, 132, 152